TOPIC:

Hi,

I am using the DDRC-24 for both active crossover and Dirac Live correction. I am trying to understand if I am using Dirac properly; it seems to be degrading impulse response in my system (see attached impulse response screetshot). As you can see, the impulse is more concentrated in the "before" state than it is in the "after" state, although the "after" state does seem to put overall more energy in the beginning of the impulse, even if it's smeared over more time:



I will say that Dirac does improve the sound by reducing coloration, so I'm not unhappy with the subjective result. It's more that I am wondering if there is still further room for improvement.

My speakers are JBL 4344 Mk2, which are 4-way dynamic speakers each with a 15" woofer, 8" midrange, 1" HF compression driver with short horn and acoustic lens, and a slot tweeter for the top octave. The speakers have an internal passive crossover on the top three drivers at 340, 1300, and 8000 Hz. The speakers have a switch for "normal" passive crossover, which assumes a single amp driving the entire range and a "bi-amp" setting which by-passes the 12dB/octave LP filter @ 340Hz on the 15" woofer and a 6dB/octave @ 340Hz HP filterstage on the 8" midrange (which still leaves 12dB/octave HP filtering to protect it). So, I'm using the "bi-amp" setting with two amps, and I've set up the DDRC-24 with a 12dB/octave LP and 6dB/octave HP filter for the crossover, which emulates the native crossover.

The speakers are about 6 feet apart (from the near edges of the baffles) and 1.5 feet from side walls.

Here is the spectrum response:



The thing which jumps out at me is the big dip in the "before" response around 120Hz. I have suspected that this is related to side wall proximity. JBL recommends a minimum of 3 feet, and I am much closer, but moving the speakers further from the walls brings them too close together, so it's a compromise given the room's dimensions.

Any ideas of whether the "after" impulse measurement indicates a problem and what I could do to correct or improve it? Could it be related to fixing up the dip at 120Hz? My understanding of Dirac is that changes in the central impulse response would be related to driver positions and baffle diffraction effects. If my calculations are correct, the time "smear" in the "after" impulse reading corresponds to about 8" or so of sound travel, which makes me wonder if we are seeing some interaction of the mid-range and woofer, which are spaced by about that much. The other thing I'd wonder about is the "acoustic lens" assembly attached to the high-frequency compression driver horn, intended to disperse the output so that it has directivity comparable to the 8" midrange below it. This acoustic lens is used in the "before" impulse measurement, so it alone isn't causing the problem, but maybe what it does is confusing Dirac?

I've tried numerous other active crossover filtering, including BW 24, BW 48, LR 24, and LR 48. They all show more or less the same kind of effect on the impulse response.

Many thanks for your help!

- Ken

Interesting. I don't have an answer but it might be helpful to zoom in on the impulse more. You may wish to try using just a single-point measurement with the mic closer to the speaker if you want to investigate. (Or use REW.)

Unrelated, I would suggest taking out the 10dB boost at 20 Hz, as that is a ported cabinet and you are boosting below its tuning frequency. (Won't affect the impulse response.)

Thank you, John. Here are zoomed-in details for the before/after impulse graphs.


Before:


After:

I've done some more testing. First, I was able to rule out effects of the acoustic lens by trying the same test with and without the lens attached--this had no effect on the impulse response. I tried running with the speakers' native passive crossover (eliminating the variable of the external active crossover), and got the same result. I even tried narrowing the spectrum of correction and eliminating Dirac's work on frequencies higher than 250 Hz, and this, too, had no effect on the impulse response shape. So, it's none of those things.

In part, I think I've been thrown off by the default display of impulse response as an average. If I click "all" I get more detail, which does show a more ideal shape of the "after" response:



Also, I noticed that if I take just one (the first--center) mic reading, then I get the expected ideal response pulse. The "smearing" is at least related to the gathering of the usual 9 samples in different positions on my sofa. I did try narrowing the width of mic positions to be closer to the center, and this changed the shape of the impulse response considerably:



I am wondering now whether what I am seeing is at last in part the result of reflections from the sofa back/headrest, which is roughly 8 or 18 inches behind the mic at the various test positions.

Do representatives from Dirac participate in this support forum? It would be great to get some feedback from the experts!

Thanks,

- Ken

You may want to try using a single-point measurement...

Are the speakers near the walls?

Yes, I tried the single point measurement and it did produce a more focused impulse response. That suggests to me that the "smear" is due to the best-effort compromise of the Dirac algorithms to produce a pleasing result in more than one point. (I didn't bother listening to a single-point measurement since it would be an unreliable correction.)

The speakers are about 18" away from side walls. They are very wide, so figure the mid-range drivers are about another 6" further than that. JBL recommends at least 3' distance from side walls but I don't have a big enough room to do that.

Could you post it? (The single point measurement.)

I'm really just curious, the impulse response uncorrected looks ... unusual. But hard to tell much.

Here's the result with a single sample. Looks like exactly what I would expect. So, the question is, then, why this isn't happening when I take the usual 9 samples?

I just listened to the result of the single sample correction. It certainly sounded different, mainly because deep bass (< 60Hz) was more than 6dB stronger. (It appears that the center reading is sitting in something of a null for that register, so the other readings moderate the correction.) Just listening to a few samples of music is not enough for me to meaningfully characterize the other changes subjectively, but it might have been a bit brighter as well. As an experiment this was interesting, but I know that this isn't how you are supposed to use Dirac.

I also tried collecting the usual 9 samples with the mic positioned well above the sofa's headrests, to see if their reflections were involved. Again, the aggregate correction impulse response looks like the ones I showed before, so headrest reflections don't seem to be the issue.

I'm suspecting it's an issue with the way the "average impulse response" (whatever that is) is displayed... try measuring (single point) at different positions to see?

Hi John,

I think I got to the bottom of this. The challenge here seems to come from the behavior of the horn-loaded HF driver, which creates an unusual impulse response. I separately contacted Dirac Tech Support and they were puzzled by the overall speaker "before" impulse response, but couldn't say why it was happening. I did more testing on restricted frequency ranges and found that the issue is isolated to this one horn-loaded HF driver:



The UHF driver doesn't seem to have these anomalies:



Combined, however, they give the two-pulse signature that dominates the full-range impulse response picture:



So, I think that Dirac is doing the best it can with a strange sound source. I would be interested to hear about others' experience with horn-loaded systems. This type of narrow horn/acoustic-lens assembly is completely obsolete, and perhaps this is why. (It was used in the early 70s and was already an anachronism when the speaker was launched in 1996, but traditions die hard.)

Thanks again for your help!

- Ken

This is the horn? I guess that explains why its impulse is delayed

Yes, I believe those are the 2307 horns. I looked them up: they are 7.75" deep, which is about the amount you'd expect from the impulse response graph. The reason, I suppose, that the impulse responses are different at different horizontal sample positions is that the UHF driver is offset horizontally from the HF horn by a few inches, so the delta varies by +/- 6" over the sample zone on the sofa. The wider the sweet spot, the more that the HF/UHF correction is frustrated. I believe that time correction is not done (very) much on the high frequencies, so perhaps it doesn't matter.

I guess the horizontal alignment of high frequency drivers is something else you don't see these days either. I still doubt the value of the "average impulse response". It shows the initial impulse much lower in value, which is not the case in the single point measurement.