Devil in the Machine

Now, on precisely the day after my biggest audio DIY project in 35 years--replacing the Low Frequency transformer in the Right Acoustat--I find that it probably wasn't necessary to do so.

I did all the required experiments beforehand, and they all came out the way that showed the LF transformer replacement.  But, in retrospect, it was all apparently coincidence.  The devil in the machine, the machine of my entire system--or perhaps just the Krell FPB 300--had me fooled.

It appears that the capacitor replacement was the "real" fix to the right speaker--the fix that stopped the fairly regular amplifier shutdowns.  I did that fix on Christmas Day, and then enjoyed the system without further shutdown incidents so far.

However, just after New Year's day, I observed a new problem.  The right channel of the amplifier, which normally runs 10F cooler, was running up to 15F hotter than the left channel, and reaching the highest temperatures I've ever seen on the Krell, just playing FM radio at moderate levels (actually, in retrospect, they were "moderately high" levels).

I then tried running the speaker with the LF transformer disconnected, so it was only playing the circuit in which I had replaced the capacitor.  At first, I did notice that the amplifier was drawing more than the expected wattage at startup in the morning.  I was thinking "aha, there is still a problem with the HF circuit after all."  But, then the wattage went back to the normal levels (630W average long term, playing FM radio, both channels playing apparently at plateau level "2").  And normal behavior was then observed for two days.

So I switched back to the LF circuit, by plugging it back in, and disconnecting the HF circuit, by clipping the wire from the capacitors feeding the variable resistor.

With this change (and also, I wasn't noticing, the warmer weather) the right channel of the amplifier started "overheating" again, running hotter and hotter each day until Friday I was seeing 194F again, just as I had just after New Years, when the left channel was the typical (for it) 169F.

To follow up the capacitor replacement I did on Christmas Day, the next day I had re-cabled the Krell (and the Hafler 9300) to run through my modified QSC ABX amplifier switching box.  And for that purpose, I had twisted all new speaker cables for each amplifier to feed the ABX box, and hooked up the old Canare 4S11 cables to run from the ABX box to the speakers.

It occurred to me that the new (and old) cables, and switching box, might be causing some kind of amplifier parasitics.  So, I did the test for that, I ran my old low cap twisted pair Teflon speaker cables, bypassing the ABX box and the 4S11.  That did not help, I was still getting the high temperatures on the right side.

So, I switched the cable over to the other speaker, so I was powering the left speaker with the right amplifier channel (while the other channel had shorting input plugs and was disconnected from speaker cables).  Now, there was no overheating problem in the right amplifier channel all day Sunday.

So I had all the evidence I needed, from a perfectly consistent and complete set of experiments, to show that the problem was being caused by the LF circuit in the left speaker.

After I did the replacement on Monday, I proceded to play both speakers on their normal respective channels, and noticed the right channel of the amplifier was running about equally warm as the left, but not getting above 170F.  Perhaps this was just because I am now running both speakers full range.

So, even the "experiment" after the transformer replacement showed I had done the correct thing.

So all was well, until Tuesday morning.  Then, over the course of 3 hours, the right amplifier channel heated up to 190F all over again, while the left channel was loafing around at about 160F.  On the way to 190F, however, sometimes the left channel seemed "normal" at 160F or so, then other times it was in the 170's.

Monitoring the amplifier wattage (around 730W) at idle, I tried various things, including disconnecting the speaker connections from the amplifier altogether and plugging the shorting input plug into the right amplifier channel.  Nothing seemed to help.  I ultimately let the amplifier idle for 3 hours, and it fell only slightly to 630W (with the left channel at 120F and the right channel at 170F, this suggests correct idle "level 1" in the right channel and "level 3" plateau in the left).

Now this "overheating" problem seems to be some malfunction in the Krell itself, and something I've seen long before, mostly in the left channel.  As long as the amplifier doesn't actually shut down, perhaps the best thing is to just let it do it's own thing.  Perhaps the cold weather affected some sensor in the Krell and it's temporarily or not so temporarily misbehaving.  But sounding great!

At this point, apparently the thing to do is just let the system play.

The right digital filter

The correct digital reconstruction filter for digital to analog conversion, is the closest approximation to sinx/x that you can make.  That is, the steepest brickwall filter, the kind that been the mainstream basic from the beginning.  The high end that's always tried to make a "better" filter, inevitably slower, is wrong.  See the discussion between Dave Moses of Wadia and David Rich of The Audio Critic and the famous Professor Stanley Lipschutz, starting on the page marked "pdf 6".

