Fixed L-Pads

[dmtp]

I have a somewhat general question about L-Pads with fixed resistors. I am still working on my LE14A/LE85 project. I measured the relative sensitivity of the two drivers and find the LE85 about 19dB higher than the LE1`4A (no fixed L-Pad, 16 ohm variable L-Pad set @ 12 o'clock). The JBL L200B network (which I am using as a starting point) uses an L-Pad composed of 2.5 ohm series and 5 ohm parellel. This does not provide enough attenuation for my horn. In researching options, I have found two resources with different options:

The pi speakers article on cross-overs which was excellent http://www.pispeakers.com/Speaker_Crossover.doc shows the 'L' backwards from the JBL (which I believe is standard) version with the series resistor coming AFTER the parellel. It also tends to have a series resistor larger than the parellel one. Unfortunately, they do not give formulae to calculate values.
Vanc Dickerson's Cookbook gives the formula for a fixed L-pad. Unless the labels of R1, R2 are swapped (which is why I refer to parellel and series rather than 1-2), the series resistor comes out much larger than the parellel one. (For 16 ohm, 16dB attenuation, the numbers are sereis= 13, parellel = 3) I tried the L-pad as calculated in the cookbook and the attenuation seems about right, but I am wondering what advantages there might be to using the pi speaker approach (they claim increased damping of the horn) and how one might calculate values. Also I wonder that JBL uses a small series resistor and larger parellel resistor in most of their designs.
Finally, pi speakers bypasses their series resistor with a small cap (which varies between 0.47 and 5 uF). Not sure how this value is calculated. The L200B accomplishes this bypass with 1 uF cap and 0.16mH coil and a fixed 5 ohm resistor. Does anyone really understand how this circuit works or how the values are calculated? I would like to add a little more high end boost (above 10k) but at this point it would be blind guessing at alternative values.
Thanks for all your help. This forum is really great!

[Robh3606]

Hello Mark

If I remember correctly you use a resistor value that 2x the drivers to help minimize impeadance changes that the networks see's. As far as the bypass circuit it was used in the 4430/4435 as well. If you look, it is connected at a point before any attenuation takes place. You have a capacitor, inductor and resistor in series. You get a resonant circuit with the resistor changing the circuit Q. With no resistance you get max output. The higher the resistor value the more you drop and broaden the peak just like the Q/Width adjustment in a parametric EQ. I would go with the L200B network values to start and you could put another potentiometer in series, only connect 2 legs, adjust it for the best balance. Then you can either leave it in so you can adjust the level like the 4430 or measure the value and drop in a fixed resistor.

Read this it will help explain things better than me. It's a great reference. The Profile for the 4430/35 also explains the HF Bypass as well.

http://www.audioheritage.org/vbulletin/showthread.php?p=76021#post76021

http://www.audioheritage.org/html/profiles/jbl/4430-35.htm

Rob:)

[Zilch]

Click #122 in the right column here:

http://www.bcae1.com/

Are you trying to eleminate the adjustable L-Pad, or just want more attenuation before it?

Either case will require knowing the actual impedance of the component combinations to calculate accurately and retain the original conditions.

*****

See Fig. 7.155 and related text in Dickason for what the bypass cap does in filter contouring.

The LCR resonant circuit is tuned like a notch filter to establish a more specific range of frequencies to be bypassed. The calculations should be similar. Instead of shunting a notch to common, you're passing it on to the driver.

I've only recently come to understand that the resonance of that LCR generates some amount of actual boost. What the limits of that are, and how useful it is, I don't know.

The fundamental compensation comes from contouring the differential SPL available from the HF driver/horn combination, attenuating more it at the lower frequencies.

******

I'll re-read the Pi Speakers documents. Been a while. Maybe I'll "get" more of it this time.... :thmbsup:

[Earl K]

Finally, pi speakers bypasses their series resistor with a small cap (which varies between 0.47 and 5 uF). Not sure how this value is calculated. The L200B accomplishes this bypass with 1 uF cap and 0.16mH coil and a fixed 5 ohm resistor. Does anyone really understand how this circuit works or how the values are calculated? I would like to add a little more high end boost (above 10k) but at this point it would be blind guessing at alternative values.

- A capacitor strapped across a series resistor ( the resistor is used to attenuate the mids below the reactive impedance of the cap ) forms a HF shunt around the resistor. The cap, allows frequency specific signals to flow around the resistor ( signal flow increases with an increase in frequency ).

An example ;

Stick a 16 ohm resistor in front of a 16 ohm driver and the total resistance the amplifier now "sees" is 32 ohms. This higher resistance will limit/reduce all signal flowing into the driver .

Add a 1 uF cap paralleled to the 16 ohm resistor and you have formed a "HF signal bypass", routing directly to the driver . This "HF bypass" will become ineffective once the sum of the reactive capacitance ( measured in ohms ) along with the drivers' impedance, becomes greater than the sum ( of resistance of the inline/series resistor added to the drivers' impedance ) . This is because voltage, like water, follows the path of least resistance . Once the total resistance of the capacitor "bypass" becomes more than the resistance of the simple resistor, then signal will "prefer" to flow through that route.
- ( I'm using "sums" here because only then can one use "vector addition" to properly account for the phase rotations that are inherent within Xc . These phase differences prevent the straight addition of the impedances / though I've conviently ignored this fact for the purpose of these demonstrations )

Xc = 1 / ( 2 * pi * f * C )

Xc is stated in ohms'
Frequency is stated in hertz
C is stated in Farads ( so divide uF by a million to restate your cap value into Farads )

Also , from the above formula ;

f = 1 / [ 2 * pi * ( Xc + R ) * C ]

R = drivers resistance
Solving for "f", means a 1 uF cap strapped overtop of a 16 ohm resistor and feeding a "real" 16 ohm driver will start conducting signal around the 16 ohm resistor, somewhere around 4974 hz .

If you think you know where you want the HF boost to start, one can solve for C by using ;

C = 1 / ( 2 * pi * Xc * f )

- In reality, matching the boost point exactly to where the horns' frequency response starts to droop is pretty tough.

- In my experience ; it's easier to "oversise" the bypass cap by a factor of 2 ( thus lowering the boost point ). Then carve out ( all the excess midrange that sneaks in with this larger cap ) by using a fairly broadband ( series LCR strapped in parallel ) that is targeted solely at the excess midrange content ( while leaving the HF content ) . This approach will usually allow for a couple of extra db in the HF frequencies . Another additional bonus to this approach is that the large midrange frequency area effected by the LCR notch has "frequency specific damping" applied to it ( while leaving the damping of the HF & UH,F untouched . This is a voicing trick that helps one think he has a tweeter in the system. :D


,,, snip Does anyone really understand how this circuit works or how the values are calculated? I would like to add a little more high end boost (above 10k) but at this point it would be blind guessing at alternative values.

- The N200b , HF bypass that you have asked about has an extra wrinkle from the simple RC bypass. That 0.16mH coil is in place to form a resonant boost circuit ( a series LCR , strapped in series with the load ) . My calcs. show the "boost" to be centered around 12,600 hz . The "Q" of the boost is indeterminate to me , because I'm not sure what value to use as "R" in the denominator. Using a "pure" 5 ohms in the denominator gives a higher "Q" boost as compared to using the "loop" impedance formed from the driver back to the 2 conjugate resistors ( 1 conjugate being the ground leg of the variable Lpad ) .
- The use of these resonant LCR circuits to achieve boost in certain frequency bands seems to have fallen from favour ( with JBLs' network designers ). The 4430 used this same ( resonant LCR ) approach / but that circuit is now 25 years ??? old and was the last example of its' genre that I can find . A study of Greg Timbers network designs will show that he alternatively, uses other types of "compensations" to achieve "extra" HF boost.

regards

[Zilch]

-This is a voicing trick that helps one think he has a tweeter in the system. :D[Heh, heh.... ;) ]

[Earl K]

Are you *** Mark*** trying to eliminate the adjustable L-Pad, or just want more attenuation before it?

Either case will require knowing the actual impedance of the component combinations to calculate accurately and retain the original conditions.

- Zilch, why don't you do Mark a favour and measure the "working circuit" impedance of a le85 with everything "driver-side" of the fixed Lpad ( ie; none of the 3 pole/passives in place ) ? ( You could have done that for Todd when he was looking for help ;) It would have nicely rounded out the study of that circuit )

- Depending on the actual rotation of the variable Lpad , one might have a ( MF and lower ) load of ( say ) 15 ohms ( from the comp driver/conjugate & variable Lpad ) with the 5 ohm conjugate ( parallel ) with a 2.5 ohm buildout ( series ) resistor. That gives me a working impedance of around 6.25 ohms (to be used by the 3 passives working in the crossover region ).

- Once the actual working impedance is known for that part of the circuit / then a new fixed Lpad ( with more attenuation ) can be designed that still mantains the "correct" circuit impedance .


:)

[Zilch]

O.K., tonight, probably.

I calculated 6.3095 Ohms, so we're close....

What we're talking about:

[Earl K]

What we're talking about:

- Almost! Run a WT2 impedance study of the network / without the 3 passives in place.
- Ie ; get rid of ( for the impedance test ) ; the 16.5 uF and 24 uF caps, plus lose the .8mH inductor .

- The "HF compensation" bypass network is also superfilous to the needs of this exercise .

<> :)

[Zilch]

Yup. I get it. :thmbsup:

[Might use CLIO, tho, depending on how complicated it becomes.... :p ]

[Earl K]

Does anyone really understand how this circuit works or how the values are calculated?

(a) The resonance point for your series LCR is found by ;

Fo = 1 / 2 * Pi * [ squareroot of ( L*C ) ]

Fo is expressed in hertz
L is expressed in Henries ( so divide a mH value by one thousand )
C is expressed in Farads ( so divide a uF value by one million )

example;

Fo = 1 / 2 * Pi * [ squareroot of ( 0.00016 * 0.000001 ) ] is
Fo = 1 / 2 * Pi * [ 0.000012649 ] is
Fo = 1 / 0.000079477 is
Fo = 12, 582.30303 hz

(b) The "Q" of a series LCR is found by this equation ;

"Q" = [ squareroot of ( L / C ) ] divided by "R" ( or the circuit resistance ) .

"Q" is expressed as a number
L is expressed in Henries ( so divide a mH value by one thousand )
C is expressed in Farads ( so divide a uF value by one million )
R is expressed in Ohms

example;

[ squareroot of ( 0.00016 / 0.000001 ) ] ÷ R , is
[ squareroot of ( 160 ) ] ÷ R , is
[ 12.64911064 ] ÷ R ,
if "R" = 5 ohms then "Q" = 12.64911064 ÷ 5 giving a resonant boost "Q" of @ 2.53

12,583 hz divided by this "Q" of 2.53 gives a filter bandwidth ( with 3 db down points ) , 5080 hz wide ( @ centered at 12,583 hz ) .

(c) Mark , theoretically one could choose a different Fo ( center ) point for this resonant LCR circuit / but that would require choosing a different capacitor value and then a different coil value .
- The capacitor value would need to be chosen first , since the capacitor has the dual job of determining the lowest frequencies allowed through the LCR bypass & "boost" circuit .
- The coil value is then chosen with 2 other criteria in mind .
(i) That coil is still a coil and after the resonance of this circuit has died down , the coil will actually start to attenuate the HF going through this circuit . So one needs to keep that in mind
(ii) An appropriate value coil must be chosen that will cause a resonance with the previously chosen capacitor at the desired frequency .

That formula is

L = 1 / [ 4 * pi^ * frequency^ *C ]

C is expressed in Farads
^ is a short form for "squared"



I would like to add a little more high end boost (above 10k) but at this point it would be blind guessing at alternative values.

- See above if you really want to get into it . There's a bit more math to learn before you can get started .

- Personally, I'd get rid if the 5 ohm resistor and the .16mH coil in that circuit and then take note of what happens.
- Going to a capacitor that is valued in the range of 1.5 to 2 uF , will move the F3 point down to 8490 or 6366 respectively . You'll then need to adjust your notch filter accordingly / once you start allowing more MF to HF content into the bypass circuit . The good thing here is , that you've already played around with notch filters / so / correcting anamolies ( in the presence range ) with that notch filter won't be such a daunting task .

- I'd also "graduate" to something more accurate than a RS SPL meter when measuring relative levels. IME, they aren't very reliable beyond 10 or 12 khz .

:)

[dmtp]

Click #122 in the right column here:

http://www.bcae1.com/
Great little tool! It comes up with the same 13 ohm series and 3 ohm parellel that the formula does (but much easier). Assuming the 13 ohm is "towards" the input, that will give 16 ohm to the preceding circuit, what happens when you put it "after" (or toward the horn) ala pi speakers? Obviously need different numbers!

See Fig. 7.155 and related text in Dickason for what the bypass cap does in filter contouring.
Ooops! I have 5th edition and the highest figure is 7.112!

[dmtp]

Wow! Earl K really DOES understand this circuit! The numbers certainly seem good - boost from 10k-15k. Of course, I put a variable resistor (1/2 L-pad) instead of the 5 ohm so I guess that as I turn that up (lower resistance), the boost gets higher, but the bandwidth gets narrower, right? (I think this is part of that "no free lunch" thing??)
I'll try the same circuit with just the 1 uF and see what I get.

[dmtp]

(a) The resonance point for your series LCR is found by ;

Fo = 1 / 2 * Pi * [ squareroot of ( L*C ) ]

Can I use these same formulae to calculate values for a series LCR to put across the voice coil to attenuate the med freq as Earl K was suggesting?

[Earl K]

Of course, I put a variable resistor (1/2 L-pad) instead of the 5 ohm so I guess that as I turn that up (lower resistance), the boost gets higher, but the bandwidth gets narrower, right? (I think this is part of that "no free lunch" thing??)

You've surmised correct . Reducing "R" gives a higher Q resonance / resulting in less bandwidth of boost at the centre frequency ( boost frequency ) .

:)

[Earl K]

(a) The resonance point for your series LCR is found by ;

Fo = 1 / 2 * Pi * [ squareroot of ( L*C ) ]


Can I use these same formulae to calculate values for a series LCR to put across the voice coil to attenuate the med freq as Earl K was suggesting?

Yes, you can, but ;

- First you need to determine what value of "R" will give the correct amount of attenuation ( for the desired frequencies ) within that part of the circuit .

- That means learning how to make "AC impedance measurements" for that specific part of the circuit . I call that , measuring the "Loop Impedance" .
- How to "Take AC impedance Measurements" will be covered in your "Loudspeaker CookBook " ( somewhere ) . You should plot out the impedance curve onto graph paper .

- To calculate "R" you need to know a couple of things ( apart from the loop impedance ).
(i) The "Q" of the bump you are trying to flatten.
- You need to measure and then plot out the FR of the whole horn circuit / so that you can easily "see" the centre of the "bump" along with the bumps' 3 db down points . Once they are identified; determine the bandwidth of the bump by subtracting the F"low" ( low frequency point that approximates a 3 db down point from the highest peak ) from the F"high" point ( same criteria ).
- These 2 points need to be approximately centered on either side of the bumps peak .
- Divide this number into the frequency of the peak and you'll get the "Q" of bump.

example ;

- There is a 5 db high peak centered at 6500 hz . The upper 3 db down point on this bump occurs around 10000 hz . The lower 3 db down point occurs around 3200 hz . 10000 minus 3200 = 6800 . 6500 ÷ 6800 = .9559
- The "Q" of the bump is @ .96 . Slightly less than 1 . A "Q" of 1 equates to an octave wide bump.

(ii) How many db of cut are needed to flatten the bump.

- The good visual response chart ( made or obtained from an RTA ) tells one that 5 db of cut are needed ( hypothetically, within this example ).
- Lets say that you measured a loop impedance of about 8 ohms for this frequency area. Then the question becomes what value of "R", in parallel with 8 ohms, gives one 5 db of attenuation in the "bump" area . The answer is @ 10.3 ohms. I would start with 10.5 ohms as the "R" in my series LCR . Zilch has a nice simple formula ( if you need it ) for determining db attenuations ( when only the load resistance is known / as well as the desired attenuation in db ) Hopefully he'll share it and save me the typing . :D
- The you need to multiply this derived "R" by the "Q" of the bump . The product of these two will give you a number ( slightly larger than 10 in this example ) that represents the electrical "Q" of the LC filter that you need to build .
- You can get an idea of the coil size by this formula ;
L = ( our newly derived ) Product ÷ 2 ÷ pi ÷ frequency ( the Fo )
L is an answer in Henries ( multiply by 1000 to get mH )

Example ;

10 ÷ 2 ÷ pi ÷ 6500 = 0.000244854 Henrys ( Henries ? ) or 0.2448 mH . This is the ballpark size of the coil needed for this "Q" of notchfilter to give 5 db of attenuation within an 8 ohm circuit loop . I'd start the design process of my LCR notch by using a .25 mH coil .

