Thursday, August 1, 2024

And now, the left channel

I was hoping to get off easy on the low to middle frequency alignment I did last month.  Since the left channel was "flat" all I needed to do was fix the right channel.

Not so fast.  Now the two channels sound just as different as before...actually way more different.

The left channel has more bass and highs.  The bass I can easily explain, I wasn't quite as committed to the "flat" paradigm three years ago when I last did it.  And it's gotten messed with a lot since then.  I don't know what's going on in the highs, that needs investigation.

Also the levels were radically different.  I had increase the attenuation added (with the newer speaker) from 1.0dB to 1.8dB to bring the A weighted SPL's into line.  The unweighted SPL's were even more different.

I'm going to take on this task in August, to re-EQ the left channel and make both channels match.

And that's just in the Listen mode (now Listen24).  Maybe in future months I fix the Background mode similarly.

I'm not sure how much worth it is listening to when the channels are so different, but I'm going to try.  Perhaps I did do a better job last time, though I doubt it.  Last month I was much more systematic and better equipped than ever before, and I wrote most of the key decisions down in this blog (if I didn't give the spectrum graphs for all of them...actually I largely lost track so I went with only what I needed to figure out to make the story clear, sometimes re-doing measurements if needed).

Much of the apparent full channel change is from the flatter bass and how the flatter bass is also achieved with a 4dB lower sub level than I started with.  Or maybe it was 6dB everything considered.

The frequency responses are:

Left Channel


Right Channel

Just after readjusting, the Right channel bass is far flatter.  But both of these measurements were now made in the "head center" position nominally chosen for stereo measurements.  This does not appear to change the right channel response significantly.  Here is the earlier Right Channel measurement made in right ear position:

Right Channel at Right Ear position

It's almost identical, perhaps slightly smoother in some spots below 1kHz, but the biggest difference is that it appears the highs are more extended than in the center.  Audibly the highs on the right are significantly muted and the Head Center measurment actually shows that most clearly.

Anyway, it appears to me that head center is fine for adjusting my bass, very little difference to right ear there.  The difference above 2k is much greater, though head center seems to show anomalies most clearly if not accurately, so I'm not sure which is closer to "the truth."

Merely moving an omni microphone to "ear position" doesn't create ear-identical results.  However, intuitively, it would seem to be closer than the center position.  The chief advantage of the center position is that it treats both channels fairly.

Measuring the highs accurately, especially above 10kHz, is best done with calibrated microphone.  And then there are issues, such as should I remove the chair or not.  The microphone seems more affected by chair than either iphone or listening.  It's far more problematic to set up, and (at least doing things my way) it would only slow things down because getting plots takes longer (but are more accurate and detailed).

The way I do things, fine tuning each parametric EQ, requires running response curves every minute or so to see the effect of changes.  So it's most easily done with something fast.

***** August 8

That's it, I believe in Continuous Improvement.  Though maybe that's another way of saying it's never done just "right."  This is investigation not just application.  Even if I were to stick with one app, say REW, there would be updates, I could change to different equipment (as I'm now using a combination of miniDSP's and Behringer DEQ's and the now 15 year old right sub amplifier was replaced a couple years ago skipping forwards several generations of such).  Even tiny changes in living room layout.

Anyway, this time I am trying to be more thorough and to document nearly every decision.  And to get the background understanding of things right.  Last month I learned about tangent and oblique room modes and determined that the leading resonance at the listening position is a tangent (or possibly oblique) mode which is almost unavoidable with the speaker in the corner.

One of the things I can't explain is the relatively greater high frequency rolloff in the right channel.  They have the same high frequency EQ's (or were supposed to, checking now I see that got messed up, but not change that would make the right HF lesser in fact the opposite, it seems to be missing an EQ that must have gotten erased last month, which will be explored here).  

They also have the same type of interface, the so-called "C" mod.  The left older speaker was originally an A version with the big interface box.  It was modified by one of the best respected interface experts to the C mod, but keeping the original transformers (they're not the medallions).  It is supposed to be perfectly fine to have C mod with the original transformers because the C mod actually lessens the load on the easily saturated HF transformer, and the purpose of the medallion transformer change is to improve their saturation characteristics (and not their windings or frequency response, as far as I know).

The right newer speaker has a factory original C mod interface with medallions (identifiable by the red lettering on the back saying Medallion, on the B mod the lettering is blue).

That was either from my original 1+1 set or the "replacement" interface I purchased at some point.  It has the original factory caps not the "ultimate" interface (and sadly I only did one) with a Solen PP capacitor replacing the factory electrolytic.  That is in the right speaker in the laboratory which was simply too hard to get to.  I was going to use it but haven't had time to move all the stuff--including my internet modem--out of the way to get to it.

The HF balance was adjusted by guesstimate and readjustment until it seemed about right, or the best I could do out of 3, to match the other channel.

My attempt to match levels via resistance checking was foiled by the fact that the C mod has the interface in parallel with the attenuator, and the transformer is close to 0 ohms DC. 

The correct way to match attenuator settings by measurement would be with an HF tone like 14,000 Hz into the input and check the level at the primaries (or the secondaries with bias fully discharged).

Alternatively, the AC impedances of the attenuator could be measured.  My Keithley meter can do that (it has an internal oscillator) but I've never done it before.

I see now that the reason the HF control affects mainly the highs is because the HF transformer itself contributes less UNTIL the LF transformer rolls off sufficiently.  Then the HF contribution is greater.  The point at which that emergence of the HF transformer part varies with A vs C mod (C mod seems to put the emergence of the HF signal sooner).  There may also be some "complex" effects boosting the very high peak frequency of the output.

It's not actually that the LF transformer is contributing nothing, at most frequencies it is also contributing some.  The big difference between the C mod and previous ones is that it is a true "crossover."  There is no resistance in series with the series capacitors.  The B mod and earlier are a "shelving" network, which reduces the bass frequencies only a limited amount, my linear guesstimate would be about 10dB.  So they are more likely to cause saturation in the LF transformer.

However, the B mod and earlier DO have somewhat more bass output because the HF transformer, a few dB between 100 and 200 Hz (exactly where I've had to boost the panels on the right channel).

So the difference in the highs could be at least partly related to the HF control not being identical (though I made a tiny effort to make them so).

****

But the other thing is, it seems that I have accidentally removed the 5200 cut on the Right Channel.  My treble EQ, which was tuned by ear a few years ago to sound most natural (which includes the amount of side reflected audio which perceptually boosts highs in a small room) has 3 wide cuts, at 1k, 5k, and 9k.  These overlapping cuts fall on speaker irregularities and produce a fairly smooth high frequency tilt which reaches 1dB around 1.5 Khz and continues from thereon until the speaker itself rolls off.

Well, and I think I remember when it happened, I erased the 5kHz cut from the right channel.  I think it happened when I was adding a 500 Hz cut, and thought it was already there (but in fact it had been the 5kHz cut previously).

I see the original 5kHz cut in both the Listen23 and Listen24 memories of the DEQ, just not in the latest settings, were they saved in X ???  I resaved them today in Aug.

Those latest memory(ies) include all the latest adjustments up to 1kHz that I worked on last month.  Strangely Listen24 is missing several.

It appears the original cut, as saved in Listen23 and Listen 24, was -4.5dB at 5200.  But the cut in the right channel is only -3.5dB.  I'm not sure when that -3.5 was dialed in, it might have been with the previous panel which was replaced in December.

So instead I dialed in 5200, 1/2 octave, -3.5 into the right channel to tentatively match the left, even though the right channel had already been overall measuring generally lower in the highs.  I've saved that both as Aug (knowing it has only that one change to what I had during the last measurements in July) and as Listen24, the new not quite ready Listening position standard, in the EQ for the panels.  These are not reselected automatically (so called Boost mode actually selects the Listening Position EQ for the subs, but not the panels) because I figure these equalizations are not so listening position dependent.  (Perhaps I will make separate listening and background EQ's for the Panels in the future but I have not done so yet.)

This change should not contribute anything to the steeper rolloff above 14kHz, but it might make the speakers sound more similar under pink noise, which was also disturbing me.

I am simply undoing a "mistake" though it might make the measurements look worse than the last ones until other changes are made.

Anyway I did a re-measurement of each channel in a somewhat different way, with the chair behind removed.  The listening chair makes a huge difference when running programs like ARTA with a calibrated mike.  Strangely, it seems to make relatively less difference with pink noise measured by an iPhone RTA.

I always wanted to modify the listening chair, removing the upper part and replacing it with a small headrest.  When I bought this chair (2008 I think) I intended to do that right away.  That's why I didn't buy a more expensive chair, though this one (when new) felt almost as nice.  But I didn't want to buy a mega expensive chair and then try to modify it to be an "audiophile" chair by being open above the shoulders, except for a small headrest.  But I never got around to it, and now I'm not sure I could afford it, and the chair is kinda shot anyway (needs new springs).

These measurements show about the same things as before.  Here's the new left channel measurement, showing the same lumpy bass as last time, followed by the earlier measurment:

Left channel, chair moved to right wall

Left channel, chair behind mike (as before)

Weirdly, it looks like the one with the chair behind is flatter, especially in the 250-1000 Hz range.  I wouldn't adjust the critical midrange this way if I were using this chair, I guess that's right, though I often believe things like that the brain can subtract out the effect of the chair, so maybe measuring without the
chair is more "natural."

In most places, it doesn't seem to make much difference.  Outside the bump around 350 Hz, the other difference that stands out in the chair-to-the-side measurement is a curious bump around 7kHz.  Those two bumps could actually be caused by reflections from the chair which is now to the side, so I have
to be careful assuming that is "more" the way it is than with the chair, perhaps because I equalized it before with the chair behind.

Here is the right channel, new and previous measurements:

Right channel, new measurement


Right channel, old measurement

Now the old measurement was done with the 5200 Hz cut accidentally canceled.  It was restored in the new measurment, and the change seems clearly for the better.  Now rather than a double hump in the highs, there is the intended gradual tilt downwards from 1kHz.  And strangely the extreme highs do look flatter this way, I really can't explain why the very highest bar has gone down to like nothing in the old measurement but not in the new.  It's not a chair effect because I see no change in the highest bar in the left channel measurements.

