Home Theater Forum and Systems banner

1 - 15 of 15 Posts

·
Premium Member
Joined
·
129 Posts
Discussion Starter · #1 ·
I decided to start a new thread as an offshoot of the "limitations of Java" thread. I learned something new (having missed it in PM's introduction to impulses), that the "typical" spikes in negative front of the impulse represent harmonic distortion. Probably because I haven't noticed those spikes since I linearized my system with Acourate and Acourate Convolver. So I thought I'd present some before and after pictures here for your entertainment and discussion.

All these are measurements of the left front loudspeaker, which is a Revel Gem paired with a JL Fathom F112 subwoofer. I have two subs, each devoted left and right, making this a full range stereo system.

Prior to January 2013, I was using an analog crossover between the JL and the Revels, at the end of that period, 48 dB/octave! First, attached, is a picture of the 2nd, 3rd, and 4th harmonics taken at that time. The SPL averages 77 dB. Second, attached, is the same picture using Acourate's Neville-Thiele steep linear phase crossover and Acourate Convolver. Distortion reduction is at least 10 dB and some harmonics as much as 15 dB!
 

Attachments

·
Premium Member
Joined
·
129 Posts
Discussion Starter · #2 ·
Next a picture of the log impulse responses. First with the analog crossover, then with the digital crossover and drc. Notice that the impulse is at nearly 1 second because of the latency of Acourate Convolver. I don't see evidence of the harmonic distortion bumps as seen in the analog crossed-over system. Perhaps the harmonic distortion is below the noise seen in this type of graph.
 

Attachments

·
Premium Member
Joined
·
129 Posts
Discussion Starter · #3 ·
Next is a picture of the comparative frequency response with the analog crossover. The woofers were corrected by a Meyer analog equalizer. The mains were uncorrected.

Followed by Acourate DRC, which does the whole range. This was taken with a 500 ms Hann window and 1/6 octave smoothing. I offset the before frequency response by 10 dB for clarity.
 

Attachments

·
Premium Member
Joined
·
129 Posts
Discussion Starter · #4 · (Edited)
Last is a picture of the waterfall before and after Acourate Convolver. It's much better behaved after the DRC, but you can see the transition between the effectiveness of the equalization and where the room decay predominates, particularly at 40 Hz at around 60 ms if I read the graph correctly, which isn't too bad! Still, I'll take anything I can get, until I get a Helmholz resonator that's large enough :).
 

Attachments

·
Super Moderator
Joined
·
1,673 Posts
Hi Bob,

Cool stuff. Couple questions for you: What is the sub/mains XO frequency? Do you have a room noise measurement (nothing playing)?
I see you're in FL :T Any AC going?
Lastly, maybe you could do a FR/distortion sweep of the mains only with the analog vs digital XO? Is the analog a LR type? I suspect the large reduction in the distortion in the 200hz range is the mains reduction in excursion due to the true steep filtering and knee of the digital, but we'll see.:)
 

·
Premium Member
Joined
·
129 Posts
Discussion Starter · #6 ·
Hi Bob,

Cool stuff. Couple questions for you: What is the sub/mains XO frequency? Do you have a room noise measurement (nothing playing)?
I see you're in FL :T Any AC going?
Lastly, maybe you could do a FR/distortion sweep of the mains only with the analog vs digital XO? Is the analog a LR type? I suspect the large reduction in the distortion in the 200hz range is the mains reduction in excursion due to the true steep filtering and knee of the digital, but we'll see.:)
Hi, AJ!

1) XO is 80 Hz

2) I will have to take a room noise measurement and put it up.

3) AC is running and it's INAUDIBLE. You can't even hear the rush of air or measure its presence or absence.

4) I can no longer run the old analog crossover. It's in the garage. But I can play the mains alone either through Acourate or not or the subs alone through Acourate or not if you'd like. If I run the mains without Acourate they have their own natural rolloff at 60 Hz. If I run the subs without Acourate they'll go up to about 200-300 Hz and roll off on their own. Will any of that help?

That is my hypothesis, that the large distortion reduction is due to the steeper low pass crossover on the subs so they run even more within their linear region. So we could test that by doing a "no acourate" distortion run on the subs alone. I'll make that a project for the next rainy weekend :).
 