To find "C" use this formula ;
( again, ^ is used as a marker to denote that "something" is squared )

C = 1 / [ 4 * pi^ * freq.^ * L ]

C is stated in Farads
L is stated in Henrys
( pi^ = 39.4784176 )
example ;

C = 1 / [ 4 * 39.4784176 * 6500^ * .00025 ]
C = 1 / [ 4 * 39.4784176 * 42250000 * .00025 ]
C = 1 / [ 416990.7859 ]
C = .000002398 Farads, ( multiply by one million to get an answer in uF ).
C = 2.4uF

So; a 5 db cut, centered at 6500 hz, about 1 octave wide, in an 8 ohm circuit, should be obtainable with an LCR made up of a 10.5 ohm resistor / a .25mH coil / and a 2.4uF capacitor / all wired in series to each other and wired in parallel across the driver / after some inline padding of sorts ( such as an Lpad or a simple inline resistor ) .

:)

[Earl K]

,,,, snip,,,,The JBL L200B network (which I am using as a starting point) uses an L-Pad composed of 2.5 ohm series and 5 ohm parellel. This does not provide enough attenuation for my horn. In researching options, I have found two resources with different options:

- Forget the "pi" forums' option for now. Reversing the order of the buildout resistor to the conjugate ( parallel ) will only confuse you at this stage . The 20 ohm resistor that is already across the driver provides quite a lot of broadband damping .

- The existing fixed Lpad provides around 4.5 db of attenuation .You can make up a fixed Lpad ( with slightly more than 10 db of loss ) by using a 2.5 ohm parallel resistor , preceeded by a 4 ohm buildout resistor . ( 4.1 or 4.2 ohm buildouts ) would be better if you can conjure them up .

- In case you haven't realized it yet, the working load impedance of your network that your CLC passives look into, is somewhere in the area of the low 6s' ( 6.25 to 6.5 ohms ). Zilch is working on that answer right now .
- It's not 16 ohms .
- The inline fixed Lpad drops the working circuit impedance :o:

:)

[Zilch]

http://audioheritage.org/vbulletin/showthread.php?p=104615#post104615

I'll measure stuff after dinner. Made a run to the City to meet up with a forum member. :thmbsup:

[THAT had priority.... :yes:]

[Zilch]

With PE 16-Ohm 50W L-Pad.

Z800:

L-Pad -2 dB 17.36 Ohms
L-Pad @MID 14.78 Ohms
L-Pad +2 dB 11.68 Ohms

If we take "MID," it's calculated to be 6.2461 Ohms.

[That was a LOT of work just to prove Earl off by 0.0039 Ohms .... :p ]

[Zilch]

Ooops! I have 5th edition and the highest figure is 7.112!You KNOW what you gotta do, then.... :D


Can I use these same formulae to calculate values for a series LCR to put across the voice coil to attenuate the med freq as Earl K was suggesting?Zilch will take credit for suggesting that @ #3.

[Earl won't mind one whit.... :p ]

[Robh3606]

Looking through this what horn are you using???

Rob:)

[Earl K]

Hi Rob,


Looking through this what horn are you using???

Assuming you're talking to Zilch ;

- The answer to your question is in the header portion of each graphic .
- ie; It's a H91 .

<> :)

Thanks Zilch !

PS ; Unfortunately ( for the sake of clarity ) these impedance graphs were made without the "fixed" 4.5 db Lpad in place . The average 15 ohm value used , ( which Zilch refers to ) is used to calculate the working circuit impedance.

ie; Adding a 5 ohm parallel resistor and 2.5 ohm buildout resistor onto the 15 ohm load ( amplifier-side of the variable Lpad and driver/horn combo ) gives the lowish 6.xx ohm value that was previously referenced as the working circuit impedance. ( This 6.25 ohm value is the impedance value that must be used to correctly calculate the values of the hipass network .)

[dmtp]

Looking through this what horn are you using???

I am actually using a homemade 500Hz cut-off Tractrix horn.

[dmtp]

- In case you haven't realized it yet, the working load impedance of your network that your CLC passives look into, is somewhere in the area of the low 6s' ( 6.25 to 6.5 ohms ). Zilch is working on that answer right now .
- It's not 16 ohms .
- The inline fixed Lpad drops the working circuit impedance :o:
Yes, I had just (independently!:D) figured that out - that explains why none of the values made any sense when calculated against 16 ohms! Any idea (or speculation) on WHY JBL would do that? It would have been just as easy to create fixed L-pad that maintained 16 ohms.

[dmtp]

Referring to Zilch's graphs (Thanks for all the work - this thread is really turning into a textbook on XO design!), I see the Le for the LE85 is a NEGATIVE number in each case. How do you get a NEGATIVE Le? Is that, like an imaginary coil?:blink:

[Earl K]

,,,, snip ,, Any idea (or speculation) on WHY JBL would do that? It would have been just as easy to create fixed L-pad that maintained 16 ohms.

- My speculation ( for one reason ) is that by changing impedances within separate portions of the same network, JBLs' crossover designers, can economize on the overall price of parts.
- Ie; Smaller Coil sizes can be used when "passives" are working into lower impedances ( though cap sizes do need to go up ). All in all , the price of copper ( by the pound ) likely overshadows the extra cost of the necessary larger size ( Mylar or Electrolytic type ) caps .

- There may also be a "damping" preference . I know that a low value parallel resistor does offer more circuit damping ( at least to my ears ) when used with a compression driver .


:)

[Earl K]

Yes, I had just (independently!) figured that out - that explains why none of the values made any sense when calculated against 16 ohms! Any idea (or speculation) on WHY JBL ,,,,, snip, snip

(A) Okay, now that you realize that the actual load impedance ( that the crossover portion of the N200b ) "sees" is actually in the neighbourhood of 6.3 ohms / calculate ( through your text-book formulas ) what the F3 point is for the following ;

(i) 16.5 uF cap = gives an F3 of ??? hz into a 6.3 ohm load
(ii) .8 mH coil = gives an F3 of ??? hz into a 6.3 ohm load
(iii) 24 uF cap = gives an F3 of ??? hz into a 6.3 ohm load

(B) Then calculate what the textbook values are ( caps & coils ) of a 3-pole Butterworth network ( HP portion only ), crossing at 800 hz / or 900 hz or 1000 hz ( ??? ) .

(C) Once you've finished point (B) , do the same exercise of calcs. for F3 ( as in point "A" ) for the new values of "textbook-derived" passives .

;)

[Earl K]

Hi Mark


The L200B accomplishes this bypass with 1 uF cap and 0.16mH coil and a fixed 5 ohm resistor. Does anyone really understand how this circuit works or how the values are calculated? I would like to add a little more high end boost (above 10k) but at this point it would be blind guessing at alternative values.

- Okay, back to this original post .

- Increase the value of the 1uF cap ( in the LCR bypass circuit ) to as high as 3uF . During this exercise ( in empirical study ), bypass your 0.16mH coil with a piece of wire. Your variable pot. will effect the F3 point of the cap in question when you rotate it . Increasing the cap to as high as 3 uF will likely mean you will need to broaden/deepen your 6K or 7K notch filter to get rid of excess level in this area .

- Once you have done this, and are happy with the amount of extra Hf your getting / then feel free to add back in an appropriate resonating coil in your series LCR. Be warned though ; to maintain a 13K center point for your boost ( with a 2.5uF cap , for instance ) leads to a tiny coil . @ 0.06 mH for a 2.5uF cap . If the R = around 5 ohms, the "Q" of this boost is around 1 octave wide. Such a low "Q" will likely result in little real boost / but go ahead and try it .

:)

[Earl K]

Hi Mark,

The next trick to getting a tiny bit more HF & UHF into that horn is to create a ;
"Frequency-Dependant, Fixed Pad ".

How to Construct this sort of pad ?

- It's simplistic to implement. One just buries a small inline coil ( JBL likes to use .2mH ) underneath the parallel resistor of a high loss , fixed Lpad.
- One end of the coil is connected to the parallel resistor / while the other end of the coil is connected to the common / ground buss .
- The 10 db pad I suggested you build will be a starting point / though 15 db pads are better for this exercise .

How does it work ?

- Any coil builds up Inductive Reactance ( resistance ) as AC frequency is increased.
- The formula is as follows;
XL = 2 * pi * L * Frequency

L is stated in Henries
Frequency is stated in Hz .

- This ( inductive ) resistance is "added" ( though it's not a straight addition ) to the value of the parallel resistor ( in the fixed pad ). This higher value conjugate ( parallel ) resistor, essentially lessens the overall attenuation of the pad, with increasing input frequency. At 10K, ( and if the addition was not vector based ), we would add 12.6 ohms to the value of that 2.5 ohm parallel resistor ( found in the 10 db pad ).

- So roughly , at 10K the pad is attenuating only 3.8 db ( versus the original 10 db ) . 15K would be around - 3.4 db .
- At 5K, attenuation is around 4.77 db ( versus the original 10 db design of the straight pad ).
- At 1250hz , attenuation, calcs out to be - 7.16 db .

As stated, the above numbers are misleading, because I haven't used vector based addition to arrive at the real value for the combined resistance of the coil and resistor combo . The real resistance will be somewhat less .

Anyways, this is a "cheap an cheerful" way to steal a few more db back for the horn driver ( when compared to 1000 hz ) .

The down side ? The per octave db increase is small; Here about a single db per octave. Though a comparison of 15K compared to 1K looks better since it's a gain in the 4db range.
- The other downside is that this slight increase in HF also pulls with it a bit more midrange / which will likely need to be further notched out by your midrange LCR "trap" filter .

:)

[Robh3606]

The other downside is that this slight increase in HF also pulls with it a bit more midrange / which will likely need to be further notched out by your midrinage LCR "trap" filter .

Why would you use an LCR trap in this case with this driver and horn combo??? Why wouldn't you use a simple series cap to do the midrange attenuation to flatten the curve above the mass break point?? These are not like the newer horns and drivers where you need to use LCR traps to do basically parametric EQ on the driver horn combo. Are Tractrixs???(spelling??) CD type horns or a mix between Exponential where the directivity changes with frequency so the EQ is not as straight forward as a true CD horn????

Rob:)

[Zilch]

Unfortunately ( for the sake of clarity ) these impedance graphs were made without the "fixed" 4.5 db Lpad in place.Hadta scrounge up the correct R's this morning.

Z800 = 6.2417 Ohms, measured.

The impedance the actual filter faces is quite stable with this topology. Compare to that of just the driver and horn in #18, above, top.

Le is calculated at 1 kHz, and the impedance is dropping at that frequency, so it's negative.

Le at all sampled frequencies is given in the WT2 full sweep data file.

[Seems no matter what, Fs remains the same.... :p ]

[dmtp]

(B) Then calculate what the textbook values are ( caps & coils ) of a 3-pole Butterworth network ( HP portion only ), crossing at 800 hz / or 900 hz or 1000 hz ( ??? ) .
If my calculations are right, a 3rd order Butterworth @ 1000Hz =
C1 =16.84 reaonabley close to 16.5
L1 = 0.75 reaonably close to 0.8
C2 = 50.52 about 2x JBL's 24!
Now what does THAT do???

Taken separately, (based on 1st order Buterworth)
16.5uF ~ 1500
0.8 mH ~ 1250
24 uF ~ 1050
but I really don't know what any of this means:blink:

[Zilch]

Why would you use an LCR trap in this case with this driver and horn combo???Hi, Rob!

It was in the precedent thread on this project, a "broad peak at 7 kHz."

http://audioheritage.org/vbulletin/showthread.php?t=12365

The neato horn is shown there, too.... :thmbsup:

N200B voltage drive curves are here:

http://audioheritage.org/vbulletin/showthread.php?p=43157#post43157

[dmtp]

- That means learning how to make "AC impedance measurements" for that specific part of the circuit . I call that , measuring the "Loop Impedance" .
- How to "Take AC impedance Measurements" will be covered in your "Loudspeaker CookBook " ( somewhere ) . You should plot out the impedance curve onto graph paper .

OK, I know how to do impedance measurements and set up a little switch box to do it. And I even graduated from graph paper to computer plots. Using a spreadsheet, I used 10^x to get a logarithmic frequency scale in 1/6 octave steps (1000, 1100, 1300.....); measure the impedance at each step and then plot a graph of x=freq, y=ohms which looks like it were plotted on standard semi-log paper (which BTW I cannot find ANYWHERE to buy!)
I measured the impedance of the LE85/horn and found a peak of 50Ohms @ 400, down to 16 ohms @ 200 & 560, fairly flat for a bit then sloping down from 16 @ 2200 to a value of 2.7 @ 20k. (Hmmm, is this falling impedance with rising frequency the NEGATIVE Le in Zilch's data? Normally the Le of a woofer causes impedance to rise with increasing freq.???)
Anyway, I'm not sure what parts of the circuit to include in my measurements for calculating the 'loop impedance'. Do I start at the amplifier end of the circuit and include the 18dB XO? or do I skip over that and just measure the L-pad and bypass?

[Earl K]

Hadta scrounge up the correct R's this morning.

Z800 = 6.2417 Ohms, measured.

The impedance the actual filter faces is quite stable with this topology. Compare to that of just the driver and horn in #18, above, top.

Ahhhh, Most excellent !!!

Thank-you Zilch !


**** Finally, this forum has good graphical evidence of the levelling effects ( to impedance ) that the fixed pads give to those older, wonky variable types. ****

Many Thank-Yous !

:)

[Earl K]

OK, I know how to do impedance measurements and set up a little switch box to do it. And I even graduated from graph paper to computer plots. Using a spreadsheet, I used 10^x to get a logarithmic frequency scale in 1/6 octave steps (1000, 1100, 1300.....); measure the impedance at each step and then plot a graph of x=freq, y=ohms which looks like it were plotted on standard semi-log paper (which BTW I cannot find ANYWHERE to buy!)

- Good Stuff Mark !

- Okay, since you're going to plot it within a computer program / try to do so as both a linear plot and a logarithmic plot .

- The reason ? In a linear plot, the in between values are much easier to approximate . Here's an example of a fairly "tough" to read log plot . "Tough" in the sense of approximating values that fall in between the recognizable hash-marks . Approximating the 3 db down "skirt" values from this plot / is a "bitch" . This plot is from "Project May". Even my estimation of the "peaks" of the "bumps" don't correlate very well with the official numbers stated .
- A linear version of this plot would be easier to read / and therefore identify the peaks ( through sight ) .


:)

[Earl K]

Anyway, I'm not sure what parts of the circuit to include in my measurements for calculating the 'loop impedance'. Do I start at the amplifier end of the circuit and include the 18dB XO? or do I skip over that and just measure the L-pad and bypass?

- Build the complete hipass circuit / excluding your LCR trap ( the one presently notching around 6000 hz ). Connect the network to the source ( amp ) and the load ( horn/driver combo ) . The amp can be on and idling .

- Measure the ac-impedance of the "loop" ( connecting the test leads ) to the drivers' terminals .


:)

[dmtp]

- A linear version of this plot would be easier to read / and therefore indentify the peaks ( through sight ) .


:)
The other alternative that I do, is once I identify a peak on one of the "standard" frequencies in my plot, I simply run the Freq generator up and down until I find the true peak - even if it isn't one of the freq I would normally plot. I then do the same thing to zero in on the -3dB points. I do this for tuning bass enclosures, but I guess I can do it for the horn too. Probably this weekend before I get to do more measurements.

[Earl K]

Probably this weekend before I get to do more measurements.


Okay!

Post your FR & "Z" plots once you've made them .

:)

[Zilch]

Finally, this forum has good graphical evidence of the levelling effects (to impedance) that the fixed pads give to those older, wonky variable types.I had to go look to see if we kept the fixed attenuation in the "Keeper" crossovers. :blink:

We did.

[Whew.... :p ]

[Zilch]

but I really don't know what any of this means:blink:Look at the measured voltage drives I got for N200B:

http://audioheritage.org/vbulletin/showthread.php?p=94678#post94678

Ignore the UHF, which was separated out (the bypass boost loop removed and lowpass added) to make it three-way in that iteration.

The stock N200B HF sim with bypass HF boost in place is here:

http://audioheritage.org/vbulletin/showthread.php?p=43157#post43157

[Earl K]

If my calculations are right, a 3rd order Butterworth @ 1000Hz =
C1 =16.84 reaonabley close to 16.5
L1 = 0.75 reaonably close to 0.8
C2 = 50.52 about 2x JBL's 24!
Now what does THAT do???

Taken separately, (based on 1st order Buterworth)
16.5uF ~ 1500
0.8 mH ~ 1250
24 uF ~ 1050

but I really don't know what any of this means :blink:

Hi Mark,

- The reason I asked you to do those calcs. is so that you begin to understand that your 3-pole HiPass is not a standard Butterworth type filter. It's best thought of, as a HiPass "bump-filter" that provides a bit a gain before the 18db per octave, attenuation kicks in. The "bump" is achieved through resonance and occurs within the filters'' working / "knee" area.

- Earlier I had commented that your usage of a N200B network was a wise starting point, because I'm guessing your "Tractrix" horn may need a bit of low-mid bump ( EQ ) before crossing over to the le14a. It's just a hunch and since we haven't seen any "raw" FR plots ( of the horn/driver combo ) / I don't know if the hunch is right or wrong .

- Anyways, if you don't need/want this EQ "bump", you can get rid of it by implementing the standard values for a 1000hz, 3-pole Butterworth filter ( ie. essentially, double the value of your 24uF cap ) .