Anyway, by measurement at least, I've settled the issue of whether it was a mistake to remove the 5200 Hz filter from the right channel.  Yes it was a mistake, that filter belongs there and makes the treble look much flatter and like there isn't actually a problem there.  (It strangely 'sounds' more extended too, listening to the pink noise anyway. )  

The right channel now looks flatter in the highs, but the bumps above the tilted bass line actually represent quantity of highs more like 1kHz, so it's hard to say which channel is more "right" except by listening tests.

And it's not meaningful to do that until the issues in the bass and lower midrange in the left channel are sorted out as they were in the right channel.  (Though I could also listen to material on the two different kinds of bass presentations now, or wait until I have it all dialed in for both channels and then compare to LISTEN23.)

I noticed that two channel "in phase" pink noise does actually seem pretty well centered despite the obvious differences between the channels.

One problem with music listening tests comparing one channel to the other is that I have very few mono recordings, and nothing in my system offers a "mono" switch.  No, wait, the Tact actually has a mono switch!

A quick spin of a bass test disc (Bass Erotica) makes me believe I am correct about the benefits of the flatter bass now in the right channel.  It sounds better and improves headroom.  The mono switch works well but I also have to spin down each channel from 0 (or -1.8 in the left channel) to -99.9 to test one channel at a time and that's the bothersome part.  No instant A/B is possible.

But also the way I dialed down the subs and panel DAC recently...that was a bad idea because now I can't play loud enough without clipping the digital output of the Tact.  So I'm going to dial everything up by a few dB again so I can play Tact at lower level, not using digital gain as much either.

One thing I could do differently is add some digital gain at the DEQ for the panels.  The Behringers do not seem to produce inter-sample-overs.  They appear to operate synchonously.  The place where the iso peaks were becoming a problem was at the input to the miniDSP's 

So I've done this now and resaved into LISTEN24 on the panel DEQ, a +3dB "Gain Adjustment", and I turned the Emotiva DAC for the subs from -11 to -8.  Then later I added 2dB more, with these adjustments:

Sub DAC: -6
Panel DAC: +10.5
Panel DEQ: +3

This now has 2dB more maximum output than June 2024 (before the Tact digital output clips).  But the Emotiva is set 1dB lower because the gain compensation of the DEQ for 3dB of gain.

I was also inspired by the more punchier low bass in the left channel to increase the boost at 35 Hz to +6 in the right.  I knew that measured flatter but wanted to avoid that much boost.  But it seems to be needed for the bass to come alive and better matching what sounds right about the left channel.  The left channel still sounds bloated elsewhere in the deep bass, possibly around 60 Hz.

But looking at the response curves, with the chair removed (new measurements) the dip in the right channel bass moves up from 35 Hz to 65 Hz.  In the old measurements, 65 Hz was the top of a gradual peak.  Which is correct?  I am now thinking the chair behind the microphone is the correct way to measure, just as if I were sitting there, even though it messes up some kinds of measurements irreversibly (such as the time alignment measurments, those need to be done w/o chair).

Testing the +6db boost at 35 Hz in the right channel, there is no significant bulge created even without the chair behind the microphone.  It's pretty hard to tell the difference between +4 and +6 on the spectrum graph, but it looks and sounds more solid in the bass with +6 and not boomy.  It's possible a more correct approach would be to add some Graphic EQ boosts there and elsewhere, but for now this seems to work.  Previously I had +6dB boosts dialed into the graphic EQ at various places, but for the New EQ, I made that all flat and so far haven't used any GEQ's, resulting in a much simpler total filter function.  There is now only one boost in the bass, previously there were several.  (However, a few more might be needed now.)

Right +4dB at 35 Hz



Right +6dB at 35 Hz

(Both are with chair on side wall near mike.)

Since 35 Hz is a positional depression at the listening position, boosting with that central frequency never sounds boomy.  This is the glory of parametric EQ.

It is sometimes said you can't boost a null.  That may be true (though last month I seemed to make one go away with EQ changes) but you can boost a minor depression.  You may not get as much as you like, but it is being boosted.  I think it's like this: when you boost a depression the boosting is lossy so you get less than what you put in.  When you boost a peak the boosting is augmented so you get more than what you put in (and likewise, you take more out than you think with cutting).  That explains why I've often thought the spectrum response scale was wrong based on effects I've seen.  Dealing with peaks you seem to get larger changes than expected from cuts.  (I'm not sure if this is correct given that dB's are relative.)

Though one might wonder that if flatness were the goal, I should have cut 22-25 Hz instead.

You would almost think you could get to anywhere either with cuts or boosts, like moving a whole range of graphic sliders up or down.

But it's not like that.  Depressions are real and not just relative things.  Sucked out regions sound overdamped.  Bringing everything down to the same level just makes everything sound overdamped.

The correct thing, if you can find it, is to cut peaks and boost depressions, and not more of either than matches the actual situation.

To me, the right bass just sounded wrong, overdamped around 30 Hz, whereas the other channel, didn't sound overdamped there, just way more boom elsewhere.

It still does remain to be seen what the best room curve is, whether really flat does it.  I'm currently allowing a bit of unchecked room gain below 30 Hz, thinking that sounds (and feels) the best.*  But I'm not actually boosting there in any way, I'm boosting above that for a bit so there's a broad plateau all the way down from 40 Hz, a solid first octave.  And it sounds that way, despite having far less deep bass boost and room gain than the left channel.  The left channel sounds lumpy and bumpy just like the response curve now looks.  (And before last month, to think it looked the flattest!)

(Below 30 Hz, I'm getting a variety of whole house resonances, that make the listening room sound like a much larger space, a benefit of having two open passageways to the living room.  I'm actively damping some of this with an eTrap at the first corner of the hallway.)

I'm noticing a little peak popping up when I boost 35 Hz.  But I think that is not actually 35 Hz it's 39 Hz, where I have a small cut that may need deepening now that I've increased the nearby boost.

*****

So now I've hooked up the left channel sub to the chairside EQ the way I had done last month for the right channel.  At first glance, it looks pretty flat.  There's no Mount Everest at 100 Hz anyway.

Left Channel Sub, No EQ

But there are a number of issues.  There are peaks around 45 Hz and 71 Hz.  The crossover cutoff is interruped by a huge resonance around 140 Hz so it's as if the crossover doesn't cross over until there.  (I'm assuming for now the FIR implemented crossover, 8th order at 125 Hz, is working OK.)  Above 140 Hz the crossover is impressively steep, but there may be some additional resonance around 300 Hz worth tamping down.

On sweeping the modes at 45 and 71 Hz (approximately) are quite obvious.  The resonance at 140 Hz is unclear because it almost seems (just as it looks in the RTA) that the crossover occurs there.  I had issues at all these same frequencies in the other channel, though the 45 Hz one (which I actually tuned to 43 Hz) is only -2dB and I may even have to cancel that for flattest response (currently there's a 42-55 Hz depression in the right channel).  It doesn't look that big here either but it looks like a broad rather than extremely narrow filter might work best.  In the right channel there's too much other stuff going on to use anything but a narrow filter and even that might have to be canceled.

And weirdly, the combined left channel response (way above) shows no issues from 125-200 Hz.  It may well be that the sub is filling in for the panels, not by design but by accident.  There's a 140 Hz room mode that sucks out the panels and augments the boxed sub because they excite it differently.

Anyway, it looks like it would be best not to mess with the 140 Hz without looking at the combined response.   





Basically, the biggest issue has moved from about 100 Hz in the right channel to 140 Hz in the left channel.

The slightly sagging 20 Hz is a concern.  I do need to check the plate amplifier to be sure the correct "Extended" mode is selected.  But perhaps I want to keep it the way it is now anyway.  The left sub is in the ultimate hard corner (not a passage way)...of the entire house.  Perhaps having it tuned wrong prevents the mother of all whole house resonances (and why I did it that way).  But output seems surprisingly weak at 20 Hz on sweeping and I don't like that.  All that potential lost.

Speaking of which, the combined response doesn't show the peaks in the sub response, it shows different ones at about 35 Hz and 58 Hz.  So while I will try to correct the peaks here, I'll then check what cancelling the correction does in the combined response.

*** August 10

It's curious that the bumps in the combined response don't match the bumps in the sub response.  However we're looking here at the sub response with no EQ's.  The bumps in the combined response were using the sub response WITH eq's.  So it is the old EQ which effectively created those bumps.  I doubt they were caused by contribution from the panels.

The new EQ will be simpler I hope, small notches for the bumps caused by modes at 43 and 71 Hz, combined with some sort of attack on the huge resonance at 140 Hz so long as it doesn't depress 120-200 Hz in the combined response.

With much less notching in the sub response being done, it's likely the sub level will need to be lowered just as the right level needed to be lowered.

Studying the C mod interface schematic, consider that the two ends of the attenuator (plus resistor if applicable for retrofit versions using rotary attenuator) are tied to the two ends of the small transformer, which are effectively a short to DC.  Well then it's clear that measuring the slider resistance to ground will reach a maximum somewhere in the precise center of the attenuation resistance.  Going either way will decrease the resistance until it's moved all the way to one terminator (or as close as it gets) as the smaller of the two parallel resistances will be at its very smallest.  So that's why it looked "non-linear" when I was trying to adjust before.  For each value lower than peak resistance, there will be equivalent points on both sides of the center having that value (in principle anyway).

With my steep 125 Hz digital crossover, it may well be that I'd be better off with B mod because the LF transformer saturation might not be as big an issue because less LF is being passed by the transformer.  OTOH, I'm playing at higher levels made possible by that, so it's not certain.

With B mod I might be able to eliminate, or at least reduce, the wide boost I give to the right panel at 138 Hz.

Even using my 3 inch closed cell foam pad, I can't just dive onto my knees to adjust the the equalizer as I'd become too accustomed to doing.  My left knee, injured in March, got sore again today from doing that recently.  I need to carefully bend down and sit on the pad, or get down on knee very slowly if at all, even on the pad.

Other than that, the getting up and doing stuff part of audio experimentation appears to be good for health and weight loss.  The writing it all down part, which I like better, is better for the soul but not much else.  Even then, when caught up in measurement and fine adjustment I still can't write it all down.

One thing I'm doing better this month from the outset, is since I have the temporary EQ wired in on a parallel optical path, I can easily switch back to "normal" listening and listen to my automatically generated playlists in background, which I'm doing right now (and it's been all great stuff today, including Beethoven's Seventh, which I remembered I had played cello in High School Orchestra).  The last half of July was almost music free even after I finally got around to doing the optical connections...by the time I could I barely (k)needed them because I was doing all remaining adjustments in front.

