PDA

View Full Version : Crossover help



dmtp
09-23-2006, 11:11 AM
I am building a system with an LE14A woofer in a ~4cf cabinet and an LE85 driving a 500Hz Tractrix horn. For the initial try at an XO, I copiesd the L200B which has a second order roll-off for the woofer and a third order for the LE85 plus a high freq boost. I tested the upper section separately and found it did, indeed, roll-off at about 18dB/oct below 1000 Hz, but there was a broad peak centered at ~7000. I tamed that with a series LRC filter (6uF, 0.1mH, 30ohm) and got a nice flat response. The problem is that when I hook up the bass section, all of a sudden the nice 18dB roll-off of the horn goes crazy and it still has significant output (down only 5-6 dB to 100 Hz! I tried running the woofer full range (no network) and the same thing happens. I am getting feed-through from the woofer into the horn somehow screwing up the response!
HELP!!!:banghead:

Zilch
09-23-2006, 11:32 AM
Hi, Mark.

Take the LCR out and see if you still get interaction.

I'm thinking it matters where you place it in the circuit.

Measure the voltage drives if you have the facility to do that.

I worked with L200B circuit in the "Keeper" crossover project without a problem, though I never added any notches.

Post your schematic, please....

dmtp
09-24-2006, 08:39 AM
OK, here are some images I hope will help. I have compared the schematic (sorry for it being hand drawn) to the actual XO too many times to count. I'm not sure if you can trace it or not. The #10 screws sticking up are for attaching leads from input and to output. The original construction (shown in hipass below) did not have the 2.5/5 ohm fixed L-pad. This was added as can be seen on the (admittedly messy) pic that shows the XO with both L-Pads sitting on the speaker. I have tried it with and without the notch filter and the results (in terms of roll-off below 1000) is the same (you can see the notch filter was just held in place with clamps for testing. My test set-up consists of an HeathKit freq generator, an old Marantz amp, and a Radio Shack SPL meter. (Not exactly high tech, but I get repeatable results within 1-2 dB.) I also have a set-up with a DVM to measure impedance. I have not tried measuring voltage output, but I don't see why I couldn't follow the JBL standard set-up. I guess that is next.
Images follow:
#1 - low pass schematic (note I have listed combined cap values and eliminated by-pass caps for clarity)
#2 - low pass pic
#3 - high pass schematic
#4 - original high pass pic (lacks 2.5/5 fixed L-pad)
#5 - notch filter schematic
remaining on next post!
#6 - pic of final high pass XO with notch filter and L-Pads
#7 - test set-up (freq generator and amp with XO sitting on top)
#8 - test set-up (speaker and SPL meter)

dmtp
09-24-2006, 08:41 AM
Here's the rest:
#6 - pic of final high pass XO with notch filter and L-Pads
#7 - test set-up (freq generator and amp with XO sitting on top)
#8 - test set-up (speaker and SPL meter)

Earl K
09-24-2006, 09:57 AM
Hi,


I am building a system with an LE14A woofer in a ~4cf cabinet and an LE85 driving a 500Hz Tractrix horn. For the initial try at an XO, I copied the L200B which has a second order roll-off for the woofer and a third order for the LE85 plus a high freq boost.
Overall, IMHO a pretty wise choice .

HELP!!!
Okay, here goes .


I am getting feed-through from the woofer into the horn somehow screwing up the response!

Yes, your measurement results will be skewed if you are measuring the horn response ( & looking for your 18 db per octave rolloff ) with any woofer connected to the low-pass portion of your network . Acoustic "feed-through" ( as you call it ) will fill in the measured horn response ( below 1000hz ) / even when your test mic is in close proximity to the horn .


The problem is that when I hook up the bass section, all of a sudden the nice 18dB roll-off of the horn goes crazy and it still has significant output (down only 5-6 dB to 100 Hz! I tried running the woofer full range (no network) and the same thing happens.