- If you don't believe there's a builtin resonant EQ "bump" / then look at these voltage drives ( courtesy of Giskard and Zilch ) .

:)

[Earl K]

Hi Mark,

- Here are 2 pics of voltage drives that quite clearly indicate that the amount of HF drive-energy, available to the N200b is less ( about 5 db ) when compared to the N3135 ( 4435 ).

- I believe it's that inline 5 ohm resistor that's keeping the N200B from providing the HF levels you are after. Once it's removed ( as well as the coil ) and you increase the size of the 1uF cap to something close to 1.7uF or 2 uF / you should obtain more HF . ( As mentioned earlier; you can resize the coil and reinstall it if you want to try to recreate a "LCR boost" )

- The most significant point of these V-Drives (for me anyways ), is that the N3134/N3135 allows the HF ( in the HiPass section ) to return back to "0" level ( no attenuation ) .

- Notice the "bump" in the hipass section of the N3135/N3134 . ;)

- These voltage drives are courtesy of Giskard ( with N200B, Lpad" data provided by Zilch ) . The second pic is mostly all N200B V-Drive .

:)

[Zilch]

Mark has not yet stated whether his objective is to eliminate the variable MF L-pad or not, i.e., we don't know his purpose in desiring to incorporate more attenuation in the fixed one. Doing that will provide more "gain" to the HF bypass, as well.

The problem I encountered with using the stock N200B crossover two-way was that I could not get enough gain without doing something like that, because just turning the adjustable L-Pad down to achieve it ALSO attenuates the HF boost as its parallel leg approaches zero resistance. See the white curve in Giskard's SIM, immediately above, which is what occurs at 8:00 setting, 7:00 being "off." That's the most compensation "gain" I could achieve with the stock circuit.

I'm suggesting Mark's adaption of the the circuit might best increase the attenuation at the fixed pad, but retain the adjustable one to provide a higher impedance (in the upper portion of its operating range) into which the bypass loop may operate effectively; removing it will force the HF bypass to "see" the relatively low impedance of the fixed pad, presently 5 Ohms, in parallel with 20 Ohms and the driver impedance, presumably to become even lower when additional attenuation is incorporated there.

Further, for safety's sake, I would retain some small fixed resistance in the bypass loop, perhaps 2 Ohms or so, as, under the condition of both controls being set to zero, it will otherwise see a direct short to common, limited only by the DCR of the bypass inductor. N3134/5 accomplishes this isolation (and also isolation of the MF and HF adjustments,) with a series resistance after the MF adjustable pad, another option, perhaps, in lieu of increasing attenuation at the fixed pad.

Note also how the slope of the compensation increases with the increased gain, in particular in the area where Mark is presently using a notch filter. We don't know the FR response of his horn/driver combination, but it's looking to me as if the increased gain achieved via increased attenuation at the fixed pad may also provide more optimum compensation for his setup.

At the time I was using these two-way, I did not want to mess with modifying my factory stock N200B crossovers to make them work with 2344(A) horns for quasi-4430. I had paid good money for them on eBay.

When I subsequently opened them up for other reasons, I discovered they were trash (well, one of them was, anyway) and would require rebuild for any future use, anyway. Thus, the "Keepers," I built from scratch with much improved quality....

[Earl K]

- Good Points about the potential for a short circuit !


Mark has not yet stated whether his objective is to eliminate the variable MF L-pad or not, i.e., we don't know his purpose in desiring to incorporate more attenuation in the fixed one. Doing that will provide more "gain" to the HF bypass, as well.

- True enough /though the "gain" is only "by comparison" .
- I stand by my observation that the HF bypass circuit needs to directly "see" the driver. Your short circuit concerns can be still addressed .


The problem I encountered with using the stock N200B crossover two-way was that I could not get enough gain without doing something like that, because just turning the adjustable L-Pad down to achieve it ALSO attenuates the HF boost as its parallel leg approaches zero resistance. See the white curve in Giskard's SIM, immediately above.

- Yes, I understand this point. The variable Pad ends up attenuating the complete curve in one fell swoop.
- One fix to that is to add resistance directly after the variable Lpad / effectively turning it into a "T" pad . I've done that with a few "bench" projects of this sort . The 3134/5 also uses this approach / remove its' 20 ohm ( inline ) resistor and those networks now have the same "problem" .




I'm suggesting Mark's adaption of the the circuit might best increase the attenuation at the fixed pad, but retain the adjustable one to provide a higher impedance (in the upper portion of its operating range) into which the bypass loop may operate effectively; removing it will force the HF bypass to "see" the relatively low impedance of the fixed pad, presently 5 Ohms, in parallel with 20 Ohms and the driver impedance, presumably to become even lower when additional attenuation is incorporated there.

- Right, the fixed Lpad needs a 15 ohm load (to the driver side of it ) for the HF bypass to actually work ( and for the "Fixed Lpad" to rpovide the correct 6.3 ohm load the preceeding passive components. Therefore the variable Lpad needs to remain ( or at least an inline resistor of similar value ).

- My concern is if the attenuation of the fixed Lpad is increased too much ( and subsequently ) the variable Lpad is opened up too far, as in "wide open" / the reflected impedance the fixed Lpad "sees" will drop down from 15 ohms towards 7.5 ohms ( 20 ohms plus 12 / the conjugate with the drivers' "Z" ) . This downard change in load impedance will bugger everthing that precedes it within the circuit. This can of course be fixed with a simple resistor of appropriate value.



Further, for safety's sake, I would retain some small fixed resistance in the bypass loop, perhaps 2 Ohms or so, as, under the condition of both controls being set to zero, it will otherwise see a direct short to common, limited only by the DCR of the bypass inductor.

- We are on the same page here / except I would add the "safety" resistor directly after the 16 ohm variable Lpad ( creating a "Tpad" ). Altec does this as well / BTW / in the N1201a. ( The 20 ohm conjugate may need to be changed to a different value ). Tinkering in this area still means having to maintain a 15 ohm load ( for the preceding fixed Lpad ) .


Note also how the slope of the compensation increases with the increased gain, in particular in the area where Mark is presently using a notch filter. We don't know the FR response of his horn/driver combination, but it's looking to me as if the increased gain achieved via increased attenuation at the fixed pad may also provide more optimum compensation for his setup.

That's not gain / it's midband attenuation ! :D


At the time I was using these two-way, I did not want to mess with modifying my factory stock N200B crossovers to make them work with 2344(A) horns for quasi-4430. I had paid good money for them on eBay.

When I subsequently opened them up for other reasons, I discovered they were trash (well, one of them was, anyway) and would require rebuild for any future use, anyway. Thus, the "Keepers," I built from scratch with much improved quality....

I also noted, that at the time, you were still staring at that goofy 4LED thingy you called an "RTA" . Like Giskard said " You've come a long way " .

<> :)

[Zilch]

I also noted, that at the time, you were still staring at that goofy 4LED thingy you called an "RTA" . Like Giskard said " You've come a long way " . :)Heh, heh.

Well, it served to immediately convince me that I needed adequate instrumentation if I was going to accomplish anything worthwhile in this.

Thank you again, Mr. Widget, for loaning me my first RTA!!! :thmbsup:

[Zilch]

We are on the same page here / except I would add the "safety" resistor directly after the 16 ohm variable Lpad ( creating a "Tpad" ). Altec does this as well / BTW / in the N1201a. ( The 20 ohm conjugate may need to be changed to a different value ). Tinkering in this area still means having to maintain a 15 ohm load ( for the preceding fixed Lpad ) .Indeed.

Notwithstanding wonky L-Pads, the adjustable wants to "see" 16 Ohms. With the 20-Ohm conjugate, an 8-Ohm series resistor accomplishes that. I don't understand why JBL used 20 Ohms between and 8-Ohm L-pad and and 8-Ohm driver in N3134/5.

In any case, adding that in stabilizes the L-pad (somewhat), and the new "input" impedance measures 6.3511 Ohms at 800 Hz (below).

For the record, I ran impedance curves on four LE85s on H91 horns. The one I have been using for these determinations is the red one (middle). It's not an "oddball."

For convenient comparision, I've reposted the WT2 impedance measurement on it immediately below the CLIO determination (bottom).

[I'd call them, uhmmm, "similar." :p ]

[dmtp]

Here are some FR plots (hope the pdf format works, not sure how to save graphs as jpgs)
Legends:
LE14A XO is close miked in box
LE85 XO is at same voltage as LE14A using standard L200B XO with fixed 2.5/5 L-pad in circuit, 16 ohm L-Pad set to 12 o'clock, and boost variable resistor (1/2 of 16 ohm L-pad) set to 5 ohm
1 ohm is same with variable high pass set to 1 ohm
1 uF is with coil and resistor by-passed
1.5 uF is just a 1.5 cap as high boost

[Earl K]

Hi Mark

- Yes, I can "read" this pdf . Those graphs actually do make some sense, given the scenarios you just tried.
- Though, now I'm wondering somewhat about the overall health of the diaphragms in your le85s / mind you / I have already expressed my reservations about believing an SPL meter from RS ( over 10K ) .

- I'll get back to you later today with an alternative schematic to try out.


- ( I assume these measurements were taken with your 6.5K "trap filter" , "out of circuit ". Correct me if I'm wrong )

<>

[dmtp]

Yes, this was without the trap.
I have impedance measurements for the LE85/horn combo, but I'm not sure what else would be helpful to measurefor impedance.
I have no way of knowing about the RS SPL meter (it does have A and C settings, but I forgot what they meant), on the other hand , my ears give out much above 14k anyway so perhaps it is moot. (Must have been those Moody Blues concerts in my youth!):D

[Earl K]

I have no way of knowing about the RS SPL meter (it does have A and C settings, but I forgot what they meant),

Oy Vay !

- Those "A & C" filter settings have different amounts of LF & HF rolloff ( for reasons of obtaining specific types of noise measurements / for the enforcement of civic bylaws ) .

- Put it on the "C" setting . That gives the broadest frequency spectrum.
- BTW; the "official" C weighting is already 5 db down at 10K ( according to my books ).
- I have know idea what the RS SPL meter is actually "weighted" to.
- This is why "designing" networks with these things as a test mic is a really bogus idea .



on the other hand , my ears give out much above 14k anyway so perhaps it is moot. (Must have been those Moody Blues concerts in my youth!)

- My hearing also plummets at around 14K / mostly age related . I only "sense" pressure beyond that point .


- I'll get back to you later today with an alternative schematic to try out.

- Here's the schematic for you to build from what I hope is mostly existing parts .
- Maybe Zilch would like to run this through his Spice program. ( I really don't mind the help :D )

Some Points to ponder ;

(i) I've included my best guess for the values of the "Series" LCR strapped in parallel . Use your variable pot. to dial in the notch. A .25mH coil can be cludged together from your present .1mH and .15mH coils ( by connecting the two of them in series ). Don't stack them ! the mutual inductance from stacking will give a coil value that is smaller than .25 mH.
- If the "Q" of this filter is way off / you'll see it in new FR measurements. Let me know and I'll help you redesign for a different "Q" .
- This LCR is tuned to 6,366 hz ( Oh by the way ) / mostly because it should be able to use your current parts .

(ii) The attenuation obtained from the fixed Lpad has been increased to over 8 db.
- The 3.8 ohm resistor value can be easily obtained by wiring 2, 7.5 ohm resistors in parallel ( 3.75 is close enough ). This is likely a good idea anyway / to double the power handling of those buildout resistors .

(iii) The 8 ohm resistor that is immediately after terminal 2 of the variable Lpad is a isolation/safety resistor.
- Feel free to replace this value with something between 4 and 8 ohms ( whatever is on hand ) .
- The use of this resistor is mandatory for this newer circuit arrangement .

(iv) The 2 uF "bypass" cap is really just a guess now. Do some empirical studies by substituting in values from 1.5 uF up to 3 uF.
- The larger this cap gets / the more ( excess himid ) you're going to need to cut ( with the notch filter ).

(v)Oh yeh, for the time being, I've left out the "buried" coil under the fixed Lpad that would be needed to create a frequency dependant Lpad .

Let me know how this all goes .

EDIT ; ( some hours later )
- Mark don't build the circuit in this post . Build the last variation below in post # 54 . Zilch published this, this afternoon .
- ( Keep the fixed Lpad the way it presently is, with original values ) .
- Add my series LCR / strapped in parallel from this post ( only if you need to still tame the 6500 hz bump )
- Keep your 16 ohm variable Lpad in place and just add the 8 ohm resistor directly to terminal #2 of the pad .

:)

[Zilch]

Adj = -6 dB
Notch = 10 Ohms

[Earl K]

Zilch


Notwithstanding wonky L-Pads, the adjustable wants to "see" 16 Ohms. With the 20-Ohm conjugate, an 8-Ohm series resistor accomplishes that.

- That's the "R" value that I included in the modified N200b . ;)


I don't understand why JBL used 20 Ohms between and 8-Ohm L-pad and and 8-Ohm driver in N3134/5.

- Actually, the 3134(s)' 7.5 ohm conjugate when paralled with the AC load of the 2425h/2344 horn/driver will average down the AC impedance spikes to around 5.3 ohms ( looking at the big 18 ohm spike seen at around 1100 hz ). - So logically, a resistor value of around 3 to 5 ohms would have made more sense ( if the goal was strictly to provide some resistive isolation and impedance matching for the 8 ohm variable Lpad ).
- But, I think those were just secondary goals with the primary goal being to create the proper conditions for a variable "bump filter" to work from.

- Giskards' Voltage Drives that I have reposted above, clearly show the variable bump filter in some "wild" resonant action .


;)

[Earl K]

Adj = -6 dB
Notch = 10 Ohms

Thanks ! :)

[Zilch]

Well, here's more to consider, using the original bypass:

I like killing it above 20 kHz. ;)

No joke, actually. I suspect part of what makes two-way "boosted" compression driver HF sound nasty is subharmonics of overdriven UHF above what we can hear.

I've thought that since I first saw 311xA and N200b voltage drives.

That's why I'm showin' it out to 40 kHz.

[Zilch's theory #117.... :p ]

[Earl K]

Thanks Zilch.

- It's amazing how effective that LCR boost actually is . Mark, I'd keep it in based on what Zilch just posted.

- I'm pretty sure that ( MarkT ) does need a notch filter of sorts .

- I'm concerned that the circuit ( both variations ) now look like there's too much overall attenuation .
- Is that why you went back to the original values for the fixed pad ?

- I figure the drivers' sensitivities are really only a 12 -14db spread ( when considering the different impedance of the drivers and size of that horn ) .
- What do you think ?



:)

[Zilch]

I think those were just secondary goals with the primary goal being to create the proper conditions for a variable "bump filter" to work from.Also, N3134/5 has no fixed padding. The 20 Ohms is generating the requisite attenuation, it would appear....

[Zilch]

- I'm concerned that the circuit ( both variations ) now look like there's too much overall attenuation .
- Is that why you went back to the original fixed pad configuration ?Yes. I'm showing ~15 dB differential with -6 dB on the adjustable, so at "zero," it'll just be ~9 dB. We're in the range.

While Mark's original question here was how to get the additional attenuation he believes he needs, putting in that 8 Ohms gives ~3 dB more, so I kept the original fixed L-Pad values.

He can easily adjust it there, if required, now that we know the impedance at that point is 6.35 Ohms.

I'm not certain he's going to need a notch, since the midband attenuation slope is steeper, but that's easily added as well.

I'm going to try and resolve the question of how good a mic the RS meter is, also. If it's good, then his driver isn't generating VHF, or his horn is killing it. :(

LE85 plays WAY better up top than Mark measures....

[Ian Mackenzie]

Also, N3134/5 has no fixed padding. The 20 Ohms is generating the requisite attenuation, it would appear....

Zilch, R5 in the schematic is used to trim the level of the driver overall, the values in both 3134 /35 are different, 10R in the 3135 to increase the level to match the 2234's, only a couple of db.

[Zilch]

Zilch, R5 in the schematic is used to trim the level of the driver overall, the values in both 3134 /35 are different, 10R in the 3135 to increase the level to match the 2234's, only a couple of db.Thanks, Ian. I see that now.

These references are to the N3134/5 schematics:

R4 sets the compensation differential, and R5 then establishes its overall balance with the woofer(s). :thmbsup:

[It's NOT merely a gratuitous conjugate.... ;) ]

[Earl K]

I'm going to try and resolve the question of how good a mic the RS meter is, also. If it's good, then his driver isn't generating VHF, or his horn is killing it.

- That'll be very useful .


LE85 plays WAY better up top than Mark measures....

- Yes, I remember ( from the Q&D 4430 & the Widget studies ) . I wonder if Mark has seen all the relevant info on the le85


I'm not certain he's going to need a notch, since the midband attenuation slope is steeper, but that's easily added as well.

- Yes, he'll find that out soon enough . If Mark does needs it, this thread ( on padding ??? :p :blah: ) now contains, virtually all the formulae I know for manipulating LCR filters ( apart from determining what value of "R" should be used ) .

- I'll admit, I was surprised the 2 uF cap extended its "energy adding effects" down so low / well down into the 2K region of the midband ( I guess it was due to that the extra padding that was put inplace ). All along I've been saying the Notch Filter was going to need to deal with some extra energy up in the "presence" frequencies / but I wasn't expecting this much extra energy down lower / a properly designed LCR wouldn't reach that low ( assuming the Q was correctly calculated ) .