With regards to the infamous pre and post "ringing" of the brickwall filter, David Rich emphasizes that it's not actually "ringing" or a resonance.  Prof. Lipschutz says it does not represent "time smearing".  Likewise for the ripples in pre-cutoff frequency response, which--which are very tiny nowdays due to oversampling anyway.  No information is being added earlier in time or later in time. The correct reconstruction, and there is only one, is the best possible given the initial brickwall filtering and sampling.  This is the one that minimizes error, and it is the sinx/x brickwall filter.

In my words, it represents the information lost in the initial brickwall filtering and sampling.  That information is lost, gone, and any attempt to systematically fill it in is going to fill in elsewhere where not needed.  So it is that slow filters lose the high end frequency response (as, apparently, the Wadia filter being discussed above) and then add grundge (noise plus IM) from aliases which are not being filtered out.  The tiny inaudible ripples are removed at much more audible expense.

And most recently I was reading Archimago saying the same.

What about MQA?  Well, first one has to qualify that MQA is many things, including a companding codec that encodes the some of the benefit of higher sampling rates into lower ones, a very clever trick.

Beyond that, I don't know that much about the digital filters used in MQA.  The format claims to have the ability compensate for the time response for coding at both ends.  That sounds, in principle, ideal, but I'd be sure it isn't perfect.

And MQA as a streaming format cannot be "perfect" as it is necessarily information losing compared with the straight digital at the same rate.  MQA claims the benefits encoded into this information space far exceed the losses.  So, in short, it doesn't even try to be a "perfect" codec, it's a perceptually enhancing codec that replaces some of the original information necessarily and replaces it from other information gleaned from the original encoding and/or metadata about the encoding.  I'm not in principle opposed to such things, I have always enjoyed HDCD, and remain a fan, though the looser in HDCD is the person who doesn't have the HDCD decoder.  MQA is an updated version of the same kind of thing, however the losses to the decoder-less user are less, and the benefits to the decoder bearing user greater.  Perhaps.

Saying all that, MQA is no cultist thing, and I'd be inclined to use quite fast filters, if not the fastest possible.  Good old HDCD swung in the appropriate content-dependent filter at moments notice, in addition to volume compansion.  MQA more cotinuously adjusts to its target mathematics, which I believe is the semi-brickwall filter producing no "pre-ringing."  That's probably unnecessary, but pretty harmless.  If indeed "more-perceptual" information from the higher rez is being restored, this would make up for the lossy encoding and suboptimal filter.

MQA could never work for me as a 'final' codec.  I can't in any useful sense use MQA digital at the input to my DSP's or at my ultimate amplifier feeding DAC's.  And as far as I know, there is no MQA-to-highres converter, the same limitation that plagued HDCD.

All the same, I've been planning to add MQA, it might be sonically beneficial actually in some cases, and analog resample to 24/96, as I do with HDCD, SACD, and so on.

(*Of course, in cases where MQA encoding is used, one wants the MQA decoder, just as with HDCD.  Sadly for me this always requires an analog resampling.)

Reconsidering the Breeze and Emotiva

I've figured out how to optimize the DC offset on the Breeze 6010D, and that is to adjust the pots near the input to zero the unbalanced output offset (or as close to zero as possible) with the volume control full up, and then adjust the pots near the volume control when the volume control is at roughly the "unity gain" position of about 1:30.  This should be done with shorted inputs, and since I now plan to mainly use the unbalanced outputs, those should be optimized.  However, the offset at the balanced output tracks that at the unbalanced pretty well anyway, with about double the total offset and often less than that.

In the left channel, this gives me unbalanced offset 0.05-0.10mV (virtually zero) all the way to 10:00, then the offset starts rising and rises to about 2mV with the volume control full off.

In the right channel, I couldn't zero the volume full up position, so that remains at about 4.9mV.  As you turn down the volume, the offset quickly falls to near zero for the middle range, and then rising to 2mV at volume full off.

If alternatively I adjusted the controls near the volume control to zero the output at zero volume, I get offsets like -12mV in the midrange.  The factory had adjusted (or not) the offset to 12mV at volume full up and 5mV in the midrange.

Sadly there is no way to adjust the balanced input distortion to make it better than the factory 0.0068%.  But, as I last reported, using the unbalanced input, distortion is as low as 0.0003% at the balanced outputs.

To do that last measurement I ran my computer balanced outputs into the Emotiva, and then used the processor loop output of the Emotiva, which seems to add zero distortion, making 0.0003% measurement possible.

