To start the month of June out with another big but long, very long, delayed project, I decided to try to tackle the time alignment of the living room system using my newly learned tool REW. This is, of course, my number one serious system, with subwoofers, "full range" electrostatic speakers (which are crossed over at both ends by me btw, I am not a minimalism true believer obviously, though I tried the opposite approach for many years), and super tweeters, with each crossover and numerous parametric EQ corrections, each "way" handled by a separate Behringer DEQ box in pure digital, then to identical DACs and suitable amplifiers. The speakers for each channel are clustered but they are not coincident, so at the listening position each contribution would be offset by the different distances, if not other factors. (Previously, I didn't much consider these "other factors," but if there's one lesson to be learned this time, those other factors may be the important ones.) I can easily compensate for the different delays by adding additional delay in the DEQ boxes to compensate, making the total delay for each way exactly the same as experienced at the listening position.
Ever since the living room system got "serious", when I got the Acoustats and the TacT 2.0 RCS digital preamp, I started using the Tact to measure, adjust, and align my system.
I didn't like the actual room correction part of the Tact system (however, I'm not claiming I gave it a totally fair trial either, but I did try it a couple times, though I am predisposed against such "automatic" things). But the frequency response measurement done by the Tact prior to generating corrections generally seemed quite good, with the strange quirk that you needed to look at two different spectras to get the whole picture--which could only exist in your mind. The two different spectras actually both covered the entire range, but with different emphasis. Why it did this, or what the two spectra really meant, was never clear to me, but you did clearly get more bass resolution in the LF response curve than the HF response curve.
But for time alignment, the Impulse Response the Tact showed was not very clear in showing the correct time alignment. It was also quite hard to interpret, once again with two different versions. But twiddling the delay knobs on the DEQ's for each frequency range, I felt, sometimes, like I had, probably, made it better.
It was still very troublingly equivocal. I went through this exercise several times, and always meant to make a complete post showing all the complexities, but in the end, I kept away at it, like the proverbial gambler, hoping for the big win instead, the method that would clearly and intuitively reveal the correct time alignment, but never catching anything that was completely intuitive and self evident.
So each experience was like the the battles described by Stephen Crane in Red Badge of Courage in the sense I got so lost in it--the fog of war. Which was good in some ways, but not so good in developing an ultimate understanding and methodology of how to do it, or even an honest feeling of certainty that I had finally done it right this time. Just that at some point, each time, I found some little fit somewhere in the impluse response, where I at least imagined it showed the bass, midrange, and highs all starting at the same time, and I called it done, without coming close to being able to prove it was right, I just didn't want to look at it anymore.
Each time I thought, maybe, my final little trick needed a writeup on this blog. But I could never get myself to do it.
I recall that in the end, the most trustworthy way of aligning the subs and the panels was to make separate impulse measurements of each, and get the signal starting time (shown on a fixed scale by the Tact, relative the the signal it is sending) of both to be the same. It was easy to see the starting point of the panel transient. It was not so easy to see the beginning of the bass impulse, because it starts very small and blends into the LF noise of the room and microphone.
The time scale wasn't fine enough to do this well for the super tweeters, and the crossover settings changed the appearance of the initial 0.1ms of the impulse quite a lot. (I had many different crossover settings to choose from earlier on because I started using the DCX crossover units instead of the DEQ digital eq units. Now I can only cascade 12dB/octave low or high cuts to achieve the crossover filters I want. I switched to the DEQ's mainly because only the DEQ's have full digital I/O and I'm no longer stuck using their lousy built-in DAC's.) At one point I felt I had "nailed" the timing and the crossover of the panels and supers for a near perfect impulse start. But, I could never get myself to get the pictures online, also fighting with different kinds of smart phones and the demise of the picture sharing service I had originally used. And now I no longer use the DCX which was far superior wrt time alignment and the crossover selections, but I felt I couldn't live without digital output.
Other methods turned out worse. I could never get one of my old scopes to proper lock onto pulses, and the pulses themselves led to problems with my amplifier (then, the Krell), my Acoustat fuses and transformer, and differing with different DACs.
So now, I am leaving the 2001 Tact system behind (not really, in fact I think I'll try it again soon) and using the thoroughly modern, up-to-date, and free REW, everything is easy and fine, right???
Well, no, I'm back in a trench, but now it's even more unfamiliar.
Sorry to disappoint, but this is merely the INTRO! I'll get the the details, and pictures (finally) I hope, in future installments...