:)

[Earl K]

I'm going to try and resolve the question of how good a mic the RS meter is, also. If it's good, then his driver isn't generating VHF, or his horn is killing it.

- Well, I couldn't wait. I just did a rough FR check on one of my RS SPL meters / it's approx @ 6 db down at 10K .

- With such a dramatic HF rolloff, here's no telling what MarkT(s)' network/le85 is really giving him .:(

- Also, I've edited post #50 ( regarding the network build ) .

:o:

[Zilch]

And 6 dB up at 6.3 kHz, looks like.

I don't think Mark needs a notch filter, rather, a good pressure mic for measuring....

[dmtp]

I am overwhelmed by the responses! Thanks for all the help, but it DOES look like my next step needs to be something better to do my FR measurements with! (I hve mostly used the RS meter for LF in the past and it seemed OK.) With fear of getting a lengthy side discussion going (if so should we move it to a new thread?), what would you suggest I consider for my next purchase? Here are a few parameters:

I am not a destitute college student (that was back in the 60's when I built pseudo AR-4a's in my dorm room)
OTH, this is strictly a hobby so I am not ready to shell out $$$$ for pro grade instruments.
I would consider computer based solution, but my computer is a long way from where I work on the speakers (in the basement) so getting everything together would be a lot harder than with separte equipment.TIA for your suggestions.

[dmtp]

To clarify some of the discussions, here are 3 more SPL graphs (unfortuantely still take with the RS meter).. Again, sorry for the pdf. If anyone wants to convert to jpg and post I would be thankful.
Plain is just driver/horn
Conjugate is LR (2.0 mH, 14 ohms) with the 20 ohm parellel resistor
20 ohm is no LR, just the 20 ohm.
Previously posted curves had NEITHER the LR nor the 20 ohm conjugates
I will try to come up with better instrumentation before confusing things further!

[Zilch]

What would you suggest I consider for my next purchase?Behringer UltraCurve Pro DEQ2496, available at Guitar Center, for example, for $250. You might be able to get them to throw in the $50 ECM8000 mic and some cables for that. You'll also need a mic stand with mini-boom, about $35 more.

So, for ~$300, you can get set up with a good stand-alone semi-pro test/measurement system. I believe those who have followed my exploits in these forums will attest it has carried my investigations a VERY long way. :thmbsup:

I have plastered these forums with pics from its distinctive orange display. It's a 31-band equalizer as well as a 61-band RTA, with "AutoEQ," which will show what adjustments need to be made to achieve any desired response curve.

Transcribing the FR curves of driver/horn combinations into crossover optimization software, you can achieve quite good results via simulation. With a little more ($25) input protection hardware, you can measure crossover voltage drives with the same setup, too

I also like the AutoEQ because it lets me hear what the system is going to sound like before I put in the effort to tweak a crossover for it. Sometimes the answer is "Don't bother...." :p

[Earl K]

MarkT;

Read this thread (http://audioheritage.csdco.com/vbulletin/showthread.php?t=8859&highlight=Audio+Test+Software) for some ideas and opinions on test gear. Here are a couple of quickies ;

(i) Hardware ;

- One of the simplest to acquire ( and cheap to boot ) is an older Behringer DSP 8024 ( or the older DSP 8000 ) . These are two channel digital EQ units that happen to also have a decent builtin, RTA section . They "output" sine waves, white noise & pink noise at user selectable levels . They have a builtin preamp ( with phantom power for condensor mics ) on a standard 3 pin XLR connector . They eBay in the low 100s' these days .
- The necessary test mic / can be the inexpensive Behringer ECM 8000 / which is @ $ 50.00 from most musician supply stores .
- The only "problem" with these stand alone hardware units is that "sharing" with others, the FR that is seen on the screen requires taking an actual photo with a digital camera / importing the pic / then sizing the jpeg to fit into this forums size rules / and then of course, posting it . ( Having a software based system , just about always circumvents this tedious task and allows the screen captures to be manipulated in other more meaningful ways ) .

(ii) Software Based ;

- Buy an inexpensive notebook computer / then load it with TrueAudios TrueRTA (http://www.trueaudio.com/) / then get a decent "full duplex sound card" and use the Behringer test mic with an inexpensive preamp to feed the "duplex" soundcard . This is more money / but a better approach in the long run . The 1/6 octave or 1/12 octave versions will be the most useful .

- A slightly more expensive alternative to the somewaht "simplistic" capabilities of "TrueRTA" is ETF(s)' ETF5 Measurement Software. (http://www.etfacoustic.com/) From what I have seen Jack Bouska accomplish with it / and at only $ 150. ( software only ) / it appears to be the bargain of the year ( though I haven't used it ) . Since it has waaaay more capabilities that TrueRTA / expect the learning curve to be longer . The company also sell a test mic/preamp combo that seems to be reasonably priced ( though I don't know any of the specs on the hardware to really comment further ) . Consider getting some private advice from Jack B. about this product .

:)

PS ; When you get something that can capture semi "accurate" FR plots / then we can proceed further . In the meantime , build Zilches last schematic / and a notch filter if you guess that it's necessary. :o:

[Zilch]

Here's your FR curve from the PDF. To post it here, I have used the Adobe "Snapshot Tool" to capture the portion I want to the clipboard, which I then paste into "Paint," and save as JPG for upload here.

Your results look quite good. We now know the peak at ~6-7 kHz and the HF rolloff are likely artifacts of the RS microphone. You need better measurement tools (a "bigger flashlight," as Giskard once suggested to me,) to go much further with this....

[dmtp]

I really need to thank Earl K for measuring the RS meter and preventing a lot of wasted effort on everyone's part. I am now reminded of a previous project - a hexagonal speaker designed to 'match' our bedroom furniture utilizing a 12" floor facing woofer and an MTM arrangement using Altec car speakers (I know they aren't REAL Altecs, but I got them for 10 cents on the dollar when Leachmere went out of business). Anyway, the 4" mids and 1" tweeter came with a fairly sophisticated (for car stereo) XO which I used as a starting point. Using my same RS meter, I found a broad hump @ ~6k and designed an LCR trap to tame it. Again, I used a pot for the resistor to adjust the cut. Funny thing was, in listening to music both my wife and I preferred the XO without the trap. At the time I figured it must have had something to do with room acoustics or whatever, but now I guess it was just that our ears told us the meter was lying but I didn't realize it!

[Earl K]

I really need to thank Earl K for measuring the RS meter and preventing a lot of wasted effort on everyone's part.

You're welcome / but / it actually was Zilch who did the definative measurements that show how "wonky" that SPL meter is and how it just is not up to scratch ( for this sort of design work ) . :D

[Zilch]

When you get something that can capture semi "accurate" FR plots / then we can proceed further . In the meantime , build Zilches last schematic / and a notch filter if you guess that it's necessary. :o:I'm gonna argue against the older Behringer RTAs in comparison to the DEQ2496. They're 1/3 octave resolution (31 band) and just don't provide enough detail. I have them both, and only use the DSP 8024 as a signal generator anymore.

*****

As y'all might have surmised, I built up the crossover last night, and will try to post voltage drives sometime today. It works well, with good adjustability, though there's some interaction between the MF and HF adjustments. I don't have Mark's horn to measure with it, tho, for actual FR results. The "shape" of the compensation contour is gonna matter at that level.... :blink:


I am overwhelmed by the responses! Thanks for all the help....Just havin' fun figurin' this stuff out here.... :thmbsup:

[dmtp]

You're welcome / but / it actually was Zilch who did the definative measurements that show how "wonky" that SPL meter is and how it just is not up to scratch ( for this sort of design work ) .
Sorry about that
THANKS ZILCH!

[Zilch]

MF control at -2 dB (9:30), Mid (12:00), and +2 dB (2:30)
HF control varied Min (16 Ohms), -2 dB, Mid, +2 dB, and Max (0 Ohms)

[Zilch]

HF control at Min (16 Ohms), Mid, Max (0 Ohms)
MF control varied -2 dB (9:30), Mid (12:00), and +2 dB (2:30)

[See the "Q" change?]

[Zilch]

HF control is series leg of 16-Ohm L-pad, i.e., pins 2 and 3.

It certainly makes LE85 sizzle on H91. :p

[Gonna try some CD horns, now.... :yes: ]

Note: This is a virtually fruitless endeavor with H91 horn/lens, in L200B, for example. The vertical dispersion narrows so drastically as to become little more than a horizontal slit of VHF:

[Zilch]

O.K., how 'bout 2344A, then?

Just happen to have a couple of those around. :D

+/- 2.5 dB O.K.?

I'd say there's new options for playing 16-Ohm LE85 most sweetly with this.

Expanded scale showing available HF adjustment range, bottom.


:thmbsup:

[Zilch]

As built:

[dmtp]

I am happy to report that I have today oprdered the Beheringer ECM8000 with a preamp to use with an old (P3) computer I have laying around. Haven't decided between TrueRTA and ETF. It is not so much the cost ($100 v. 150) but the steepness of the learning curve that worries me.
Anyway, while waiting for my new toys, I started playing with the SPL plots making guesstimate adjustments for the RS meter, but focusing on the XO point instead of the high end. (I used the formula 20log((10^A/20)+(10^B/20)) for combined response.) It appears that I will have quite a notch around 600. I remember one thread where the builder "fixed" this by adding a 16 ohm resistor in series with the woofer, but noone thought that was a good idea. Looking at the separate curves, it is obvious that the woofer is rolling off too soon. This appears to be due to the 24uF cap. Keeping the coil @ 2.25, I found three choices for XO freq/cap:

Butterworth, freq = 625, cap = 28 (closest to what we have)
LR, freq = 900, cap = 14
Bessel, freq = 775, cap = 18On the assumption that the hi-pass is a third order Butterworth (somewhat 'tweaked' by the second cap value as already discussed), is there a compelling reason to use a Butterworth alignment (second order) for the woofer, or will L-R or Bessel work as well? I am much more familiar with L-R XOs and it looks like a 900 XO would be a lot better. I think I'll try different cap values and see what I get.
Attached image:
Orange = LE14A with XO as in Zilch's post above
Purple = LE85 with 6300-6db, 10k+10dB and smoothed
Turquoise = same dropped 2 dB
Red - combined

[Zilch]

I've always played it with 2235H. I recall deciding in the end that the 600 Hz thing wasn't real, upon measuring from further back. That's all documented in the "Keeper" thread.

Adding resistance in series with the woofer is a BAD idea. The amp loses control. It's best practice to use a minimum DCR inductor for the filter, even, to keep the damping as low as possible. It just makes no sense to go and add in more.

*****

Bear in mind that it's the acoustic rolloff that ultimately matters, i.e., the driver(s) behavior in combination with the filter. I can run that for you here; plenty of LE14s around. Tell me again what is the box size and tuning you're using with it so I can get close to what you have.

Like you, I worried and fretted about all kinds of stuff relating to that crossover design. In the end, it all just came together and worked with 2235H:

http://audioheritage.org/vbulletin/showthread.php?p=94678#post94678

My analysis was that the driver rolloff combined with the second order filter to make more like a third order rolloff. Note also that there's a bump in the LF lowpass; it's not a textbook alignment. The crossing is at -6 dB. I betcha Earl can advise as to how to tweak that as well, if necessary, once we get the measurements for him.

I'm not understanding your scale on the y-axis there in #77, above. Still 3 dB/div?

[dmtp]

I'm not understanding your scale on the y-axis there in #77, above....
Yeah, I missed that. Somehow it lost the second digit. Major dividsions are 3 dB, minor 1 dB. Numbers should be: 70, 73, 76, 79, 82, 85, 88, 91, 94, 97, 100, 103
My box is ~3.9cf. minus bracing and horn ~ 3.6 net. Two 3" PVC vents x 6" = 34 Hz by theory, impedance measurements suggest 36.

[Zilch]

It's "Plug-'n'-Play" with LE14H-3 (below). Both controls set to "Normal," i.e., middle.

I'll set up Citation 7.4 (3.5 cuft.) with LE14A tomorrow, probably.

[Earl K]

Hi Mark


I am happy to report that I have today oprdered the Beheringer ECM8000 with a preamp to use with an old (P3) computer I have laying around.

- That's good news .
- Don't forget about the soundcard requirements for any of these software packages.
- Does your computer have an adequate card ? ( Sound Card requirements are posted at software companies websites ). If you don't see your card listed / email them and ask if it's supported / ie; usable with their software .


Haven't decided between TrueRTA and ETF. It is not so much the cost ($100 v. 150) but the steepness of the learning curve that worries me.



- Ask Jack Bouska about the learning curve for ETF5.
- Also ask him what power of computer does he figure a person needs to adequately handle the processing overhead of this software . A P3 might not be adequate .
- You may not need a MLS based system for your casual use / only you can decide that. I'm biased towards "TrueAudio" as a company / so I am not impartial . I'd likely buy the 1/6 octave version ( 70.00 ) for TrueRTA to start with / and then upgrade later . Each version is upgradable with a simple entry code / ie ; the "lessor" versions are full-blown TrueRTA packages that are just partially disabled / meaning an upgrade is a simple email "call & response" away .

:)

[dmtp]

I'll set up Citation 7.4 (3.5 cuft.) with LE14A tomorrow, probably.
So, do you guys just have all these drivers and cabinets sitting around somewhere? Must amount to a small warehouse (and a not so small fortune!) I thought my addiction was bad - I build 1 or 2 new systems each year - but frequently they are relatively inexpensive ones for my kids or out by the pool, etc. My current 'reference' system is a slightly modified version of Siegfried Linkwitz's Phoenix system bi-amped with a 400W ss amp for the 4 woofers and a 70W tube amp (modified Dyna 70)for the mid/hi's. I can't wait to see how the JBL's compare!

[Earl K]

Looking at the separate curves, it is obvious that the woofer is rolling off too soon. This appears to be due to the 24uF cap.

- It is most likely a combination of both the coil value & the presnt cap value .
- For a 7.5 ohm load / figure out the individual F3 points for the values of your 2 passive components .




- Keeping the coil @ 2.25, I found three choices for XO freq/cap:

1. Butterworth, freq = 625, cap = 28 (closest to what we have)
2. LR, freq = 900, cap = 14
3. Bessel, freq = 775, cap = 18

- You've lost me here with your version of "reverse-engineering " against a single known value . But really , that doesn't matter now . :D

- Remember these "Filter" relationships ;

(a) 2-pole Butterworth Transform ; the F3 of each individual pole ( coil & cap ) when compared to each other, create a spread that is 1.0 octaves apart / & they are not equally arranged around ( above & below ) the resulting F3 point .

(b) 2-pole Bessel Transform ; the F3 of each individual pole ( coil & cap ) when compared to each other, create a spread that is 1.5 octaves apart / & they are not equally arranged around ( above & below ) the resulting F3 point .

(c) 2-pole LR Transform ; the F3 of each individual pole ( coil & cap ) when compared, create a spread that is 2.0 octaves apart / & they are in fact equally arranged around ( above & below ) the resulting F3 point .

- Now, for a 1000hz crossover point , determine ( for each transform type ) the necessary values of the coil & cap ( on a 7.5 ohm load ) .
- Then calculate the F3 of each derived value against the 7.5 ohm load .
- Then look again at the F3s' for your existing passives' .
- These comparisons will start to illuminate the lowpass situation and possible remedies .



On the assumption that the hi-pass is a third order Butterworth (somewhat 'tweaked' by the second cap value as already discussed), is there a compelling reason to use a Butterworth alignment (second order) for the woofer, or will L-R or Bessel work as well? I am much more familiar with L-R XOs

- Forget the "dogma" ( & everything that this implies ) that is inherent in using the names of the standard transforms.


and it looks like a 900 XO would be a lot better. I think I'll try different cap values and see what I get.

- That's a good idea .
- Try a bunch of different values for the coil & cap / keepng in mind that you are trying to fit together 2 different "transforms" that were never designed to fit together.
- You're looking to achieve "the best fit" given the ( unchangable ) circumstances that the two drivers are physically separated and voice coils are way out of alignment.
- Don't be afraid to flip polarity on the horn driver when looking for better energy summing within the crossover area.

- You are now into an area of empirical experimentation . ( Good Luck ! )

:)

ps; Why do all this you may ask ? Because real-world drivers don't output "flatlines" as FR / therefore every situation has some extra "EQ" built into the choice of filter slope .

[Zilch]

So, do you guys just have all these drivers and cabinets sitting around somewhere? Must amount to a small warehouse.... It's mostly an array of "stuff," collected over time at minimal expense. I only put real money into leading-edge components not otherwise available.

The LE14H-3s I measured above are running in L55 cabinets (2 cuft) that forum member Don C was taking to the dump. Do I care what they look like? Not one whit.

For the purposes of this "hobby," what matters is having the requisite platforms available for convenient experimentation. If there's some result that I really like, I'll build new cabinets from scratch.

There's exceptions, of course. L200, for example, is not only an excellent research platform, it's a good-looking 5 cuft. cabinet that can be developed through upgrade into quite a nice finished system.