But, instead if I use the "Main Preamp" unbalanced output of the Emotiva to feed the Breeze unbalanced input, distortion rises to 0.0016%.

So what I see now, is what I've never actually been able to measure before, thanks to the Breeze, I can see that the unbalanced output of the Emotiva is not nearly as low distortion as its balanced output (0.0017% vs 0.0005% as measured).

Meanwhile I measure only 0.0006% THD from the unbalanced output of the Breeze (and 0.0003% from the balanced output), so long as I use only the unbalanced input of the Breeze (fed by the Emotiva processor output) yielding this reversal of previous beliefs:

The Breeze through it's unbalanced input and either output has lower THD than the Emotiva through it's balanced input and the matching output.  Where the Breeze fails to deliver superlative performance is only through it's balanced input.  I believe that's because the gain of the two amplifiers in the balanced input circuit is not precisely matched or adjustable.  The AD797 IC's in the Breeze appear to be about as good as the LM 4562's used the Emotiva, which is not surprising.  The OPA211 might be still better.  BTW, I believe the difference between chips such as these, and even more pedestrian chips like 5534 and derivatives, is not audibly significant.  The 0.0062% I am protesting here is probably of no audible consequence either, just lacking sufficient bragging rights.  -80dB distortion, which is 0.01%, is probably sufficiently low regardless of harmonics, and as it turns out the 0.0062% here is mostly 2nd and somewhat less third order, and their multiples, so perhaps of less audible harm than other harmonic distortion profiles.

Other than non-perfect DC offset, still basically acceptable, and likely to be near zero at near unity gain settings (as that's how I adjusted it) the Breeze looks like an excellent gain adjustment preamp for ABX, provided only the unbalanced input is used.  This does somewhat limit it's utility, for abx, for example I could have a pair of Breeze's feeding the Krell and a different amp through all-balanced and all-unbalanced chains made possible by the Emotiva Stealth DC-1's combination of balanced and unbalanced outputs.

The relatively poor balanced input performance makes in not useful as an instrumentation preamp for my test bench.  For that purpose, I currently have no alternative to the Emotiva, and best using it's processor loop for testing equipment unbalanced, or balanced output to change level to balanced inputs.

In a pinch, I could use the balanced I/O of the Breeze to adjust Krell level so visiting power amp could use the unbalanced output of the DC-1.  I actually believe the balanced input not to be audibly imperfect, however, all the same, I'd personally avoid it

No, I hadn't fixed the Acoustats

Everything was going so well.  The sound was the best ever, and there had been no shutdown since I put the right speaker back together with both transformers connected, with the new electrolytic capacitor.

Then, after 4 hours playing, it was waay past my bedtime, but I decided to measure the temperature of the amp.

Things were not well after all.  The right side of the Krell, powering the newly "restored" Acoustat, had heated up to almost as high a temperature I've ever measured on either side of the Krell, 194F.  And the right side has reliably run 10-15F cooler, not hotter, for reasons I'd always assumed were due to airflow caused by the air inertia of circulation of my always circulating HVAC system.

194F was what I measured on the right side, and a mere 178F on the left side (typically for 4 hours or more).  This was not at all normal for my system.  (It appears, btw, that the Krell FPB 300 runs plateau bias at the highest plateau demanded, UNTIL the temperature reaches 180F.  Then it shuts off all but the two highest bias levels, but it can still go higher if there's enough excess current draw, at the lower bias levels.  I observed this behavior a few years ago when I always had the Krell running through a AC wattmeter.  I believe the ultimate thermal shutdown of the entire amp occurs at 200F.  All these numbers are way beyond most power amplifiers, but the Krell can do these temperature levels because it was designed to, and the transistors are special high temperarture parts.  Inverters typically run at temperatures like these and higher.  The inside of the amp is cooler than the heat sinks too, by some clever design.  When the right channel was ultimately "broken" in November 2016, before I shipped it in for "Capacitor Service", I recall seeing temperatures on the left side approaching 200F just before shut down.  But that was the ONLY time I had ever seen that, or gotten thermal shut downs.  Even when the amp was "normally malfunctioning" as it had been, actually, since the day I got it, I didn't see temperatures above 189F on the left (while the right was 150F or so).  Since the 3/2017 capacitor service, all day L/R temperatures of 169F/159F were typical, an enormous improvement.

Because it was easy to do, I disconnected the LF transformer first.  For some reason I also temporarily disconnected AC power on the Krell (I still wonder if this forces it to "recalibrate" on the next startup).  When it started up, power went up to 1200W fairly quickly.  I was immediately convinced this proved the HF transformer was the one causing the excess current draw.  However, after awhile, the power settled down around 650W, and all seemed normal.  The next day the startup was normal, simply going from  the standby of 70W up to about 650W and holding there.  And both days the right side temperature did not exceed 160F, even after the apparently excess current draw during the first day.

I decided I'd better check the LF transformer after all.  I hate to do this because this test requires cutting a wire, and then resoldering it.  Or I supposed I could "resolder" the wire, but fortunately there's enough slack I can just cut the wire to start, which greatly simplifies the start of testing.

And then, it was clear by the first day of operation that the LF transfomer on the right side was causing the excess current draw, probably at some peak voltage, forcing a higher bias plateau in the Krell.  And sure enough, power use was more like 830W, and the right side of the amp would begin exceeding the right side in less than an hour, reaching as high as 170F in 60 minutes, and 194F (just as I had seen before) in less than 3 hours, when the right side was a typical (for it) 168F.

To be sure nothing has happened to the amplifier, I hooked the right channel of the amp to the left speaker (I will shortly be moving the right interface away to the test/rebuild bench anyway).  And I'm seeing normal operation--for the right channel of the amp--with about 145F at one hour, not 170F.

So...I've got a transformer replacement job to do.  When I complete this soon, which I must, it will be the biggest job I've actually done (and not just contemplated...like my idea for an external Acoustat crossover box for the right channel, where I'd put the Solen capacitor and fuse) since about 1982, when I built an Audio Research inspired phono preamp from scratch.  Well, almost scratch, as I gutted a Scott preamp, which I should not have done, in retrospect.  However, my phono preamp was stellar, easily besting an Audible Illusions, for example.  I never compared to a real Audio Research, but I had incorporated many of the ideas we had gleaned from that and other preamps at Audio Directions.

Actually, building the Preamp was the biggest audio project I've ever done by far and way, and it's almost inconceivable to me now that I did it.  (I no longer have the power supply for that preamp, which was crucial as it provided 6 seperate 400V regulated B+ supplies, plus 2 regulated 12vdc supplies for the filaments.

To be fair, the Acoustat transformer replacement is a much smaller job that building a preamp from scratch.  I should be able to finish this by the end of the month, if not the end of this 3 day weekend.

Have I finally fixed the Acoustats???

It's now been 6 days since I replaced the electrolytic capacitor in my right Acoustat with a new 50uF 100V capacitor from Madisound.  There have been no failures so far.  Alas, it appears both my subwoofers are now defunct, so I've been playing the Acoustats full range, which I concede has virtues, including being a more complete test of the speakers...  Sound has been marvelous, though I also realize now some of the excess bass reverberation in DJ voices in NOT because of my subwoofers.

Months of previous testing suggested the problem with that speaker was not in the Low Frequency circuit, which has the LF transformer, where failures are often said to "usually" occur.

I had been thinking, therefore, for quite some time that the failure might be in the HF transformer.  But Andy Szabo, the AcoustatAnswerMan says that is so rare he has never seen a failed HF transformer.

Then I had also been thinking the failure might be in the resistor.  But unlike the resistor in the left speaker, this one does not have a hairline crack in the visible wire contact.  So I had not suspected a fault in the right resistor.  (The "cracked" left resistor seems to be fine according to all tests done so far.)

But it did seem to me, even when my first interface failed in 2010, that a likely failure point might be the electrolytic capacitor.

Although it seems at first blush a failure in the 47uF capacitor wouldn't cause the amplifier to detect a "short" (in fact, the load would be considerably reduced) I reasoned earlier this year that any such intermittant failure might cause a kind of "flyback" phenomenon when connection is restored that temporarily stored energy is then reflected back at the amplifier in an instantaneous transient, causing the amplifier to "see" what looks like a short for the briefest of durations, such as 10 msec.  The Krell might be able to decide to shutdown in less time than that.  This was my reasoning why the cracked resistor might be causing a problem (although, it could cause OTHER problems if the crack was in the portion going to ground).  And the same reasoning would apply to the electrolytic capacitor.

That is now what I now think was happening with the right speaker.  A momentary failure in the capacitor was being seen as a short by the amplifier.

The left speaker was fixed two months ago by connecting the LF transformer to the correct winding for 1+1 operation, suggesting there might be a fault in 2+2 tap I had connected to.   I didn't really think that changing this would fix the problem, but so far it seems to have done so.

