Setting up 3 way JBL horns for correct polarity

[Doogster]

Hi chaps

I had a 2 way JBL horn system which I've now expanded to 3 way. It now sound way better (I had some sort of peaking/beaming with the 2 way, but the new setup has cured this). However, I'm a tad confused as to exactly how to wire it all up. Here are the gory details:

Bass - JBL 2235 15 inch woofer in 5 cu ft reflex box.
Midrange - JBL 2441 (aka 376) on 350Hz tractrix horn.
Highs - JBL 2404H tweeter.

Crossover is a 12db passive line level LCR design for bass (low pass) and midrange (band pass). XO for the tweeter is just a 3.3uF cap in series with the tweeter.

Crossover -3db points are 700Hz and 5kHz.

Amps are a huge mofo Perreaux SS for the bass, a 45 SET for the midrange, and a crappy receiver for the highs.

I'm confused about which polarity to use for each driver. Does a 12db PLLXO do anything to the signal that requires phase reversal of some of the drivers at any point? What about the tweeter? Is there an easy way to determine the polarity of the midrange driver and tweeter (it's easy with the woofer)?

Cheers. Doug

[Zilch]

You haven't mentioned the most significant variable, the vertical alignment of the acoustic centers of the drivers.

Assume it's the voice coils, as indicated by the front plates of the drivers.

The optimal phase is largely dependent upon the depth of your midrange horn....

Do you have an RTA?

[Doogster]

You haven't mentioned the most significant variable, the vertical alignment of the acoustic centers of the drivers.

Assume it's the voice coils, as indicated by the front plates of the drivers.

The optimal phase is largely dependent upon the depth of your midrange horn....

Do you have an RTA?

Hi Zilch

I intend aligning all three drivers so their voice coils/diaphragms are in line - I am still experimenting with this aspect of things at the moment.

Unfortunately I do not have an RTA. I only have a Rat Shack SPL meter which I understand is somewhat inaccurate.

Can you recommend an RTA-less way of aligning horns?

Cheers. Doug

[Zilch]

Can you recommend an RTA-less way of aligning horns?

An alternative "listening" method is described in the 5235 crossover manual.

[sourceoneaudio]

Doogster
Where are you pulling the signal for the tweeter from? If you are running a band pass on the mid you can't pull it from there. The signal would have to come directly from the input so you are getting all HF information. If you are running a 12db per octave x-over on the mid, and 12db on the LF the general rule is to run the mid or next 12db network out of phase. I have discussed this with other members on the forum, and pulled the information from my RTTI training manuals. The final result with no RTA is your ear.. If your amp is not stable to low loads just be careful of impedance dips, so as not to damage the amp. I don't think I would worry about the tweeter to much @ 6db per octave, until you upgrade the network to a 12 or 18 slope, 12 to keep it matched is the way I would go.


J/S-S1A

[Doogster]

Doogster
Where are you pulling the signal for the tweeter from? If you are running a band pass on the mid you can't pull it from there. The signal would have to come directly from the input so you are getting all HF information.

Hi

Thanks for your reply. I am taking the signal for the tweeter straight from a separate power amp. The only thing in between the amp and the tweeter is a cap that introduces a 6db XO at around 5kHz.

Do you think a 6db slope is not steep enough for the tweeter? I intend building a dedicated tweeter amp (tube of course) that has very little output above, say, 1kHz. If that's the case, can I still get away with a 6db XO for the tweeter?

Cheers. Doug

[sourceoneaudio]

Doogster
Hello, I would go no less than a 12db network. It is much safer and you'll end up with less garbage getting passed through to the driver. (cleaner sound reproduction) 6db networks can tend to be harsh.

Here is some parts and a schematic for you
Get your part from www.partsexpress.com

Second Order Butterworth Alignment
C1= 2.81uf Parts Express PN. 027-416 2.7uf & 027-400 .10uf This will = 2.8 uf
L1= 0.36 mH Parts Express PN. 255-030 .35 mH

L-Pad (100 watt) if needed (if the tweet horn is to bright) 260-265

The caps recommended are Metallized Polypropylene, good quality, and the coils are 20 gauge air core. They have other caps better quality but cost tends to go up qwik. I use these caps, and I'm very happy with the quality and sound reproduction that comes out of them.
Make sure to order you free catalog when placing your order.

