EV Dx38 active xover questions/observations

[DrWho]

Also, for a lot of the intellectual stuff you

said, you ought to direct that towards Mike, or at least someone who

better understands it than I.

How bout a simplified explanation? By turning off the time delay you

introduce new phase errors - the crossover kinda breaks meaning you

can't really do a direct AB comparison between time delay on and off.

That's why I suggested comparing a system optomized with time

compensation and one optimized without it - and that'll be easier to do

when there is no delay between presets.

It's kinda like go-karts....you really can't compare driving skill of

big fat momma chick versus her scrawny young child. The kid is simply

going to have better lap-times until you start getting into vehicles

where the weight difference isn't an issue [] Maybe you can give big

momma a slightly more powerful motor to achieve the same power to

weight ratio, but chances are the powerband of the beefy motor is going

to be different (thus no longer an AB comparison). The phase response

in this situation parallels to the extra weight.

Btw, I just today realized that we're using a Dx38 at the Illini Union.

I usually don't notice the specific model #'s of the gear I work with,

but all this talk about this unit brought it to my attention. I've been

in love with this unit since last spring and after reading some of

these posts about how the unit works I have even come across a few more

features that proved extremely beneficial tonight. Having the option to

set the global system delay is just plain awesome (ton easier than

adjusting each channel individually). We had 14 foot worth of stage in

front of a large wall and were able to put the front of house in

alignment with the reflection of the monitors off the wall (24ms) which

dramatically improved the sound. So thanks Coytee for all the

discussion of your playing around - I learned something mega cool [H]

So about that wife...where does a single guy like me go shopping for one that's a keeper? []

[Coytee]

She's got a (single) younger sister but I still think from your perspective, you'd still have to like older women! [6]

[Al Klappenberger]

RoyBoy,

The only absolute is that you might learn something about filter design.

Al K.

[Al Klappenberger]

Edgar,

Maybe you might be able to give me some input on passive crossovers. It sounds like you know a few things about active ones. Your comment about having the phases 90 degrees apart at the crossover is what I have come to think may be the right thing too. The extreme-slope networks I build are inverted images of each other and are 180 degrees apart at the crossover, which is a very narrow band of frequency because of extremely sharp skirts. This says it would only be necessary to phase it right a one spot. It would seem to me that the difference in time dealy between the woofer horn and high frequency horn would throw that 180 degree relationship off. It would have to fall plus or minus from 180. With luck it might hit 90. If not it would be easy to add inserttion phase to either filter in 45 degree steps.The 6 dB crossover loss is automatic becasue the actual zero-loss crossover is the half power point (3.01 dB). The losses in a passive network add to the 3 db putting it pretty close to the nice 6 dB number. What do you think of that theory?

Al k.

[Edgar]

Your comment about having the phases 90 degrees apart at the crossover is what I have come to think may be the right thing too.

It depends upon the application. In a home situation, where the listener is in the near field of the loudspeaker, a 90 degree phase difference at crossover probably doesn't hurt much because the resultant main lobe of the response gets sent into the floor or ceiling. In a sound reinforcement application, though, that main lobe gets pointed at a section of the audience, and is very audible to them.

The extreme-slope networks I build are inverted images of each other and are 180 degrees apart at the crossover, which is a very narrow band of frequency because of extremely sharp skirts. This says it would only be necessary to phase it right a one spot.

Essentially true, because the response of the non-dominant driver falls so steeply that it quickly becomes negligible.

It would seem to me that the difference in time dealy between the woofer horn and high frequency horn would throw that 180 degree relationship off.

Correct.

It would have to fall plus or minus from 180. With luck it might hit 90. If not it would be easy to add inserttion phase to either filter in 45 degree steps.The 6 dB crossover loss is automatic becasue the actual zero-loss crossover is the half power point (3.01 dB). The losses in a passive network add to the 3 db putting it pretty close to the nice 6 dB number. What do you think of that theory?

I don't completely understand what you are proposing. Determining the response at the crossover frequency (or anywhere else, for that matter) is pretty simple -- express the magnitude and phase of the drivers as complex numbers, and add them together. The magnitude of the sum is the magnitude of the combined response; the phase of the sum is the phase of the combined response. If each of your drivers has an amplitude of .7071 at crossover, and they are 90 degrees apart in phase, then their sum will have a magnitude of 1.0. If each of your drivers has an amplitude of 0.5 at crossover, and they are in-phase, then their sum will have a magnitude of 1.0. If they are 180 degrees out of phase, then you'll have to invert polarity on one of them. This is done in some Linkwitz-Riley crossovers, for example.

Greg

[Al Klappenberger]

Greg,

It's nice to run into somebody who knows something for a change!

Could you elaborate about how the main lobe goes to the floor or ceiling. I am not well versed on acoustic (or even electrostatic and electromagent fileds) develop or interact.

********

The Linkwitz-Riley crossover, in L-C form at least, will sum the two channels to a very nice flat response, in a computer, where all the components are perfect. Each filter is 6 db down at the crossver. An even order Butterworth sums to +3 dB in the same computer. BUT: in the real world, the parts have losses that add to the attenuation at the crossover frequency through each filter. This loss reduces the summation amplitude. If you pick the "Q" if the parts right, it could be made to sum flat as well, in the real world. The idea is an agressive filter (pole placed with one finite zero) to drop the response through each filter like the side of a barn. This will require higher Q parts than a gentle slope filter. The idea is the same. In the 400 Hz extreme-slope filter I build I have managed to get the loss at the crossover so that it is slightly over 6 dB. I think this gives me the Linkwitx-Riley sumation, at lest in amplitude. The phasing would be a function of the delay difference between adjacent drivers. Once that delay difference is know, the insertion phase of the filters cold be adjusted by adding elements to add skirt sharpness at the same time. That is the idea anyhow.