·
Super Moderator
Joined
·
1,673 Posts
Hi Bob,

Well, I was viewing on a much smaller screen earlier today, making thing a wee bit more difficult to see.
Now things make even less sense. :)
Actually, about the only thing that does make sense is the increase in output..and corresponding increase in distortion, round 17-19hz.
I think there are just too many variables involved. The magnitude differences in distortion doesn't seem to make sense with the different filters and yes, some EQ, although what you are measuring is simply pressure at one spatial point. I assume this is at the LP? What is the distance?
I can't help but think something in the analog chain is creating some distortion. But I also think LF distortion sweeps like this outside an anechoic space should be viewed cautiously.
Since the analog XO is mothballed, perhaps you could do a sweep using the Fathoms internal XO/EQ capability (set at 80hz of course), with acourate bypassed and then how you have it now.
I really don't expect much of a difference, at the magnitude response should correlate heavily to the distortion generated. The low pass filter slope should have minor effect at 80hz.
I am not aware of any mechanism for external DSP to lower distortion that should be inherent to the transducer/output.

I'll take a peek into the "Java" thread.

The DRC is obviously smoothing the FR, but I would be very careful with any EQ above say 500hz and I would also love to see the nearfield (<1m) response of both sub and mains that resulted in that smooth response.

but you can see the transition between the effectiveness of the equalization and where the room decay predominates, particularly at 40 Hz at around 60 ms if I read the graph correctly, which isn't too bad! Still, I'll take anything I can get, until I get a Helmholz resonator that's large enough :).
Not sure what you would need any such thing (unless joking as smiley might indicate). I haven't looked for a while, but last I checked (mid 2000's), human ears had significant difficulty perceiving any such (40hz) LF "ringing" in the time domain. Pressure, yes, remove any peaking via EQ.
A read for you: http://www.genelec.fi/ht/documents/publications/aes116th_2.pdf
I'll have to dig if anything newer has invalidated this.

cheers
 

·
Premium Member
Joined
·
129 Posts
Discussion Starter · #8 · (Edited)
Dear AJ: 1) In my copious free time (not!!!) I will try to do some single-variable tests and measurements to prove or disprove my hypothesis. The dramatic (10 to 15 dB) decrease in distortion is so large we also have to prove repeatability of the distortion measurement from moment to moment, mike position to mike position etc. without even changing any other variable. Well, when I find the time. They say, "if you want to get something done, ask a busy person."

AJ, how would you suggest I present the room noise measurement? I suppose I could show the real time RTA of REW...

Oh, another element of the analog system that I do not wish to recreate is that the woofers were located in line with the mains because I could not implement proper time delay compensation. The woofers are now in the corners and the mains are located about 3 feet closer to the listener.

2) Regarding the high frequency smoothing and high frequency target curve. This is all due to the excellent FIR filtering in Acourate and Acourate Convolver. The proof is in the listening. Hey, if a loudspeaker system can have all kinds of anomalies in the drivers and crossover, why can't a digital filtering system smooth those out? My answer: Absolutely no reason why not. Done it, six times so far in six different rooms, and the cure has always sounded better than the disease. Provided you use a transparent, high quality platform for the calculation and reproduction. For that I nominate Acourate, which has:
1) 65k-long FIR filters. Steep linear-phase crossover.
2) 64-bit floating point arithmetic
3) Psychoacoustically-correct variable windowing and modal estimation
4) is properly-dithered to 24-bits on its output to the DAC
5) Does NOT require or use an ASRC. It works at the native sample rate of the source.

This translates to a sonically transparent system which does not deteriorate or change the depth, imaging or transparency of the original source. I spent a year proving that to myself before mothballing an analog-based subwoofer-mains crossover and subwoofer equalizer.
 

·
Premium Member
Joined
·
129 Posts
Discussion Starter · #9 ·
Dear AJ: The mechanism I am suggesting for lowered distortion is that the Neville-Thiele linear phase crossover is EXTREMELY steep and keeps the woofer within its linear region and doesn't ask it to reproduce frequencies outside of the range.

We shall see if we can get to the bottom of this (pun intended). I can do a controlled distortion test of the sub measurement only, with the mains speakers turned off, with the Fathom's internal XO versus Acourate XO. Sweep to say 400 Hz and you can also compare the steepness of the XO.
 

·
Super Moderator
Joined
·
1,673 Posts
They say, "if you want to get something done, ask a busy person."

AJ, how would you suggest I present the room noise measurement? I suppose I could show the real time RTA of REW...
Hah, can certainly relate to that.:)

Yes, RTA mode would be fine.

Oh, another element of the analog system that I do not wish to recreate is that the woofers were located in line with the mains because I could not implement proper time delay compensation. The woofers are now in the corners and the mains are located about 3 feet closer to the listener.