L101, on the other hand, there's not much to do but stack a horn on top and admire the fretwork grille and marble top, which is fine, too.... ;)

http://audioheritage.org/vbulletin/showthread.php?p=92383#post92383

That's Citation 7.4 on the right there, with a refoamed flea-market LE14A in it. I got that (the cabinet) in a late-night exchange with forum member Ti Dome in an underground parking lot in Sunnyvale. It was raining, as I now recall. Citation 7.4 barely fit in the venerable Zilchmobile. :hmm:

[Scroll up to post #414 for a closer look....]

[dmtp]

- For a 7.5 ohm load / figure out the individual F3 points for the values of your 2 passive components .


My lord I'm getting an education - far better than Dickerson;s book!
OK here goes:
L=2.25 = 500Hz
C=24=700

- Now, for a 1000hz crossover point , determine ( for each transform type ) the necessary values of the coil & cap ( on a 7.5 ohm load ) .
- Then calculate the F3 of each derived value against the 7.5 ohm load .
- Then look again at the F3s' for your existing passives' .

OK, 7.5, 1000Hz
LR
L=2.39 = 500Hz
C=10.6=2k
Butterworth
L=1.69=700
C=15=1400
Bessel
L=2.07=600
C=12=1800
Current
L=2.25=500
C=24=700
Obviously the cap in the current design has a lot lower F3 than any of the standard XOs. That would account for a dip between 500-1000 (or tame a peak).
But even if one was going to use a standard XO, how would you choose between the LR/Butterworth/Bessel? Don't they all accomplish the same thing? I'm learning, but still have quite a ways to go it seems!

[Earl K]

OK here goes:
L=2.25 = 500Hz
C=24=700

Actually my calcs for a load of 7.5 ohms , put the 2 F3(s) at ;

L = 2.25 mH = @ 530. hz
C = 24 uF = @ 885. hz

- The point to this is, that anytime the difference between the two is less than 1 octave wide , you've built up a "bump filter" of some sort. Through ringing/resonance they have a bit of gain in the region of the F3(s), along with an accelerated transition to the final slope. Bump filters also have compromised group delay properities ( as well as transient response ) .
- Why use one ? To get the builtin EQ ( gain ) .
- Who ( what networks ) used them ? Just about all the older 43xx series of 3 & 4 ways ( with 2307 horns ).
- Do you need one ? I don't know without seeing accurate FR studies of the le14a. Since JBLs' published FR drawings show a bit of rising response from 300 hz and up / your project may not need any "boost' in the crossover region ( though you still need to fill in that "frequency hole" / if indeed it actually exists ).


But even if one was going to use a standard XO, how would you choose between the LR/Butterworth/Bessel?

- In your case, by an educated guess .
- ie ; get us some accurate response studies of your le14a in its' cabinet & we'll help you "guess".


Don't they all accomplish the same thing? I'm learning, but still have quite a ways to go it seems!

- No, they all have different characteristics in the transitional / crossover region .
- Ultimately ( maybe a couple of octaves later ), they do achieve similar attenuation slopes .



:p

[Zilch]

I don't know how 1 kHz got into this; it's an 800 Hz crossover.

Are you wanting to move it up to 1 kHz and I missed that part?

[Earl K]

I don't know how 1 kHz got into this; it's an 800 Hz crossover.

Who are you asking Zilch ?

- From my perspective / the actual Fc point is just a bit of a side issue / though I would prefer to mantain a 800 hz area crossover point .
- ( IMO ) A more important concern is whether a person feels the need to design with "bump-filters" to achieve a decent Fc .
- For instance; Does the le14a really need a bump filter in this region ?
- It could be that the typical Tractrix horn and Oblate Spheroid ( like the older 2307 ) need some EQ "help" in their lowest octave ( assuming typical DIY sizes such as we've seen here lately ) . Id' like to see all the raw FR studies before making sweeping generalisations .
- That's the reason I'm still talking :blah: It's a bit of a 2 way investigation ( maybe 3-way )

Are you wanting to move it up to 1 kHz and I missed that part?
- No, I'm just working towards creating some network design perspectives and protocols .
- So far I'm guessing you haven't picked up on my "hints" for designing a 3-pole / "bumped" / 800hz HP ( ie; start with a 1000hz Butterworth HP & then migrate the bottom pole upwards ) .

:)

[Zilch]

Who are you asking Zilch ?:)Uhmm, it was a "rhetorical" question.... :p

I assumed the 1 kHz "illustrative," but began to think I might have missed something, is all.

Re: The "bump," we've already got that, both HF and LF, I believe, needed or not.

I'll get the LF response this evening, with and without the filter. It's hard to accomplish anything real during the day here. ;)

Gotta get it up off the floor if the results are gonna be meaningful.

The ugly truth of crossover design, below, using "old school" Vector spring breadboading pins on perfboard. No soldering.

[SOUNDS good, tho.... :D ]

[Zilch]

1) Nearfield (~1/4") response of LE14H-3 and LE14A. L55 port is on the back; Citation 7.4 port is next to the woofer, hence extended LF response; mic is "hearing" a bit of the port output. Bigger box is part of that, as well. It's NL200B LF filter, of course.

2) Hi res version of same.

3) Zoomed to the area of interest.

4) L55 DUT.

5) Citation 7.4 DUT.

Zilch deems them "Similar." :)

Filter gives a 1 dB "bump" 100 - 200 Hz.

Where'd THAT come from, L200B? :p

[Zilch]

1) I can tell you, not much here sounds sweeter than this system.

2) Citation 7.4 is 3.48 cuft. tuned to 29 Hz.

Mark, you may want to try your box tuned down to ~30 Hz, too.

I think I'll build per John W's uber-cool implementation.

Just picked up a pair of LE14H-1s and had 'em rehabbed.

Gotta do SOMETHING with them. :p

Thank you for this opportunity to play LE85s and LE14s two-way again.

[Gonna go listen more, now.... :yes: ]

More: As the time stamps indicate, I was up pretty late listening, and did a couple of additional measurements:

3) The MF adjustment control range, HF constant at "Normal."

So, if the MF is at Max to balance with LF, is there still enough HF to go with? Yup, if flat is boost enough:

4) Range of adjustability, both controls Min and Max in comparison to "Normal."

[Earl K]

:D Nice Work Zilch !


1) I can tell you, not much here sounds sweeter than this system.

- Now,,, I hope you realize that you've tricked yourself into wiring up ( and enjoying, you say ) a 2344 horn playing below it's "accepted-best practises", LF cutoff.:applaud:
- Some around here may start to panic, if you use this as an example to challenge their current paradigm .
( You better unwire it and forget about it. ) :p

- Of course, this is all due to the power of "bump filters" / not that I'm an advocate of them you understand . ;)

- A question ; Why didn't you place the HF(s)' variable pot at the beginning of the circuit making it a true Drive Control ( a la , the 3134/5 network) ?
( The way it's now arranged, it also adjusts the F3 points for the preceding LC parts. )


:)

[Zilch]

Some around here may start to panic, if you use this as an example to challenge their current paradigm.POO on 'em, and the JON-BOAT they rid in on, too!! :p

http://www.jblpro.com/pub/obsolete/2344A.pdf

If we moved it to 1 kHz, they'd just be sayin' that's too high for LE14, instead....


A question: Why didn't you place the HF(s)' variable pot at the beginning of the circuit making it a true Drive Control ( a la , the 3134/5 network) ?

The way it's now arranged, it also adjusts the F3 points for the preceding LC parts. :)Well, I don't think it matters, but I'm not above proving myself wrong in the interest of a better design. ;)

You mean F3 of the bypass loop, or F3 of the basic highpass? I need to know where to look.

[Easily remedied using these fancy spring breadboarding thingies, if so.... :p ]

[Earl K]

Hi Zilch


You mean F3 of the bypass loop, or F3 of the basic highpass? I need to know where to look.

[Easily remedied with these fancy spring breadboarding thingies, if so.... ]

- I'm referring to the position of "LP-2" in the bypass loop .

- I'm not saying, that it's in the wrong postion . I am saying that at that position, with it offering variable resistance ( in parallel with the midrange loop below it ), in a small way it also changes the working impedance that the 1uF cap ( and .08mH coil ) "sees" . No biggy :p

:)

[Zilch]

- I'm referring to the position of "LP-2" in the bypass loop . As luck would have it, it's actually wired as you suggest would be optimum. Here's the measurements, both ways:

1) LP2 "in front"
2) LP2 "behind"
3) Highpass knee
4) HF response

If F3 shifts, it's negligable, apparently.

I've just ordered parts to build a pair. :thmbsup:

$225.98 for two. :banghead:

[Zilch]

Lest any doubt their VHF capabilities:

Six red-sealed LE85s of varied vintage and indeterminate usage histories, below.

4712 - Red, 7429 - Grn
15567- Cyn, 16881 - Vio
18172 - Blu, 22222 - Org

The signature of the 2344A horn is apparent (bottom). [Notches, anyone?]

The filter "bump," too.

I was using the red one for this work.

[Gonna switch to the cyan 'cause it's better lookin'.... :thmbsup:]

[Zilch]

Work in progress:

[hjames]

Missed the original discusion a few days ago about
no easy source of graph paper, then I found this yesterday online (from work!)

The answer: you can print your own!

http://www.pdfpad.com/graphpaper/

[dmtp]

No, I didn't drop off the end of the earth. It is just that work and other obligations keep getting in the way!:banghead:
Anyway, I finally got a reasonable test set-up going and found some good weather (if COLD) to take everything outside for a semi-anechoic measurements.
First my gear: an old PIII computer with TrueRTA, an old (vintage) Marantz amp, Beheringer EM8000 mike with UB802 preamp. Now, some pics:
1. LE14A with XO
20188
2. LE85 with XO showing L-pad @ min, mid, max
20189
3. Expanded HF
20187
4. Effect of HF boost
20190

[dmtp]

I tried to upload more files, but for some reason I can't????
I'll try again later.

[dmtp]

Here are some measurements of the full system:
1. On blocks measured @ 3' (note difference in LF/HF levels and big notch @ 300)
20195
2. Measured with speaker flat and mike 3' above (sort of 1 pi without digging a hole)
20196
3. Ground plane measurement (speaker and mike both on ground)
20197
This certainly looks the best. Not sure how it or others relate to real world. Now I have to finish of the speakers, put in foam, FG, etc. Seal all the leaks and then test them in the living room.
Stay tuned!

[Zilch]

:hmm:

Our LE14A nearfield results are certainly similar.

Looks like you need to attenuate the HF by an additional 10 dB or so. Try doing that with a fixed L-Pad just before the driver. I can't imagine your horn is that much more efficient; something else is up, maybe.

The VHF appears to be dropping off above 12 kHz. Measure something else to see if you're getting accurate results up there. If so, either your horn is killing it, or your driver isn't generating it like it should. Do you have any 2344(A) to try?

Waiting on some 24 uF caps for my crossovers. Once they're done and tested, I could send them to you for a week or so to compare with yours, maybe....

[dmtp]

No other compression drivers around. I'll hook up the other LE85 and see what is does. I did test my current system in the living room (the one in my avatar) and it is pretty much flat out to 20k.
What are you using for a schematic for your XO? Any changes from what I origianally posted? I am using Solen caps bypassed with AudioCap Theta's.

[Zilch]

What are you using for a schematic for your XO? Any changes from what I origianally posted? I am using Solen caps bypassed with AudioCap Theta's.
http://audioheritage.org/vbulletin/showpost.php?p=127396&postcount=76

Put R4 (8 Ohms) in there, at least....

[dmtp]

Are you using an 8 ohm LE85? In addition to the 8 ohm R4, I notice you have L3=0.08 mH vs my 0.15. C6 is the same @ 1 uF. I also do not have R5 (20 ohm).
Or do R4/R5 form another L-pad that is further attenuating the LE85? And how does the HF bypass feeding into the 'knee' of the L-pad effect that? Hmmm, with R5 parellel LE85, the HF bypass would be 'seeing' about 8 ohms anyway, wouldn't it?
I'm sure I have 20 & 8 ohm reisitors. I'll see if I have a 0.15 mH or another 0.08 mH coil to try out your circuit. ( I was following the 16 ohm version posted elsewhere - not sure I still have the thread reference) I also want to try the other LE85 as all my testing to date has been done with the same unit.

[Earl K]

Hi Mark

This is a bit discouraging. :o:


( Zilch )Are you using an 8 ohm LE85?

- Mark, the stock ( original ) diaphragm in the le85 was only available in one impedance ( no matter what the foilcals say on the back of the le85 drivers ). It's closer to 11 or 12 ohms .
- We "regulars" know this because the topic comes up pretty regularly / / I guess it's not that surprising that you're unaware of this .
- JBL ( at some point ) changed ( how they "interpreted" the drivers' "impedance curve" ) and then they changed the "listed" impedance .




In addition to the 8 ohm R4, I notice you have L3=0.08 mH vs my 0.15. C6 is the same @ 1 uF. I also do not have R5 (20 ohm).

- I'd recommend just building your hipass section according to the schematics that Zilch has published.
- The response curves he has published look a lot better than what you've achieved on your own.
- He has done extensive testing of his circuit and has "shown" the type of response one can expect to get ( if you follow directions ).
- What more motivation do you need to accept the results of his hard work ?



Or do R4/R5 form another L-pad that is further attenuating the LE85? And how does the HF bypass feeding into the 'knee' of the L-pad effect that? Hmmm, with R5 parellel LE85, the HF bypass would be 'seeing' about 8 ohms anyway, wouldn't it?

- There's nothing wrong with wanting to understand what you're building .

- One can argue that R4/R5 form another "Lpad".
- Alternatively, one can argue that R4 coupled with the preceeding variable Lpad, now form a "T" pad followed by a single "load resistor".

- Either perspective is correct. The important point to remember is that all these resistive elements represent attenuation and offer some circuit isolation .




I'm sure I have 20 & 8 ohm reisitors. I'll see if I have a 0.15 mH or another 0.08 mH coil to try out your circuit.

- Good ! Use the 0.08 mH coil. That'll at least make me feel that I wasn't wasting my time when I posted the formulas for "resonant LCRs" ( much earlier within this thread ).



( I was following the 16 ohm version posted elsewhere - not sure I still have the thread reference) I also want to try the other LE85 as all my testing to date has been done with the same unit.

- I believe that you've needlessly confused yourself into thinking that there should be 16 & 8 ohm versions for this crossover.
- So far ( from my perspective ) there's just a single usable version for the le85. That happens to be Zilches' version ( though I feel the lowpass may still need some tweaking / we'll have to see ) .


:)

[dmtp]

OK, I couldn't find a 0.08mH coil, but I found a 0.10. Everything else matches Zilch's XO within reason (7.5 v 8 ohm, etc.)
Here are my results (note: this is just the LE85 on my tractrix horn out of the cabinet which is getting cherry veneer right now. Full XO with 7.5 ohm R for LE14A):
1. Here is the best results after tweaking both the l-pad and the hf boost. If you take 88 as the average, it is +3/-5 out to 20k.
20333
2. Interestingly, here is the graph out to 50k. Note sure why the big rise above 20k - I can't hear that anyway:o:
20331
3. Finally, here is the effect of the hf boost at min (blue), mid (magenta), and max (green). Not very much effect.
20332

[Zilch]

Hello, Mark.

The LE85s I used are all original and red wax sealed. As Earl states, they all have the same "16 Ohm" diaphragm, no matter what it says on the foilcal.

Your results look excellent, now, except for the rolloff above 12 kHz. I'm suspecting it's the horn if the HF boost is indeed ineffective there.

Do you have a standard horn you can try? My initial work with your crossover in this thread was with H91.

I get a similar 5 dB rolloff followed by a peak just above 20 kHz on 2344A horn:

http://audioheritage.org/vbulletin/showthread.php?p=127977#post127977

[Earl K]

Hello, Mark.

The LE85s I used are all original and red wax sealed. As Earl states, they all have the same "16 Ohm" diaphragm, no matter what it says on the foilcal.

I'm suspecting it's the horn if the HF boost is indeed ineffective there.

Do you have a standard horn you can try? My initial work with your crossover in this thread was with H91.

I get a similar 5 dB rolloff followed by a peak just above 20 kHz on 2344A horn:

Yes, Double-Checking results on a few different horns is always a good idea ( as a "reality-check" ) but ; I believe the "HF boost circuit" isn't functioning properly. Let me explain.

- As a laff ; ( multiply 159.155 by 7.5 and divide by .1 ) The answer ( in hz ) is 11,936.63 . That's the "lowpass" F3 of this .1mH inductor "working" into a 7.5 ohm circuit .
- ie; Everything in the HF boost circuit ( after @ 12K ) is "attenuated" in a 6 db per octave rolloff ( with a .1mH coil in a 7.5 ohm circuit ) .

- Now ; ( using the same formula divide 1193.6625 by .08 ) , The answer ( in hz ) is 14,920.78 . Now the "lowpass" F3 of that coil ( in the "HF boost circuit" ) has moved up to about 14,921.00 hz . That's about were I would place it , if not a little higher / such as 16K .

:o:

PS : Mark, it would be helpful if you published your most recent HiPass schematic ( which you just used for these recent TrueRTA measurements ) .