"Keep the music playing" is a pretty good default.  I have a no-words music sub collection for that purpose.


***** August 12

A new handheld mirror arrived yesterday but I didn't have time to use it until today.

I checked the right subwoofer settings.  The right sub has an older (but not the original) plate amplifier which has no bluetooth control but instead a menu knob.  (The very original PB13 Ultra had no menu, just toggle switches.)  The Subwoofer tune was set to the middle of three settings: 20Hz, 19Hz, and Sealed.  (Or maybe it was not 19 Hz but 16 Hz, it was hard to read upside down with a mirror, but I could not see it as other than 19Hz even though that makes little sense.)  The sub has one port closed so the middle setting is correct.

Nothing else was activated.  Phase was at 0 degrees.  The level was set to -7 which I lowered to -10 because of the likely effect readjustment is going to have as on the right channel.

***** August 14

Take a look at the absorption curve of a well regarded bass trap  (the 244, the B4 isn't nearly as good).  The absorption declines rapidly below 80 Hz, is less than 4dB at 63 Hz, is about 1dB at 40 Hz.  (And all that presumes adequate coverage, which could mean line a majority of the wall and ceiling space, though a lot can be done at the hot spots.)

I never used to think much about modes above 80 Hz, and now I remember why.  I used to cross over at 80 Hz!  For one thing, I needed to avoid resonances and a limit on dynamic range from them just above 80 Hz, it seemed.  (That may have mostly been with earlier speakers.  My current selected pair from 2 pairs seems free of resonance and rattling issues.)

I was obsessed with issues from 45 Hz, where there was the mother of all resonances, a few higher frequencies like 56 and 71, and some lower suckouts.

How are you supposed to do anything with 45 Hz with bass traps that can only attenuate 1dB there???

I throw this out to refute the nonsense that "you don't need EQ, just acoustic absorption."  At frequencies below 80 Hz, bass trapping becomes less and less efficient.   (There might be some traps that do better than the 244's, but likely cost a lot more.)  And then consider the trapping needed even for a small "living room."  (A tiny closet sized "studio" might be affordible.)

However, now year later, I seem to be discovering something different.  In my latest right channel tuning, it wasn't 45 Hz which was the biggest peak, it was 100 Hz.  That had to be pulled down by over 15dB.  Well at 100 Hz, even bass traps like the B4 do pretty good.

It's curious the long list of things that have changed.  While I had noticed earlier that 45 Hz cancellation was virtually no longer needed in the right channel and had reduced or minimized that, I'd never filled in the box for what the biggest resonance was until last month.  What caused everything to change so much:

1) I crossed over below 100 Hz, so wasn't stimulating the 100 Hz tangential/oblique mode(s).
2) Speakers moved from less than 3 to over 4 feet from wall.
3) Changed from 1+1's to 2+2's, then changed to second 2+2 on right in 202
4) Moved listening position from near back of room to fairly close to speakers ("giant headphones")
5) Replaced SVS PB13 plate amplifiers...new ones have more power.
6) SVS Ports and Port Tunings (I've been through a lot of variations, now stuck at Extended)
7) iPhone RTA refinements, new oscillator


I evolved away from 80 Hz crossover around 2021 (12 years after I started subwoofering this system) when I began really trying to optimize low frequency response including the effects of wall bounce (quite noticeable) and dipolar roll off.  Even moving 4 feet from wall I could not get the wall bounce suckout much lower than 115 Hz.  (My then right 2+2 also had an intolerable internal resonance and rattle not much below that.)  With a very steep crossover at 125 Hz, I found I could bypass wall bounce suckout and speaker rattle.

I still like that decision because it seems I can fine tune every 1/6 octave better with the subs anyway (though, sadly, they need a lot more fine tuning).  I'm not planning to explore the possibility of a lower crossover again.  What I am trying to do is really and finally get things right for this crossover AND the current set of speakers, room layout, and everything, which I really had never done right before.

My initial experience of the newly tuned right channel is "Wow!"  The transparency is incredible top to bottom compared to before, and with the 35 Hz boost increased to 6dB it has just enough punch too.  It measures flatter than I've ever seen on RTA, but even finer tuned sweeps were used to find the exact best frequency for each PEQ, so it also sweeps the best.

Sadly transparency always seems to increase when bass is decreased, and then it takes longer to notice that more bass gives it more body and realism.  When I increased the boost from 4dB to 6dB that seemed to hit it, but it might need more and elsewhere.

*****  August 15

Last night and this morning I swept through the subwoofer and found (again) primary resonances at 43, 71, and 144 Hz.  Following my now standard approach, I adjusted notch filters at each modal frequency so that it was almost but not quite back to baseline by ear, and widened each notch so that it swept the most evenly by ear and did not remove anything from adjacent depressions.  It was helpful as in July to have frequency counter connected, but after issues with Tact analog output seeming insufficient, and last night not seeming available at all (has the Tact analog output gotten fried?) I hooked the counter (actually my Fluke 8060A) directly to the sync output of the oscillator which works much better.

(Note that as before this oscillator doesn't allow me to sweep very slowly or evenly.  And I spent a few minutes and not days on each filter.  So it may not be exact, but at least it tries to be close to the actual resonance frequencies unlike Graphic EQ approaches.

It seems by these means (and the recent attenuation of the right sub by 3dB) I have gotten nearly to the required response for the sub only, and to flatter than before in the full spectrum response.



Left Sub before Adjustment

Left Sub with 43,71,144 Hz notches

The notches are basically working though they could be increased, and possibly some more filtering around 45 Hz.  I am constrained there because, unlike in this smooth looking 1/6 analysis, there are narrow depressions all around the 43 Hz notch filter.  So I can't make it wider, though I could possibly make it deeper.  Likewise it looks like I could deepen and widen the filter at 144 Hz.  Now response is at least beginning to fall at 125 Hz (it was peaking there beforehand instead of falling...see the bar just above the "5" in 125) but it should be falling much more.

But the combined response suggests another problem, it looks like the sub level is still too high.

Left Channel after new sub EQ

There's a a 355ish peak that needs to be tamped down in the panels, but compare the 1k level with the bass "plateau around 31.5 Hz.  It looks like the sub needs to come down 2-3dB more.

The bulge in the crossover region will be lessened both by reducing sub level and by enlarging the 144 Hz sub filter.

Then the worst thing in the bass will be smoothing the 43 Hz and environs, which has to include sweeping to be mindful of not increasing narrow depressions.

After doing more sweeping, I decided that widening the filters was not good, it merely created more depressions.  So instead I added 3 more filters, at 49, 104, and 129 Hz.  This enabled me to improve the sweeping to sound good.  I note that the right channel also needed the 104 and 129 Hz filters, I'm not exactly certain that the 49 Hz resonance is a thing or shifted considerably because of the 43 Hz filter.

This improved the sub and overall response.  Now it doesn't look as much like the sub level needs lowering.  Strangely though I still haven't reduced the 125 Hz peaking in the overall response.

I've now photoed this set of filters for reference because it's starting to look good.


Left Sub with 6 notch filters

Left Channel

First 6 PEQ adjustment

I think the sub level still needs to go down another 1-2 dB, but it's hard to judge until I fix the now glaring peaks in the panel response.  I see now there's a second one just below 1kHz too.  That's similar to what I needed to do in the other channel.

*****

I couldn't resist another set of sweeps to see if I could correct the remaining bass issues.  Primarily I needed to pull down the 129 Hz notch enough to fix the acoustic crossover which was barely down a couple dB (instead of 6dB required by Linkwitz-Riley) at 125 Hz, also the broad peak around 43 Hz and the narrow depression around 90 (for which narrowing the notch at 104 Hz helped a bit).

Well when I first measured it, for no apparent reason, bass below 63 Hz just rolled off.  Could it be interaction with the things I set, perhaps I made some big mistake?  I couldn't see any mistake, so I started resetting the previous values, first the amplitudes, then the bandwidths.

Fortunately, before I'd changed a single bandwidth (the critical part you need to do by sweeping) I realized my mistake.  I'd accidentally pressed the weighting button and advanced it to A weighting.

I set the amplitudes back from memory and spectrum results.  It looked very good.

Left sub after some changes to notch widths


Left channel after sub notch changes

Second 6 Left Sub PEQ adjustment

I was amazed at how flat the bass now looked, but the glaring issues in the panels looked even worse, so I decided to look at them.

WRT the rise before 1kHz, some of that was a deliberate boost in that region.  In some earlier situation (I have no idea what changed this) 1kHz was the highest, but a region just behind 1kHz wasn't as high so I boosted it.  Now the situation is exactly reversed, so I removed the boost at 853 Hz.  It could use some more cutting there actually.

I identified the huge bulge as 355 Hz, and put in a new notch there which helped a lot.  I'm worried now if I increase it more the lower midrange will be all sucked out, for which the current 355 Hz resonance residual is a bit of relief.  It's now looking pretty flat though could be much better.

Left channel after sub and panel eq changes

Left panel EQ


***** August 17

Knowing the two issues in the panels, I dove right in to sweeping.  The resonant point just below 1kHz was 800 kHz, so I put a notch there.  The lower resonant point appeared to be around 355, but as I notched it out it seemed to get higher, so it ultimately landed at 371.  But then an adjacent resonance appeared around 440, so I put a notch there.  If I tried to remove the 440 it popped back up, and I couldn't make the 371 Hz filter any wider or stifle the upper 200's.  I ended up juggling the two filters to what looked best on the RTA.

It swept pretty good.  The RTA, however seemed to show a peak at 630 Hz.

Right Channel with mystery 630 Hz peak



I went through the same steps, tuning the 630 Hz filter by sweeping for uniformity.  Then the RTA was clearly showing 630 Hz sucked out, so I dialed until it was back to zero, and that looked best.  I can only guess there was some ambient noise or something.

The resulting RTA is now so flat I don't want to do anything more.  When it looks this good, you don't mess with it without a better reason.

Another sign it's time to quit is when you add a filter and then have to completely remove it.  I was almost at that point with the 371/440 Hz filters, but i ultimately seemed  seemed best to have both (check the pudding).

The sub level could possibly be reduced 1-2dB, but I'm not even sure about that any more.  

