Required Total Bandwidth of Audio System

The correct answer is, at least two times the highest you can plainly hear.

Why two times?  Actually, the highest you can plainly hear is dependent on two things, the power levels involved (and we don't want to go there) and...the bandwidth of the test system.

As the bandwidth of the test system improves, so does your apparent bandwidth.

What's important is not that your bandwidth is "met," but that any further improvement in your apparent bandwidth is not achieved by turther improvements in the audio system.  Thats...as good is it needs to be.

This happens around twice the desired bandwidth, because when you have two separate sytems...the audio system and your hearing system...to achieve any particular total bandwidth over both those entities, requires each half to have twice that bandwidth.

Although my hearing goes to about 16kHz, an appropriate audio system total bandwidth is 40kHz to allow for the nominal human limit of 20kHz.

And since audio amplifiers are only one part of the system which must achieve 40khz bandwidth, it's appropriate for audio amplifiers to have at least 100khz bandwidth or maybe higher depending on rest of system.

Marantz AV8805 surround processor revealed

I've decided it's barely worth considering a surround sound processor which does not have the ability to control the target curve for each speaker.  In the Marantz line, the Audyssey App is supported by the AV7704, AV7705, and AV8805 units, and not earlier models such as the AV8802a.

My favorite audio tester, Amir at Audio Science Review, measured a Marantz AV8805 and did not like what he found in this $4500 unit:

At 4 Volts RMS at the XLR main outputs:

1) 91.2dB SINAD at 1kHz (0.0027% THD)
2) Shocker: THD rises dramatically at 5kHz and above, above 1% just above 10kHz

Also he also thought the jitter spectrum looked relatively busy compared to his favorite DAC, but not so much as to be an audible issue.

Now, in my estimation, Amir did something a bit unfair to Marantz, but often addressed by commenters.  Amir didn't bother to do any measurements at the official AV8805 Rated Output Level, which is 2.4 volts.  He said that was simply a crutch that Marantz used to get better measurements.  4V is the defacto industry standard level for XLR for consumers, and Marantz should be expected to perform well at that level like any top product.

Furthermore, in long discussion which compared the specs and measurements of other Marantz processors done by others, it appears that the 91.2dB SINAD is consistent with measurements done by others at different levels.

For example, Secrets of Home Theater found the AV8802(a) to have less distortion at 2.0V, and more distortion at 5V, which was still below the apparent maximum output.  Amir reiterated many times that he turned the level "down" to 4 volts, though he never indicated what he turned it down from.

This other review of the AV8805 didn't do a single measurement, mainly talking about features and giving subjective reactions to various discs.

After much discussion, it appears that the Marantz AV8802, AV8802A, and AV8805 have the same output circuit and cound be considered comparable.  The specs are identical, and in fact the specs are also identical for the AV7705 and AV7704.  Therefore, conveniently, a review of the audio raw performance of one of these units is equivalent to a review of another.  Therefore, despite many people complaining about Amir only doing one measurement (not true) and calling it a day (apparently true, he shipped the unit out to make room to walk around in his laboratory), we can fill in the gaps by looking at measurements done by others.

Though it's not great, I don't think the 0.0027% distortion is that big a deal either.  Sure, my Emotiva Stealth DC-1 (another product Amir does not like) measured 0.0003% THD in my testing, which is about 10x (20dB) better, and Amir's favorite dacs (I now use the cheapest one he recommended for measurement purposes) do even better, but few recordings are going to be made that well and no speakers would even come close.  Many highly regarded audio products have no better than 0.1% THD, which is often suggested as a limit to the audibility of nearly all distortion harmonics.

I think, and what unfortunately to this moment has not gotten enough discussion is the #2 Shocker.

Amir did some further investigation of this.  First, he ran a 10kHz signal and found a ginormous distortion peak approaching 1% at 34kHz.  Since this is not an even multiple of 10khz, this got him thinking it might be an aliasing artifact.  So, he runs a white noise signal and then discovers only 3dB rolloff at 22.05kHz, where (according to Shannon et all) there should be infinite attenuation from a reconstruction filter.  Acceptible levels of attenuation were only ultimately achieved near 40kHz.

Amir looked in vain for any discussion of digital filtering in the manual, or ways to select different digital filtering options.  Failing to find such, he pronounced the AV8805 flat out broken.

Such slow reconstruction filters are not at all unusual on audiophile gear nowadays.  There's quite a strong following for no filters at all (NOS).

Now the only piece that leaves me wondering about this, even after reading every post in a 24 page discussion (which was quite interesting in itself) is whether the high levels of HF distortion would essentially disappear if Amir had stuck to the Rated Output level of 2.4 volts.

Amir simply believes that such an output level is not worth bothering with, 4 volts is the defacto standard, he tests everything balanced at 4V, and if a unit can't do 4V well that's simply another way that it's broken.  Commenters suggest that 2V is more than adequate for many amplifiers and challenged him to suggest one amplifier where this would not be true.

Here's one aspect which wasn't discussed.  We can't simply tell the maximum input voltage an amplifier can use from its rated input, sensitivity, and/or rated power.  Most power amplifiers have considerable headroom.  So even if it only takes 1.5V input for such-and-such an amplifier to reach rated power, it may be able to handle input peaks up to nearly 3V without clipping.  This is a good reason for preamplifiers AND surround processors to have at least 4V output, as Amir suggests.

Amir does note that the level of the 10kHz distortion products including aliases does go down at lower levels.  And therefore, he admits, the audibility of this problem probably isn't that huge either.  But it represents unacceptible performance for a flagship product.  I agree.

But I'd also just like to know, if the 2.4v output level fixes things.  It seems OK in the AFAIK identical audio circuitry of the AV8802a, as tested by Secrets.  They measured 0.002% THD at 10kHz with 2 volt output with 44.1kHz sampling, and lower levels with higher sampling rates.  That suggests the Marantz would easily meet it's specified 0.008% THD from 20-20khz at the rated output level which Amir refused to test.