[dmtp]

I am totally imbarrasssed, but in the interest of science I will expose my stupidity.
First, the circuit I am using is Zilch's except for the value of L3 in the hi-pass circuit (I didn't have a 0.08 on hand).
http://audioheritage.org/vbulletin/showpost.php?p=127396&postcount=76
Now the studip part. I just realized that I had the ground wire from the horn connected to the POSITIVE post of the XO instead of the NEGATIVE one! Everything else connected as in the schematic above. What is REALLY WIERD, is that connecting this the RIGHT way, results in a less flat response curve!
20336
The aqua is the 'crossed' version and the blue is the 'correct' one! I don't understand this at all!:blink:

Now, here is what I am getting with the correct wiring and varying the value of L3. I used what I had on hand to expirement - not to doubt Zilch's hard work, just to see what happens. Anyway, 0.15 was the coil I had originally, 0.10 was sitting around, 0.06 is 0.15 in parellel with 0.10 and 0.05 is two 0.10 in parellel. I also tried a straight wire instead of L3 as EarlK suggest way back when. interestingly, with just the 0.1uF cap, LP2 has no measurable effect at all.
20335

[Zilch]

C6 is 1.0 uF, not 0.1 uF no?

If you actually have 0.1 uF in there, the HF bypass loop isn't doing zip....

*****

With the driver connected to (+), it would be powered by the drop across the network. Difficult to contemplate, but you can see there's no highpass, 'cause the driver's playing -20 dB down at 250 Hz.

Correctly connected, it looks like you've got too much compensation dialed in, is all. Get the HF boost bypass working properly, then re-adjust....

[dmtp]

C6 is 1.0 uF, not 0.1 uF no?
Yes, my typo. Here are the results of my HF boost with L3=0.06 which is as close as I can come until I get a new order in. As you can see, just about 3 dB variation is all I get. Zilch got ~ 10dB http://www.audioheritage.org/vbulletin/showthread.php?t=12420&page=7
20358
Any ideas? I don't quite see how the horn would effect freq this high (1" throat measurements taken @ 3' on axis), but I do have an H91 I can try for comparison.

[dmtp]

OK, well the H91 does have better response above 10k
20359
Here is the Tractrix taken at the exact same settings
20361
Here is full range comparison
20360
While it is lacking over 10k, I like the tractrix's better response in the 2k-4k region
Any ideas on flattening out the tractrix from 5k-15k? Would I need two notch filters to avoid making the 9k+ notch worse?

[Zilch]

Let me take some measurements with H91 on the crossover this weekend.

I would expect that H91, being an exponential horn, would track the voltage drive on axis.

I'm saying, let's first verify that everything is working as expected with H91, which we both have, then deal with your Tractrix, as required.

What do we know about Tractrix, generally? Exponential? Constant directivity? Over what frequency ranges?

As you suggest, the compensation contour may need to be tailored accordingly, but notch filters are not the only, and perhaps not the optimum, approach for that.

Again, as earlier, I raise the question of intent. Since we're now concentrated upon VHF response, the desire would be to use this horn two-way, i.e., with no tweeter. Is that correct?

[dmtp]

What do we know about Tractrix, generally? Exponential? Constant directivity? Over what frequency ranges?I am certainly not the best one to explain a tractix horn, but I will give it a try. A tractrix horn is neither exponential nor constant directivity. It would be closest to a catenoid. Tractrix is a unique geometry such that the wave front is always hemispherical. At any point along the curve, the length of the tangent of the curve to the axis is a constant. For any given throat diameter and frequency, the horn has a fixed length and mouth becasue at the mouth the curve becomes tangential to the surface and cannot grow further. Some folks (Bruce Edgar in particular) think that the unique characteristics give it a "better" sound, especially in the midrange.
In terms of frequency response, I did some experimentation with cone drivers and tractrix horns and found that a tractix horn is effective (i.e. has a higher SPL output than the driver alone) for about 4-5 octaves at which point the reponse falls to what it was for the driver without the horn. I have never seen anything published on the frequency range of a tractrix horn.

Again, as earlier, I raise the question of intent. Since we're now concentrated upon VHF response, the desire would be to use this horn two-way, i.e., with no tweeter. Is that correct?Yes, this is to be a two way system.
Finally, I did a few more measurements to be sure my results were not skewed by mike placement, etc. The graph below is the average of 7 readings taken with the mike varying from at the mouth to 5' away and with the horn, sitting on the (carpet) floor facing up and sitting on a chair edge facing horizontal.
20366
This was done with the high boost totally disconnected, LP1 @ mid. Perhaps the peak at 7-8k needs taming, but it is only a few dB, and the 'notch' @ 9k now looks less impressive as well. If I could just pump in some boost over 14k, it would be darn nice.

[Zilch]

If I could just pump in some boost over 14k, it would be darn nice.Easy, and you know how to do that from this thread.

The C in the HF boost loop establishes where it kicks in, and the L sets where it rolls off at the top. Do your calcs, put in the desired values, and hook that boost back up.

Looks like you need ~10 dB of "assistance" up there. I believe the circuit will provide that. Set LP1 to balance with your LF, then dial in as much boost as is required with LP2 to bring the VHF up to match the MF.

Yes, it looks like you'll need a notch at 7.5 kHz. You know how to do that, too, now.... :thmbsup:

[Robh3606]

Some folks (Bruce Edgar in particular) think that the unique characteristics give it a "better" sound, especially in the midrange.


FYI Bruce uses a single cap on his 350Hz Tractrix horn driven by a 2441. That gets him good out to 10K of so. Just seems like you guys are going to an awful lot of work. Have you tried a simpler aproach with these???

Rob:)

[Zilch]

Have you tried a simpler aproach with these??? Naw, gotta BUILD stuff!!


***Zilch konks Robh3606 with a 28.5 uF ClarityCap***

[CONVERGENCE]

Something wrong here . Are you deaf ? Nothing beyond 10Khz. This horn should go up to 30K at 100 DB.
Get your money back from JBL.

.................................................. ....................................

[dmtp]

Easy, and you know how to do that from this thread.OK, I'm doing SOMETHING wrong here. Based upon formulae earlier in this thread
http://www.audioheritage.org/vbulletin/showpost.php?p=126752&postcount=4
I derived the following:
Lowest freq for boost = 12000 (my choice)
Z=7.5 (12ohm LE85 parellel with 20 ohm)
therefore C=1.0uF
Peak boost = 16000 (my choice)
therefore L= 0.10mH
desired bandwidth = 8000 (my choice) (boost 12000-20000)
so R = 5
Here is what happened:
20381
The boost is spread from about 3k or 4k to 20k (or beyond)
WHAT did I do wrong????

[Zilch]

Can't be right. C6 is already 1.0 uF.

Earl's gonna have to resolve this one.

Me, I'd be trying lower value C6's.... :)

[Earl K]

The boost is spread from about 3k or 4k to 20k (or beyond)

- That ( curve ) looks quite normal when the Variable Lpad ( MF ) that controls the midrange content ( in the horn circuit ) is dialed down too low / and the HF "bypass circuit" is allowed to run too hot .

WHAT did I do wrong????

- I'll suggest that there is a significant conceptual problem that needs to be addressed first.
- Also, for the moment, consider not using in this HF bypass circuit, the series LCR (" for boost" ) .
- Just use a plain cap ( no coil ) until you understand how a simplified "HF Bypass Circuit" works and how you can manipulate it .

(i) The overall "HF Boost Circuit" ;
- It's a bit of a misnomer because generally , passive circuits do most of their work through attenuating signas .
- A "HF Boost Circuit" is really a "HF Bypass Circuit".
- It needs to be designed to work "hand in glove" with the MF attenuation circuit .
- ie; Dialin too much attenuation into the midband and ( & by extension ) a wide open HF bypass circuit will appear to offer way too much HF content ( or HF/MF content of the wrong type ).

- The typical HF bypass circuit ( with no LCR , just a C and an R ) allows one to send the bypassed High Frequencies directly to the horn driver at a predetermined F3 point ( say 20K hz which is chosen through the value of the capacitor within that bypass circuit ).
- Everything one hears ( through this HF bypass circuit ) / is
below the chosen F3 point,. It's down in the caps rolloff area ( the "stopband" ) and this HF content is being attenuated at a rate of 6 db per octave .
- It's worth repeating that typically ( in JBL "bypass" networks ), one is usually listening to HF content that is below the F3 point of the cap ( Altec is different ). A lot of the audible HF content is down in the capacitors' "stopband" ( which is below the caps F3 point ).

(ii) An Example:
- Since "F3" is already 3 db down ( chosen as 20K hz for this example ), another octave lower ( 10K hz ), the HF bypass cap has "allowed in" frequencies ( in the 10K area ) which are 6 db lower ( than 20K ). This gives total HF attenuation ( again at 10K ) of 9 db.
- Now, if one has dialed in 15 db of midband attenuation ( like you've probably done in the variable Tpad ) the HF bypass circuit will continue to addin MF to content at the constant attenuation rate of 6 db per octave . So one will get this unwanted "MidRange content , 15db down (in the 5 to 10K region ) more or less because that is where the "Tpad" kicks in it's contribution of frequency content . ( Believe it or not , I've tried to simplify this and may still need to edit a lot of this ). :o:

- The above example outline the general HF bypass concept . It still doesn't touch on all the wrinkles that come into designing a realworld, "working" circuit .


I derived the following:
Lowest freq for boost = 12000 (my choice)
Z=7.5 (12ohm LE85 parellel with 20 ohm)
therefore C=1.0uF
Peak boost = 16000 (my choice)
therefore L= 0.10mH
desired bandwidth = 8000 (my choice) (boost 12000-20000)
so R = 5

- As far as I can recall, you still haven't run ( & graphed & published ) an AC impedance study ( measured at the horndriver, with a fully built network in place ). I did see the woofers' impedance curve ( with Zobel ). Therefore, one doesn't really know what the circuits' "working impedance" is at that particular spot ( just after the "T" pad ). We still don't know what AC impedance the HF bypass circuit actually "sees" . So 7.5 ohms is still just conjecture.
-Therefore , choosing new values for the passives in your HF bypass circuit is still a crap shoot .

(iii) "R"
- The "R" ( as is typically used in a HF bypass circuit ) is there to limit the drive voltage of the circuit. It's usually accomplished through a 0 to 30 ohm pot. ( & is usually located at the start of the HF bypass circuit ). This resistance, controls the voltage flow through this part of the circuit.
- Since you referenced my first post in this thread that only just touched on basic LCR theory, I'll assume you've tried to work "R" into your LCR resonance formula . I wouldn't do that in this case. Think of the "R" in a HF bypass circuit as mostly controlling the voltage flow through the circuit .
- OTOH, "R" is a very critical component of the LCR formulae when one is deriving the needed component values for an effective series notch filter ( strapped in parallel across the load ) .

What remedies do I suggest ?

- Build one HiPass circuit "exactly" as Zilch has outlined . But be aware, this may not be your final circuit . This will be a worthwhile exercise to see if you can duplicate Zilches circuit / as well as his published FR measurements . You gotta "walk before you can run" .

(iv)Comments ;
- It's really a mystery to me why your results are so drastically different than Zilches', especially when you state that your circuits are virtually identical to his.
- Also, considering some of the early measurement and wiring gaffs that I'm aware of ( from the last couple of months ) / I'm inclined to think the fault lies somewhere in those areas . That's why I believe it's a much better idea to see if you can simply duplicate Zilches efforts ( before tweaking this circuit for your particular horn ).
- He's published many graphs that show what happens when he turns the "drive" pot up or down ( in his HF bypass circuit ) . His circuit works. When he adjusts the ( HF ) drive-pot , only frequencies from 10K and up are adjusted.
- I also know that he has properly balanced the le14 woofer to the MFs' variable Tpad ( since he also published those studies ).
- Therefore I know all parts of his circuit, balance out well / and appear to be virtually finished .


:)

[dmtp]

That ( curve ) looks quite normal when the Variable Lpad ( MF ) that controls the midrange content ( in the horn circuit ) is dialed down too low / and the HF "bypass circuit" is allowed to run too hot .I didn't state, but the L-pad (MF) is set at mid position, the circuit has both the fixed 2.5/5 L-pad and the following 8 resistors. The HF bypass had the 16ohm Lp2 but I set it at a measured 5 ohms to match my 'calcualted' R.

- The typical HF bypass circuit ( with no LCR , just a C and an R ) allows one to send the bypassed High Frequencies directly to the horn driver at a predetermined F3 point ( say 20K hz which is chosen through the value of the capacitor within that bypass circuit ).OK, if I understand this right and if we accept 7.5 as the working impedence (more later), and an F3 of 20k, that would mean a C=1.06 = 1/(2*pi*Xc*F) which is what we had. I will try putting Lp2 before C and try measuring that.

As far as I can recall, you still haven't run ( & graphed & published ) an AC impedance study ( measured at the horndriver, with a fully built network in place ).I'm confused as to exactly what to include and where to measure? Do I take my measurements at the terminals of the LE85 with the freq generator driving through everything (except the HF bypass)? What about the Lo Pass (LE14A) part of the network? Is that attached too? And I assume I still calibrate the setup by substituting the 10 ohm R for the LE85 and set my DVM to read 10mV through the whole network?
I did the LE85 alone and it is shown below:
20384

[Zilch]

Here's the impedance curves from earlier in this thread. The elements are out of the circuit:

http://audioheritage.org/vbulletin/showthread.php?p=126871#post126871

See the headers for what they are.

[dmtp]

Interesting. As Bruce suggests, the tractrix seems to have far less variation than the exponential horn in terms of peaks and valleys. (of course I only started @ 500 so did not include the resonant peak - I was worried about driving the LE85 below 500 - probably needlessly. My curve is also based on descrete measurements @ 1/6 octave so might have missed some.

[Earl K]

MarkT;

- Please confirm that your le85s still have their original "16 ohm" diaphragms.
- Are the red wax seals still in place ?
- Measure the DC resistance of the coils and post the result .




I'm confused as to exactly what to include and where to measure?

- Well, not to worry, nobody ( to the best of my remembrance ) has ever done this before and posted the results to this forum .
- ie; it's new ground that's never been covered to date .
- Measure at the compression drivers' terminals .



Do I take my measurements at the terminals of the LE85 with the freq generator driving through everything (except the HF bypass)?
- Yes, the freq generator drives signal through the horn portion of the network .
- Take measurents with and without the HF bypass in place .


What about the Lo Pass (LE14A) part of the network? Is that attached too?
- At this time, it's not necessary to include the woofer portion .


And I assume I still calibrate the setup by substituting the 10 ohm R for the LE85 and set my DVM to read 10mV through the whole network?

- The calibration process is the same ( just as you've described ) .


I did the LE85 alone and it is shown below:

???? Really ????

- Your posted "Z" curve is quite close to what I would expect to see from a measurement that was taken with the HF bypass "shunt" in place ( over a 16 ohm Lpad ) / but feeding into a 8 ohm diaphragm ( 5.??? ohms actually ), & with the 20 ohm "load resistor" strapped across the driver . ( & Measured before the variable Plad ). Hence my first piece of direction within this post .
- Since "my scenario" doesn't match your written text, I'll have to believe that there's a problem with your impedance measuring methods.

- Look at Zilchs' impedance studies HERE (http://audioheritage.csdco.com/vbulletin/showpost.php?p=126871&postcount=18)
- Yours' should approximate one of his from that batch of "Z" runs. The HF ac impedance isn't greatly effected by the difference between your Tractrix horn and that of a H91. Again , " I'll have to believe that there's either a problem with your impedance measuring methods " or you aren't measuring a real le85 diaphragm . There are many impedance curves of le85s on this forum ( most done by Zilch ) and none of them decline to 3 ohms at 20K ( or even 5 ohms at 10K ) . In fact, those curves typically climb after 10K unless there's alot of resistive padding in place .



:)

[dmtp]

Just use a plain cap ( no coil ) until you understand how a simplified "HF Bypass Circuit" works and how you can manipulate it .It will probably be a couple of days before I can get the impedence maeaurements, but I did some more SPL measurements.
I am using the entire network as published by Zilch earlier "as built" with 7.5 ohm resistor for LE14A EXCEPT NO HI BYPASS.
This was taken with the MF L-pad @ mid, 16ohm section of Lpad2 ahead of 1uF cap. The curves are NC, min, mid, max;20421
This is the HF detail
20423
Now with the MF down to 9:30 (NC, min,mid,max)
20422
Now with MF @ max (NC, min,mid,max)
20424
Obviouly the amount of upper mid is greatly effected by the MF control position as EarlK suggested it would be, but the HF (>10K) is still pretty minimal.

[dmtp]

- Please confirm that your le85s still have their original "16 ohm" diaphragms.
- Are the red wax seals still in place ?
- Measure the DC resistance of the coils and post the result .
Original as far as I know. red wax seals intact.
DCR = 6.9 on each

[Earl K]

Interesting. As Bruce suggests, the tractrix seems to have far less variation than the exponential horn in terms of peaks and valleys. (of course I only started @ 500 so did not include the resonant peak '''snip,,,,,, My curve is also based on descrete measurements @ 1/6 octave so might have missed some.

- I believe this "tractrix" assertion of yours is based on "wishfull thinking" as much as anything concrete. ie; I don't believe your impedance studies .

- It would really be worth your while ( and ours ) , if you would practise duplicating Zilches' AC impedance measurements. Use your le85 ( on a H91 ). Adhere to the circumstances employed by Zilch ( in post # 18 ) . ie ; match the various pad settings , & use ( or don't ) the 20 ohm load resistor .



:)

[Earl K]

Original as far as I know. red wax seals intact.
DCR = 6.9 on each


Okay, thanks for that confirmation .



:)

[Earl K]

Mark,

Re; the new FR studies in post # 127 .


I am using the entire network as published by Zilch earlier "as built" with 7.5 ohm resistor for LE14A EXCEPT NO HI BYPASS.

A recap of conditions ;
(i) - Built with the same values as Zilches "modified" N200 .
(ii) - Built without a HF bypass circuit in place .
(iii) - A 7.5 ohm dummy load standing in for a real woofer .

Observations;

(i) The MF variable Lpad is not working correctly.
- It should be equally attenuating all frequencies, down to the hipass crossover point of @ 850 kz ( not just the higher frequencies ) .
(ii) Since it's obviously not accomplishing equal attenuation we need to find out why .

Some Remedies ;

(i) Remove the 7.5 or 8 ohm resistor that was inserted after the variable Lpad / making it into a "T" Pad .
- Redo your measurements with that resistor out of the circuit .

(ii) Make sure you've wired the 16 ohm variable Lpad, correctly into the circuit .
- With the shaft pointed towards you, the terminal numbers are ; # 3, #2 , & then #1 .

:)

PS ;
This was taken with the MF L-pad @ mid, 16ohm section of Lpad2 ahead of 1uF cap. The curves are NC, min,

- This statement contradicts the line above it ( which I have quoted at the beginning of my post ).
- ie; You actually have built a HF bypass circuit .
- Therefore, once again , I don't know what circuit conditions where actually employed for these pics . :o:

[Zilch]

I'm looking at Earl's model calculations in this post:

http://audioheritage.org/vbulletin/showpost.php?p=133189&postcount=109

which is telling me the L should be .06 mH (0.0597) for 20 kHz,

and if you're using 0.1 mH, you're shutting down the boost at 12 kHz (11.937).

I'm gonna go do the measurements I promised now....

[dmtp]

Quote:
Originally Posted by MarkT
This was taken with the MF L-pad @ mid, 16ohm section of Lpad2 ahead of 1uF cap. The curves are NC, min,

- This statement contradicts the line above it ( which I have quoted at the beginning of my post ).
- ie; You actually have built a HF bypass circuit .
- Therefore, once again , I don't know what circuit conditions where actually employed for these pics . :o: Today 01:09 PMSorry for the confusion. The bottom curve is with NO HF BYPASS (i.e. NC= not connected). I then did three curves with the HF BYPASS consisting of a 16 ohm section of L-pad followed by just a 1.0 uF cap as suggested. L-pad set @ min, mid, max. So we have 1 curve with NO HF BYPASS and three with variable R preceding the cap. The same is true for the following graphs showing MF @ 9:30 and max. Four curves - lowest is NO HF BYPASS, then min, mid, max on 16 ohm
If you compare the green (bottom) curve of each of the three posted graphs, you will see that the L-Pad #1 DOES change the entire curve from 95dB, to 100dB to 105dB (eyeball average). Again, sorry for the confusion on the posts, I probably should have shown the three MF settings on one graph. I was trying to see if the position of the MF did effect the HF effect as EarlK had suggested earlier. Of course, it did. I think everything is working as it should except very little effect of HF bypass.

[Zilch]

This is LE85 on exponential HL91, which approximates your quasi-exponential Tractrix.

1) Disconnected HF boost, what's it look like with no compensation? Familiar, is what. Red

2) Boost reconnected, MF and HF at mid setting, also familiar, Orange

3) Same, but kick the HF boost to max, Green.

4) Change L from 0.08 mH to 0.06 mH. Yup, VHF rises. Blue

5) How 'bout pushing it a bit further, to 23,873 Hz with 0.05 mH. Yup, a bit better, but the returns have diminished. Violet

6) O.K., now let's work the low end. Change C from 1.0 uf to 0.5 uF. Good. Cyan

7) Can we lower it more? 0.33 uF pulls down the VHF more, nearly back to where we started with no boost, so that may be too low. Black

So, use L = 0.05 mH and C = 0.5 uF, then apply your notch filter to the result. That's about the best we're going to get with this circuit and your driver/horn combination. Try that and post your result.

There's other topoligies which might be used to "tailor" compensation above 12 kHz only if you find this simply doesn't work for you.

I'll go listen now....

Edit: Sounds fine, but quite beamy in the vertical. I gotta be smack on axis to hear the cymbals' sizzle. Full range, bottom. We accomplished exactly what Earl predicted would be achievable, i.e., 9 dB of boost at the top. :thmbsup:

I did try one of the alternative topologies on hand here, AM6212/00. It virtually superimposes the cyan curve.

[dmtp]

Thanks, Zilch. Although I haven't gotten to trying to duplicate your impedence measurements yet (Work getting in the way again!), I am glad to see that you have essentially duplicated my results, so it looks like we are dealing with the same things.
I agree that the cyan curve looka best and it just so happens that two 0.10mH coils can make 0.05 and two 1.0uF caps can make 0.5 so I have on hand enough to duplicate that for a trial (may be a few days).

So, use L = 0.05 mH and C = 0.5 uF, then apply your notch filter to the result. That's about the best we're going to get with this circuit and your driver/horn combination. Try that and post your result.
For the notch (LCR || LE85) would you just focus on the range from ~ 5.5k-9.5k or would you go al the way to 14k? I'm worried that going to 14k will exagerate the existing notch @ 9.5
I'll measure on my horn before building LCR, but it looks like ~4 dB centered @ 7.5k with a bandwidth of 4k (which is a Q of ~0.5, right?)

[Zilch]

Try both ways on your final setup. I'd use one leg of an L-pad to vary the Q and ascertain the optimum resistance value. I'm guessing the broad notch will give a good result, as the twin peaks are symmetrical. If not, consider dual narrow notches, though that's gonna be tough to get right, them being so close.

Again, my results are on HL91, not your horn. Try some polar measurements, too. See what your HF beamwidth is, maybe 10°, 20° and 30° off axis.... :thmbsup:

[dmtp]

I had a break at work and was working on the calculations for the LCR, but I have a question: How do I calculate the value for R? I realize that the easy thing to do is use the 16 ohm side of an L-pad, but to calculate L [L=(Q*R)/(2*pi*F0)] and then C=1/(4*pi^*F0^*L), I first need a value for R.
In a previous post EarlK said Zilch had a neat formula for calculating the value of R that would give 'x' dB attenuation when parelleled with a known impedance (which I am assuming would be the measured impedance of (LE85 || 20 ohm) at the resonant freq. Could you share this formula? Thanks!

[Zilch]

Clearly, we've departed substantively from the original N200B, isolating and assuming control of its HF boost bypass loop, and altering its frequency range of influence. In retrospect, now observing that the response of LE85 on HL91 is basically flat out to 12 kHz, what was going on in the original? Zilch wants to know, 'cause it makes no sense. :banghead:

Answer: Not the compensatory VHF boost via midband attenuation we're practicing with it, rather, rising "forward" MF and HF to 10 kHz, and rolloff at 12 kHz. The single control balances an essentially constant and very familiar "Quintessential Rock and Roll" response contour with the low frequency driver (top).

EDIT: Compare to L200's LX-16, bottom, particularly 7 - 12 kHz....

[Zilch]

Could you share this formula? Thanks!The formula is for calculating dB attenuation from L-pad resistances and driver impedance, and is linked earlier in this thread.

In this case, however, the resistance establishes the "Q" or gain of the filter, i.e., the depth of the notch. Nominally, it's set at the impedance, but adjusted up or down to accomplish the desired effect. That's why I suggest choosing it empirically, and one leg of an L-pad is a convenient variable resistance to use for making that determination....

[Earl K]

Zilch,

- Okay, I have a couple of hours before my next fitup .


Answer: Not the compensatory VHF boost via midband attenuation we're practicing with it, rather, rising "forward" MF and HF to 10 kHz, and rolloff at 12 kHz. The single control balances an essentially constant and very familiar "Quintessential Rock and Roll" response contour with the low frequency driver:

- Was that FR plot obtained with the 2308 lense on or off ?

-The 2308/L91 lense, does tend to flatten out these rising response curves .


:)

[Earl K]

An Observation ;
- We're attempting to build an LCR series notch filter ( strapped across the load ) that has an electrical "Q" which matches ( as a recipricol ) the realworld "Q" of the observed "bump" found in the horns FR response .
- IME, even with all the theory in place, this a process where one employs "best guess" / then measures / then adjusts LC values / and then remeasures / repeating until the desired results are obtained .


In this case, however, the resistance establishes the "Q" or gain of the filter, i.e., the depth of the notch. Nominally, it's set at the impedance, but adjusted up or down to accomplish the desired effect. That's why I suggest choosing it empirically, and one leg of an L-pad is a convenient variable resistance to use for making that determination....

- Zilches suggestion has much merit . It applies a variable "R" and allows one to see ( & measure ) the results immediately ( using any adhoc LCR notch filter ).



- I had a break at work and was working on the calculations for the LCR, but I have a question: How do I calculate the value for R? I realize that the easy thing to do is use the 16 ohm side of an L-pad, but to calculate L [L=(Q*R)/(2*pi*F0)] and then C=1/(4*pi^*F0^*L), I first need a value for R.
- In a previous post EarlK said Zilch had a neat formula for calculating the value of R that would give 'x' dB attenuation when paralleled with a known impedance (which I am assuming would be the measured impedance of (LE85 || 20 ohm) at the resonant freq.
- Could you share this formula? Thanks!

- IME, "R" can be calculated / and the result will be an "R" that is still just a guesstimate ( for a realworld LCR, working within a realworld circuit ).

- The difference between a "wild" guess ( at "R" ) and a half-assed accurate guess ( of "R" ) relies on knowing the real "working impedance" for that portion of network / ( where the LCR is going to be applied ).
- So far , only conjecture places that number at 7.5 ohms .

FORMULA ?
- FWIW: Zilches formula involves using antiLogs . So;

Divide 4 by 20 ( with 4 db, looking to be the amount of attenuation needed to flatten the 7800 hz bump / with no HF bypass circuit in place skewing the picture ). Take the antiLog of this answer . It's 1.5848931922 . Divide this into 7.5 ohms. This answer is 4.732180084 . This is the "new" frequency dependant load impedance that must be obtained to drop 4 db of level in that part of the circuit . How do we drop the ( 7.5 ? ) load resistance to 4.73210084 ohms ? Answer ; Parallel 12.82285398 of resistance ( within the LCR ) across the guess at 7.5 load . ( The DCR of any inductor chosen for use, becomes a portion of this R" ) .
- Conclusion ? 10 to 14 ohms of conjugate resistance should level out a 4 db high bump ( if the working impedance is really 7.5 ohms ) .

- You can measure the 'Q" of the real-world "bump" as I outlined earlier in this thread ( hopefully one has taken an accurate enough picture of it ) . If the 3 db down skirts ( of a 7800 ) bump are in the neighbourhood of 3000 to 3200 hz wide ( point to point ) then the bump will have a "Q" of around 2.5 ( 7800 ÷ 3120 ) . 10 times 2.5 is 25 ( & 14 times 2.5 is 35 ) . This ( 25 to 35 , "Q" ) becomes the range of the theoretical "Q" which the LCR must be designed to offer ( into a 1 ohm load )

(A) A Workable Approach to getting a real LCR filter ;

- I like Zilches suggestion of just dialing in "R" to find what out what value of "R" actually attenuates the bump by the desired amount of db . So ;

(i) I suggest building a series LCR notch filter ( strapped in parallel across the load ) that's close to resonating/notching within the area of interest . To my eyes , that's 7800 hz ( if that double-headed bump was expanded into one with an "imaginary peak" ) .

(ii)
- Since this is really just a "test" LCR to help us derive "R" , a person can use any LC combination ( within reason ) as long as they result in the necessary Fo point ( resonating frequency ) .
- The necessary working formulas were posted earlier in this thread ( that will allow one to chose a coil size first / & then derive the correct cap size ) .


- A .56 mH coil resonating with a .75uF cap has a Fo frequency of 7766 hz .
- I chose these values below based on some "intuition". ( intuition, meaning ; 25 ÷ 2 ÷ Pi ÷ 7800 giving 0.000510112 Henries / or .51 mH . Using 35 ( as the target Q ) would have resulted in a .71 mH coil . :p ( One can actually buy a .56mH coil . )



(iii) Passives values of .75uF & .56mH ( Offers a Resonance Point of 7766 hz ).
-These two passives will "theoretically" offer a ( full "notch" ) "Q" of 27.325 ( into a 1 ohm circuit ) . 27.325 ÷ (10 to 14 ohms , our range of "R", I guesstimated ) should offer a final filter "Q" that's in the range of 2.7 ( to just below 2 ).
- As Zilch suggested, wire up a portion of a Lpad that gives usable resistance ( 0 - 16 ohms from terminals 3 and 2 when using a 16 ohm variable Lpad / or / whatever range of resistance is available when using terminals 2 & 1, usually 60 ohms to a low of .5 ohms ?? ) .
- "Q" is also a short form for "Quality Factor" . One will discover the realworld "Quality Factor" of their chosen components by building with them / and then measuring the notch results .
- If the notch depth or width isn't "as expected or calculated" / then this will indicate that the values of L&C need shifting. This is realworld. The "Quality Factor" for these component types now have to be factored into the final equation . This usually means buying better quality capacitors ( they have "Q" that affects how well they resonate ), as well as using larger value coils ( larger than what text-book theory actually suggests ) .
- Using larger coil sizes will drive up the theoretical "Q" but this allows for ( balances out ) the lower ( ie; sloppy ) "Quality Factor" inherent in real world inductors .
- One can also study JBLs' network schematics to get hints of what size of coils they found worked( in their real world circuits ). Of course we don't know the "Q" of the bump they were notching / so this study is of limited value . But it is still worth the study .


Mark , I hope this isn't all too confusing :)

PS ;
- First do this exercise with the HF bypass circuit disconnected .
- The HF bypass circuit has the capacity to completely skew the initial results . In this projects case / your HF bypass circuit is not finished / and in reality, should be "fixed" prior to mucking about with LCR notch-filters .

[Earl K]

Mark,

About your HF bypass circuit ;

If you are convinced that you haven't miswired something and if you are convinced that the MF variable Lpad offers enough attenuation to the whole horn circuit , then ;

- Start reducing the value of C ( in the HF bypass circuit ) . And for now get rid of any drive pot and get rid of the resonating coil . ( You can attempt to add them back later in the process ) .

(i) If the HF bypass circuit ( when the MF is properly balanced to the le14 ) has bled into the final FR ) too much 5 to 10 K information or lower / then simply reduce the value of the cap by a similar amount ( in octave terms ).
- ( 5 to 10K represents one octave of information ).
- Halving the size of a cap shifts its F3 up by one octave.
- By extension , shifting the F3 up one octave will reduce all content below the F3 by a comparable amount. ie ; what was at say minus 6db will now be at minus 12 db ).

(ii)
- If the HF bypass circuit is "contaminating" 10K by adding an addition 6 db of unwanted information / then shifting the F3 point of the bypass cap upwards 1 octave will, reduce this contamination by 6 db.
- As I said earlier, everything we are hearing/seeing ( the net contribution for this HF "bypass" circuit ) comes from below the caps F3 point / therefore shifting the F3 point ( of that cap ) determines how much level ( of any one frequency ) is added into the final FR plot ( until the original MF signal swamps out the HF contribution ) .
- So everything "contributed" ( from this "bypass" circuit ) is electrically filtered at an attenuation rate of 6 db per octave .


- So try a .5 uF ( bypass cap, if you are presently using 1 uF ), then remeasure and post your results .


:)

ps ;
Can't be right. C6 is already 1.0 uF.

Earl's gonna have to resolve this one.

Me, I'd be trying lower value C6's....

- Yeh, what he said !

- I've lost track, is C6 the same cap I'm referring to in the bypass circuit ?
( I don't keep this schematic taped to my computers' monitor :p )

[Zilch]

Was that FR plot obtained with the 2308 lense on or off ?Lens on. I measured the flattening earlier, but didnt' save it, of course.... :blink:

Edit: See below. :D


I've lost track, is C6 the same cap I'm referring to in the bypass circuit ?
( I don't keep this schematic taped to my computers' monitor :p )Yup.

[[**Looks around for Earl's spycam....**]]

[dmtp]

- So everything "contributed" ( from this "bypass" circuit ) is electrically filtered at an attenuation rate of 6 db per octave .So if I understand it correctly, the best we could hope for from this circuit is something on the order of 4dB from 12k-20k (assuming I keep increasing C6 until there is no "boost" @ 12k). If OTOH, I allow the "boost" to bleed down into the 6k-12k range, I will have a bigger "hump" to attenuate with the LCR, but I could get more like 10 dB of "boost" @ 20k (by dropping the "zero boost" point by an octave and therefore gaining 6 dB.)

When next I get a chance to play (probably Sat.), I will first get high detail SPL from 5k up and then try to dial in the VHF boost needed (using just the cap, no L, no R). Once that looks good, I will re-evaluate the 5k up SPL to determine the height and width of the "bump" to make up the LCR. Depending on how well I can make up these circuits with what I have on hand, I'll post results as I go.
I'll also try to get the impedence measured with the H91 and the tractrix for comparison to get a better idea what the imdepence we are dealing with REALLY is. (To be sure, I will measure LE85 alone and LE85 || 20 ohm, right? or do we need to look at the WHOLE network?)

[dmtp]

Answer: Not the compensatory VHF boost via midband attenuation we're practicing with it, rather, rising "forward" MF and HF to 10 kHz, and rolloff at 12 kHz. The single control balances an essentially constant and very familiar "Quintessential Rock and Roll" response contour with the low frequency driver (top).

Do I correctly interpret this to mean that the SPL response we are seeing is what was considered "desirable" for a "Rock & Roll" speaker? "The JBL sound" with "forward" MF? Hmmmmm

[Zilch]

Do I correctly interpret this to mean that the SPL response we are seeing is what was considered "desirable" for a "Rock & Roll" speaker? "The JBL sound" with "forward" MF? HmmmmmYes, one element of "West Coast Sound." L200B apparently implemented it to somewhat of an extreme in the HF (see below). You'll need boomy bass and a couple of other things to complete the package.

From the frequency response curves, the original L200(A) was a better emulation; it's as if some marketing type liked the sound even more with the lens off, thus defining L200B. Let's not pretend any of this analysis provides insight into the actual product development process, tho. ;)

We coveted juke box sound. Anybody have the FR curves for '60s & '70s vintage Seeburgs?

[There was a tweeter in there somewhere, I'm CERTAIN.... :p ]

[Robh3606]

Do I correctly interpret this to mean that the SPL response we are seeing is what was considered "desirable" for a "Rock & Roll" speaker? "The JBL sound" with "forward" MF? Hmmmmm

Well that depends. Look at the difference in the amount of boost you are seeing over the same range as the posted 4311 graph. Worst case with the pads set to 0 at about 4K you would see about +2/+3 dB. You are seeing quite a bit more in the posted graphs. That extra 2/3dB rise over that range is going to make a big difference in how they sound. The 4311 could sound forward. You don't want foward to turn into shrill.

Rob:)

[Zilch]

In terms of frequency response, I did some experimentation with cone drivers and tractrix horns and found that a tractix horn is effective (i.e. has a higher SPL output than the driver alone) for about 4-5 octaves at which point the reponse falls to what it was for the driver without the horn. I have never seen anything published on the frequency range of a tractrix horn.Perhaps forum horn theorists can fill in the details, but I'd suspect the "rub" in attempting to use a 500 Hz Tractrix for extended VHF is severely compromised power response, as occurs with HL91 in the vertical.

That's why I'm asking dmtp to do some rudimentary polar response measurements once the boosted VHF is working. The hemispherical wavefront may behave differently up there....

[Robh3606]

as occurs with HL91 in the vertical

Hello Zilch

Actually that would be both axis. The lense is what changes the pattern otherwise it's a round horn so it would be symetrical. You should read this thread.

http://www.audioheritage.org/vbulletin/showthread.php?t=12967

Rob:)

[Zilch]

For horns with exponential, or Tractrix flare, the most rapid horn wall curvature changes occur towards the mouth of the horn, with the sides exhibiting less curvature near the throat. This means that waves will only diffract into the full angular coverage, defined by the exit angles of the mouth, for wavelengths which are longer than the axial length of the horn. For shorter wavelengths, the coverage pattern approximately conforms to the angle of the horn walls at the ½ wavelength axial position. For most of the other horn curves in fig 7, the angles range from 8°-12° at an axial distance of 3”, meaning that these horns (exponential, Tractrix, hyperbolic and spherical) will be unable to support a beam width of more than +/- 10° at 9kHz and above.The refractive lens mitigates it in the horizontal, but at expense to the vertical:

http://audioheritage.org/vbulletin/showpost.php?p=127369&postcount=74

[Robh3606]

Exactly so you end up with a focused beam on axis. Off axis you get nada above say 10K.


The refractive lens mitigates it in the horizontal, but at expense to the vertical:

Why at the expense??

Rob:)

[Zilch]

I was hoping Tractrix might defy some of that, but it looks in theory like the horn and driver throat size must be precisely optimized for a successful two-way design with extended HF.

The other thing I'm observing is that LE85/HL91 isn't all that flat in frequency response through the region they're commonly used here. I'll measure a bunch of 'em before concluding that, tho. Could be they'd benefit from a notch or two.

I don't have an agenda in this; I'm just following up on what's appearing in the routine course of doing these investigations....

[Zilch]

Why at the expense??No free lunch. You want horizontal, you pay with vertical. :p

I had earlier concluded that extending the HF response with HL91 beyond these limits is consequently a fruitless endeavor.

Well, for me it certainly is. I've kinda grown to like 50° vertical beamwidth, but a 20° listening slit is, well, "untenable" thus far....

[Robh3606]

No free lunch. You want horizontal, you pay with vertical. :p

Not so look at the conical lense. The serpentine and slat lense have minimal effect in that plane. You essentially go back to the horns directivity. Look at the directivity curve for the horn and horizontal pattern they mirror each other. They didn't plot the verticle but it would follow as the other plot does. That 22 degree and less wedge works in well with Jack quoted post.

http://www.lansingheritage.org/html/jbl/reference/technical/lens.htm


Well, I was hoping Tractrix might defy some of that, but it looks in theory like the horn and driver throat size must be optimized for a successful two-way design with extended HF.


This is Deja-vu. We going back to the 4430??;)


Well, for me it certainly is. I've kinda grown to like 50° vertical beamwidth, but a 20° listening slit is, well, "untenable" thus far....

Yes I tend to agree. One of the reasons I like the 2344 and the PT1010


Rob:)

[Zilch]

Not so look at the conical lense. The serpentine and slat lense have minimal effect in that plane. You essentially go back to the horns directivity. Look at the directivity curve for the horn and horizontal pattern they mirror each other. They didn't plot the verticle but it would follow as the other plot does. That 22 degree and less wedge works in well with Jack quoted post.

http://www.lansingheritage.org/html/jbl/reference/technical/lens.htm I noticed that the DI didn't support what I was saying, but chose to ignore it. :p

[O.K., I get that now.... :thmbsup:]


This is Deja-vu. We going back to the 4430??;) Naw, it's de riguer - PT Waveguides at ZilchLab. :yes:

For a symmetrical waveguide, the little work I did with the Engebretson indicated it "promising." Maybe I'll continue some with that:

http://www.ddshorns.com/catalog.php?page=ENG190Pro

If it's good, someone should document the profile for DIY.

["Family".... ;) ]

[dmtp]

OK, I finally had a chance to do some SPL plots. As follows:
1. No HF bypass (green)
2. 1.0 uF only (no L, no R) (magenta)
3. 0.68 uF (yellow)
4. 0.33 uF (orange)
20642
Now with the smaller caps, we get less MF contamination, but we also get not enough VHF boost, so let's try a larger cap:
1. No HF bypass (green)
2. 1.0 uF only (no L, no R) (blue)
3. 1.5 uF (aqua)
4. 1.5 uF in series with 0.05 mH (magenta)
5. 1.5 uF in series with 0.1 mH (yellow)
20643
To my eye, curve 4 looks the best as 5 gives up some VHF, so it makes more sense to me to go with a 1.5uF cap in series with a 0.05 mH coil and then try to tame the resulting MF/HF bump with an LCR across the LE85. Using a blow up of the above curve, I determined the F3l = 6300; F3h = 15000 (although very steep here). (F3h-F3l)=8700 width, "midpoint" = F3l+((F3h-F3l)/2) = 10650, but visually the midpoint measures out nearer 8750. I think I will try an LCR based on that freq. (with about 10dB total from 4k to 18k.)

[dmtp]

To answer Zilch's questions about polar response, I took measurements @ 0, 10, 20, 30, 45 degrees off axis. (The actual degrees may be a bit off, I was using a plastic protractor held on the horn's mouth with a stick to the mike.) I am not sure about the absolute dB as I am sure the distance to the mike varied as I moved it off center, but at least the shape of the curves should be correct. (I do not know how to do a polar plot from these curves.)
Green = on axis
Magenta = 10 degrees off axis
Yellow = 30
Aqua = 45
20644

[Robh3606]

Looks to me like your best curves are without any HF boost. You may have to suck it up and either take what your getting or add a 2405. You can try what they did on the 4430 with a series LCR but it looks like you have a brick wall at 15K.

Rob:)

[Zilch]

Your L & C tests confirm my findings. As you raise the C, the boost starts at a lower frequency. As you lower the L, it cuts off at a higher one.

My approach to it was to start the boost where it's needed, above 10 kHz, and stop it just above 20 kHz. Try the combination I found worked best, 0.5 uF and 0.05 mH. That gives ~9 dB of boost between 10 and 20 kH, the max you're going to get out of 6 dB/octave topology, plus ~3 dB from resonance.

You're already too high at 10 kHz. It doesn't make a lot of sense to me to boost that even higher and then yank it back down with a deeper notch.

*****

Your beamwidth tests indicate that the VHF is concentrated in a 20° "window," as discussed above. For most accurate results, the mic has to be at the same distance for each reading.

I use the "Move the mic" approach, too, with string used to hold the distance to mic constant. Another approach is a lazy susan under the speaker to rotate it about the acoustic center. I have a motorized rotary table for that, but haven't built it up yet.

To plot in true polar coordinate format requires a level of precision in positioning I have not achieved here. That's also multiple plots at different frequencies, a different way of looking at it....

[dmtp]

OK, I have some interesting results. First the calculations:
Based upon a 10dB hump, I calculated R= 3.5
With F0=8750, F3h=15000, F3l=6300, I got L=0.05, C=5.37. I had a 0.06mH coil so used that with 5.5, 4.0, 3.0 uF caps.
First, here are the nulls (R=0) to show the center freq of the above combos:
green=no LCR (but WITH hipass = 1.5uF in series with 0.05mH, no R)
blue = 5.5uF
aqua = 4.0uF
yellow = 3.0uF
20645
This pretty much confirms the F0 expected.

Then for each cap, I adjusted R for what looked like the best curve. Here are the results:
green=no LCR (but WITH hipass = 1.5uF in series with 0.05mH, no R)
aqua = 5.5uF, 2.1R
blue = 4uF, 2.5R
magenta = 3uF, 1.5R

20646
To my eye, the magenta (3uF, 1.5R) looks the best, but it now has an 'M' - two humps and a notch which I think indicates I need a broader notch to pull down the humps on each side. That would require a smaller Q, but a smaller Q won't "pull down" the original hump enough. So now what?
Here are some 'families' of curves.
For 3.0uF, R varied 0, 1, 1.5, 2, 5
20649


For 4.0uF, R varied 0,1,2,3
20650


For 5.5uF, R varied 1,2,3 (I forgot to include the R=0, but you can see it above)
20651

None of the curves seem to get broad enough to flatten things out. I'm not sure where to go next.:blink:

[dmtp]

Looks to me like your best curves are without any HF boost.Looks like Robh might be right. Here is a recap of boost with just a cap
green=0
blue = 0.4uF
Aqua=0.5
Orange = 0.68
Yellow = 1.0
Magenta = 1.5
20653

Per Zilch's suggestion, here are the results with 0.5uF cap
blue = no boost
green= just 0.5uF
magenta = 0.5uF with 0.05mH (Zilch's choice)
yellow = 0.5uF with 0.1mH
20654
pumping in anything above 14k just raises 11-12k too much. I tried notching them out, but couldn't come up with anything worth posting as by the time I got the 10k notch tamed, I lost most of what I had gained above 15k.
I think at this point I amo going to work on getting the speakers finished and in the room and then maybe play with the XO more after listening to music.
(Now there's an idea!):applaud:

[Zilch]

You got within +/- 3 dB with a couple of those options.

That's damn good, I'd say....

[Zilch]

Complete:

And mirror-imaged, to boot.

[Ooops, not the L-pads, tho.... :p ]

[dmtp]

Finally had time to do impedance plots on my LE85 driving both the tractrix and H91 horns. I discovered that my Freq. generator falls off in output at high freq. so this time I recalibrated for every single measurement! Numbers are much more as expected, but it does show far less swings for the tractrix. Later maybe I can repeat with the ||20R and/or the rest of the network. (Still not 100% sure of how to attach drive/measurement for that one.)

[Earl K]

Finally had time to do impedance plots on my LE85 driving both the tractrix and H91 horns. I discovered that my Freq. generator falls off in output at high freq. so this time I recalibrated for every single measurement! Numbers are much more as expected, but it does show far less swings for the tractrix. Later maybe I can repeat with the ||20R and/or the rest of the network.

- Thanks for taking the time to correct the impedance plot you get for your le85 driver .
- FWIW, many horns that have large aspect ratio mouths ( in comparison ) to the H91 , show a reduction of those impedance ripples ( or impedance reflections ). It's worth looking at Widgets' impedance plots of his H4003 (?) horn clones ( I think that plot is posted in the public forums ) .


(Still not 100% sure of how to attach drive/measurement for that one.)
- I've mentioned this a couple of times so;
- I'm going to have to leave this to someone else since I'll be "toast" till sometime Friday due to a strenuous work schedule .

A few OBSERVATIONS :

(i) Mark, you've really put your TrueAudio software through its' paces. More so ( to date ) than anyone who has posted their results from using that software .
- Congratulations on that !.

(ii) You've shown ( & are aware of ) many of the competing approaches that occur when one is searching for enhanced HF response .
- I've drawn different conclusions from what you've just recently stated. ( I'll expand on that comment later in the week ) . I guess in the meantime, as you say, it's time to get the speakers loaded and playing in your living room .

- FWIW ; I use a round mouthed exponential horn on a Altec 288-8K ( that supports 800 hz fairly well ). The horns rising Directivity Index ( DI factor ) means I need to use a series LCR ( strapped across the driver ) to flatten a quite symetrical peak ( centered around 6200 hz ) . I also use a pretty simple HF bypass circuit that does ( somewhat ) overdo the HF content ( and therefore, "contaminates" the octave below /as you have run into / & forces the need for a bigger notch )

regards; :)

[dmtp]

- FWIW, many horns that have large aspect ratio mouths ( in comparison ) to the H91 , show a reduction of those impedance ripples ( or impedance reflections ). It's worth looking at Widgets' impedance plots of his H4003 (?) horn clones ( I think that plot is posted in the public forums ) .

Yes, I suspected that was the problem with the H91 - the mouth is too small to support 500 Hz. One of the nice things about a tractrix horn is that it has a defined mouth that cannot be exceeded whereas an exponential horn can be extended to any size mouth and is still expanding.

- FWIW ; I use a round mouthed exponential horn on a Altec 288-8K ( that supports 800 hz fairly well ). The horns rising Directivity Index ( DI factor ) means I need to use a series LCR ( strapped across the driver ) to flatten a quite symetrical peak ( centered around 6200 hz ) . I also use a pretty simple HF bypass circuit that does ( somewhat ) overdo the HF content ( and therefore, "contaminates" the octave below /as you have run into / & forces the need for a bigger notch )

The directivity issue is the main reason I decided to "finish" the project before finalizing the XO. These speakers will be firing 'into' a curved reflector panel ala paragon/metrogon. This will obviously help dispersion, but I may be more concerned with power response than on axis response and may need to leave in a fair amount of on axis HF rise to get good HF response in the listening area. I now know how to 'get' and how to 'tame' the HF response with quite a lot of options.
Thanks to everyone - especially Zilch and EarlK - for all the help. When I finally get things finished, I will post picutres and response curves. In the meantime, I may come back to the well for help as things progress.
Again, thanks to all!

[dmtp]

OK, so it has been a long time, but the speaker is finally together and in the den. Yes, an HD TV on the wall will be an improvement as will finishing the grill covers, but I thought you might like to see the results of my efforts. Immediately following this, I will post the sonic results!

[dmtp]

For those more interested in the sound than the looks, here are my final plots. Note that these are taken with the mike placed right at the cneter of the curved front touching the front. There is no way to eliminate room effect, so I adjusted the "midrange" L-pad by ear. It is set @ -2dB which is about what these curves look like to me. Any higher and it sounds too bright. As is, they are GREAT!
FWIW, the 500 Hz dip must be a room effect or measruing artifact. I was worried it was a crossover notch, but I tried reversing the leads of the LE85 and it created a deeper notch soemwhat higher in freq. When measured half way between the center curve and the speaker itself, there was a deep notch @ 200 Hz. Perhaps this summer I'll take the assembly apart and measure one of the speakers out doors, but for now, I'm happy listening.
Thanks to everyone for all the help, this truly was a "community" project!

[Zilch]

That looks mighty fine, I must say! :thmbsup:

[timc]

Very nice work. Hope it sounds as good as it looks :)




-Tim

[Skywave-Rider]

Wow, it's beautiful! Would love to hear the imaging on that. Recently read about the Paragon.

[spwal]

reading is one thing, hearing one is quite another! :p