2) Regarding the high frequency smoothing and high frequency target curve. This is all due to the excellent FIR filtering in Acourate and Acourate Convolver. The proof is in the listening. Hey, if a loudspeaker system can have all kinds of anomalies in the drivers and crossover, why can't a digital filtering system smooth those out? My answer: Absolutely no reason why not.

This translates to a sonically transparent system which does not deteriorate or change the depth, imaging or transparency of the original source. I spent a year proving that to myself before mothballing an analog-based subwoofer-mains crossover and subwoofer equalizer.
Yes, DSP is very handy for time domain.

Any EQ, digital or not, must be applied judiciously. With a non-coincident design, especially so. Remember, EQ applies axially to one point in space. We have two ears..and HRTF. Big problem.:)
I understand Acourate goes beyond that, but EQ cannot correct globally, the polar response of the speaker must be smooth natively. It would be interesting to see what is happening globally as you eq to your LP. Gated NF response of the system in a 60 degree window, with the Acourate EQ applied, would be interesting.
Luckily, in your case (at least horizontally) the Revel should be very good here. i.e., the off axis will track the on.
If the "JJ" you refer to in the other thread is who I think it is, you've struck gold. Read very carefully what he has to say. Especially about EQ above 500hz.

No doubt your getting excellent sound, you're starting with excellent sources. How is the inter-channel consistency (REW measured?) with the Acourate?

cheers
 

·
Premium Member
Joined
·
129 Posts
Discussion Starter · #11 · (Edited)
Dear AJ:

JJ has actually been to my studio and marveled at the sound in this digitally-corrected room. I have no fears equalizing above 500 Hz when you know what you are doing. Concerns include (as you say) polar response, sidewall reflections (read Floyd Toole's book), floor reflections, ceiling reflections, etc. The ETC in Studio A is a reflection-free-zone. After the first millisecond, it is clean to below -30 dBFS for the first 6 ms. and below -20 dBFS thereafter except for one spike at 36 ms. that's "only" at -18. Current wisdom is that a reflection-free zone is considered below -15 dBFS for the first 15 ms. So this room is exceptionally beyond that, with a cathedral ceiling that goes up to 23 feet high in the back, and side walls that are far enough apart to place any sidewall reflections behind the listener's head. It's hard to even see a floor bounce in the ETC. There are many other considerations, but after a year of experimenting with and proving the efficacy of Acourate, I am sold on DRC from 20 Hz to 20 kHz as well as the excess phase correction in Acourate that has allowed the creation of an impulse that removes the typical "tweeter, midrange, woofer" signature in the impulse response and replaces it with a remarkably tight pulse with much less duration or overshoot. Of course the proof of the pudding is in the listening; the results are transparent, accurate, musical, with superb depth, dimension, transient response to wake you up, and clarity. It's a joy to listen to this system from morning through night, and I have to work in here. Oh, and it passes the LEDR test with flying colors. AJ, all the objective and subjective listening tests say "come on in, the DRC water is fine".

Attached, REW-measured ETC of the left front loudspeaker and REW-measured impulse response of that loudspeaker taken a year ago with the old analog crossover followed by a year later with full Acourate correction.

Also, attached, is the psychoacoustic frequency response as determine by Acourate using a variant on JJ's theories that has a sliding FFT window combined with transient response simulation that reduces overcorrection. In fact, to my ears, I have never had to recorrect for any low frequency corrections that have been done by Acourate. Every other system I have tried has overestimated the corrections that must be applied. And you can see the corrections to the Revel's response above 500 Hz in the graph, all with the desire to bring it closer to the target and remove the ratty anomalies caused by any practical midrange and tweeter driver in pretty much any enclosure I have analyzed. At 1 kHz, the FFT window is 15 ms. and it gets shorter above, so the HF response is closer and closer to that which is direct from the speaker, which also according to JJ's theories (and I concur) reflects what the ear hears and discriminates.

John M. has it on his plate to try to implement a sliding window, which to my mind is a revolution. It is important in order to see the frequency response detail above 1 kHz. Much of it is obscured in the typical REW plots that you can make when you concentrate on the low end. Currently REW's frequency response plots can either be accurate below 200 Hz or so using, say, a 500 ms window, and above that, a much shorter window.

You asked about interchannel consistency. If I recall correctly, the IACC is better than 86% as measured by Acourate. I have to look at that again when I take another reading. But passing the LEDR test pretty much also verifies interchannel consistency in the critical midrange through the upper midrange, as well as immunity to your HRTF concerns.

The combination of the window and the smoothing is also necessary. The devil is in the details. That's why acoustics is still a science and an art.


BK

Hah, can certainly relate to that.:)

Yes, RTA mode would be fine.



Yes, DSP is very handy for time domain.

Any EQ, digital or not, must be applied judiciously. With a non-coincident design, especially so. Remember, EQ applies axially to one point in space. We have two ears..and HRTF. Big problem.:)
I understand Acourate goes beyond that, but EQ cannot correct globally, the polar response of the speaker must be smooth natively. It would be interesting to see what is happening globally as you eq to your LP. Gated NF response of the system in a 60 degree window, with the Acourate EQ applied, would be interesting.
Luckily, in your case (at least horizontally) the Revel should be very good here. i.e., the off axis will track the on.
If the "JJ" you refer to in the other thread is who I think it is, you've struck gold. Read very carefully what he has to say. Especially about EQ above 500hz.

No doubt your getting excellent sound, you're starting with excellent sources. How is the inter-channel consistency (REW measured?) with the Acourate?

cheers
 

Attachments

·
Premium Member
Joined
·
129 Posts
Discussion Starter · #12 ·
One more thing. I made my HF target slope purposely with a little more rolloff than the Revels produce at their -0.5 dB tweeter setting (which I found the most linear) because subjectively it sounded just a little bright, so you see the resulting HF response just a little below the average original Revel response. And while we're at it, I'll ask (respectfully), "why wouldn't you want to smooth the ratty response between 1k and 3k"? If you can do it without artifacts... without phase shift, with 64-bit float precision dithered to 24 bits... that's the miracle of this FIR-based approach. The only sonic cost so far I have been able to detect is latency. Everything else is a plus.

But as you say, if you don't start out with a superior room, superior loudspeakers, superior treatment, you can really screw up the sound trying to equalize it. That was the conventional wisdom, but I have defied it and survived :).
 

·
Super Moderator
Joined
·
1,673 Posts
One more thing. I made my HF target slope purposely with a little more rolloff than the Revels produce at their -0.5 dB tweeter setting (which I found the most linear) because subjectively it sounded just a little bright, so you see the resulting HF response just a little below the average original Revel response.
I see that. You're still pretty close to the B&K curve. Looks about right to me.

And while we're at it, I'll ask (respectfully), "why wouldn't you want to smooth the ratty response between 1k and 3k"?
Because, unless I misunderstand the system you are using, we have 2 pinna, HRTF and significant post processing. Not one pressure sensing mic. I would like to see 2m anechoic measurements of the Revel (can't find anything related to their native response). I'm curious as to why that now corrected dip is there in the first place.:)

But as you say, if you don't start out with a superior room, superior loudspeakers, superior treatment, you can really screw up the sound trying to equalize it. That was the conventional wisdom, but I have defied it and survived :).
Indeed. Hopefully you are also aware that Toole found that home listening rooms and studios have different needs....and preferences do vary.
Did JJ try to convince you that 3 more channels and a 7 mic spherical array would be nice?:)

cheers
 

·
Premium Member
Joined
·
129 Posts
Discussion Starter · #14 ·
Dear AJ: Yup, I'm not interested in improving the response of ratty speakers in a ratty room. But even the best loudspeakers with standard analog (minimum phase) crossovers have their issues as you know. Acourate helps to remove many of the remaining issues and turns the loudspeaker into a more ideal transducer. I have extremely critical ears and have been in this business professionally since 1972 and been an audiophile since forever. My first serious loudspeakers were Dahlquist DQ-10s. The transformations which Acourate can perform on a pair of good loudspeakers (including above the bass region) in a good room are sonically remarkable. Interaural cross correlation improves tremendously because even the most precisely-made loudspeaker systems have small (measurable and audible) differences between channels. Impact and transient response improves tremendously (measurable and audible) because of the linearization of the phase response and minimization of the width of the impulse and its common overshoots and ringing. Harmonic distortion (I claim) is reduced if using Acourate's crossover between subs and mains (yet to be 100% proved).

But still, the Revel Gems have a bit of beaming. The HF response is brighter on axis than on the side, making the speaker remain pretty much a one-man speaker for the most part. I spoke to Kevin Voecks (the excellent designer) about this and he said that this is the case for the Gems but not for their larger products. What do I do about it? I linearize for the near-anechoic high frequency response measured using a sliding FFT window as implemented by Acourate and originally recommended by Jim Johnston in his seminal research papers: AES preprint #7263, #8314, and #8379. The result: For a center-located listener, the sound is, to put it mildly: glorious. Other listeners enjoy a more linear response, but the imaging tends to pull towards one loudspeaker or the other, as it does anyway by the law of the first wavefront, only a bit more with the Revel Gems than with speakers that have a more linear polar response.

Acourate does measure the response only at one point in space and has an innovative method of dealing with bass anomalies and room modes below the Schroeder frequency but does not have any special compensation for polar issues or reflection issues other than the fact that the ear itself responds increasingly to the direct sound from the loudspeaker as frequencies increase. And in fact, JJ advocates that if you use too wide a window, you will be misinterpreting the true psychoacoustic response of the loudspeaker. My experience agrees.

So I have to say, tentatively, that if you start off with a pair of good loudspeakers in a properly treated room, deal with the early reflections properly, then response correction above the bass region is not only useful, it is downright ear-friendly. Not one critical listener who has come to this room has said, "your high frequency response is wrong." Instead, they say, "wow, I've never heard anything so musical and also so revealing at the same time".

If, however, you have reflection problems in your room, and Acourate attempts to correct for high frequency anomalies that are directly related to reflection problems rather than the anechoic response weaknesses of the loudspeaker itself ---- I cannot predict what the result will be. I do not have enough experience with Acourate in bad rooms to know when the "cure" will begin to sound worse than the "disease"! And to tell you the truth, I don't really want to find out. A bad room is a bad room.
 

·
Premium Member
Joined
·
129 Posts
Discussion Starter · #15 ·
It's been a year since I posted and started this thread. As you may recall, the claim I made was that the analog crossover, not being steep enough, was causing more harmonic distortion in the drivers by running the loudspeakers outside of their optimum range. But since I put the old analog crossover and analog equalizer to bed, I was not able to confirm my measurements or whether the SPLs were comparable.

Recently I realized I could easily resurrect most of the old analog system, at least enough to prove my point, calibrate the test microphone and ensure that all measurements are made on a level playing field. The left front speaker is under test. What I did for the analog xo test is to keep the same pair of DACs I am using with Acourate Convolver, but patch a full range, unaltered digital signal into both DACs simultaneously. This means that the Revel Gems are receiving a full range signal and the JL Fathom woofers are also receiving a full range signal. I'm depending on the natural rolloff of the Revels and the built in 4-pole low pass filter in the JLs to combine the two signals.

Attached here and in the next few posts are some powerful measurements demonstrating the superiority of Acourate Convolver with a steep linear phase crossover, not only in frequency and impulse response, but also in harmonic distortion. At 75 db SPL, the distortion difference is not significant, but at 85 dB SPL there is significant improvement in THD.

First a comparison on the impulse responses. The impulse response with the DRC (Acourate Convolver) has been corrected for phase and time response. The corrected system is considerably tighter, more unipolar (positive going) and sharper than the uncorrected analog system, which shows considerable over and undershoot and evidence of the individual drivers with opposite polarities. Sonically it's no contest, the digitally corrected system has a much more impressive impact and transient response.

Next a comparison of frequency response, showing the significant improvement at all frequencies from 20-20 k. The digitally corrected system sounds much smoother, musical and more accurate.

Next, comparison of distortion. Black is THD, Brown is third harmonic. At 75 dB SPL and 85 dB SPL. Let's look at the distortion between 40 and 50 Hz, the lowest range of the 4 string bass, visually taking the center of the displayed curve.
At 75 dB SPL, the analog XO's distortion is around 42 dB, 33 dB below the fundamental, or about 2.2%
At 75 dB SPL, the digital XO's distortion is also around 42 dB, so this is a tie.

At 85 dB SPL, the analog XO's distortion is around 52 dB, 33 dB below the fundamental, or about 2.2%
At 85 dB SPL, the digital XO's distortion lowers quite rapidly between 40-50 Hz, with an average of about 60 dB, so it is actually worse

However, if we look at the distortion below 30 Hz, the digital XO is significantly better, 20 dB better in fact. Is this a win-win situation? I think you have to examine the graphs in detail to reach a conclusion, also keeping in mind that it is very rare to have bass levels at this kind of amplitude anyway.

Oops, I see errors in the labels at the top of some of the graphs. Well, you should be able to tell which is the digital xo/DRC by virtue of its smoother and tighter frequency response

Comments are invited, if you're still around, A.J.!




I attribute the reduction to the fact that the Revels are receiving a full range signal and they are not meant to try to reproduce extra low frequencies below their limit of 60 Hz. I could only prove this claim 100% by using an analog high pass filter to the Revels.
Let the games begin.
 

Attachments

1 - 15 of 15 Posts
Top