It was clear from the unclear computed Impulse "Step Response", this was going to be another hard slog. But it was also frustrating to again see what I'd long presumed to be a rear/reflection/cancellation at about 230Hz. After a few measurments for the time alignment proved frustrating, and there was a curious and alarming notch in the computed step response at a couple ms, I though maybe I could fix that by something that had long been a no-no-not-again: moving the speakers. (Note: it didn't fix the notch in the impulse, which remains unexplained.)
Yes, moving the speakers away from the wall.
Now some time long ago I had established the current position as the maximum distance away from the wall that was consistent with actually being able to walk through the hallway that adjoins the living room right at the front where the speakers are.
Any more, and I'd have to move the speakers back every night, and I am not about to do that sort of thing, I think, ever. I wouldn't want "oh, I can't play the stereo now because I have to move the speakers" ever have to get in the way of actually using the speakers, though I find endless other excuses.
Well things may have changed slightly since that last major repositioning, which was 2015 or so. One bookcase has been removed and the remaining bookcases moved back, so, viola, I can move the speakers a little further out into the room, as recommended by dipolar speaker experts, fans, and even Acoustat themselves, which suggested a "minimum" of 3 feet which I was just skirting at 33 inches, thinking myself to be an expert who could cheat a little.
So, I tried moving the right speaker out from the wall about as far as I could (currently limited by the speaker cables) about 8 inches. It made a huge difference, and the notch at 230Hz notch completely disappeared (or at least appeared to) in the spectrum.
Backing up, I can now consistently measure each change (and this consistency means almost as much as getting true accuracy, which I don't necessary claim but try my best, using the correction curve with my calibrated microphone).
I found that as little as 2 inches out pretty much fixes the 230Hz notch, leaving only a small dip, and 8 inches obliterates it and makes a peak near 200Hz instead. I settled on 6 inches out, which seemed to be acceptable for room passage also.
Now this 230Hz notch has been on my mind for a couple years now, and I had been planning to attack it using an entirely new set of EQ units, feed a separate 230 Hz bandbassed signal into the subs, etc. I was just about to start ordering a new DEQ, since the previous spare has now gone into adjusting the back surrounds in the kitchen system--level mostly.
But if this approach works, of course, it is better (and simpler too). There are some issues remaining, it's possible I didn't display the spectrum with enough resolution to see any remaining notch.
I then discovered the left speakers didn't have the 230 Hz notch problem at all. But I dutifully moved that Acoustat out by exactly the same distance to match it's partner, which actually didn't seem to change the response much on that side.
Listening has also confirmed, this is a huge change for the better. And it really didn't require REW at all, but REW gave the final kick in the pants, and it also showed how easy the problem was actually to fix, something that might not have been as obvious by other methods (such as using the 1/6 octave RTA in my smartphone, which is what I've mainly done the past few years, but it often isn't consistent enough to produce firm conclusions).
So, yes, I'm chalking up this enormous yet super simple improvement primarily to REW, and just part of using REW on the living room system for the first day, after attempting to do the time alignment again appeared so frustrating.
The second day, I did get back to the time alignment, which continued on the third day as well.
I am now using a separate audio interface with my laptop, a Focusrite 2i4, so that I can get a "loopback" which is required for REW to compute delay times. I thought it would be as simple as reading off the delay times after measuring each driver.
Well indeed the delay times computed for the Acoustats and the supertweeters seemed perfectly accurate, about what I expected, and it seemed possible to do it this way.
However, the subs were again a sticking point. The delay time computed for the subs appeared to be about 3 ms higher than anything I had ever dialed in before.
I took a look at the actual sub impulse response, and noticed something very peculiar. The impulse response of the subs begins with two humps, and for some reason REW decides that the beginning of the subwoofer response is the beginning of the second hump. It seems to ignore the first hump completely. If I examine the entire dynamic range, the first hump seems to arise out of the background noise, so it does not appear to be an artifact.
Now, curiously, the beginning of the first hump is about 3 ms EARLIER than the panel response. I do not understand why this would be true, and all the explanations I've come up with seem a bit shakey.
After a vast number of additional tests, however, I've decided that where the alignment of the bass needs to be done is in fact at the beginning of the first hump, not the second hump that REW identifies.
Listening seems to confirm this as well, never has the "you are there" feeling been so solid, and percussive instruments and basses sound much more real now.
One test that clinched it for me was noticing that if I started the panels and subs as I did before, based on the distance between them, the step response looked like a tower on top of a small mountain. I presumed the small mountain to be the bass response, which was apparently already under way.
Actually, the trick method for doing the final adjustment was simply this: I adjusted until the straight line of the beginning of the step response went as far down as possible. This actually makes the initial step as long as possible, but you can't get that result by looking at the height of the step, since the height is normalized it seems always the same. But you can see how low it goes beforehand, and where it goes the deepest, you have dialed in the correct time alignment.
The point where the mountain preceding the vertical rise of the step response went away, was exactly the point where the first hump in the woofer-only step response starts. And if I turn on the supertweeter also, you can see it begin right at that same point also. So I have confirmation using REW in two other ways.
I plan to publish the pictures of all this soon, and also I'm going to continue with other tests, programs, and techniques until a clearer understanding of this arrives.
But meanwhile, once again, I think I've found the magic, and far better than ever.
Ever since the living room system got "serious", when I got the Acoustats and the TacT 2.0 RCS digital preamp, I started using the Tact to measure, adjust, and align my system.
I didn't like the actual room correction part of the Tact system (however, I'm not claiming I gave it a totally fair trial either, but I did try it a couple times, though I am predisposed against such "automatic" things). But the frequency response measurement done by the Tact prior to generating corrections generally seemed quite good, with the strange quirk that you needed to look at two different spectras to get the whole picture--which could only exist in your mind. The two different spectras actually both covered the entire range, but with different emphasis. Why it did this, or what the two spectra really meant, was never clear to me, but you did clearly get more bass resolution in the LF response curve than the HF response curve.
But for time alignment, the Impulse Response the Tact showed was not very clear in showing the correct time alignment. It was also quite hard to interpret, once again with two different versions. But twiddling the delay knobs on the DEQ's for each frequency range, I felt, sometimes, like I had, probably, made it better.
It was still very troublingly equivocal. I went through this exercise several times, and always meant to make a complete post showing all the complexities, but in the end, I kept away at it, like the proverbial gambler, hoping for the big win instead, the method that would clearly and intuitively reveal the correct time alignment, but never catching anything that was completely intuitive and self evident.
So each experience was like the the battles described by Stephen Crane in Red Badge of Courage in the sense I got so lost in it--the fog of war. Which was good in some ways, but not so good in developing an ultimate understanding and methodology of how to do it, or even an honest feeling of certainty that I had finally done it right this time. Just that at some point, each time, I found some little fit somewhere in the impluse response, where I at least imagined it showed the bass, midrange, and highs all starting at the same time, and I called it done, without coming close to being able to prove it was right, I just didn't want to look at it anymore.
Each time I thought, maybe, my final little trick needed a writeup on this blog. But I could never get myself to do it.
I recall that in the end, the most trustworthy way of aligning the subs and the panels was to make separate impulse measurements of each, and get the signal starting time (shown on a fixed scale by the Tact, relative the the signal it is sending) of both to be the same. It was easy to see the starting point of the panel transient. It was not so easy to see the beginning of the bass impulse, because it starts very small and blends into the LF noise of the room and microphone.
The time scale wasn't fine enough to do this well for the super tweeters, and the crossover settings changed the appearance of the initial 0.1ms of the impulse quite a lot. (I had many different crossover settings to choose from earlier on because I started using the DCX crossover units instead of the DEQ digital eq units. Now I can only cascade 12dB/octave low or high cuts to achieve the crossover filters I want. I switched to the DEQ's mainly because only the DEQ's have full digital I/O and I'm no longer stuck using their lousy built-in DAC's.) At one point I felt I had "nailed" the timing and the crossover of the panels and supers for a near perfect impulse start. But, I could never get myself to get the pictures online, also fighting with different kinds of smart phones and the demise of the picture sharing service I had originally used. And now I no longer use the DCX which was far superior wrt time alignment and the crossover selections, but I felt I couldn't live without digital output.
Other methods turned out worse. I could never get one of my old scopes to proper lock onto pulses, and the pulses themselves led to problems with my amplifier (then, the Krell), my Acoustat fuses and transformer, and differing with different DACs.
So now, I am leaving the 2001 Tact system behind (not really, in fact I think I'll try it again soon) and using the thoroughly modern, up-to-date, and free REW, everything is easy and fine, right???
Well, no, I'm back in a trench, but now it's even more unfamiliar.
Sorry to disappoint, but this is merely the INTRO! I'll get the the details, and pictures (finally) I hope, in future installments...
It was clear from the unclear computed Impulse "Step Response", this was going to be another hard slog. But it was also frustrating to again see what I'd long presumed to be a rear/reflection/cancellation at about 230Hz. After a few measurments for the time alignment proved frustrating, and there was a curious and alarming notch in the computed step response at a couple ms, I though maybe I could fix that by something that had long been a no-no-not-again: moving the speakers. (Note: it didn't fix the notch in the impulse, which remains unexplained.)
Yes, moving the speakers away from the wall.
Now some time long ago I had established the current position as the maximum distance away from the wall that was consistent with actually being able to walk through the hallway that adjoins the living room right at the front where the speakers are.