Digging a bit deeper still, Secrets also tested 10khz up to 5V with 24/96 data, with distortion at 0.0036%.  Unlike Amir, Secrets did lots of different tests with 10kHz and with pairs of frequencies.  Curiously, however, they did NOT report a test for 10kHz with 5V at 44.1kHz, which might speak to the high levels of distortion Amir saw at 4V output with 44.1kHz signal and attributed to aliasing.  At the higher sampling rates, distortion reduces at 2V also.

Looking at these high frequency and IM tests, the profusion of products does indeed look like it may fit an aliasing theory.  It still bothers me how it seems fine at 2 volts (and possible 2.4 volts) but then distortion products including aliasing products just take off.  It seems to me that aliasing products should be there all along and increase linearly.

It was a similar output level limitation that led to my finally ridding myself of my two Integra Research RDC-7's today.  When I bought these a few years ago I was hoping they'd make excellent DACs, which I could even use in my laboratory.  My first naive measurements at 4V showed such massive distortion I figured my first RDC-7 was broken.  It was only later while testing my second unit I discovered that distortion looked basically OK (just OK, perhaps about 0.005%) until 2V balanced or 1V unbalanced.  Then, distortion products began rising linearly from the noise floor.  At first I thought it must be clipping.  But clipping distortion products would rise much more quickly.  It almost seem like some kind of soft clipping was being applied.

All these high end surround processors are THX licensed.  This means they pass proprietary tests by THX.  We don't know exactly what these tests are, and what they required, because they are not public information.

I wonder if THX virtually demands a kind of soft clipping be introduced in the pre/pro above 1v unbalanced and 2v balanced, in order that amplifiers or speakers themselves not be the clipping elements.

Anyway, the way the shocking issue appears in all the information presented does not suggest that soft clipping alone is the source.  The products that appear in high frequency and IM tests look a lot like aliases in some cases.  However, the rapid rise when levels exceed 2.4v suggests that aliases alone are not responsible.  Then it appears the issue may be the combination of two things, the very slow reconstruction filter AND a form of soft clipping that starts just above 2.4V.

Knowing all this, I've somewhat lost interest in the AV8805, which no longer looks like a reference product and isn't worth the high price. (If you want reference, apparently the Trinnov is the unit to get, at 4 times higher price.)  However, the AV7705 may be OK for the features, and knowing that the limitations aren't likely ever to intrude on the auditory experience.

Amir correctly blames the lack of competition in the pre/pro world to the introduction of HDMI.  Apparently only 2 companies make the require HDMI chips, and they aren't interested in supporting small time manufacturers.

I've never liked the idea of squeezing my audio signal through HDMI, and now it appears (increasing distortion, jitter, etc) that avoiding this would have been good.  Given the current market, however, it's unavoidable.

Much of this impasse has also been caused by the myth of perfection.  Everyone wants "perfect" digital inputs and outputs to things, as these are considered "lossless."  One thing that's lost but barely remembered is openness.  Analog inputs are fully open, transparently, and not subject to being locked down by greedy bastards.

If instead of demanding digital I/O, we had stuck with analog I/O and just made it better, we would not be trapped by greedy chipmakers and other proprietary schemes.

I strongly believe that digital conversion is virtually transparent, especially at 24/96, so there would be no harm in sending analog audio signals from CD player to preamp, which could convert them to digital for processing, and back to analog for amplification.

This is the kind of thing I have done for over a decade now for stereo.  In some cases, curiously, I find the resampled analog to be BETTER than digital.  In not one single case have I found it to be noticeably worse.  Sometimes I just use the direct digital because it's supposed to be better, but I've never actually heard that.

But if reference high end products can get away with 91dB SINAD (and far worse at high frequencies), the -120dB added noise from resampling is just not a big deal, and if we still had analog interfaces there could be endless competition, even from people like me.  (I recently found a "open source" surround processor project.  It hasn't gone anywhere in 5 years, probably due to the HDMI issues.)

I had been thinking the Integra Research RDC-7's would give me this kind of analog processing.  For quite some time I was blocked by the fact that they use a snake multichannel input connector, and I never got around to testing the multichannel path because I didn't want to risk getting the wrong kind of snake for it.  Well, it turns out I could have seen the issue from looking at the schematic, which reveals that the multichannel inputs are not digitized, they are simply passed directly to the output with volume adjustments.  This is the same limitation as on my 2005 Yamaha HTR-5790 receiver.  So the glorious Integra Research would not move me one inch forwards, and possibly a few backwards.

I understand that there were some earlier processors by Yamaha and others that did digitize multichannel inputs, at least up to 5.

But to get with the program, and play with all the new surround toys, I'm still tempted to get the AV7705 at a decent price.

TIme Alignment Answer from Linkwitz

Linkwitz explains why I had to delay the signal to the subwoofers to get time alignment with the panels.  At the end of this incredibly informative page on Crossovers.

The woofer's natural highpass behavior causes a phase lead which is probably far from zero at the crossover point and therefore affects the addition of the woofer and midrange outputs.  This can be corrected by placing a first order allpass in the midrange channel which simulates the highpass phase shift of the woofer.

I don't think I like his solution better than mine (I delay the woofer channel instead of phase shifting the midrange channel) but his analysis is crystal clear despite being totally counter-intuitive.

One normally doesn't think of woofers as being a high-pass device, especially subwoofers.  They are supposed to go lower than we can hear after all.  But lower than we can hear, is still infinitely higher than DC.  At some point above DC the woofer HAS TO cut off (except for fan-type subwoofers, and they require massive amounts of power and cannot be made cheaply).

AND, here's the rub.  The lower the low cutoff, the greater the phase lead!  So, woofers will naturally lead higher frequency drivers, since the former have lower low cutoff.

My solution of delaying the subwoofer channel can work perfectly at a single frequency, but not other frequencies because the problem may not be that the woofer is simply delayed...it has a highpass phase lead, which means the delay varies with frequency.  In that sense, Linkwitz' solution is better, it brings the woofer and midrange into alignment at all frequencies.  Mine might be  better because I'm not introducing more allpass behavior in the midrange.