Right Channel after adjustments

Right sub 2dB lower (-12dB)

I'm going to go with the -12dB setting for now, it gives a very flat response from 1kHz down to the deepest bass.  It doesn't smooth out the crossover region as I had hoped (or at all) but it doesn't make anything worse either, everything looks better balanced.

New (and old) Left panel PEQ's



Left Channel GEQ's

Oops.  I forgot to dial out the left channel GEQ's before this alignment.  And now it appears they indeed are interactive.  The boosts above 350 are exactly where I'm cutting.  The cut at 200 is now a low spot.  This probably need backing out in those parts, the boosts below 200 Hz are probably helpful and should be left alone (there is a boost there in the other channel as well).

I dialed in the new PEQ's into the normal sub DEQ so I could listen to both channels with the new EQ.

I listened to Emerson, Lake, and Palmer title album, and Grouse.  I had never heard such transparency.  But bass seemed too weak so I dialed up the bass in both channels by 3dB (at the DAC) and that was too much, 2dB seemed too little, 2.5dB more bass seemed just right. It was a setting coincidentally of -2.5.  I think I should dial up the subs so I can dial the DAC down again.

This is clearly an improvement over the previous adjustment before I started changing it in July but I think it could still use some more work in both channels, and both channels should be brought closer together.

Left Channel


Right Channel

That strange 630 Hz is back in the left channel.  Refrigerator was not running but it could have been the RO water filter operating, except that the RO filter noises don't at all seem to be around 630 Hz.

The bass looks higher in the right channel now, though it's hard to judge because of differences in the midrange just below 1kHz.  There's a tiny depression right at 125 Hz in the left channel now, I had previously been struggling to remove a peak there.

***** August 18

Having changed my mind late last night (actually 4:30 am) that this was good enough (I see now clearly it wasn't) I hardly know where to resume.  I remeasured both speakers:

Left Channel before further changes

Right Channel before any further changes



It's pretty clear the right sub is now lower in level relative to the left and compared to their respective panels.  I was unsure if the flat "room curve" was best but now I strongly believe it isn't, the bass sounded too weak that way.  

Looking carefully at these spectrum, picking a frequency unaffected much by position and EQ like 20 Hz, the distance between that and 1kHz is about 8dB in the Right channel, but only 6dB lin the left channel.  Strangely the "overall" also shows this kind of gap (though balance is dialed in favor of right channel already in the Tact).  Ignore the sub 1kHz peak in the right channel, that needs to be notched out in the right panel EQ as effectively as I just did in the left.

So I'm going to try to dial back in the 2dB I recently subtracted from the right sub yesterday in the false believe it was bringing the two closer together, now that I see clearly it wasn't.

Since the microphone position might vary from one setup to another, and ambient noise varies all the time, it's always useful to repeat measurements for comparison fairly quickly and without moving the "microphone" in between things that are being compared (currently my iphone).  (One should always remember that this is not being done in a perfectly quiet or even identically quiet environments as different noises are happening all the time.  Also my microphone positioning is hardly perfect even with my new iphone holder which keeps slipping and/or causing my microphone stand to slip.  It's a constant battle to keep it more or less in the correct place.  And all these measurements have fairly high internal variations as well.  The RTA can easily have 2-3dB variations from one run to another in particularly "volatile" bands.  It's my believe the truth emerges from all the data, not just one test.)

One thing that's not clear from the Analyzer is that subsonic bass frequencies are nevertheless stimulated more by the left sub than the right.  And one maddening thing is that I cant set PEQ's lower than 20 Hz on the Behringer DEQ 2496.  One ought to be able to set PEQ's as low as 5 Hz.  Anyway I might have to set a subsonic shelf or a wide bandwidth filter at 20 Hz as I have done before, especially if I make left sub louder.

While it's cool to be able to change both channels with the level control at the DAC, I'm also wondering if that is necessarily the best thing.  It's interesting that my 125 Hz crossover has the same frequency as the bass turnover frequency on many "tone controls".  But getting that effect by boosting a different driver is not adding any "resonance," whereas if the bass is truly sucked out it may require an injection of "resonance."  So tone controls will give "fatter" sounding bass, whereas level shifting the bass driver may give unnaturally "tight" sound.  (It was sounding very tight last night, and though it's hard to imagine myself saying this, perhaps too tight.)

So while it's convenient to set the bass level with the sub level control on my system, it actually might be better to use an EQ boost instead, and optimize the sub level via other means.

Anyway, whatever the correct level is, I can easily change both subs at once using the DAC output level control.  The key thing is that both subs be at an equivalent level, I can tweak the level of both later (and more easily).

I'm going to shift the levels again by 4dB to preserve headroom and change ability

Current values:

Left Sub -12
Right Sub -14
DAC -3.5

Planned change

Left +6
Right +4
DAC -4

Result

Left -6
Right -10
DAC -7.5

(Note that this can't strictly be compared with what I had originally because I've changed the gain structure.  At the same input level and master attenuation, the output is now 2dB louder than before.  Also I've changed the port tuning to extended on the right, it was set incorrectly to normal.)

These changes made no significant difference (it's constantly fluctuating btw) in the right channel, but the left channel with the now relatively boosted sub shows it better matches the bass rise in the right channel.

Sadly it did nothing for the 125 Hz depression, so I decided I wanted to fix that.




Sweeping the region showed a broad depression below the EQ notch at 129 Hz all the way down to 120 Hz or so.  It was, no wonder, strongly affected by the notch filter at 129 Hz.  The bandwidth had to be made ridiculously narrow to avoid that and even then didn't, and a boom appeared in the lower 130's.  So I decided to notch out the 130 boom, which helps all the way down to 129 but leaves the critical 120-128 region untouched.  This fixed the 125 Hz depression.

But then a larger gap opened up around 90 Hz.  That was strongly affected by the 104 Hz notch and weakly affected by the 71 Hz notch.  I made the 71 Hz notch as narrow as I could before a boom started arising around 85 Hz.  I tried using a very narrow 71 Hz filter and filtering the 85 Hz separately with a very narrow filter, I worked hard on that,  but it seemed it couldn't be done without further depressing the 90's.   So I changed the 71 Hz notch from 1/3 octave to 1/4 octave, just the right amount to soften the 80's boom and leave the 90's untouched.  And I moved up the 104 Hz filter, it seemed there was now more boom around 109 Hz so I put a filter right there.  (Actually, the prior 129 Hz filter was possibly somewhat supressing the 109 Hz boom, making 104 Hz previously seem like  a better notch frequency.)

I also tried messing with the GEQ's for the panel, trying to cancel the ones below 100 Hz (typically -0.5dB or +0.5dB).  In every case it made the response worse so I ended up resetting them as they were before.


Left channel readjusted to remove 125 depression

New Left Sub PEQ's

The result has a very solid bass plateau below 160 Hz.  This seems to be the kind of thing that sounds best as a room curve in my room (and to me).  It also more or less matches the countour in the right channel which was adjusted last month (but also boosted yesterday when listening revealed it needed stronger bass).  Once the two sub channels match, further level changes to both can most conveniently be made at the DAC.

Not much left to do in the left bass, I think the main remaining problem in the left (and right) is the sub 1kHz peak.

To do THAT properly, I need to reprogram the chairside EQ for the panel EQ's, and hook it up that way, a significant task that may not be done today.

But I could not resist the temptation to get one more thing done Sunday night.  I was going to sweep the right channel, but since it was the left channel that was hooked up to the oscillator, I went after the sub 1000 Hz resonances there.

I noticed that the 630 Hz peak was reappearing in the RTA, and I shut off the pink noise and I could see nothing at all there in the background noise.  Nevertheless I tried to equalize out peaks very close to 1000 Hz first but they were very problematic.

Finally I did go after the 630 Hz peak, and doing that seemed to change things for the better elsewhere.

Slow oscillator scans show something very different from the usually fairly serene RTA's.  There is an endless set of hills and valleys every small fraction of an octave.  Now some of those peaks are much higher than others, and that is what seems to drive variation in the RTA.  When I use PEQ's, I aim just to damp one peak at the maximum point of that peak.  I can make the PEQ wider if an only if the downsides don't become much worse depressions.

While I'd think that merely equalizing down a small fraction of an octave like 2% wouldn't make much difference in a 1/6 octave RTA, it usually does.  And not only just local to one peak, it seems to make peaks elsewhere nearby either go up or down, OR become more or less amenable to PEQ's tamping them down.  It's as if a big peak is telling nearby peaks "don't give up, I'll be here for you."  But once you squash it, other peaks can't hold on anymore.

I came up with an adjustment that looks much better on the RTA, with a broad plateau from 160-4000 Hz marked only by a few small irregularities.  (Actually, I intended the 2-4kHz region to be slightly more depressed than it is, and the rolloff about 4kHz is mostly deliberate tuned by ear.)

Left channel with 654 and 798 Hz filters

Left Panel PEQ's (August 19)

I wish I had realized that deep in the mix there are 1khz and 2700 Hz BOOSTS! (!)   I find it hard to read the display when I am sitting in front of it (much easier to read the pictures) especially without glasses on (which I mostly had on doing this).  I was struggling to cancel a peak near 1kHz and almost added in a cut exactly where there is a boost.  I should have experimented with cutting or eliminating that boost.

The 2700 boost may have started out as a cut.  Perhaps I turned it into a boost by accident, or for matching the right and left speakers.  That's going to have to be rethought in comparison with the other speaker and actual music listening.

Anyway, the result looks good enough and sweeps good enough now to move on to bigger irregularities in the right channel.

Going through my photo history, I see I have two earlier photos from this year (!) which show the 2723 Hz filter as a cut, not a boost.  This is really looking like a knob twisting mistake.  It's quite often for me to turn the knob on one of my two DEQ's when I had intended to turn the other, and it doesn't help that I can't read the lower display at all without glasses so I can't see if I messed something up easily, or tell the difference between plus and minus.

Here's the midrange PEQ photographed only just over a month ago in mid July, and the 2723 filter is a 3dB cut.  I think this photo was taken just before I started adjusting the right channel bass (and well before I started adjusting the left), as a permanent record.  There IS a 1khz boost but it's only 1dB and it's broad.  I think this was because I decided previously to raise 1khz rather than suppress the sub 1kHz peaks.