AL K.

[Edgar]

Could you elaborate about how the main lobe goes to the floor or ceiling.

An example is best. Let's say that the low frequency and high frequency driver responses are 90 degrees apart, and that the drivers are physically separated by some distance -- the HF driver is physically above the LF driver. If the listener is on-axis, the two responses may sum to 1.0. But at some vertical angle above or below the axis (usually below), the phase difference between the drivers is exactly cancelled by the distance between the drivers, so they now add in-phase. At some angle in the opposite direction they add out-of-phase, again because the physicsl distance corresponds exactly to the phase difference. In a living room, the listener is almost always within a few degrees of the vertical axis, so this effect is only noticed in reflections off the floor or ceiling. In a concert hall, though, that axis of "in-phase", and its corresponding axis of "out-of-phase", shine like searchlights over the audience, moving around because phase varies with frequency.

In the 400 Hz extreme-slope filter I build I have managed to get the loss at the crossover so that it is slightly over 6 dB. I think this gives me the Linkwitx-Riley sumation, at lest in amplitude.

Only if both amplitudes are equal, and both are 1/2 the passband amplitude.

The phasing would be a function of the delay difference between adjacent drivers. Once that delay difference is know, the insertion phase of the filters cold be adjusted by adding elements to add skirt sharpness at the same time. That is the idea anyhow.

Oh, I think I understand what you're trying to do. If your amplitudes are correct, but your relative phases are not, then yes, you could conceivably use delay to move the relative phases to 0 or 180 degrees at the crossover frequency. This would eliminate a peak or notch in the overall response at the crossover frequency. But it would do little to smooth the overall phase (or delay, if you prefer) response of the system across all frequencies.

Greg

[DrWho]

wow, what the heck are you doing designing crossovers if you don't understand what's going on with the acoustics?!? The ONLY data that matters is how the response of the acoustic system behaves - not how nice the electrical world looks. Sorry for the vent, but the acoustical summing is what it's all about. []

Why not try using some software that takes the real electrical behavior of the driver into account along with the measured response of the driver?

http://www.linearx.com/products/ProductsTop.htm

[Al Klappenberger]

Greg,

I gotcha! I think you are describing ordinary "comb" effect, only vertically. I usually think of it in the horizontal plane.

Yes, the two filters would be phased up to add at the crossover on-axis, but off-axis you would still get the comb effect. That would only happen over the frequency range where there would be enough energy from both drivers to interact. The thinking here is to make that frequency "window" extremely narrow and just live with it. There is just no way to generate actual time delay with a passive network. Everything is a comprmise it seems!

I have been doing this at 6Khz to cross between a midrange and a tweeter. I made no attempt to phase them considering the wavelength is so small compared to the physical separation between the two. I notice a big improvement in stereo image. A smearing I have noticed for years went away too! I am just thinking it could actually be taken a step further in a woofer / midrange crossover. This could give you a good stereo image, very minimal lobing and a virtually flat 8 Ohm (or whatever) resistive load to the amp over the entire frequency range. The actual resonance of the woofer drive would be the only expection.

Al K.

[Al Klappenberger]

Who,

You are absolutely right, the acoustics and the crossover should be combined into a single science. This is what I hope to do someday. The actual facts are that most of the stuff that is out there is so elementary that it seems that anyhting goes. At least this is what I see in the home entertainment loudspeaker field. In the theater and professional fields things might be better, but I doubt it. My background in serious filter design shows me that the crossover designs are very poor from a pure filter design standpoint. What I need is a serious coustic guy to work with to complete the "marrage" of the two sciences. Greg is the kind of guy I am looking for. Too bad we don't have a real life project to do it with. Like the Jubilee for example (HeeHaw).

Al K.

[Deang]

"I am not well versed on acoustic (or even electrostatic and electromagnetic fields)..."

Uh oh -- I see some humility creeping in.

[Edgar]

I gotcha! I think you are describing ordinary "comb" effect, only vertically.

Essentially, yes. If the delay is constant you get a comb filter. If the delay varies with frequency, you get "searchlight" lobing.

The thinking here is to make that frequency "window" extremely narrow and just live with it.

Yes, that is one way to deal with it. The other is to use broadband transducers and crossover topologies that are in-phase at all frequencies, like Linkwitz-Riley or Bessel-Derived Matched-Delay Subtractive.

There is just no way to generate actual time delay with a passive network.

You can approximate it with a string of allpass filters.

Greg

[Edgar]

What I need is a serious coustic guy to work with to complete the "marrage" of the two sciences. Greg is the kind of guy I am looking for

Um, I'm not an acoustics guy. Signal Processing is my bag.

Greg

[Al Klappenberger]

Greg,

"Searchlight" lobing is a term I have never hear before. Could you expalin that? Does it simply mean the lobes move around? That would make sense.

Dean,

Humility! ME! NEVER HAPPEN! Whatt'ya trying to do? Mess up my image!

Al K.

[Edgar]

"Searchlight" lobing is a term I have never hear before. Could you expalin that? Does it simply mean the lobes move around? That would make sense.

It's just a term that I use to describe it. If you have low-order (shallow slope) crossovers, and the relative phases of the LF and HF drivers vary with frequency, then the angle at which the two sources add in-phase also varies with frequency. If you think of that as a lobe of higher amplitude response, then it moves around as the frequency is changed, like a searchlight.

Greg

[Al Klappenberger]

Greg,

Ok.. Thanks for the expanded vocabulary. Buzz-words like that will help make me sound smart!

Seriously though, I think the sharper skirts may not help with that. I think It would even make it worse at the crossover. The point of maximum phase slope is usually the 2 dB point of most filters. That's the group delay peak. That would be within the "interaction window". It would simply be the trade off for better performace elseware.

Al K.