So, six months wondering in the deepest audio desert may have come to an end.  Now, among other things, I need to get the subwoofer plate amplifiers replaced.  A second time in left subwoofer...however that might be fixed by replacing the fuse which I haven't tried yet.

Now, many previous times, I thought I had fixed the Acoustats, not so much by fixing the acoustats but by changing something, like putting the subwoofers on a different circuit, only to find later that each such change didn't prevent another amplifier shutdown a week later.  I'm only 6 days into testing the speakers with the present level of changes, so there's still some possibility of a failure.  But because this has been a holiday vacation week for me, I've done enough testing now, the cure seems pretty well established.

Emotiva Preamp better than Breeze "6010D"

[See final update.  Distortion of the Breeze Preamp is comparable to the Emotiva so long as the Balanced Input on the Breeze is not used.  Using only the RCA inputs, the distortion is 0.0006% or less through the RCA outputs, through the balanced outputs setting a new record at 0.0003% or less.  The DC offset can be nearly zeroed in all conditions at the same gain level, but changes with different gain levels, though I've adjusted it to 3mV or less provided the volume control is not advanced beyond the 3PM point, which I should never need to do.  At max, it's 14mV, which may not be OK for all amplifiers.  This is acceptable for use in gain adjustment for blind testing so long as only the RCA inputs are used and volume kept below 3PM.]

Clearing off the audio bench, again, I finally figured out the R/L mismatch issue that's been preventing me from doing computer measurements for the last few months.  One of the TRS to XLR balanced cables has gone bad.  (The old one was a Belden from Blue Jeans, my usual cable supplier.  TRS is not their usual thing.  Perhaps not XLR so much either.  Their specialty is premium 75 ohm Canare RCA's on Belden cable, and I've never had one like that in hundreds fail.  I took a break right here to order a replacement set of Mogami Gold TRS to XLR cables from Sweetwater.  Will be going that kind of route for TRS cables from now on, and perhaps XLR's as well.  The Mogami XLR's look better strain relief'd than Blue Jeans does.  I've now decided I like Gotham Audio the best for AES XLR, but they don't make TRS to XLR cables.  Mogami has a different "CORE" line of "professional use" cables but they are not the optimal shielded cable, which I need for my instrumentation bench.  Mogami Gold ought to be durable enough, I'm not doing daily gigs in beer joints.)

So, to finally measure the Breeze Preamp using RMAA, I borrowed a TRS to XLR from my keyboard.

First I measured my Emotive XSP-1.  It is impeccable.  Excellent construction with excellent design, including full balanced operation made possible by discrete resistor based attenuation with 0.5dB steps.  This is the kind of technology you used to have to buy $6k (now $20k) Krell preamps for.

Actually I first measured the residual.  It's 0.0003% distortion.

When the Emotive is hooked in, the distortion rises to a mere 0.0005% distortion.  This suggests that the Emotiva actually has about the same THD as my residual, around 0.0003%.  That's what decent (not stellar) engineering can do nowadays.

Then I hooked in the Breeze in place of the Emotiva.  Distortion rises to 0.0068%.  Despite being based on the fabled German MBL 6010, which was measured as having less than 0.002% distortion by Stereophile at 2V, and using even better IC's, the even more fabled AD797's.

I was hoping to use the Breeze as my measurement preamp, but clearly it's not good enough for that, while the Emotiva is.  I'm not thrilled about using it in my living room system either.  The plan was to use it to equalize the gain for ABX purposes.  In fact, the original plan was to use it to convert the unbalanced output of the Denon DVD 5000 (which I was then using as my midrange DAC) to balanced for running the Krell FPB 300 on a different AC circuit, for which balanced operation is a really good idea.

Well, I ultimately decided the Denon DVD 5000, which measured 0.0038% distortion, though having much better S/N than the DVD 9000 for some not yet understood reason, wasn't clean enough.  It doesn't have the ultimate "good" IC's opamps that are available now, it used an earlier generation which aren't very well regarded anymore.  So I've switched to using all Emotive Stealth DC-1's, which have AES inputs, XLR outputs, and measured as good as good or better than I am able to measure, 0.0003% distortion.  Emotiva uses the good IC's, and uses them in textbook quality engineering.

So, if a DAC with 0.0038% distortion isn't clean enough, a preamp with 0.0068% distortion certainly isn't either.

