Long day's journey into DSP EQ

[suttondesign]

Just thought I'd share with the forum my EQ journey switching out of the Linktwitz dipole (521.4) into a good, DIY 2-way (Heissmann DXT-MON-182). I wanted to keep the open-baffle bass module and my two potent dual-opposed SEAS 9" and 11" subs that take the strain off the OB bass modules, which bottom-out below 30hz in a decent-sized room.

I was having a lot of trouble integrating these three. The monitors sit directly atop the OB bass module, so it was pretty much an issue of matching level and keeping the Linkwitz system EQ. That has proven to be the case. And inverting the monitors, which caused a big null around 120hz, showed that the phase was correct the first time (whew). But trial-and-error also showed me I needed to toe in the OB bass modules along with toeing in the mains. Not sure why.

The biggest problem has been integrating the two big subs. I was getting some big nulls at first and a huge bump at 60hz, and then after trying various things, like high crossovers, comb filtering. But using REW and tweaking over and over, I surprised myself by finding that the answer lay in crossing over the big dual-opposed subs at 60hz instead of down lower. And I added a bit of delay to one of the subs. This drastically smoothed out the response below 120hz (the crossover from OB to monitors). Not sure why the Linkwitz OB bass modules and the separate subs play well together around 60 hz (I'm baffled, so to speak). But measurements don't lie.

A bit of further tweaking to each sub based on what each was contributing further smoothed things out. It's still a room, so there are peaks and valleys, but nothing nearly so bad as before.

So, as DIY goes, it's darn good, if a bit complex what with 2 plate amps for the subs, 4 channels of amplification for the OB bass modules, 2 outboard mains crossovers, and a MiniDSP flex 8. But, you, know, why not.

 

[skankhobag]

A couple points:

• Your crossover from the main speakers to your subs should be at least one octave above the mains’ low frequency resonance. That bottom octave will be the area where the mains distort the most and the subs perform the best.

• Using both an open baffle and closed box in stereo give you somewhat of a cardioid response for your left and right channels (dependent upon where they’re placed). My thought is that you’d want the dipole subs placed closer the side walls since their radiation pattern nullifies side output.

Consonus Audio (https://www.consonusaudio.com) has an active speaker with a 13” sub and a 9” woofer with overlapping output. The 9 incher smooths out a couple room modes.

 

[suttondesign]

i cross the mains at 120hz, well above the area in which they strain. distortion is extremely low for the mains.

moving the ob bass modules is not feasible.

 

[KSTR]

I was having a lot of trouble integrating these three.

This I can imagine.

While trial and error works, sort of, however it may often be better, more rewarding, to use a highly systematic approach.
That is, you start with defining reasonable acoustic target curves (magnitude and phase) for all three systems so that they perfectly integrate by design once those target functions are really met.

A suitable target could be a 4th (or 6th) order Linkwitz-Riley between sub-bass and bass @~50Hz and a LR4 (or LR2, depends on the bass roll-off of the mains) between OB bass and mains @~120Hz. Linkwitz-Riley (== matching phases) is strongly preferred here because it is the most tolerant for residual phase and time-of-flight mismatches.

I personally am not a fan of having so much XO down low as that makes the low bass always lag behind from a ton of group delay (but then again the issue can properly addressed by FIR phase-unwarping of the source signal, upstream of the XO). I see there is no other way in your case with the given building blocks.

((
Side topic: The all important thing to note is that the slopes/shape of those target functions is not the simple textbook Linkwitz-Riley with 4th order roll-off because the XO frequencies (and the sub-woofers' system highpass behavior) are narrowly spaced and thus phase response of the adjacent XO (or the roll-off) distorts the simple LR target response.
Textbook target functions only work when the next XO (or highpass) frequency is at least a decade or so away.
In practice one would use the allpass version of the adjacent XO point's response and multiply it into the textbook target. Sometimes one can also get away with using other simpler functions like shelves etc. Another core principle of XO design is once you have matching phases, the response automatically is correct Linkwitz-Riley at the moment when the response is globally EQ'd to flat. That's why matching the phases is paramount and the first step.
))

Then you "only" have to bend the raw response (magnitude and phase) of the the three subsystems to match the targets and you should be 90% there.
The problem of course is that you don't have anechoic measurement of the raw responses. For the mains and the sub-bass you could use nearfield (@cone) measurements and for the OB bass one could use nearfield as well, then multiplied with the analytical modification to turn it into far-field OB response with the typical dipole roll-off and peak.

While the inverted polarity check is useful, it tells only part of the story. Even with a deep null you don't know how phase develops left an right of the XO frequency, and more importantly, you don' know if the match is off by a mulitple of 180degrees.

(Edited some stuff).

 

[dualazmak]

Hello, OP @suttondesign...

I assume my post here wound be somewhat of your reference and interests...
Summary of critical factors in integration of subwoofer(s) and main woofers in our individual acoustic environments

Also, I wrote here as follows;
No matter whatever steep-slope (or mild-slope) crossover filters (LPF for subwoofer, HPF for woofer) would be used, we always have the Fq-zone where sub-woofer and woofer do sing together in almost same gain. The careful optimization is needed/indispensable, therefore, on selection of XO-Fq, XO-filter-type (BW, LR, Bessel, etc.), both-side slopes, phase matching (smooth phase continuation, invert or not), time alignment (group delay), gain matching, etc., etc.
I believe the objective and visual observation of tone-burst sine waves of suitable Fq "in the XO Fq zone" and the 3D color representation (X:Time, Y:Fq, Color:Gain/SPL) thereof, both microphone-measured at listening position, would be most suitable and reliable for such XO optimization between sub-woofer and woofer, as I have done in #495, #503, #507.