Left panel PEQ's on July 13, 2024

***** August 20

I've dialed in the GEQ, PEQ, Delay, and Gain settings from the panel DEQ to the chairside DEQ and hooked that up to the panel DAC where it can be switched from the normal AES/EBU.  This allows me to experiment with chairside EQ as if it were the panel EQ

To make sure this was done correctly, I did before and after acoustic RTA's.  The right channel seemed close enough right away for me to declare a match.  The left channel had at least one minor error, which I fixed, and it almost seems to match now.  I've quadruple checked everything and can't see why there would be any difference.  In the right channel it's surely just a measure of natural variability.

Right Channel (seemed to match from the start)

Right Channel, front DEQ

Right Channel, chairside DEQ

Left Channel didn't match well, I went back and corrected one error, still didn't exactly seem to match but I quadruple checked everything, and this *could* be normal variation now, all the main features are matching with some fluctuations (and the measurement late last night looked better).  The bigger fluctuations here seem related to bass below 125 Hz, and there aren't any PEQ's in that region only some GEQ's (which I've been trying to eliminate).  I checked the GEQ mode, it is set to True Response for both DEQ's.

Left Channel, panel EQ


Left Channel, chairside EQ

Last night's measurements might have been made with chair slightly farther back than normal.  I corrected that today before doing these left channel measurements.

I'm going to go ahead with this arrangement with the caveat that when changes are coped to the front EQ they should be re-checked there with at least RTA if not sweeping.

It's actually more disturbing that my fine tuning last night didn't make the 200-1k response as smooth as it seemed last night in the left channel.  That might have been because the chair was a few inches back.  So it may not be done yet.

*****

There wasn't much time after a long movie last night, so I thought I would just try something simple:  remove the left channel GEQ cut (-0.5dB) at 80 Hz.  This was no doubt adding to the depression at 90 Hz I believed.  Actually, it was so small it should not make any measurable difference I thought.  And then it would be good because it makes the overall filter function simpler.

Although I'm not 100% sure because of the high variability of my RTA methods (see last few graphs above) it appears I was wrong about not making measurable differences.

Left with 80 Hz 0.5dB cut in panel

Left with 80 Hz 0.5dB cut in panel

What mainly seems to happen is that a peak 70-80 Hz gets 1-2dB larger.  Meanwhile, 90 Hz is barely increased, it's hard to see though it might actually be increased a tiny bit, but dwarfed by a much larger peak below it, this is clearly worse.

So I'm going to keep the 0.5dB cut at 80 Hz for now.

Most of the behavior here is determined by the sub EQ, this 0.5dB cut may have not been added to correct the overall response so much as the panel-only spectrum, and the fact I know the Acoustats have resonances in the 80 Hz region.  Further significant improvements might accrue from readjusting the sub EQ in that region.

This experiment tends to reinforce my view that when you boost or cut a resonance, the effect of that boost or cut is magnified, and the effect of EQ is diminished on a depression.

**** August 21

I decided the next most important thing was the apparent error I made in setting the 2700 kHz cut into a boost by accident on the left side.  I already found that as late as mid July last month, that filter was a cut, and then somehow mysteriously (through some act of wrong knob twiddling and/or incorrectly reading display because I was too lazy to get up and get my reading glasses) it got turned into a boost.  

And correspondingly there was an anamalous plateau where my design manual says there should be a cut because of the small size of the listening room (the Linkwitz-Gundry dip) and the ear's heightened sensitivity to input from the side aimed down the ear canal.

Well it looked like a mistake in every way, but I thought the ultimate test would include at least one kind of listening, since this is a broad EQ.  I compared current left and right channels listening to pink noise, then made the change, and listened again.

I felt changing the boost back to a cut brought the channels much closer in sound, about 50% closer, with 50% of the additional variation unexplained.  I could not make the channels sound the same by readjusting any of the 3 top level curving cuts at 2700, 5200, and 9200 (which are again identical in both channels--they were based on listening tests done in stereo a few years ago, somewhat also influenced by measured results and believing the basic principle that smoother is better, if not necessarily horizontal).  

I tried big changes in the upper 3 EQ's and nothing seemed to change what part of the sound needed changing.  The remaining differences seem to lie elsewhere.  And two big candidates are the remaining 100 Hz hump in the left channel that I was just talking about, and the bump prior to 1000 Hz in the right channel--which will be my next focus.

Right channel

Left channel after changing 2700 Hz boost back to cut

The channels are getting a much more similar profile than when I started.

Despite thinking I'd tackle the right channel I did the reverse.  I tried to eliminate the 1kHz 3dB boost in the left.  That clearly made 1kHz depressed compared to 500-1kHz.  I swept and found the residual peak at 820 Hz, so I moved the 798 Hz filter upwards and deepened it slightly to 4dB.  That looked pretty good, but then I tried boosting 1kHz even more and widening the boost.  That seemed to make everything look better, though I called it quits at 4.5dB boost at 1khz at 1/2 octave.  (Wider boosts are better in general, narrower cuts are better in general to at least to 1/6 octave, but whatever works best is best.)

Left with Changed 820 and 1kHz filters

Now what used to look like tall peaks look like fluctuations, within a narrow envelope from 250 Hz to 2kHz followed by mostly a very gradual roll off which starts nicely just before 2kHz.

(One of the measurements I made with very similar settings was ruler flat in the range or at least very smooth.  It was apparently pure luck as I failed to save it and could not come close later.  The above "looks" are very robust.  I can do it time and again and get the same curve just about now.  But it gets worse strangely as I lower or narrow the 1kHz boost.  The valleys down to 250 Hz get deeper making it look much less flat.)

Are 1kHz boosts good?  Well they have an interesting history.  The B110 woofer used in the LS3/5A has a 15dB peak at 1kHz.  It's flattened to zero by the crossover which has 3 taps on the autoformer for matching to each driver.  It's almost as if they put in a boost deliberately to allow fine tuning.  This speaker is still highly regarded for clear midrange.  I'd note that when each speaker is fine tuned, the one that makes 1kHz not depressed at all and possibly even most slightly peaked is best...meaning in practice there is always a small peak.  I try to follow that principle, 1kHz should never be depressed.  (Though, here, I'ave allowed a slight depression as 4dB boost is a lot and I hesitate to go beyond... but I think the mystery time when it looked flattest was with 6dB or more, maybe should try that again.)

LP cutter heads also, curiously, have a very narrow 15dB resonance near 1kHz.  Filtering it out is not easy and nobody ever does (and most are unaware).

I'm in something like the reverse of the B110 situation, where my speaker (or room?) may be overdamped at 1kHz, so I'm adding boost rather than cutting it.

Not to make too big a deal about this...I've been using only a slightly smaller 3dB boost at 1kHz for years.  And if you are having boost, which seems to often be useful, digital is best.

I was unable to reproduce the graph above with 4.5dB boost as claimed above, I think it was for a different run when I used a 6dB boost which I now think works better.  I tried larger and smaller and 6dB worked best though up to 8dB was almost the same, 6dB everything around 1kHz gets out of whack.  Width is 1/2 octave.

*****

I started scanning the Right channel and found a resonance at 850 Hz which I notched out.  I was thinking that was almost good enough right there.  But while 1000 Hz was not in an obvious depression, it was generally hanging lower than the other chanel.  I expanded the 1kHz boost already there from 1dB ultimately to 5dB thinking it looked better at least that high, and was better matching the other channel, but here I also widened it to 3/4 octave which fixed the more erratic surrounding bands mostly.  But somehow this made the high frequency cuts inadequate in the right, so I increased them to restore the gradual rolloff starting just before 2kHz that had been originally intended, and once again matching the other channel.

Right channel after enlarging 1kHz boost and other changes


New Right Panel EQ Adjustments

Unlike the left channel (sadly), there are no GEQ's for the Right panels, yet.

***** August 22

You would think that having a 6dB EQ boost at 1kHz would be audible.  But, shockingly, it's less audible than you might expect on music.

I've dialed in 3 memory options in chairside EQ, one with the boosts as I designed them last night, yielding pretty much flattest response through midrange, one with no boosts, and one in between.

I can switch between them instantly using the memory menu.  The difference is clearly audible listening to pink noise (but it's basically impossible to say which one is actually better).  On music, it's pretty subtle.  I often wasn't believing the switch was made at all.  (Listening to Santana Supernatural btw.)

I think the full boost gives the most clarity and transparency.  Sometimes I wonder if it's a bit too much "in your face."  But previously I've often thought my system was sounding "too laid back" so perhaps this is a good correction.  Sometimes the boost gives "goosebumps" which is great of course.  But I fear possible listening fatigue.

Note that the measured response isn't showing a hump there even with full boost, instead it shows the flattest ever response.  When the boost is removed, there isn't much of a depression either just a subtle sag with nearby peaks standing out more.

For now I'm sticking with the full boost, which I think is most accurate.  But I hope to keep the chairside EQ around so I can switch if I feel the need to.

It's worth noting that while I'm not opposed to the use of EQ (obviously) and even to correct the sound of particular recordings, I've rarely (if ever!) found it to be necessary.  When some recording has an issue, I investigate and if I make a change at all it tends to become a permanent change.

I've rarely had a chairside EQ to experiment with this, but I did in the summer of 2022 when I first rolled out the "Chairside EQ."  I hoped to make chairside EQ a permanent feature, but that particular EQ unit died 6 months later (currently awaiting installation of new power supply caps, which I purchased in 23).  The unit I have now was a spare I bought at some point.

Now that I've whipped together a new "Listening Position" EQ for both channels I can say it sounds pretty good in background also, just a bit of excess bass at very low frequencies which gives a nice loudness compensation for background.  I still plan to make a background EQ again which would suppress at least some of the excess deep bass at other house locations.

At the listening position, it's awesome and mesmerizing.  It's transparent with the deepest and most effortless bass.  It's more 3D than ever, everything is more solid and real.  And it' wasn't bad before either.

I'm going to keep both panel and sub EQ's in the chairside as memories, and switch from one to the other as needed for testing.  (I've now loaded the new panel EQ's into the front DEQ, and plan to switch to bass fine tuning to cure that lump at 100 Hz in the left channel in my next stage.)  It might also be nice to have a more or less blank EQ for use as general "tone control" or when I don't know which I'm going to want to start changing bass or treble.  But whatever changes are made on such a "tone control" may not map easily to the applicable sub and panel DEQ's in the front, especially in the upper bass.