With a reasonably steep crossover, I don't believe the mismatch in allpass behavior is such a problem generally (at least in digitally crossed over systems like mine that can easily apply delay instead), and especially in my system where the "midrange" panels can go nearly as low as the subwoofers.  However, it's worth keeping in mind.  It's an incredibly important insight.  It seems to suggest that my friend who was often criticizing speakers with recessed woofers especially as not being "time aligned" may have been all wrong.  Also explains why one of the least costly "transient correct" speakers of all time, the Spica's, have deeply recessed woofer.  I vaguely recall, back in the day, that a defender of Spica or some such mentioned the woofer lead issue, but I never understood it before at all.  I thought it was the crossover that was causing the woofer to "lead," and in that case the distances still needed to be equal so the crossover would work correctly.  But the problem is, it's not the crossover that causes the woofer to lead, it's the difference between the woofer and the midrange drivers in themselves, their differing low frequency cutoffs.  Ideally both should have identical low frequency delays, but they don't.

However it still troubles me that this "lead" might not be applicable entirely to transients.  That's where my intuitions regarding how filters work seriously breaks down.

It seems to me that information cannot travel through a system in negative time, and that information does not travel through woofers hugely faster than midranges.  Any kind of "phase lead" must represent some kind of information loss, like the front part getting lopped off.  But in saying that, I still can't see how this would apply to the woofer/midrange situation.

Perhaps it's because I'm not thinking of the information loss to the midrange--it is losing the lowest frequency information (because of it's own highpass behavior at least) relative to the woofer.  The lopping off of this low frequency information is possibly causing it to be delayed MORE.  The woofer is being delayed LESS because it is not losing as much low frequency information.

I don't really know how to understand this yet, but it does seem that to work properly in a system with a crossover, the woofer must be delayed even if this means it doesn't start contributing information to the listening position until a later time than the midrange.  And that is what one sees in a decently engineered but nevertheless allpass speaker impulse...in which the high response starts and rolls into the lower responses.  But how would it work with transient perfection?  Perhaps the greater phase correction is needed for the woofer itself than the crossovers.

It seems that electrical engineering has many different abstractions, and issues arise with mixing the abstractions the wrong ways.

Although we are talking about the low pass function of a woofer (Kellogg-Rice dynamic woofer), but actually, as far as the abstract high pass function nature of it, we could just as well be talking about a capacitor in a series-capacitor followed by shunt resistor circuit (following some AC generator), which is another high pass function if we idealize it slightly.

In that circuit, current must lead voltage in the capacitor.  For abstraction purposes, we assume the resistor has no parasitic capacitances and inductances, and as such it samples the voltage at its terminal instantaneously.  (Somehow when people talk about the capacitor in the voltage leading thing, they rarely talk about the load.)

But what does this mean for information traveling from the generator to the load?  I'm thinking out loud here.

At the moment the generator applies a voltage, that voltage appears across the capacitor and the load. But the capacitor is having none of it.  Instead, it all appears across the load.  As current flows through the load, it flows into the capacitor and therefore a voltage appears across the capacitor AFTER it has appeared across the load.  The capacitor is soaking up the signal AFTER it has started. So in the long run, DC or a Step would be quenched all into the capacitor and none into the load.

In what sense, then, is the signal "leading" into the resistor?  In the sense that the maximum rate of rise for the load is at 0 degrees, whereas the maximum rise in the AC signal itself is at 45 degrees.  As soon as the input signal has started, the rate of change in the load begins falling, as would happen at 45 degrees.

The voltage at the load...which is the output of the low pass filter, is therefore 45 degrees ahead of the input signal.

However, none of this means that start time of signal is changed.  It simply starts at 45 degrees rather than 0.  One may be tempted to draw sine waves like arches but they aren't.  The first 45 degrees is a gradually inflecting upwards, starting like nothing.  OTOH, the output of a high pass filter starts at maximum upwards movement.

Now, for a larger capacitor this leading time is longer.  So the high pass filter retains the maximum upwards characteristic longer.  I still find it baffling that this means the larger capacitor "leads" more. It seems to me more that it holds on to the lead longer.

CIVIL: Capacitor: I leads V, and by 90 degrees, Inductor: Lags.

The Capacitor is maximally charging at 0 degrees into the input signal, which means the current is at maximum at that point, which means the voltage across the load leads the AC input signal by 90 degrees, because the voltage across a resistive load has the same phase as the current flowing through it, and because the current flowing through the capacitor must be the same as the current flowing through the load.

This is not to say it's starting any earlier in response to any signal, but that it's starting at the 90 degree point instead of zero for AC waveforms.

I'm still not clear why the lower high pass filter would lead more.  However, if the lower high pass filter leads by 90 degrees of it's cutoff frequency, that would be "more leading" time.  I just don't see how it's cutoff frequency comes into play here.  It should be just 90 degrees.  The capacitor is maximally charging at 0 degrees, i.e. when the signal starts.

And even then, it's still even more baffling how the lower high pass filter leads more then the higher high pass.

The "more leading time" seems to be the correct reason and answer.  The leading DOES have to do with the cutoff of the RC circuit.  It leads by 90 degrees of that cutoff frequency, even if that frequency is no where in the input, that same effective leading is applied to everything along with the associated cutoff.  It's weird but I think I'm beginning to get it.

One again this has little or nothing to do with how fast the input starts, but the phase angles are going to contaminate the step response unless the leading is accounted for...this tends to mean delaying the "more leading" woofer...and it does seem desirable that it should be delayed, and also so that the impulse response "rolls into " the bass, rather than the bass in any way contaminating the leading edge of the impulse.  In fact that contamination is what happens is the woofer is not delayed, as the speaker with the lower highpass cutoff is going to have pre-mature phase angle from a dirac impulse--actually beginning to tend downward as the midrange is still going up, causing cancellation of the leading edge.

The exception would be designing each range of speaker for identical highpass response.  Some very quirky designs do that, not that I'm endorsing them in total, but it's easy to understand the desirability of it now.