[Earl K]

Hi,

I'm confused about which polarity to use for each driver. Does a 12db PLLXO do anything to the signal that requires phase reversal of some of the drivers at any point?

Yes,
- Flip the polarity of the horn driver ( as compared to the woofer ).
- Do the preceeding only if you indeed have a 12 db PLLXO filter that is properly performing the crossover functions that you've described ( have you measured this PLLXO crossover to check its' performance ? ) .

FWIW :
- An analog filter will always produce the expected phase shifts ( if built properly ) whether the filter is acting at microphone , line or speaker level .

- The phase shifts are fully realized over in the "stopband" / not the "passband"( which is the one typically listened to ) . The transitional area between these 2 areas happens to be called the crossover region . In the "passband" ( a good 1.5 - 2 octaves away from the crossover point ) the phase shift is very negligible.

What about the tweeter? Is there an easy way to determine the polarity of the midrange driver and tweeter (it's easy with the woofer)?

- There's really no easy way to determine a tweeters' best fit / without some reasonably good measuring equipment ( ones' ears may qualify ) .

I've heard this method recommended ( but never tried it ) .
(i) Use an FM Tuner as a Noise Source by "tuning off station". The resulting noise is White Noise ( or Pink , I forget which it is ).
(ii) Run this noise source into both the tweeter & horn circuit.
- Start with tweeter having the same polarity as the woofer .
(iii) While listening to this noise source, flip polarity of the tweeter back & forth.
(iv) Use whatever polarity gives you the fullest "perceived" sonic signature in the crossover area .

Hint ; It's also been recommended that one can use a 18" by 18" ( or 2' by 2' ) piece of cardboard to create a sort of a mobile barrier ( planar horn surface ). Place this barrier ( in a vertical orientation ) just to the side of the tweeter & the horn bell, to help "unify" the sound of the two outputs. Play with the degree of "tilt" from being fully perpendicular ( cardboard edge towards your ear ) to 45° to 60° away from your ear ).

Crossover is a 12db passive line level LCR design for bass (low pass) and midrange (band pass). XO for the tweeter is just a 3.3uF cap in series with the tweeter.

- Your words imply that you have an inductor within your LCR type PLLXO. That means the inductor has to be a very large value ( in the "full" Henry range / as opposed to milliHenry ). I'm assuming the inductor is a potted ferrite core type / such as Marchand Electronics uses. Is this assumption correct ?
- Perhaps, you should publish some details about your PLLXO crossover ( ie; is it of your own design ? / or someone elses ? )
- What are the component values for this PLLXO ? .

[sourceoneaudio]

Hi,



Yes,
- Flip the polarity of the horn driver ( as compared to the woofer ).
.

And yes this is mentioned in my schematic calculation, sorry I forgot to mention it. Kinda of important. Pretty sure I mentioned it b4 though in above thread. Good luck and let us know how it sounds when completed.

J/S-S1A

[toddalin]

The easy way using your existing equipment is to simply play a tone at/near the crossover frequency and use your RS meter to determine which polarity on the horn produces the smoother curve (typically the louder volume). Then do the same for the tweeter.

Test tones can be from any available source such as an oscillator, computer, CD, or a musical keyboard.

[Doogster]

Hi,

- Do the preceeding only if you indeed have a 12 db PLLXO filter that is properly performing the crossover functions that you've described ( have you measured this PLLXO crossover to check its' performance ? ) .

- Your words imply that you have an inductor within your LCR type PLLXO. That means the inductor has to be a very large value ( in the "full" Henry range / as opposed to milliHenry ). I'm assuming the inductor is a potted ferrite core type / such as Marchand Electronics uses. Is this asumption correct ?
- Perhaps, you should publish some details about your PLLXO crossover ( ie; is it of your own design ? / or someone elses ? )
- What are the component values for this PLLXO ? .

Thanks for your reply Earl

In answer to your questions - yes, I have measured the PLLXO on my scope with dummy loads. It measures exactly how I intended it to (with a lot of tweaking!).

Yes, I am using large value pot core inductores, a la Marchand. I used John Broskie's Glassware Filter Designer to get the basic values. I was limited by the inductors I had on hand (I have about 50 different types I got from a salvaged ex-telecommunications filter unit). I then tweaked the design for the real world using my scope and a signal generator.

Values are as follows:

Bass low pass - 710mH and 0.2uF cap, terminated with 2k resistor.

Mid band pass - 230mH and 0.1uf cap (high pass) / 130mH and 0.0068uF cap (low pass), terminated with 2k resistor.

I only use a cap at speaker level for the tweeter. However, based on advice from others on this forum I will experiment with another PLLXO section for the tweeter.

Regards. Doug

[Earl K]

In answer to your questions - yes, I have measured the PLLXO on my scope with dummy loads. It measures exactly how I intended it to (with a lot of tweaking!).

Okay, that's great .

Yes, I am using large value pot core inductores, a la Marchand. I used John Broskie's Glassware Filter Designer to get the basic values. I was limited by the inductors I had on hand (I have about 50 different types I got from a salvaged ex-telecommunications filter unit). I then tweaked the design for the real world using my scope and a signal generator.

Good stuff / sounds like I need to get out and do a bit more dumpster diving .

- Do your inductors look like the type that Marchand sells with their XM-46 kit ?
- Not that I've mounted an exhaustive search for those things / but up here in Toronto, I've never seen them for sale at my standard suppliers .

Values are as follows:

Bass low pass - 710mH and 0.2uF cap, terminated with 2k resistor.

Mid band pass - 230mH and 0.1uf cap (high pass) / 130mH and 0.0068uF cap (low pass), terminated with 2k resistor.

- What's the input impedance of your referenced amps ( Perreaux & Set45 ) ?
( Off-hand , I don't understand the values that you've used to arrive at your goal / but then , these things are pretty finicky)

I only use a cap at speaker level for the tweeter. However, based on advice from others on this forum I will experiment with another PLLXO section for the tweeter.

- The last time that I listened to a 2404h, I concluded that it was best crossed into somewhere past the 7 to 9 Khz area. ( Widget/Zilch did do a nice study on JBL tweeters, that's somewhere in the archives ). The "FR" curves that they ran pretty well sum up why these things sound nicer when they're crossed over higher up. This "complication" would of course require a reworking of your bandpass filter. Oh well .

[Doogster]

- Do your inductors look like the type that Marchand sells with their XM-46 kit ?

- What's the input impedance of your referenced amps ( Perreaux & Set45 ) ?
( Off-hand , I don't understand the values that you've used to arrive at your goal / but then , these things are pretty finicky)

- The last time that I listened to a 2404h, I concluded that it was best crossed into somewhere past the 7 to 9 Khz area. ( Widget/Zilch did do a nice study on JBL tweeters, that's somewhere in the archives ). The "FR" curves that they ran pretty well sum up why these things sound nicer when they're crossed over higher up. This "complication" would of course require a reworking of your bandpass filter. Oh well .

Earl

Yes, my inductors look just like the Marchand jobbies.

The input impedance of both amps is 100K.

You are right - the PLLXO values I've used are not textbook. They do deviate a little from what the formulae say. However, they have been thoroughly measured so I don't doubt their accuracy.

I'll bear your suggestions on the 2404s in mind. I wanted to cross over at 5kHz because I had some peaks in the 2441 around 5kHz which I wanted to cure by crossing over where I did (that's probably not the best reason to cross over, is it!). Ideally I would have crossed the 2404s in at 7 to 10kHz and implemented a notch filter in the 2441. Or maybe the peaking I experienced with the 2441 was because I was running it without a tweeter (ie. from 700Hz all the way up to where-ever). I went with a PLLXO to simplify things. A 3 way PLLXO with PLLXO notch filter is going to take up a lot of real estate.

Do you think a 6db XO at 5kHz for the 2404 is OK, or should I use 12db as others have suggested?