It's funny how both sides are very peaky in the 600-800 Hz range, followed by a depression at 1 khz and slightly beyond.  It makes me wonder if in the Acoustat manufacturing there was a peak in one material offset with an anti-peak in another material (say, the wires vs the louvers) which over time has separated, leaving a lower frequency peak which fails to correct the depression it is supposed to.  So the lumpy region around 1kHz could be aging by this or some other means.  Or, it could be the infamous "credit card" coloration that Harry Pearson of The Absolute Sound referred to (and I often disparaged).  The two speakers are of different vintages and so the separtion between resonance and anti-resonance differs slightly, or maybe that's just coincidence.

The Acoustat is an amazing speaker, but perfect it has never been.  This has not a serious problem for the many like me who don't mind using a bit of EQ to fix things.  (Robert E Greene of The Absolute Sound did likewise, and he once mentioned that the Acoustat's needed a peculiar midrange EQ correction.  One of the original team was an Acoustat owner who probably had RE Greene's correction.)

Now, there is far more than one way to get a similar looking correction curve.  It may not matter much which one you choose, but I'd think the simpler the better, and my correction this time for the subs at least is far simpler, less in need of re-correction in the GEQ undoing what's in the PEQ.  The result seems far smoother and nicer looking in a spectrum graph everywhere than I've ever achieved.

But also as much and more of my correction is due to room modes, crossover, and subs (which excite room modes much more than the Acoustats, especially that terrible 100 Hz one in the right corner).

You're always going to have stuff like that, with any reasonable combination of room, damping materials, placement, etc.  There's always going to be something that's just too hard to fix without EQ.  And that's true for most of the issues below 80 Hz if not 120 Hz, the usual sub range.

Now the way I've done it, with a combination of oscillator sweeps and RTA and cut and try and thinking, when I have time to really do it all the way through from scratch, yields a result which is much less super optimized than what the TacT did, and what people using total phase correction systems usually do, so it's less exact but also far less position sensitive.  Other than the exaggerated 20 Hz (which is sometimes cool anyway) it sounds fine at the kitchen table, where I actually sit most of the time.

Maybe now I simply need a less 20 Hz EQ position for showing movies, when most people get exaggerated 20 Hz around the periphery of the room, and movies are often loud there too.

So the Listening Position EQ would be the new normal, and the other EQ would be Movie, which gets switch back to normal at bedtime.

*****

I quickly cooked up a new Movie EQ (though it's still called BACK and B24) by applying GEQ below 125 Hz to the newly made LISTEN24 eq.  It does an ever better job than the old BACK of suppressing room modes in the back of the room, just giving "the right amount" of deep bass.  This is stopgap perhaps until I create an entire new BACK from scratch.  However BACK (to be renamed MOVIE) is less critical anyway, so perhaps this will do.

I've reversed the storage in the sub DAC so the new default is LISTEN24 now renamed NORMAL and the alternate EQ is now MOVIE, so that if I watch a movie and go to bed the next day it will be reset to NORMAL (assuming my home control system works, which it doesn't always do).

Without the movie compensation, there is the mother of all 35 Hz peaks in the corner "guest" position.  MOVIE tamps than down to a desirable bass rise.

It helped that I started both sub EQ's this time from blank, with no GEQ's.  So the MOVIE simply
applies a bunch of GEQ cuts to each channel, with slight variations, to remove the modes amplified at the back wall especially.

Sadly I forgot to reset GEQ's when EQ'ing the left channel panel, so those are baked in, and I couldn't remove them without making things worse or starting all over (but more or less they are OK, this needs rework soon).  Anyway, the panels don't need a MOVIE mode I think, just the subs where the deep bass distribution is problematical.

Movie EQ measured at back wall seat nearest corner (which was worse):

Left Channel at back wall, Movie EQ


Right Channel at back wall, Movie EQ

For now you'll have to take my word about the monumental peak at 40 Hz at the back wall with the normal EQ, as I forgot to take a photo.  It was about 10dB higher throughout the bass, or at least as much as the reverse of the graphic EQ:

Left Channel Movie correction

Right Channel Movie correction

***** August 23

Yesterday I copied over the new EQ's into the Panel DEQ at the front of the room from the chairside DAC where I had adjusted them.  To be sure I did this correctly, I did before and after RTA's but even if I saved some of them, it all got confused so I couldn't report.

Today I intended to re-do that test and save the results, also giving another report on how good the new EQ's are (since in the meantime I put everything away to show a movie, then put it all back this morning, including the chair and microphone stands, so these are "independent" measurments).

As I was doing it, however, I immediately got confused about whether I had started from the front or chairside DEQ (selected by AES/EBU or Optical on the panel DAC).  So I had to toss out the first measurements.  Then I may have accidentally twisted the DAC level control instead of the selector, and I had to repeat my thinking about where it is now supposed to be set (I believe it was last set to +9.5).  So then I had to scrap two more measurements because of fearing I had accidentally reset the level.  Finally I made all 4 measurments, but still wasn't sure the first two were correctly identified, though I had originally believed so.

So I believe these are "correct" but might not be:

Left Channel, Chairside DEQ


Left Channel, Front DEQ

The features look the same, but with random level variations.  I'm judging this a likely perfect match.

Right Channel, Chairside DEQ

Right Channel, Front DEQ

These look even closer, and I'm judging a perfect match.  It's also notable how much smoother the right channel appears to be.  I'm thinking this is partly/mostly a speaker location issue, the left channel is in a "hard" corner while the right channel is in a softer corner that is also the front entry leading to the hallway.  (Little variation is caused by repeating measurements, more variation is caused by slight differences in setting up the chair and microphone stand between these measurements and ones done on previous days, and possibly differences in ambient noises.)

I'm doing all measurements now with the iPhone in "head center" position which is fairest to both channels (and I did not find huge differences moving it to left or right ear positions anyway).  In spite of having tuned the right channel at a different position (the ear position) it still looks smoothest at head-center position.  I don't think it is smoother because I originally adjusted it the different ear position, that's just coincidence I believe, but it shows that it didn't hurt.

What's been bugging me the most is the 100-125 Hz peaking in the left channel.  I think I can hear it as a slight midbass boominess.  But to work on that I was thinking I should copy the right channel bass adjustments to the chairside EQ and combine them with the left channel which are already there because that's where I did them.  When I started fine tuning in July on the right channel, I was doing everything directly with the front DEQ, but getting down on my knees to adjust it so much brought back a minor knee injury which led me to setting up the chairside DEQ again (and discovering I had an available spare).

But the left sub memory already in the chairside DAC should be sufficient, I am thinking now, since I'm only working on left sub channel today.

Before sweeping this time, I figured I should cancel the EQ's between 71 Hz (which is solid, I know it needs that, and only 1/3 octave, which doesn't *seem* to affect much beyond 80 hz or so) and 138.  But the 138 was fudged at one point so I turned it off too.  I fudged that to allow removing the 129 Hz filter which was sucking out 125.

In fact I think much of the lumpiness in the present curve resulted from trying to fix 90 Hz (an unfixable positional null not much affected by EQ) and 125 Hz (which was fixed, but then I did too much unnecessary fixing).

I turned off the left panel to focus on the bass, sweeping with the oscillator and the chairside EQ.

There is absolutely a need for a deep notch around 140 because the boom there is simply incredible, despite the 8th order crossover at 125 Hz which should mostly be shutting it down.  And the natural center of that boom, as I had determined many times.  So I put a deep notch right there again, at 144 Hz.

But once again, the 129 Hz notch also seemed necessary.  I simply made that as narrow as possible, 1/10 octave, which fixed the issue in sweeping without creating any adjacent weakness.

No EQ or lack of EQ at 71 or 85 affected the depression at 90.  I dialed in the filters that seemed to be needed in those locations.  Previously I had cancelled the 85 Hz filter in the hopes it would elevate 90 Hz but it did not.  Nevertheless, I made the filter quite narrow at 1/7 octave.  (I spent many minutes trying different widths for this filter.  Nothing narrower than 1/3 octave seemed to affect the depressed 90 Hz range at all.  I know these appearances may be deceiving however.)

I noticed that a shallow depression around 44 Hz was centered right on a notch filter I had there.  So I lightened that notch from -7.5 to -6 which fixed the depression.  I also widened it slightly to 1/3 octave.

The result of over an hour of sweeping and adjusting EQ's and measuring RTA's was far better looking than I've ever seen the left channel.

Left Channel after bass sweeping session

It's almost ruler flat (+/- 1dB) from 20 Hz to 71 Hz, then a positional depression I can't do anything about with EQ, then back to almost the same level from 100-140 Hz but gradually rolling off about 1dB, then returning to slightly lower panel baseline basically giving the desired room curve.  It sweeps about as nicely as it looks too, or nicer, because the ear becomes more sensitive right where the level changes around 200 Hz making it sound like there wasn't any change.  The positional depression is not as it looks, most of the 80's and 90's are not depressed at all, there's just a very narrow deep half null at 91-92 Hz.  

(*Some things could be done about the 91 Hz near null but they vary in cost.  Curiously enough, I cut the panels 0.5dB at 80 and 100 Hz when maybe I should be doing the opposite, they'd fill in better than the sub could because their output is being amplified rather than suppressed by the mode.  The positional null is probably caused by a mode, elsewhere 91 Hz would be amplified, such as at the back of the room.  No serious re-arrangements of my room, speakers, or listening position are possible, I've already sacrificed enough of my multipurpose living room to have my Acoustats 4ft from the front wall which is a good chunk of it.  I know the optimal position might be to have speakers in center of the room, but no way can I do that.  So I'm trying to this arrangement as good as I can, and with minor exceptions, I can get it very good, I've never seen flatter bass.  91 Hz is also a practically absorbable frequency, so I could try to deal with the issue that way somehow maybe.  Also, since it's a very narrow range, a specifically tuned absorber would be useful.  Any such solutions would require hundreds or thousands of dollars and many hours of brain and work power.  Compare that to just turning a knob...how I dealt pretty well with most of the other issues of which there were a lot.)

New Left Sub EQ's

As I've said before, when you get something this good, you don't mess with it for awhile (though I've often fallen into that trap).  I'm considering the sub problems solved.  The remaining issues in left channel are in the panel EQ's.  There's a bothersome 355 Hz depression which is possibly an EQ error.

And none of this required any sub boosting.  (There is quite a bit of panel boosting which might be re-examined.)