- The measured 18 db acoustic rolloff will indeed disappear when the woofer is attached and acoustically adds it's output into the test mic.
- When "searching" for " electrical interaction" between the high pass and low pass portions of a new network / connect only one "transducer at a time ( & then make your acoustic measurements ). Connect an appropriate "dummy load" on the output of that portion of the network which has no transducer attached to it . Ie; test the acoustic response ( & acoustic rolloff of the filter/horn combo ) with an 8 ohm resistor substituted ( for the le14a ) as the load ( on the lowpass portion of your network ) .


,,, but there was a broad peak centered at ~7000. I tamed that with a series LRC filter (6uF, 0.1mH, 30ohm) and got a nice flat response.

- It's very encouraging that you found a solution to this "response anomaly for yourself. :applaud:
- FWIW; your "series LCR" is usually referred to as a "Parallel LCR" oriented in series within the overall circuit .
- IME, a better choice LCR ( trap-filter) , would be a "Series LCR ( the components are connected in series with each other )"/ strapped in parallel across the load ( after the variable Lpad ) .
- The reason for this preference ? the "Series LCR" across the horn load, adds a nice bit of frequency dependant damping into the horn circuit / which I find is always necessary when using these drivers. The parallel LCR doesn't offer nearly the same level of added damping ( for some reason though I'm no E.E. so cannot really describe why ).


I am building a system with an LE14A woofer in a ~4cf cabinet and an LE85 driving a 500Hz Tractrix horn. For the initial try at an XO, I copiesd the L200B which has a second order roll-off for the woofer and a third order for the LE85 plus a high freq boost.

That's a very nice looking horn you've turned. I like the combination of transducers/horn, that you've chosen for your project


:)

dmtp
09-24-2006, 10:16 AM
WOW!
Thanks for all the thoughtful info. I will go right back and test things as you describe (now why didn't I think of that!:banghead:)
Anyway FWIW, I did try doing voltage measurements as described in the JBL "standard test fixture." Maybe my confusion is their spec of a "VU" meter. I used a DVM set to 'AC volts'. The results are confusing (but repeatable) On the lo pass XO I started at an arbitrary 10mV @ 100 Hz. The voltage rose smoothly to about 27 mV @ 400 and then fell smoothly to 3 mV @ 3000. :screwy:
The high pass was a bit more understandable. Again starting at an arbitrary 10 mV @ 1000 this time, it rose smoothly to a peak of 155.7 mV @ 12700 and then fell to 80.6mV by 18000. In the XO region, (more to the point) it fell steeply from 10mV @1000 to 0.3 mV by 450. Attaching the low pass XO had no effect on these numbers. I next tried using the real transducers instead of the dummy resistors with essentially the same results.
My concern is that even with the DVM reading 0.1mV @ 200 Hz, there was still audible output from the LE85 and it did not sound too happy. I am worried about damaging it with real world volumes. Am I just paranoid (a distinct possibility) or is this a true concern?

dmtp
09-24-2006, 10:21 AM
- FWIW; your "series LCR" is usually referred to as a "Parallel LCR" oriented in series within the overall circuit .
- IME, a better choice LCR ( trap-filter) , would be a "Series LCR ( the components are connected in series with each other )"/ strapped in parallel across the load ( after the variable Lpad ) .
- The reason for this preference ? the "Series LCR" across the horn load, adds a nice bit of frequency dependant damping into the horn circuit / which I find is always necessary when using these drivers. The parallel LCR doesn't offer nearly the same level of added damping ( for some reason though I'm no E.E. so cannot really describe why ).


OK, I'm game to try it. On the "Keeper" thread, I notice a LR across the horn which was described as a conjugate. Coud this be doing the same thing? (I haven't tried it.) If not, how do I determine the values for a Series LCR across the horn? I used formulae from Lance Dickenson's Cookbook as a startng point for mine and then tweaked from there (increased C to lower peak of trap).
Thanks again for all the help!

dmtp
09-24-2006, 11:40 AM
Earl K was, of course, 100% correct. It was all a mistake of measuring. With the dummy resister in place of the LE14A, the horn still has a nice roll-off below 1000, just like it did before.
So now I can foret about that "problem" and work on a series LCR to parellel the horn!