Cheers. Doug

[sourceoneaudio]

Using a 6db network, and if you had a way to measure it via pink noise you will see all kinds of issues at the x-over point, not a sharp enough slope. Kind of like cutting a fine cheese with a butter knife, not clean, and all kinds of S&%^ sticks to the blade. Trust me you will be very satisfied wit the out come. Much smoother and a better transition from driver to driver. Plus if your any kind of a high volume listener it is very unsafe, you will either blow up the caps or the driver which ever decides to fail first. Good luck.

J/S-S1A

[Earl K]

Hi There,

The input impedance of both amps is 100K.

- A 2K terminaton ( load ) resistor ( 1% tolerance, I'm assuming ) across a 100K load ( unbalanced ) , gives a new working load impedance of @ 1960.8 ohms .

Values are as follows:

Bass low pass - 710mH and 0.2uF cap, terminated with 2k resistor.

(A) Thanks for that ! I would have expected that the usable range of cap values would be ; .082 to .066 uF, used with a 710mH coil ( to give an @ F3 of 700 hz ) .

(B) Running the numbers for the individual elements within your low pass filter gives individual F3(s) of @ 440 hz ( coil ) & 406 hz ( cap ) .

- With these 2 F3(s) so closely stacked on top of each other / I would have expected that you would be measuring some significant peaking ( high Q resonance ) in this FR area before actual attenuation takes palce .

(C) Bear with me here while I run through some math & buildup procedures .

(i) Since the 710 mH coil is your given.

- I would first test the accuracy of two assumptions. These are the net load impedance and the real value of the 710mH coil. I would do this by measuring for the F3 point of a single inline coil ( in this case, 710 mH ). The reference voltage would be set a good two octaves away ( in the passband ) from the expected area of the attenuation . 100 hz fits this criteria .
- eg. Use a 1 volt reference set to 100 hz. Sweep up in frequency till voltage drops to .707 volts when measured across the net load ( 2K resistor in place ). This voltage drop equates to the 3 db down point.

- Assuming all is well ( with the assumed load impedances and coil value ) , then an @ F3 right around 440hz ought to be measurable.

Some Math :
(ii) I'm going to play around with math that uses reciprocals, so ;
- 700 hz divided by 440 hz ( the first given pole in the filter ) gives 1.59 .
- Its' reciprocal is .62857 ( or .63 if you prefer ) .
- 1.6 ( I rounded up ) times 700hz is 1120 hz. This is the frequency where the second poles' theoretical F3 should occur ( when measured separately ).
- .625 times 700hz is 437.5 ( this closely approximates the F3 point of our 710mH coil working into a 1961 ohm load )

NUMBER CHECK:
- 1120 times 437.5 = 490000 . The square root of 490000 hz is 700 .
- For those who don't realize it, taking the square root of the product of the individual F3(s) , ( stated in hertz ) will give the F3 point of any 2-pole filter from this class. Since 2-Pole ; LR, Bessel & Butterworth filters are all derived from reciprocals they all fit this classification.

Reciprocals Used in calcs for some Common 2-Pole Filter Types :
- The respective ( rounded ) reciprocal #s are ;
Bessel at 1.74 & .57
Butterworth at 1.41 & .71
LR at 2.0 & .5
- Ours' is going to be ; 1.6 & .625

(iii)
- One can utilize our new " weighting numbers" ( ie; 1.6 & .625 ) to derive the filters' coil & cap values .
- One can see that the resulting filter will have a transform that is somewhere in between that of a 2 pole Bessel and that of a 2 pole Butterworth filter .

(iv) Using 1.6 & .625 as our weighting ( multiplier ) numbers ( when plugged into standard formulas for deriving coil & cap values for an F3 at 700 hz ) result in component values of, 713.3mH & .0725 uF ( respectively ) .

Sorry if I bored you with this exercise in derived math / but / math doesn't lie ( though in this case its' practitioner could always be missing a relevant formula / or missing information that will skew realworld results ).

You are right - the PLLXO values I've used are not textbook. They do deviate a little from what the formulae say. However, they have been thoroughly measured so I don't doubt their accuracy.

- I am wondering outloud about those working values in your PLLXO and how they're able to work for you / though I've only looked at the lowpass circuit for the woofer.

- I'll always defer to accurate measurements.