**** August 24

A lot more happened and was written here on Friday night, this morning, and up until a few minutes ago (1:50 PM Saturday).  But then a computer restart, and it was all lost.  When I reopened the page in my browser, all the updates I'd worked on for hours were gone.  Pages of work and dozens of photos for a big update that was nearly completed.

Which is doubly curious because the problem I identified this morning and spent hours on today was ultimately fixed by doing a reset on the Behringer chairside DEQ.  It had reached a point where the display was not matching what it was doing anymore.  I was worried I was losing my mind.  And just before my computer restarted, losing pages I'd written up on this issue and the previous ones I worked on last night, I finally figured out what to do: power cycle the chairside Behringer.

So, enough on the reset thing and the weird way the display didn't match.  I'm not going to replay that looong detective story.  I've proven to myself that stuff like this does happen.  Now I know it might be good on any big change to save it, then restart and reload.  Sadly if what you get is worse afterwards, the magic was never saved and perhaps never even displayed, but at least you know what you're doing now, at least until the next glitch.

I also reported on my futile knob twisting last night trying to eliminate the tiny dropout at 280 Hz.  There are GEQ's operating there, but increasing 250 Hz only makes the dropout look worse.

Except that I decided to go with making the GEQ flat at 250 Hz rather than the 2dB cut it had been, even though the spectrum could be interpreted either way.  It didn't actually raise 280 Hz detectably, and even made a slightly bigger difference between 250 Hz and 280 Hz, but it seemed like the 250 Hz itself might best be raised slightly, there was a kind of droop and it was already lower than I think that area needs rework badly, possibly eliminating the GEQ's and replacing them with more finely controllable PEQ's.  For now, one less cut is probably a good idea.  I showed the photos before my update was wiped out but it's not a big change in either the EQ or the result.

Also worrisome, the response may no longer look quite as good as it did, almost magically, yesterday afternoon.  It could actually use a bit more fine tuning in the bass again.

*****

I opted for another sweeping session on the left panel EQ late into Saturday night.

I decided to throw out all the old GEQ boosts and start from scratch.  The weird patterns in the left channel response are at least partly because of using GEQ which is not as fine tuneable as PEQ.  On PEQ's you can dial in the center frequencies to close to 1%.  That turns out to be crucial to getting the kind of smooth response curve I've been able to get in the right channel.

And while it's sometimes said you can't boost a null, it is indeed the nulls, or at least the lowest points, where you should center each boost.  You can boost as much as you want there, and not only does it (slightly) help there, it seems to fix even distant things in the response curve, and without creating any new peaks.  If some peaks are exaggerated by this boosting, you can either decrease the bandwidth of the boost, OR you can notch out the troubling peaks.  I tested both approaches and came up with a combination of both, fairly narrow (not wide) bandwidth boosting, combined with deep notching of adjacent peaks.  But I'm also open to the idea of wider boosting.

I'm afraid to use more than 6dB of boosting.  It cuts into dynamic range and can cause other issues, so for now I've adopted that rule.

One other now is that I'm running close to the limit of 10 PEQ's per channel.

There was no getting around it, even with no boosting I had an elevated section with a peak at 250 Hz right next to the lowest null at 272 Hz.  So I tried various ways of setting two correction filters (a notch filter at 250 and a boost at 272) to find something that basically looked right.

I'm also trying to transition from a slightly higher level in the bass below 140 Hz, to a midrange level that's about 5-10dB lower.   That seems to be the requirement, for me, for natural sounding bass in my living room.  I was thinking that transition should look more like a pool slide, downwards u shaped, rather than upwards u shaped which was what I kept getting over and over (and for which I partly blame the GEQ's).

In the right channel, I achieved a perfectly smooth decrease from 20 Hz on down, leveling out a bit in the lower midrange, and down again starting before 2000 Hz.  Looking at it now, I'm amazed I achieved such a nice looking response.  I think that's about ideal.

Right Channel (August 23)


Left Channel (August 24)

The left channel is veering towards the correct tilt (although the bass looks a bit too elevated, not so much comparing 20 Hz and 1kHz, that might use a bit more difference, and there are peaks in the 63-125 range which could be tamped down more.  But it's lumpity bumpity all the way.  Still, the worst single feature is the near null at 280 Hz and I decided to try to fix that.

With the GEQ's all turned off, it looked like this (which curiously like what I was getting in the morning when somehow the Chairside DEQ had started acting weird, apparently the GEQ had disconnected itself, but the light was still on):

Left Channel, no GEQ boosts

It's mostly smoother and more relaxed, though the bass between 125 Hz and 200 has fallen too much (instead of the equally undesirable peaking with the boosts).  Anyway, it seems overall like a better place to start.

Using a wide band boost centered at the 272 Hz null, and a mere 6dB boost, things look very different:

Left Channel, wide boost at 272 Hz

There's a smooth decline from over 500 Hz down to 250, but a huge peak on the other side.  Somehow, the wide boost has stimulated the upper side more than the lower side.  (It may have to do with notch filters operating on the upper side already.)

Clearly I needed to notch out the region immediately below 250 Hz as well as boost at 272 Hz.  I tried various widths for the 272 Hz boost, and ultimately decided on a mere 1/3 octave, to keep the 250 Hz resonance from growing out of control.

Left with new filters



New (and old) left filters

This even sweeps very nicely in the middle, you hardly notice the null anymore.  What used to be jagged there is now no worse than +/- 3dB from 140 to 500 Hz.

Now certain other features are out of whack, and I hope I can fix them with existing filters or I'm stuck with the filter limitations.  (Though in principle I do have workarounds, I could also use PEQ's in the miniDSPs.)

There's currently 6dB of boost at 1kHz and that's mostly responsible for the apparently jagged peak right there. It probably needs to be brought down, but how much?  There are many other filters across the 200-1000 range that could be made larger or smaller, to raise everything up or bring it down.  I have no idea which would be more correct.  I'm pretty sure the 125 region now needs to be brought down.

*****

I tried to smooth out the 200-1khz range a bit.  I cranked down harder on the 624 Hz resonance.  By then it appeared the 825 Hz cut was unnecessary, it was already sucked out there, so I removed it.  I tried different levels of 1kHz boosting and width and settled on a two octave bandwidth (!) and +4dB of boosting, a bit less than before, giving a nice smoothing effect but also depressing the 272 Hz depression more than no boost, only relatively speaking.


Left with deeper 654 Hz filter


Left with reduced 1kHz boost

Left with wide 1kHz boost

New left EQ's

While I was measuring it, I was thinking how wonderful the wide 1kHz boost looked.  Plus or minus 3dB from 200 to 4k.  But then looking over the results, I was basically there with a smaller 1kHz boost in the previous measurement, which actually looks better in most ways.  The wide boost does seem to smooth things a bit around 1kHz, but at the expensive of raising it even more (by 4dB of course) over the lowest spot at 272 Hz.   It makes the Gundry-Linkwitz dip on the high side much too extreme too, so if I kept wide boost, I'd have to dial those cuts back (instead of increasing them, as I had to do in Right channel).

So, I need to go back to explore various 1kHz boosts.  I'm also noticing that the primary "lack" that I was trying to "boost" is not 1kHz but 900 Hz, strangely enough.  It's quite possible I need to re-center the 1khz boost to 900 Hz or something like that.   More sweeping is called for, this center boost issue is not yet settled.

When I started this re-EQ process I was still struggling to read the Analyzer display correctly.  It seems the bands that correspond to the numbers are the ones at the right end of those numbers, but I had trouble seeing that at first.  I think now it always was 900 Hz that needed boosting, there was always a local peak at 1khz.

**** August 26

Comparing the previous wide 1k boost to no boost, it's clear that no boost is better.  But with no boost, there is a small depression and now it is clear that is around 900 Hz.

Left with wide 1kHz boost

Left with no 1khz boost

I swept the region from 800 - 1000 Hz over and over and was disappointed to find there was no clear "900 Hz" suckout.  Through most of the region it sounds quite loud, perhaps partly because of increased ear sensitivity.  I had to turn the frequency knob very slowly to find slight downward ticks at nearly 1000 Hz, just above 900 Hz, and around 850 Hz.

Boosting the 990 Hz downward tick is just like equalizing 1013 as I was doing previously.  It doesn't fix the 900 Hz depression.  Boosting the 850 Hz tick is almost as good as boosting the 903 Hz tick, pretty good.  I got the very flattest result at some intermediate frequency, perhaps exactly 900, which sadly I didn't write down (any or the amplitude or width either).  But it made no sense to me to boost there because it was already elevated in sweeping.  Nor did it really make sense to boost at 850, since it's so close to the already peaking 800.

At this point it would be good to be clear on what the "800 Hz" band in the RTA means.  Does it mean 800-900 Hz or does it mean 750-850 Hz.  I would guess the latter.  That is the band I don't want to boost because it's already peaking.

For that reason, a boost in the 850-950 region seems warranted to raise the 900 band, and the 903 I've chosen at a minor downtick works pretty well (if not as good as that one I don't remember because it didn't make sense):

Left channel, boost I don't remember

One thing about the unknown boost is that despite the flatness of the 800-1400 Hz range which I haven't been able to achieve otherwise, it bugged me that 1400 Hz was trending upwards before going down, rather than trending downwards.  That was what made me dismiss it during the experimentation, but not enough not to at least photo the measurement.  That upwards trend suggests it was a wide or higher 900's boost.

Meanwhile, boosting 853 predictably raised the 800 band too much (now it's the highest of all, but only by a hair) so it wasn't right.  The 903 Hz boost was ever so slightly better in every way, with an almost flat 1kHz wideband, if not as flat as they mystery boost.

Boost at 853 Hz (?)

Boost at 903 Hz

I might have to decide what looks better in the spectrum and which sweeps better.  But for now 903 Hz boost (3.5dB, 1/3 octave) seems to best of what I know and kinda makes sense too.

Quickly trying different boost levels and bandwidths, and higher boosts such as 914 and 924, I came up with another version that looks pretty close to the "mystery" version while still making sense.

Left, 903 Hz, 1/6 octave, +4dB

Now I think I am going to rest the 1kHz situation now as it's about as good as anything else.  I haven't had to budge from the 903 Hz which is a natural near null to places higher which don't seem like they should need boosting.  The whole midrange is flatter as I'm not boosting it with any wider boosts.  And it's funny how narrow boosts, even if they sweep nicely--as this does, seem to pump air into the curve farther away at weak spots and make everything look more level.  (Speaking of which, I could pump more air into the boost at 272 but it's already at my arbitrary stop point of +6db, still I'ld like to see what it does.)