The June 2019 TIme Alignment

Finally, as promised (but similar to a long line of unmet promises), here is the report on the living room system time alignment and adjustment near the beginning of June 2019 when I finally turned REW onto the living room system in order to do a measured time alignment which hadn't systematically been done since about 2014...something like forever for this ever changing system.

In May I obtained a Focusrite 2i4 interface so I could do full loopback with REW.  I had been unable to get my historic Emu 0404 to work on Windows 10, so I just decided to get the cheap interface everyone uses now (and which apparently still works on Windows 10, with the status of other interfaces such as the Lynx Hilo uncertain).  I already had a suitable calibrated mike I bought specifically for REW a few years ago.  You cannot use a USB mike during loopback because you must use the same audio interface for both input and output, and a USB mike only has an input interface.

When you do full loopback, in theory, you can simply have REW compute the delay time from the loopback electrical output to the measured acoustical stimulus.  Then, in theory, you can simply dial in the required compensation into your digital processing units.

Using a loopback is also preferred for getting a more accurate computed step response.  REW uses a stimulus which is a sine wave sweep from a low frequency to a high frequency.  The actual response to that is hard to interpret for phase response by eye.  So then REW analyzes the response to this sweep stimulus using FFT, and then it can compute what the response would be either to a dirac type impuse (the "Impulse" response), and a step type stimulus (the "Step" response).  Because it is the simplest possible stimulus, and arguably most informative, speaker designers and reviewers mostly evaluate the step response.  It should always be remembered, however, that REW did not actually measure this stimulus as such, it merely computed that, given the way the unit under test responded to the sweep, this is the way it would have responsed to an actual step.  The computed step response will have artifacts.  But if you were measuring an actual step, it would be highly random because of noise.  The swept sine stimulus, having more broadspectrum (across desired band) energy increases the S/N over what you could measure with an actual Step stimulus.

I had long done such compensation simply based on the physical distances of the speakers as seen from the listening position.  Roughly speaking, the supertweeters are about 1 foot back from the panels, and the subs are about 2 feet back.  Assuming a foot of distance to have about 1 ms delay (I'll use that approximation a lot in this post), I could then compensate by having NO DELAY (0 ms) dialed in to the subwoofer crossover unit, 1 ms delay dialed in to the supertweeter crossover unit, and 2ms dialed in for the panels.   In practice, I intentionally add 10ms to these numbers to facilitate more on-the-fly experiments and adjustments, since you can't further reduce a delay already at 0 ms.  So, a typical adjustment based on distances might be: 10ms delay on panels, 9ms delay on supertweeters, and 8ms delay on the subs.  (In practice, I tried to make the numbers accurate to 0.1ms by measuring to the nearest inch.)

But this is all approximations until you actually do acoustical measurements, especially when all speakers are nothing like point sources.  Real acoustical measurement, not distance approximation, is what has to be done to do the time alignment correctly.  But it's so difficult, and fraught with problems, I haven't actually done it very much.  In principle, I should do a new time alignment after any adjustment of any kind.

It was failure to do actual acoustical measurements (which many subjectophiles and speaker amateurs HAVE NEVER DONE) that led to a huge mistake which I continued for several years!  I started using different DACs on each way, without realizing that the DACs themselves have delay differences, and these differences change at each different sampling rate.  Once I discovered this issue, I quickly gave up on dialing in new compensation when the sampling rate changed.  Too much work.  But that was when I decided to add 10ms extra delay to each path.  However, it turned out that some DACs could add far more than 10ms delay for some sampling rates.  I ultimately decided the best (?) way to handle that variance is to eliminate it by using identical DACs on each way.  I went with inexpensive but decent measuring and pure sounding Emotive Stealth DC-1's (now discontinued) because they had all the needed features: AES digital input, balanced and unbalanced outputs, exact level adjustment.  Audio Science Review puts this DAC down (somewhat unfairly IMO) however if I had to make the choice now, I'd go with the similarly inexpensive DAC they recommend, which has even lower distortion and noise.

This time around, my plan on simply dialing in the REW computed delay for each way was swiftly defeated by an apparent fault with the way REW computes delay for my subwoofers.  I have written a long post about this previously.  I first noticed that the delay times being computed by REW for the subwoofers were about 3ms longer than I had expected.  Then, looking at the computed step response, it appeared that REW was computing the zero point as the beginning of the second hump in the impulse response.  This appears to be an error in REW--while the second hump is slightly larger than the first, it is nearly the same height, and does not appear to be a digital or noise artifact (though there are lots of those, so it's not certain either).  It may have something to do with peculiarities in my system--the rear wall reflection is my current best theory.  This is an example of the computed step response for the Right subwoofer.  The red line shows where REW believes the actual step response is beginning:

It shocked me to see that I got this same result even with No Crossover, which I finally realized I could safely do simply by restricting the range of the sweep to below 300 Hz.  In tests a few years back I decided it might not be safe to send a full range signal--especially a dirac--to the subwoofer with no internal or external crossover being applied.  So I feared sending audio to the sub without crossover.  But this is safe when the input stimulus itself is restricted.  Coincidentally REW defaults to a range below 250 Hz, making it safe to run on most subs without crossover, until further adjustments by the user.  Here is the computed step response with No Crossover:

So, the smaller leading hump which REW's delay calculation ignores has nothing to do with the use of a crossover, which had been the leading theory until I took that measurement (and it's still hard to get out of my mind, because I remembered either lowpass or highpass was supposed to LEAD...but actually it's the highpass that's supposed to LEAD).

Along with the double hump is the extremely curious fact that the beginning of the first hump actually has less delay than the output of the panels, even though the subs are further back.  The beginning of the second hump, where REW seems to think the step response begins, has about 3ms more delay than would be guess from the physical location of the speakers.

Now, REW seemed to give me two choices: go with the officially determined "delay time" which put the sub about 6 feet back from the panels instead of the approximate 3 foot distance, or go with the time of the beginning of the first hump, which put the subs ahead of the panels by about a foot (???)