Any more, and I'd have to move the speakers back every night, and I am not about to do that sort of thing, I think, ever. I wouldn't want "oh, I can't play the stereo now because I have to move the speakers" ever have to get in the way of actually using the speakers, though I find endless other excuses.
Well things may have changed slightly since that last major repositioning, which was 2015 or so. One bookcase has been removed and the remaining bookcases moved back, so, viola, I can move the speakers a little further out into the room, as recommended by dipolar speaker experts, fans, and even Acoustat themselves, which suggested a "minimum" of 3 feet which I was just skirting at 33 inches, thinking myself to be an expert who could cheat a little.
So, I tried moving the right speaker out from the wall about as far as I could (currently limited by the speaker cables) about 8 inches. It made a huge difference, and the notch at 230Hz notch completely disappeared (or at least appeared to) in the spectrum.
Backing up, I can now consistently measure each change (and this consistency means almost as much as getting true accuracy, which I don't necessary claim but try my best, using the correction curve with my calibrated microphone).
I found that as little as 2 inches out pretty much fixes the 230Hz notch, leaving only a small dip, and 8 inches obliterates it and makes a peak near 200Hz instead. I settled on 6 inches out, which seemed to be acceptable for room passage also.
Now this 230Hz notch has been on my mind for a couple years now, and I had been planning to attack it using an entirely new set of EQ units, feed a separate 230 Hz bandbassed signal into the subs, etc. I was just about to start ordering a new DEQ, since the previous spare has now gone into adjusting the back surrounds in the kitchen system--level mostly.
But if this approach works, of course, it is better (and simpler too). There are some issues remaining, it's possible I didn't display the spectrum with enough resolution to see any remaining notch.
I then discovered the left speakers didn't have the 230 Hz notch problem at all. But I dutifully moved that Acoustat out by exactly the same distance to match it's partner, which actually didn't seem to change the response much on that side.
Listening has also confirmed, this is a huge change for the better. And it really didn't require REW at all, but REW gave the final kick in the pants, and it also showed how easy the problem was actually to fix, something that might not have been as obvious by other methods (such as using the 1/6 octave RTA in my smartphone, which is what I've mainly done the past few years, but it often isn't consistent enough to produce firm conclusions).
So, yes, I'm chalking up this enormous yet super simple improvement primarily to REW, and just part of using REW on the living room system for the first day, after attempting to do the time alignment again appeared so frustrating.
The second day, I did get back to the time alignment, which continued on the third day as well.
I am now using a separate audio interface with my laptop, a Focusrite 2i4, so that I can get a "loopback" which is required for REW to compute delay times. I thought it would be as simple as reading off the delay times after measuring each driver.
Well indeed the delay times computed for the Acoustats and the supertweeters seemed perfectly accurate, about what I expected, and it seemed possible to do it this way.
However, the subs were again a sticking point. The delay time computed for the subs appeared to be about 3 ms higher than anything I had ever dialed in before.
I took a look at the actual sub impulse response, and noticed something very peculiar. The impulse response of the subs begins with two humps, and for some reason REW decides that the beginning of the subwoofer response is the beginning of the second hump. It seems to ignore the first hump completely. If I examine the entire dynamic range, the first hump seems to arise out of the background noise, so it does not appear to be an artifact.
Now, curiously, the beginning of the first hump is about 3 ms EARLIER than the panel response. I do not understand why this would be true, and all the explanations I've come up with seem a bit shakey.
After a vast number of additional tests, however, I've decided that where the alignment of the bass needs to be done is in fact at the beginning of the first hump, not the second hump that REW identifies.
Listening seems to confirm this as well, never has the "you are there" feeling been so solid, and percussive instruments and basses sound much more real now.
One test that clinched it for me was noticing that if I started the panels and subs as I did before, based on the distance between them, the step response looked like a tower on top of a small mountain. I presumed the small mountain to be the bass response, which was apparently already under way.
Actually, the trick method for doing the final adjustment was simply this: I adjusted until the straight line of the beginning of the step response went as far down as possible. This actually makes the initial step as long as possible, but you can't get that result by looking at the height of the step, since the height is normalized it seems always the same. But you can see how low it goes beforehand, and where it goes the deepest, you have dialed in the correct time alignment.
The point where the mountain preceding the vertical rise of the step response went away, was exactly the point where the first hump in the woofer-only step response starts. And if I turn on the supertweeter also, you can see it begin right at that same point also. So I have confirmation using REW in two other ways.
I plan to publish the pictures of all this soon, and also I'm going to continue with other tests, programs, and techniques until a clearer understanding of this arrives.
But meanwhile, once again, I think I've found the magic, and far better than ever.
No comments:
Post a Comment