At this level, the 100-140 Hz band should come down (that's a bass adjustment)*, the deepest bass could come up, and the upper dips readjusted to be flatter without that local peak around 3.8 kHz, perhaps a little boost to 180 Hz which is easily done, and even 90 Hz though I think that's more touchy.  If 180 is boosted perhaps 250 could come up a tad which might relieve some of the sag now around 300 Hz, also the 393 Hz filter might be readjusted.

(*With only a narrow boost at 903 Hz, which is a natural depression unlike the 1kHz I thought it was, the midband is now sufficiently beneath the deep bass that I could lower the sub a dB or two, which might help everything, or not.  In previous curves, my wider boost brought 1kHz much closer to the deep bass level.  Lowering sub level might resolve some issues in the 20-200 Hz region, though it also might necessitate some re-adjustments.  If I lower the sub level, I could nevertheless raise the deepest bass with a tad of EQ there, which I did use already in the right channel.)

Though also, to be honest, I haven't done enough listening to conclude whether the taller 50-125 Hz bass of the left channel as it is now is worse than the right, or vice versa, and it's hard to remove that from other differences.

*****

I tried reducing the sub by -2dB (-8) and -1dB (-7).  My intuition is that -7 is the best and better than the previous -6.  (Examine these graphs from 500 Hz down because there was ambient noise...my dryer was running.)

Left with sub at -8

Left with sub at -7

I failed to do a "control" measurement at -6 at this time, relying on the previous -6, but that showed a nearly flat middle midrange around 1khz which is gone here.  That couldn't possibly be related to the sub level changes, I figured, but to be sure I ran -7 and the original -6 later when all was quiet.  Actually the sub change may have some subtle effects, but not the flat to bumpy thing.  Bumpy is apparently the norm and the earlier super flat one was luck.

Left w sub at -7 (refrigerator running?)


Left with sub at -6

Left with sub at -7

I noticed the refrigerator was running at the end of the first -7 measurement so I ran it again.  Differences were minimal.  The midrange from 500 Hz and up is nearly identical between -6 and -7, the last -7 measurement looking smoother there than the immediately preceding -6, but the preceding -7 looking just a tad rougher.  So those variations from 500 Hz and up can be discounted.  Meanwhile the bass looks towering in the -6 but "just right" at -7.

***** 

I was about to start handling some other issue, but when I went to the chairside DAC I noticed that the midrange boost filter was set to 914 Hz and not 903 Hz as I had been saying.  And after all this time I'd been bragging about the fact that I'd set the boost at a near null and not at some arbitrary frequency somewhat higher just because it makes the RTA look better.

Maybe this was the magic that lead to the super flat near 1kHz response which I couldn't exactly reproduce later.  (Note here that I can also get super flat like response often by standing in front of the phone microphone, or something like that.  If I do that by mistake I have to restart the measurement, but might have forgotten to.)

So I made that one change, back to 903 Hz, and got this result:

Boost changed to 903 Hz

Wow, that was it!  I thought.  

What I should have immediately done at that point would be to photograph the EQ's, something I'd neglected to do when I was last adjusting the boost (or I might have noticed the 914 vs 903 mistake).

But instead, I thought I'd just remember it (as I do most of these things) and tamp down the peak at 654 Hz which was seeming to now stand out more.  Then it seemed that 654 wasn't the 6dB cut I remembered but instead 4dB, so I upped it to 6dB.  That didn't look right either.

I then spent about the next half hour merely trying to get back to something that had the midrange flatness of the preceding "903 Hz boost" measurement.

Here it seems there are several phenomenon going on.  One is that the 654 boost, especially if it's made wider, and interact with the resonance in the 800 Hz band (which keeps pining to go higher), and that resonance affects everything else up to 1kHz and beyond.  AND, my attention can be laser focussed on some details while completely ignoring how others are changing.  The settings which made the midrange centered around 1kHz look flattest were often the settings that drove 654 above that flatness, whereas when it's done right, perhaps, it just blends in.

Notice in the "903 Hz boost" measurement immediately above, and the measurement immediately preceeding, that this attention issue was working.  Indeed, the 903 Hz Boost measurement has flatter midrange, but 654 stands out more.  Whereas in the preceding measurement, the 654 Hz peak looks like just another one of the bumps.  So it's hard to know which one is better anyway.  And it seemed there was no way I could get that smooth looking middle response without making 654 Hz stick out more.

As I was doing these measurements, or at least in the last 15 minutes or so, I started trying to write down the settings and the time each measurement was taken.  But I made mistakes there and this was slowing everything down immensely.  Previously I've been "identifying" each measurement only afterwards when making the next post, which works ok only for a small number of changes, and perhaps not even there.

Somehow I had dialed in wide boosts for the 654 Hz notch filter, and those really mess things up.  Another issue is that I can barely read the display on the chairside EQ without my reading glasses and I rarely bother to put them on.  And given the controls on the DEQ 2496, it is very easy to mess things up.  You can't, for example, scroll down through the amplitude adjustments if one PEQ is turned off.  When you scroll down in such a case, the cursor snaps to the right again.

Finally I got back to something fairly decent again, and I took a photo of the EQ's so I don't lose even this:

Left Channel

Left PEQ's

Now while writing this I can see from the previously photographed set of EQ's that I previously had a -8 cut at 654 instead of just -7 (though the Q was the same) and I ought to just be able to copy in the 903,1/6,+4 measurement I used in the so labeled graph which shortly followed.

Left, restored filters from 8/26



(I had to do this several times to get it right.  During that process, the EQ locked up again and started looking weird so I power cycled it again.  My phone didn't update to the cloud so I put up last night's measurements by mistake, and then boasted about how much better they were than last night, before noticing they were the same.  I finally got the EQ's correct and measured again and updated to cloud after waiting a half hour.)

Now photographed and saved (memory PB), this is the new default.



Same as what should have been the old default except I didn't save or photo it, and wasted about 3 hours over 2 days as a result.

Rule 1: When you get it pretty good, photo and save to memory before making ANY changes, or you may never get back there again.

(I already violated this rule just now, as I was going to save, I turned the 4.0 to 3.5 again by mistake, and just turned it right back and saved.  Ok, maybe that's ok.)

After saving that also in the front DEQ, I thought I'd reduce that huge 363 Hz notch (at -7dB) where there's basically a depression anyway (though it's a kind of "smooth depression" if you will, extending across the entire lower midrange).

I began to think that maybe that "smoothness" isn't such a good idea when it's unbalancing the overall spectrum, 270-500 Hz was still a low point in the entire curve, even though it looks smooth maybe that isn't really such a good thing.

In fact, if you could zoom in to single digit frequencies, or even finer, you might find that there is no such thing as smooth.  The finer your resolution, the more furious the up and down variation becomes.  So why preserve an apparent "smoothness" which is really only something related to the limited resolution of your analysis.

Anyway in copying over the numbers, I was also thinking that in principle a 7dB cut is not such a good thing.  Cuts that large should be avoided unless absolutely necessary, and once again this was a spectrum wide depression.

Well, that immediately seemed to raise the entire low spectrum part up, and I was inspired to apply similar reasoning to all the cuts in this region, at 272, 363, and 437 Hz.

When I thought I had a good new thing, I photoed it, saved it, then tried (and succeeded this time) in making it even better.

So with this change, the low smooth plateau has been eliminated, and output is roughly at the same level (with a few 1/6 and 1/3 octave depressions) all the way from 160 to 1.7 kHz where my intentional dips begin.

I'd been hoping to illustrate this with photos but the cloud isn't updating very fast again, it's been an hour and my updates have been in limbo.

Then I figured I'd go ahead and play with the boost, since there clearly was a spectrum wide depression there, although I couldn't increase the boost from 6dB, what I think is reasonable, I could shift the center point and Q based on careful sweeping, with the new raised surrounding notches it might be different.

And it was, but not by very much.  I took the new center as 282 Hz, and a 1/3 octave boost neatly raised the lowest depression so it wasn't so low anymore.

Well that was great, except now the 251, 363, and 654 filters needed adjustment.  At some points I took photos of the EQ's but kept on adjusting a bit more.  The last adjustment of the ultra critical 654 Hz notch (which determines the tilt around 1000 Hz, and whether there is peaking at 800 Hz) was a mere 0.5dB, with things markedly worse on either side.

I may have exceeded my cloud bandwidth today, it's infuriating that I haven't been able to post any of these photos yet, which make the easier for me to examine too (as well as keep track of, before I forget what what what).

I just plugged phone in as I was set to take a nap, and it updated.  So I'll skip all the intermiediate steps, and cut right the the chase, the new response I got by cutting the peaks so they all line up with the elevated region around 1 kHz, and that was after re-tuning the boost from 272 to 282 Hz and widening it to 1/3 octave.  As a result of these changes, response is flat within an envelope all the way from 160 Hz to 1.4 kHz or more, the best looking left channel yet.  No need to go back over these latest steps I think now, it's basically nailed within that region, even the slightest change messes it up.

Readjusted boost and level tuned notches

Latest PEQ's

This still isn't perfect, but I still think the improvement is astonishing.

*** August 30

It's time to start cleaning up the living room for my next Sunday party.  I copied the chairside PEQ's into the front DEQ so I can move the chairside DEQ out of the way for awhile.  I added a 282 Hz notch filter to the sub MOVIE preset because at the back of the room (movie watching) the panel boost at 282 Hz creates a peak.  It seemed I could cancel the +6dB 282 Hz boost in the panels pretty well simply by notching out -7.5dB in the subs, which is strange but appreciated for now because I only have automatic switching set up for the sub DEQ presently.

There seems to be a difference between using the chairside DEQ and the front DEQ even after making all the settings (GEQ's, PEQ's, and Delay) identical.  That will be the subject of the next post.  I need to solve that problem, then I can get back to making the left channel better.  It seems to need boosts to fix depressions around 90 Hz, 180 Hz, and 1.7 kHz.  It's going to take some work to get those right.

This is the end of this post.  It's been a wild ride as usual, but done better than ever before, and fairly well described here for the first time.




No comments:

Post a Comment