I was so frustrated by this, I decided to look at something else: the notch in the panel response around 225 Hz which I had long assumed to result from rear reflection/cancellation.  I discovered that moving the speakers further out from the wall could change this a lot, seeming to almost eliminate the problem, and I decided to optimize THIS before doing the actual time alignment.  As I changing the Acoustat position, and looking only at the frequency response btw,  I simply left the delay compensation as I had previously set it by distance approximation.  Starting from the "original" position from a few years ago, which had the 225 Hz suckout, I was getting a step response like this:

Notice the little peak at just over the top of a large hump.  That little peak is the beginning of the Acoustat step response, wheras the hump is the first hump of the subwoofer step response.

When I finally decided to move the Acoustats exactly 6 inches further out (for a total of about 39 inches from the wall) because it removed the suckout at 225Hz in the frequency response I was looking at, the step response looked like this:

You can see the little peak of the Acoustat step response moving forwards in time, as the Acoustats are being moved closer to the listening position, just about to the highest point of the first subwoofer hump.  But even moved 6 inches forwards (which is about 5 inches closer to the listening position) the subwoofer response still seems to start first.  Notice you can also see the little wiggle where the supertweeter is starting about 1 ms later than the Acoustats.  At this point I had not yet moved the supertweeters forward to match the Acoustats, which I couldn't completely do anyway, and it appears their initial adjustment was slightly wrong also, but less than 0.5 ms as you can see in the earlier graph where the supertweeter start is a bit more subtle.

And now, for something completely different, here is what the Acoustat step response looks like just by itself (with the low and high pass crossovers I am using):

It pretty much starts with a spike upwards, as it should.  Somehow the amplitude appears much greater than it does in combination with the woofer response, partly that's due to normalization and the curious nature of step response computation.

I measured and stared at the subwoofer step response for a long time, and finally made a wild guess about where the beginning of the woofer step should have been computed by REW.  I then used the computed panel delay, and MY computed subwoofer delay, to figure out a correct delay compensation.  Sadly I was not taking notes about any of this, it was just a guess, and I tried it, and this combined Acoustat and Sub alignment is about good as anything I've done since, which have only been minor excursions away from this point anyway:

The Acoustat step now merges with the woofer step so that the intial rise is a tall spike, as it should be.  It still possibly looks like there is some preceding bass response, but that's down by at least 40dB, and is probably just background noise and artifact.  I could not make the step look any better by delaying the bass much more than the above.   I tried adding a bit more delay, and the step got worse:

OK, so whatever that was, it looks like too much subwoofer delay has pushed the initial Acousat spike down into the hump in some bass noise valley that precedes the actual bass response, which you can see kicking in a few ms after the initial spike with a blast.

Comparing these pictures begins to get subjective again, I admit, and I wish I had something more solid to latch on to.  It did seem to me that I could try to maximize the length of the leading edge of the spike.  That would indicate that all drive units are contributing maximally to that leading edge.  Now, since REW always places the top of the spike at 0dB (btw, I have not yet calibrated the Focusrite interface itself) maximizing the length of the leading edge curiously means that the leading edge should "start" as low as possible.  You can see that operating in the above two graphs.  In the first graph, what I am calling the noise floor joins the leading spike at -46dB, with the top of the spike at -7dB, so I would count this as showing a leading edge of 39dB.  In the second graph, the noise floor joins the leading spike at -30dB, so the leading edge of only 23dB, clearly inferior.

What isn't so easy to interpret, however, is the the negative spike excursion below the joining point means.  I think it is mostly artifact, but also should be as low as possible indicating complete cancellation or some such.  Anyway, as the delay is adjusted up and down the distance downward changes and also a "gap" seems to open up between the preceding bass-hump noisefloor and the leading edge of the step, making even more a subjective call each time.  As I can't remember the actual adjustment for each measurement (my poor note taking on display again) I'm just going to show a bunch of them, in the order I took them, and where I stopped and why.

Trial 3 has leading edge 24dB, very poor.

Trial 4 has leading edge 30dB (but look at how low it goes also--actually to the bottom of the full scale not shown).

Trial 5 has leading edge 28dB, but deep spike again.

Trial 6 has leading edge 47dB and deep valley beforehand, best so far.

Trial 7 has leading edge 32dB, very poor.

Trial 8 has leading edge leading edge 31dB, very poor.

Stepping back in opposite direction now, Trial 9 has leading edge about 35dB.

51dB !!! This appears to be the winner, especially if you take the beginning of the initial spike at -58dB, then the initial spike is 51dB.  The supertweeter is already being added in and that might account for some of the wiggle at the start--it appears pretty well aligned too.  Alternatively, you could interpret the beginning of the spike as high as -48dB, in which case this is not the winner, but close.

With hand wave you could almost say this is almost looking like a pretty good step response, with a tall initial spike and pretty well filled in right after that.

This is clearly not as good, a step backwards, with the leading edge at no more than 40dB.

After the 11th measurement, I believe I went back to the preceding one, Alignment 10, as the best.  I can't be sure of this, because I didn't take notes, so it might have been one of the other pretty good ones, such as The First Guess.  But I'm pretty sure it was the 10th, because I could still remember the values for the 10th when I did the 11th, and the 11th was clearly a step backwards.

The curious delays in the DSP boxes to achieve the best alignment are:

Subwoofer: 6.3 ms delay
Supertweeter: 4.6 ms delay
Panels: 5.3 ms delay

So the subwoofers are delayed 1 ms MORE THAN than panels, though they are also about 2 feet further back.  That is the second issue I don't understand.

I have already ruled out the lowpass crossover as causing this.  It also probably not because of delays in the complex interface box of the Acoustats, because the delays required to align the supertweeters and the panels is about what one would expect based on distance.

This should not be confused with the first issue I don't understand, which is why REW is apparently computing the subwoofer delay time incorrectly.  This is an entirely separate issue, as I did my alignment by the graphical means shown above, and NOT by using REW's delay estimates, especially for the subwoofer.  However, there might be some connection I cannot yet fathom between the two.  Or maybe not.

Now some might find it outlandish to write so much and not say a word about how it sounds.  Of course it sounds wonderful!  My "electrostatic" bass is now even more well integrated into the sound, without being the least bit less impactful.  Bass lines are easier to follow.  And in many other ways it sounds better overall.

But you should expect that I would feel this way, and it's probably not best to fully trust the audio judgements of the audio investigator, and probably not the measurements either.

Next month it could be different.

Weird Woofer Step Response in REW

I noticed and mentioned this strange problem over a week ago, after I had decided how to do the 3 way system time alignment on the living room system.  I do not trust the Delay times computed by the Room EQ Wizard (REW) program for the subwoofer output.  I found a different way of doing the time alignment using the summed response, but more about that in a future post.  In this post I'm just going to show the problem with the REW computed delay time, which is a very serious issue in my opinion.  I can imagine many people using the delay times to time align their system, as I almost did, and it would be wrong, at least if it's like mine.  (I don't really know what is causing these weird results, perhaps it's something that mainly applies only to me.  As I am writing this post I've completed about six different experiments to try to get REW to calculate the subwoofer delay correctly, but none have worked so far.)

Examining the (computed) step response REW is showing for the subwoofer shows the problem with REW's delay computation.

Notice how the step response begins (that's where it clearly rises above the noise) with a series of humps.  The first hump is not as high as the second, and the second one is the tallest.  FWIW, the black line, which represents the Schroeder Integral, seems to start pretty much where I would consider the bass step to have begun.  But that is not where the computed delay time would lead you.

The computed delay time, which doesn't show up in this graph for no good reason, but rather in an information box for "the measurement" in the REW program, is 12.05 ms +/- 0.042 relative to loopback.  (I am taking loopback at the line loop output of the subwoofer, so it is seeing the same signal as the subwoofer amplifier.  The crossover inside the subwoofer is turned off, so this signal has already been low-passed with 24dB/octave Linkwitz Riley.  I have tried taking full range loopback and it makes no difference.)

If you find 12.05ms on the graph, it is within the rise of the second larger hump, not the first hump.

Now the impulse shown above had a long line of predecessor measurements, which looked basically identical.  I first did loopback using the full range output of my preamp, where it's the easiest to tap off the loopback for my 3-way system.  If I take the loopback at the speaker inputs of the Acoustats, it doesn't include the bass or super tweeter signals--that didn't work at all when measuring the subwoofer by itself.

So then I got more sophisticated, and took the loopback at the subwoofer itself, which has a plate amplifier.  This plate amplifier conveniently already has a balanced "loop output" which I routed bact to my Focusrite interface (it required the attenuator being straight up to not clip) which is also handling the microphone input and stimulus output.

Well then I suspected that the double hump might represent the EQ, of which I am applying 8 additional typically very narrow band and up to -12dB notch corrections to the subwoofer room response.

Not wanting to erase my evolved EQ settings by mistake, I waited until I had time to photograph all the settings in my 3 Behringer units, THEN I saved them to memory (not wanted to do that--which might cause overwriting the current settings by mistakenly selecting "load" rather than "save" until I took the pictures first), then I turned off all the EQ's except for the two HC's which represent the linkwitz riley 24dB crossover.  Then I power cycled the Behringer DEQ (because sometimes it doesn't seem to immediately respond to changes in HC or LC filters, until power cycled) with the subwoofer turned off, then I turned the subs back on.  That's what the picture above represents: No EQ except the crossover itself.

Disabling HC eq's in the Behringer is tricky.  It doesn't want to turn HC or LC filters off when you simply turn them off with the controls.  The "cut" EQ's are "sticky" and remain turned on.  I tried power cycling and even that didn't always work either.  This led to several days of confusing measurements.

Finally I figured out how to shut the EQ's off, AND how to verify the number of HC filters you have enabled (both are important!).

The verification is done by using the REW "generator" function to play a fixed 200 Hz sinewave signal at a convenient level, not too loud.  I set it to -16dB.  The actual subwoofer should be turned off, but even if it isn't, -16dB isn't bad.

The most foolproof way to ensure there are no HC filters enabled is to use the Memory function to reload the "Initial Data" of the EQ.

After that is done, I go to the Level display, choosing in particular the bargraph which has the greatest resolution.  With no EQ active, the 200 Hz tone output level (analog or digital) should be the same dB level, within a few 0.1dB's, of the input level.

When one HC at 100 Hz is enabled, the level at 200 Hz drops about 12dB.  When 2 HC's are enabled, the level at 200 Hz drops about 24dB.

Then, the second HC can usually be disabled by turning it off, then pressing the "reset EQ" button in the PEQ second page.  If that doesn't work, power cycling might help.

When there is only one HC enabled, turning it off and pressing "reset EQ" doesn't seem to work.  In that situation, the working solution is to reload the initial data again.

Until I developed these strategies, I was beginning to fear that that two HC's couldn't be run at the same time because the result did not seem to be a steeper slope than just one.  (It turned out, I was not correctly turning off the second one, so I was comparing 2xHC with 2xHC.)  Then I feared the subwoofer itself (Ultra PB13) had serious rolloff above 100 Hz even with no HC filters (and the crossover in the subwoofer itself has always been turned off).  That resulted from not correctly getting both HC's turned off, which may require that loading the inital data step I just mentioned.

Anyway, when I finally figured out how to set the crossover correctly, and verify it, with all these variations, the subwoofer step response stayed fairly similar, and in every case the "delay time" seems incorrect in about the same way.

Here's the step response of the woofer measured with 24dB/octave crossover again, this time slightly differently but the same as in the next two graphs.  Because of how I captured this image from the "Impulse" and not the "Filtered IR" tab of REW, the Schroder Integral is not available.  The leftmost red dotted line is showing where the delay ends and the step proper begins (this info is shown in this unfiltered version, but not the "Filtered IR").  It is showing the beginning of the step at the beginning of the second hump, or 12.5ms in this measurement using loopback from the sub loop output.

With the 12dB/octave crossover, the humps look thinner somehow, but the beginning of the step is still being calculated to be at the beginning of the second hump at 12.5 ms.

With NO HC filters (no crossover!) the humps are even narrower, with the first hump turning into two humps, and now the computed delay falls at the beginning of the third hump, at 15.3ms.

In no case does REW compute the delay to begin at the beginning of the first hump, but that's where my other method seems to land for the time alignment with the beginning of the panels and the super tweeters.  The step responses for the panels and supertweeters does not have this issue at all, the computed delay begins right where the first hump rises from the noise.

Attend Live Performance, Or Not?

Somehow, I have always strongly believed that it is beneficial, to mind, heart, and soul, to attend live musical performances.  Especially, what I feel is the grandest of them all, symphony orchestras, but YMMV.

It has always seemed to me that if you listen to a live type of music, its best in many ways in an actual live performance.  (Not all kinds of music are like this.  Some music is a form of art that takes months to years to fully realize in a recorded medium--it never existed "live" anywhere.)  The live performance is the real thing that a reproduction is merely a fairly limited copy of.

In a number of ways, the true quality of live music cannot in principle or practice be reproduced, even by the best systems, and certainly such systems as you or I know, don't come close.

1) Dynamic range...the awesome dynamics of a live performance are hard to believe beforehand.  Even a bit of unnecessary audience noise does not usually obscure this...the awesome low and ethereral background noise and reverberation, contrasted with the mind blowing yet effortless peaks.

2) Lack of interference from in-room reflections and modes...the reflections you hear in a live concert ARE the awesomely low ones of that concert hall itself, which could only be superimposed on the annoying short-delay ones in your listening environment that tend to obscure the various lines of music.  The best of the best is the famous Tanglewood.  It's entirely open except the front where the orchestra is, and the top, with a higher roof than can be imagined beforehand.  The result: complete freedom from audible reverberation.  Never have I heard orchestral sections with such clarity.  But generally Symphony Halls are very much better than your living room anyway, Tanglewood is just even better than that.

3) Lack of all kinds of audio distortion (any deviation from the original): including accurate pitch (now assumed with digital media but it was not always so), lack of wow, flutter, and jitter (though I think digital jitter is vastly overrated with post 1997 equipment, some earlier equipment was obviously not good, now it's superstition), noise, harmonic distortion, intermodulation distortion, (I'm not including the various TIM, SID, etc, because I believe those are properly already part of and included by harmonic distortion and intermodulation distortion) and dynamic modulations of the above (dynamic modulation of the noise level, for example, being more than just noise itself).  People think these have all been conquered.  The best electronic equipment is probably good enough already, but the inevitable speakers are not, FAR from it, regardless of hype or salesmanship.  Nobody denies that speakers can be distinguished blind by trained listeners.  And there are the microphones too, they are far from audibly indistiguishable as well, in any honest account.

Now, some may argue that audio reproduction is not perfect, but it's fully "good enough" never to benefit from live performance, which has it's hassles and irritations as well.  But it's not clear to me, how you would know that, without having spent much time at both, and when you find people who have done both, they will always say if you love recorded live music, you cannot merely experience it reproduced, you MUST experience live music live also, as much as you can.  Such people, including me, will agree with all I've said so far, and may introduce other factors as well:

4) Learning from others.  When you attend a live concert, you somehow selected the concert.  But you didn't necessarily select all of the music (or any!).  And within he musical slections, there are also infinite options: such as to play this phrase louder or softer or faster or slower or with more vibrato...  Ultimately, when you attend a live performance, you are giving up some degree of control to someone else, who has selected the music and chosen how to play it.  So--you are experiencing someone else's artistic choices.  This is good.  This is how to experience new things, to learn and grow.  Every performance is different, but that does not mean any are wrong, just different.  (Sometimes there are wrongs too, but they are so small usually I hardly notice if at all in professional performances.)  So each new concert is full of learning, as I have said, for the mind, heart, and soul.  Even if it's a work of music you've heard before in it's most definitive performance (of which, there are usually many too).  This is true with recorded music also, but you are less open to new things with a piece of recorded music you've heard before.  Even within the range of all the recorded versions of something, each new live performance is a new thing, from which new aspects and angles can be learned.  And when this is being driven by someone elses choices--it brings greater connection to the world (and we're just talking about the learning aspects of this so far) at the present moment.  (Understandings of and intepretations even of music written long ago are constantly evolving.  Your favorite iconic recording from 1958 can only reflect the understanding o that time.)

5) Gathering with others, others who share your love of music and live performance.  This is not promoted in the USA as well as it was in UK, in my very limited experience there.  When any performance ends at The Albert Hall, the bar there opens up, and people hang around for and hour or two afterwards, possibly meeting new similarly minded people.  That is how it SHOULD be, IMO.  In the USA, when the performance ends, the venue closes, and there are no additional chances to gab with others previously unknown.  However, there is the general issue in the USA that the lack of public transportation means that people should be fit to drive.  There may be generally smaller groups in the USA as well: more singles.  Anyway, the gathering thing even such as it exists is still important, I think any concert fan will tell you.

6)  Supporting the art.  Great recorded performances of live music aren't created in a vacuum, but in an ecosystem of development which is more supported financially and otherwise by live performance fees than recorded royalties, and the dispersion of this ecosystem of development, keeping it aligned with the heart and soul of every community--absolutely demands local live performances.  In effect, those who listen to recorded music alone are "free riders" taking advantage of the musicians and the development of music itself that live performance actually makes possible.  Furthermore,  even more "concertgoers" are essential in that development, by responding differentially to what resonates with them.

Now, some people aren't impressed by such things as 4-6.  They express  "I already know/have what I want" and "What's in it for me" attitudes.  These are short sighted and selfish,  and some people may be shortsighted and selfish in one or more ways.  There is nothing positive to be said about short sightedness or selfishness, however, and it would be better to help dissuade people from such mindsets, with better music education for example, for starters.  I myself am writing this post to help understand these issues and persuade as well.

A critical thing for me is (4) and things related to it, and a broader theme I've been hoping to develop:

There is isn't really only one Absolute Sound.

I mean this in a variety of ways.  Firstly, and with greatest import, we never experience anything the same way twice!  Each experience is colored by the memories and thinking that goes with it.  If we listen to something a second time, our memories and thinking have already been changed by hearing it the previous time, as well as all other thoughts and experiences in the interval.

Orchestra Maestro's often exploit this fact with a trick, they play every totally new work of music (which the audience will necessarily never have heard before, because it has never been played before) twice.  Sometimes without warning.  It is always said, and I have always experienced it, as not just a different piece of music, but totally different in a large number of ways.  Perhaps a different genre, for example.  It is then a shocking surprise that it was actually the same work of music, played the same, as much as humanly possible.

Now each new performance, or even audition, will highlight different things.  This is all the more true if there are deliberate or unavoidable changes.  This change in the highlighting of different things is not a disadvantage, it is a path to greater learning, personal resonance, and attachment.

It is claimed by some that music is art, but reproduction is just engineering.  This is wrong in many ways, but principally in that it reflect this obsession with the Absolute Sound, as it was laid down, in one way, at one time.

But the very essence of the true absolute sound, is that it is NEVER exactly the same.  So in freezing a slice of the past, and not permitting any deviation, we are also destroying its very essence.

Thus I have already argued that at least in this one way, analog reproduction, through turntable or tape, is more like live sound precisely in the fact that it is never twice the same.  Everytime it is slightly different, highlighting things slightly differently, and helping to enhance a broader understanding of the music itself, even if recorded at one long lost slice in time long ago.

Now all this being said, I do appreciate the effortless noise and distortion free character of digital recorded sound anyway.  I'm not saying we need turntables (though some people do) to stir things up.  But what I'm trying to open people's mind to, is that there is an advantage of listening to the same thing (not that other things might often be better) in different ways.  Not always on closest-to-perfection system A, but pleasant and surrounding system B, for example.  Neither one is an absolute, but they complement each other by shining light through different facets.  And certainly neither is perfect either.

Ultimately, the absolute, is only in one's mind, and it is the sum of all the experiences of something, different or similar as they may be.

I also feel strongly that as the ultimate executive designer of all my audio systems, I am an artist also, extracting from cans and making virtual performances, each one different.

I was going to say a lot about how perception itself is not a passive thing, music does not strain through us like as a feeder fish, but a constructive thing, each time we construct the sound and the meaning and from only a small part of the available information, and never exactly the same parts.  It is this ultimate construction, in our minds and hearts, that is the ultimate art of music, and it includes every being in that chain, including ourselves.

Now, certainly, in a live performance, you have a whole variety of tools you can use.  Every slight change in angle of your head, for example, brings a different set of information of which a stereo recording can only have one version.  You are choosing how to sample from a much larger sphere of information.  And your eyes and body are part of the whole sensation as well.

So the difference between recorded and live is as that between swimming in a fishbowl and swimming in an ocean, as far as the variety of direct information that is available.  A reproducing system can, at best, only produce the perfect fishbowl, and the truth is, nothing even comes close to that.  Audio reproduction is fundamentally an illusion, not true recreation of an infinitely complex soundfield.

Dolby Pro Logic IIx vs DTS Neo 6

I've been playing the various surround formats in my kitchen system and comparing them, as applied to two channel sources, primarily FM radio, classical and jazz stations.  I have an outdoor antenna, an ultimate state of the art Kenwood L-1000T FM tuner, and a Yamaha HTR-5790 receiver from 2005.

Contrary to other self-appointed experts like Stacy Spears, so far I find DTS Neo 6 to be the most pleasant sounding of the 30 or so available options.  All the various Dolby options like PL IIx Music, add some kind of hashy grain to the front stereo channels if not the rear channels as well.  I do not use a center channel speaker, I believe that is unsuited for music and refer to people like Siegfried Linkwitz to back me up.  I set in the proper listening spot for stereo, and a center channel speaker would block my video monitor too.

I suspect this is because Dolby surround processes use various complex "steering" and "feedback" algorithms to force sounds to particular speakers.  This "works" as reported by Stacy and others in moving movie soundtrack sounds to the correct speaker better than other systems.  (It might be helping that movie sound is basically made for if not by Dolby processes in the first place.)  Dolby also applies non-linear compression to the surround channels and perhaps elsewhere, so they don't "stick out."

Well, that may be fine for movies, but it's crap for music.

In contrast, it seems, and I haven't looked at the Wikipedia article yet (one of the few places where you can often find information instead of just hype and opinions) but it seems like DTS Neo 6 may be pretty much leaving the front channels alone.  Well, it would make sense that would be an appropriate thing to do for stereo music sources, especially if you are not using a center channel speaker, and you care about the sound of the music.

Actually, DTS may not be leaving the front channels alone, in the Wikipedia explanation it claims to have 12-19 channels of steering.  Well I don't hear much more than ambience, I think somehow it's less "forced" than Dolby PLIIx, and may use less audio-quality-corrupting steering processes, or simply use them less.  DTS the company was established as a "higher quality" competitor to Dolby and is occasionally recognized as such though not by all.

This is a preliminary finding, but I felt moved to get it out there because it has always seemed to me that Dolby is taking over the world, not necessarily because they are the best, and they have endless sycophants to back them up.  So I want it to be known that there are some who don't see it this way, even if their critique is not yet fully validated.

I wish someone would make a surround system that simply maps, using only phase information as primitive matrixes do, any number of input channels to any number and placement of output channels.  Without any nonlinear processes such as steering, feedback, bucket brigade delays, or Dolby Noise Reduction.

I see now that in competition with Dolby Atmos, which gives hemispheric presentation but "requires" certain layouts, DTS has DTS X, which is more flexible, lets you set up speakers anywhere within the hemisphere, and encoded sound objects in the datastream are mapped to them using an open and license-free framework.  This is great, and would be what I want for discrete...AND for simulated surround from 2 channels.  But does it also process 2 channel inputs to add simulated surround?  It appears not.  I may still have to do that myself.  But there is apparently Neuro DTS X which can translate "legacy" sources into DTS X, so maybe they have all the pieces for this now.