 

[KSTR]

@dualazmak
Yes, the shaped tone-burst method is the ultimate way of double-checking if the phases track. For both ways the waveform must be basically identical, minutely in all aspects (notably start time and exact shape) at the XO point, and left and right of the XO point it must be identical too, except for the level.
This is what matched phase and therefore matched group delay really means, all ways put out the exact same waveform at all frequencies (except for level) for a perfect summation.

 

[dualazmak]

@dualazmak
Yes, the shaped tone-burst method is the ultimate way of double-checking if the phases track. For both ways the waveform must be basically identical, minutely in all aspects (notably start time and exact shape) at the XO point, and left and right of the XO point it must be identical too, except for the level.
This is what matched phase and therefore matched group delay really means, all ways put out the exact same waveform at all frequencies (except for level) for a perfect summation.

I sincerely thank you for your kind follow-up of my above post #5; it looks we two have the same perspective and actual implementations on optimal tunings between the SP drivers, especially integrating sub-woofers with woofers, in DSP-based multichannel audio system.

 

[Keith_W]

@dualazmak
Yes, the shaped tone-burst method is the ultimate way of double-checking if the phases track.

I disagree with that. The problem with the tone burst is that the length of the sinusoid wavelet is too long, especially at low frequencies. The front of the sinusoid reflects and contaminates the direct sound before the entire sinusoid is produced by the loudspeaker, making interpretation very difficult! It is virtually useless for aligning a subwoofer to a woofer, for example. It is only useful if you have a free-field. It's probably OK for aligning midrange to tweeter, but there are much simpler methods. Like simply time shifting one driver until the phase cancellation disappears. That is, after all, the goal.

I have experimented with this. It's a fun exercise, but it's a bit too difficult to describe to people without confusing the heck out of them.

 

[KSTR]

It is virtually useless for aligning a subwoofer to a woofer, for example. It is only useful if you have a free-field.

I find it perfectly valid for that exact purpose, at least when the sub and mains are stacked (so that the contribution of room modes is fairly similar) since we compare the responses for differences rather than looking at the absolute data.
It does also work well when you have near field data (which might need some pre-processing to be useful, like splicing port and cone measurements, apply OB correction, etc)

 

[dualazmak]

It's probably OK for aligning midrange to tweeter, but there are much simpler methods. Like simply time shifting one driver until the phase cancellation disappears. That is, after all, the goal.

Hello Keith.
What would be your thoughts on my "simple" time alignment efforts using the series of different-Fq intentionally/deliberately time-shifted tone burst sequence which I shared in my post here #493 on my project thread for 1 msec precision time alignment between sub-woofer with woofer (plus other drivers)?

 

[Keith_W]

Hello Keith.
What would be your thoughts on my "simple" time alignment efforts using the series of different-Fq intentionally/deliberately time-shifted tone burst sequence which I shared in my post here #493 on my project thread for 1 msec precision time alignment between sub-woofer with woofer (plus other driers)?

My thoughts are exactly as I said. If you use a tone burst which is long, it reflects and contaminates the measurement. You end up with a stretched tone burst with extra sinusoids in it. Not to mention, the minimum-phase nature of LF drivers naturally distorts and stretches the tone burst. People think it's a simple matter of comparing the measured train of sinusoids with the reference - it isn't. Very careful interpretation is required. This is a technique for more advanced practitioners and not for beginners.

There are many ways to skin a cat, or in this case, many ways to achieve time/phase alignment. This method may be superior to REW's alignment tool for some niche cases (i.e. when you have a free-field), but it should not be routinely recommended to beginners because of the difficulty of setup and interpretation. I don't think DSP should be made more difficult than it needs to be - the learning curve is already steep, and people already struggle. When they have mastered the basics, then they can explore and have some fun.

 

[dualazmak]

This method may be superior to REW's alignment tool for some niche cases (i.e. when you have a free-field), but it should not be routinely recommended to beginners because of the difficulty of setup and interpretation.

OK, I understand you point especially for beginners...

 

[dualazmak]

.... If you use a tone burst which is long, it reflects and contaminates the measurement. You end up with a stretched tone burst with extra sinusoids in it. Not to mention, the minimum-phase nature of LF drivers naturally distorts and stretches the tone burst. People think it's a simple matter of comparing the measured train of sinusoids with the reference - it isn't. Very careful interpretation is required.

Even though I essentially understand your theoretical point, at least in my case with YAMAHA YST-SW1000 subwoofer in my listening room environment, the distortions (and the contamination by faint reflection) seemed to be minimal (better in 31.5 Hz than 63 Hz), but of course I did observe some "stretches" which I said/understood as "after shock" of the large driver cone and/or Helmholtz resonance within the large-heavy ported subwoofer cabinet.

Spoiler: one diagram in my post #493 on my project thread

Spoiler: two diagrams in my post #495 on my project thread

In any way, I agree with your point of "Very careful interpretation is required", and hence not always suitable for beginners.

 

[KSTR]

People think it's a simple matter of comparing the measured train of sinusoids with the reference - it isn't.

Well, for alignment task or double-check you compare the output of the sub with the output of the mains. Reference is not involved.
If it is involved in some other kind of compare, you are right the moment room modes start to contribute, of course.