I may just have to get Emotiva preamps for all gain stages.  My main quibble with the XSP-1 Mk 2 is it's rather large and has way more features than I need.  The Breeze seemed to be exactly what I needed but it doesn't seem good enough.  It has XLR and RCA inputs, selectable, and XLR and RCA outputs which are always running independently, and a volume control (based on a single high quality potentiometer, because the level adjustment is done in an unbalanced part of the circuit, with balanced to unbalanced converters preceding it, and unbalanced to balanced converters following it--but this is what you have to do if you are not using precision discrete attenuation like the Emotiva).

Also, I measured fairly high DC offsets from the Breeze.  Around 30mV DC.  Now, the Breeze actually looks fairly textbook also.  I wonder why the distortion measures as high as it does.  One particular point, the Breeze has pretty high 2nd harmonic, filling in the second harmonic that is virtually nil on my analyzer.

I didn't want to tinker with the DC offset trimpots which are clearly right there in the Breeze without measuring the distortion first, fearing that fixing the offset would make the distortion worse.

It's quite possible, I think now, that fixing the DC offset will also, at least to some degree, fix the THD as well, which looks like it may be generated by asymmetry at the summing point of the Breeze.

Update:

I have adjusted the offset(s) in the breeze.  It did not lower the distortion.

The two pots near the input seem to control the unbalanced input.  I adjusted those to reduce unbalanced-to-unbalanced offset to 0.01 mV or less.

The two pots near the volume control seem to control the balanced output.  I adjusted those to balanced-to-balanced offset of 0.01mV or less.  I went back and forth on these adjustments, after many preliminary attempts in which adjustments seemed to make no difference.

Then unbalanced-to-balanced has about 2mV offset, and balanced-to-unbalanced has about 4mV offset, not perfect, but not horrible.

None of this improved the distortion appreciably from 0.0065%.  At one intermediate point, I measured 0.0061%.  Adjusting the balanced output offsets seemed to make no visible difference on the distortion spectrum from the balanced outputs as I was watching it live.  All my distortion measurements are balanced-to-balanced.

Maybe I need 3 Emotiva preamps, or more.  I need at least one for the Living Room, and one for the Laboratory.

I wish I could buy something smaller and simpler, like the Breeze, with performance as good as the Emotiva.

Another Update:

The Breeze is absolutely fine for single ended inputs, except for an unfortunate tendency to change the offset as the volume control is turned.

The RCA output distortion is essentially at my residual at unity gain.  Actually the number is 0.0006%, run through the "processor loop" of the Emotiva (to convert the balanced I/O of my Juli@ card to unbalanced) measured at around 2-3V output--and I can't read the exact number because RMAA is pulsing).  At 4-6V output, distortion rises to 0.04%.  And 1-1.5V, distortion is barely changed from unity.  I would only need 3V or less with most amplifiers.

Adjusting the pots near the input seems to affect the RCA input balance.  Adjusting the pots near the volume control seems to mainly affect the balanced input balance.  Adjusting both, one can make both zero simultanously at unity gain, but both rising to 14mV at volume control maximum, and falling to just slightly below zero.  Below 3PM on the volume control, RCA output offset is below 3mV for both inputs.  For my ABX gain adjustment purposes, I should not need higher than 3PM on the volume control, which is more than a little over unity gain.

Neither set of controls seems to have any affect on distortion through the balanced input.

The balanced input measured even lower distortion.  Taking the RCA output through the Emotiva processor loop, but then taking the balanced output of the Breeze, yields distortion measuring 0.0003%, lower than I see through the Emotiva balanced outputs, in fact, 0.0003% is precisely my residual...the Breeze is not raising the distortion number...meaning it can't be known but is probably significantly lower than 0.0003%.

The source of unnecessary distortion seems to be: high outputs (above 4V for single ended, unable to do this test for balanced) and use of the balanced input.  The balanced output is fine.

It looks like I can use this for ABX gain adjustment after all, just always use the RCA inputs.

The Balanced Inputs aren't horrible, at 0.0065% THD, but it represents unnecessary distortion by today's easily achievable circuits (including LM 4562 op amps and AD 797 and OPA 211), which can achieve 0.0003% or much better.

The fact that the distortion occurs only for balanced input, and has majority 2nd harmonic, indicates that the issue is with the balancing of the input differential amplifier for the XLR inputs.  One would need to fine-tune the gain of one of the two IC's used, and there are no pots for that.  If MBL didn't use trimmers, they might have used more tightly matched resistors setting the input gains.