Earl K
09-24-2006, 12:44 PM
Mark,

- I'm out for a good portion of today / so, I'll need to get back to you later about designing Series LCR filters.

- In the meantime; if you want to electrically measure ( and document ) how well your network sections are performing / continue to measure them ( with a horn/driver combo ) as a load and an 8 ohm resistor acting as the woofer load.


My concern is that even with the DVM reading 0.1mV @ 200 Hz, there was still audible output from the LE85 and it did not sound too happy. I am worried about damaging it with real world volumes. Am I just paranoid (a distinct

It's a valid concern that's best addressed ( to set aside your fears ) by doing voltage measurements on the network sections / and then turning those voltage measurements into a measurement of real "dB attenuation".

- Measure the voltage of the horn/driver combo at 1000 hz , then 500 hz & then 250 hz ( a reminder ; these measurements are taken at the horn drivers' input terminals ). Obviously, these are 1 octave increments , chosen to help assess a filters performance on the basis of "per octave" attenuation . I'm assuming 1K is again your .1mv reference level .

- A .1mv reference reading at 1000hz should now drop to the neighbourhood of .0126 mv ( for 500hz ). This is if you have constructed a 18 db per octave filter ( between 1000 to 500 hz ) that has an actual -3 db point of 1000hz .
- Take another voltage reading at 250 hz and plug the new reading ( along with the voltage value for 500 hz ) into the following formula .

20 * Log ( voltage reading of lower frequency / voltage reading of higher frequency ) = attenuation ( as a minus number ) .

example ; 20 * Log ( .0126/.1 ) = - 17.99 db of attenuation .
( LOG of .126 = -.899629455 )

regards

Zilch
09-24-2006, 12:48 PM
On the "Keeper" thread, I notice a LR across the horn which was described as a conjugate. Coud this be doing the same thing? (I haven't tried it.)You mean L3 and C6?

http://audioheritage.org/vbulletin/showthread.php?p=93836#post93836

That's the lowpass filter shutting the horn down at 7 or 8 kHz when the UHF driver is added. It came from N7000/8000/3105/3106.

Going back and reviewing that project reminds me of what great fun you must be having with yours! Like Earl says above, you've made good design choices. 800 Hz may be better sounding than 500 Hz; I have no experience with Tractrix horns on LE85. LE14A will like either just as well.

Don't minimize the utility of your test gear. You've again demonstrated that much can be accomplished with what many members already have on hand, and with considerable precision.

Thanks, Earl, for your assisting Mark here. It'd have been at LEAST three weeks of circuit tweaking before I recognized that he was running and reading both drivers concurrently. :p

Zilch
09-24-2006, 01:18 PM
Some insight into the operation of the L200B circuit appears here:

http://audioheritage.org/vbulletin/showthread.php?p=42450#post42450

The available compensation and slope depends upon the MF L-Pad (fixed plus adjustable) setting.

Here's the original circuit for reference:

http://manuals.harman.com/JBL/HOM/Technical%20Sheet/L200B%20ts.pdf

I see you've made the VHF boost bypass loop adjustable per 4430:

http://www.jblproservice.com/pdf/Studio%20Monitor%20Series/4430LR.pdf

Cool. I've never tried that. :thmbsup:

dmtp
09-24-2006, 04:06 PM
You mean L3 and C6?
Sorry, I was referring to this schematic:
http://audioheritage.org/vbulletin/showthread.php?p=71276#post71276
The 2mH + 14 ohm across the HF.

. 800 Hz may be better sounding than 500 Hz; I have no experience with Tractrix horns on LE85.
Actually the tractrix curve has a 500 Hz cut off, but I am using a XO of 800-1000 (same as the 200B).

Zilch
09-24-2006, 04:41 PM
A different topology achieving a linear compensation curve.

The L/R conjugate acts as the highpass in that circuit.... :thmbsup: