ProSoundWeb Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 [All]   Go Down

Author Topic: SQ/AR2412 Traces  (Read 6874 times)

Nathan Riddle

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2052
  • Niceville, FL
    • Nailed Productions
SQ/AR2412 Traces
« on: July 19, 2018, 10:05:40 AM »

Finally got around to getting traces on A&H's SQ/AR2412.

Looking at the AR2412 traces I might see why SQ sounds slightly better than the 48kHz GLD/Qu/stage boxes. The phase wrap @ 20kHz & unstable performance at 20kHz is interesting. Not sure how performance hindering it is, but it's something.

SQ - Local - Main L - 0.52ms (Purple)
SQ - TB - Main L - 0.42ms     (Teal)
SQ - Local - Mix - 0.48ms      (Red)
SQ - Local - Grp - 0.44ms      (Blue)

AR2412 - SQLocal - Out - 1.13ms (Gold)
AR2412 - SQTB - Out - 1.02ms     (Green)
AR2412 - Local - Out - 1.48ms      (Red)
AR2412 - Local - SQOut - 0.88ms  (Purple)

Note, times are relative; I have no idea what the delay of my measurement setup is.
Scarlet 2i2 1-out > DUT
Scarlet 2i2 2-out > 2-in
DUT > Scarlet 2i2 1-in
1-out > Scarlet 2i2 2-in
« Last Edit: July 19, 2018, 10:16:14 AM by Nathan Riddle »
Logged
I'm just a guy trying to do the next right thing.

This business is for people with too much energy for desk jobs and too much brain for labor jobs. - Scott Helmke

Mark Wilkinson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1104
Re: SQ/AR2412 Traces
« Reply #1 on: July 19, 2018, 02:53:20 PM »

The phase wrap @ 20kHz & unstable performance at 20kHz is interesting. Not sure how performance hindering it is, but it's something.


FWIW, I've recently found when trying to use Smaart for purely electrical measurements, that it seems kinda impossible to know what's really going on with phase of the DUT at the very end of the spectrum. 
Because I think any measurement also includes sound card non-linearity.
And maybe even sample timing zero points with sound card and DUT don't completely line-up.... I dunno...

I just know that even at a 96kHz sampling rate, 1 sample is worth 70 degrees of phase at 20kHz.....so how good can 20kHz phase measurements be when both the souncard and DUT are digital and when there's no sync lock between them?
It's like there's a 35 degree +/- average error at 20kHz built in, if I'm thinking straight.......
Plus, Smaart doesn't to my knowledge allow sound card calibration for the card's non-linearites.

Coincidentally, yesterday I asked on the Rational Acoustics support forum, if there's a way to incorporate a sound card calibration.
I had been measuring some amp outputs and all the amps were showing the same tailing response up high.
So I checked the soundcard itself with REW. (which does  soundcard calibrations)
Turns out, the tailing was in the sound card (UMC404) and not the amps.....
Which means the tailing would show in acoustic measurements as well doesn't it ?
Logged

Nathan Riddle

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2052
  • Niceville, FL
    • Nailed Productions
Re: SQ/AR2412 Traces
« Reply #2 on: July 19, 2018, 03:38:31 PM »

FWIW, I've recently found when trying to use Smaart for purely electrical measurements, that it seems kinda impossible to know what's really going on with phase of the DUT at the very end of the spectrum. 
Because I think any measurement also includes sound card non-linearity.
And maybe even sample timing zero points with sound card and DUT don't completely line-up.... I dunno...

I just know that even at a 96kHz sampling rate, 1 sample is worth 70 degrees of phase at 20kHz.....so how good can 20kHz phase measurements be when both the souncard and DUT are digital and when there's no sync lock between them?
It's like there's a 35 degree +/- average error at 20kHz built in, if I'm thinking straight.......
Plus, Smaart doesn't to my knowledge allow sound card calibration for the card's non-linearites.

Coincidentally, yesterday I asked on the Rational Acoustics support forum, if there's a way to incorporate a sound card calibration.
I had been measuring some amp outputs and all the amps were showing the same tailing response up high.
So I checked the soundcard itself with REW. (which does  soundcard calibrations)
Turns out, the tailing was in the sound card (UMC404) and not the amps.....
Which means the tailing would show in acoustic measurements as well doesn't it ?

You might be onto something. And I won't pretend I know anything about all this. But my thought process is as such:

Purely considering just the results of measurements we can easily see 20kHz changes in phase/magnitude of small (in the degrees) by moving the mic minutely. How is that possible if 1 sample is truly 70 degrees @ 20kHz?

20kHz is extremely flat for a straight wire, and less so for various components, amps being some of the flattest and sound boards exhibiting some small phase distortion/difference.

Specifically the AR2412 (older purple one) measurements:
Using no smoothing or averaging one can pretty quickly discern the differences (noise floor maybe) in magnitude at 20kHz I think that's what I felt would be most noticeable audibly. Since who can tell there's a phase wrap at 20kHz when moving an inch changes the relative phase completely?

I dunno either...
Logged
I'm just a guy trying to do the next right thing.

This business is for people with too much energy for desk jobs and too much brain for labor jobs. - Scott Helmke

Mark Wilkinson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1104
Re: SQ/AR2412 Traces
« Reply #3 on: July 19, 2018, 05:01:33 PM »


Purely considering just the results of measurements we can easily see 20kHz changes in phase/magnitude of small (in the degrees) by moving the mic minutely. How is that possible if 1 sample is truly 70 degrees @ 20kHz?


Hi Nathan, I stand ready to be corrected too ...you ask a good question..

Here's my thinking... first, pls check my numbers.....

96kHz sampling means 1 sample (1 sec/ 98,000) =0.0104 ms.
20kHz sine wave has a 360 degree period of (1 sec /20,000) = 0.05ms.
So, 1 sample @ 98kHz  for 20kHz = (.0104/.0500)x 360 =75 degrees.

While we're at it, 75 degrees of 20kHz = (1130ft/20,000)x(75/360)= about 1/8 inch distance.

Ok, it seems to me, Smaart's time resolution has to be thought of as having two aspects...
One is the timing relationship, mathematically determined that probably? is continuous between the reference signal and the measurement signal.
But is seems that timing relationship is bound to being calculated off a discrete number of processing samples, represented by the delay finder time to the nearest sample.   (Note you can't adjust the delay finder less than 1 sample's worth of time)

So a small 1/8 mic movement can make for a small phase movement, because the math is continuous around whatever discrete time difference has been determined. But Smaart has to recalc constantly to the nearest sample, when more than a sample's movement is detected. So it's kinda a hop and calc situation....

I've often seen on traces that tail up or down at 20kHz, that I simply can't get zero phase at 20kHz using MTW and it's 48kHz sampling.
Bumping delay finder between two samples....it's either tail up, or tail down...unless I move the mic a tiny bit, and get lucky on the hop location.

Did I make any sense at all ?

edit: changed 98kHz typo to 96kHz


« Last Edit: July 21, 2018, 10:04:10 AM by Mark Wilkinson »
Logged

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: SQ/AR2412 Traces
« Reply #4 on: July 20, 2018, 09:38:27 AM »

FWIW, I've recently found when trying to use Smaart for purely electrical measurements, that it seems kinda impossible to know what's really going on with phase of the DUT at the very end of the spectrum. 
Because I think any measurement also includes sound card non-linearity.
And maybe even sample timing zero points with sound card and DUT don't completely line-up.... I dunno...

I just know that even at a 96kHz sampling rate, 1 sample is worth 70 degrees of phase at 20kHz.....so how good can 20kHz phase measurements be when both the souncard and DUT are digital and when there's no sync lock between them?
It's like there's a 35 degree +/- average error at 20kHz built in, if I'm thinking straight.......
Plus, Smaart doesn't to my knowledge allow sound card calibration for the card's non-linearites.

Coincidentally, yesterday I asked on the Rational Acoustics support forum, if there's a way to incorporate a sound card calibration.
I had been measuring some amp outputs and all the amps were showing the same tailing response up high.
So I checked the soundcard itself with REW. (which does  soundcard calibrations)
Turns out, the tailing was in the sound card (UMC404) and not the amps.....
Which means the tailing would show in acoustic measurements as well doesn't it ?
If you allow device aggregating, you can snap a TF of your I/O device with a loopback cable (OP to IP) and stim with the analyzers onboard generator, selecting the Gen as the reference. On v7 for me this always devolves into clock slip, but it works long enough to get a good TF of op and ip stages together. If they ain't flat enough for ya, use other IO.
Logged

Mark Wilkinson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1104
Re: SQ/AR2412 Traces
« Reply #5 on: July 20, 2018, 12:31:39 PM »

If you allow device aggregating, you can snap a TF of your I/O device with a loopback cable (OP to IP) and stim with the analyzers onboard generator, selecting the Gen as the reference. On v7 for me this always devolves into clock slip, but it works long enough to get a good TF of op and ip stages together. If they ain't flat enough for ya, use other IO.

Hi, thx. But  I'm not understanding... what do you mean by 'device aggregating'?
And how do you select Gen as reference?  I only see the Input Channels of my soundcard as available choices....either v7 or v8..
Logged

Tim McCulloch

  • SR Forums
  • Hero Member
  • *
  • Offline Offline
  • Posts: 23736
  • Wichita, Kansas USA
Re: SQ/AR2412 Traces
« Reply #6 on: July 23, 2018, 09:00:25 AM »

Adjustments to time within software is limited to a resolution of 1 sample period.
Logged
"If you're passing on your way, from Palm Springs to L.A., Give a wave to good ol' Dave, Say hello to progress and goodbye to the Moonlight Motor Inn." - Steve Spurgin, Moonlight Motor Inn

Marc Sibilia

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 75
  • Princeton, New Jersey
Re: SQ/AR2412 Traces
« Reply #7 on: July 23, 2018, 09:31:49 AM »

Adjustments to time within software is limited to a resolution of 1 sample period.

REW will do sub-sample alignment of impulse response, so it will be better for amplifier/mixer/processor measurements if you are concerned about phase at high frequency.

Quote from the manual:

The Sub-sample timing adjustment selection controls whether REW adjusts the impulse response timing to resolution below a single sample when setting t=0 at the IR peak or using the other channel as timing reference. Sub-sample timing adjustment requires a resampling of the impulse response to perform the adjustment, which slightly raises the noise floor of the measurement - however the increase is far below the noise floor of a typical acoustic measurement and sub-sample adjustment provides more accurate phase information at high frequencies.

Marc
« Last Edit: July 23, 2018, 09:36:24 AM by Marc Sibilia »
Logged

Mark Wilkinson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1104
Re: SQ/AR2412 Traces
« Reply #8 on: July 24, 2018, 09:09:39 AM »

REW will do sub-sample alignment of impulse response, so it will be better for amplifier/mixer/processor measurements if you are concerned about phase at high frequency.

Quote from the manual:

The Sub-sample timing adjustment selection controls whether REW adjusts the impulse response timing to resolution below a single sample when setting t=0 at the IR peak or using the other channel as timing reference. Sub-sample timing adjustment requires a resampling of the impulse response to perform the adjustment, which slightly raises the noise floor of the measurement - however the increase is far below the noise floor of a typical acoustic measurement and sub-sample adjustment provides more accurate phase information at high frequencies.

Marc

Yes, I've been using REW for such measurements for that reason.

But I had forgotten the VHF phase accuracy is due to the sub-sample option.
Thx for the reminder.
Logged

Nathan Riddle

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2052
  • Niceville, FL
    • Nailed Productions
Re: SQ/AR2412 Traces
« Reply #9 on: July 24, 2018, 10:21:50 AM »

If you allow device aggregating, you can snap a TF of your I/O device with a loopback cable (OP to IP) and stim with the analyzers onboard generator, selecting the Gen as the reference. On v7 for me this always devolves into clock slip, but it works long enough to get a good TF of op and ip stages together. If they ain't flat enough for ya, use other IO.

I think I found what you are talking about.

http://www.rationalacoustics.com/forums/showthread.php?1094-ref-to-gen-delay-time-problems

"Transfer Function measurements can also be referenced internally to Smaart's signal generator by enabling "Allow Multi-Device Transfer Function" option on the Transfer Function page of the main Options dialog and selecting the Generator as your reference signal source. In practice there are a few caveats to doing this, again mainly associated with delay times. For one thing, the latency of the measurement system from the output to the input becomes a factor in measuring absolute delay times. (This is generally not a factor when actively actively recording both input signals.) Another is that the latency through the measurement system may not be consistent from one measurement to the next, meaning you may need to measure the delay again each time you stop and restart the measurement and/or the signal generator. And finally if the measurement signal input and the output used to stimulate the system under test are not identically clocked, then the reference delay time will be subject to drift while measuring, and not just when restarting the measurement."

Page 100 in v7 manual:
http://www.rationalacoustics.com/download/Smaartv7UserGuide.pdf
Page 112 in v8.

I'll have to try that and see.

---

Hi Nathan, I stand ready to be corrected too ...you ask a good question..

Here's my thinking... first, pls check my numbers.....

96kHz sampling means 1 sample (1 sec/ 98,000) =0.0104 ms.
20kHz sine wave has a 360 degree period of (1 sec /20,000) = 0.05ms.
So, 1 sample @ 98kHz  for 20kHz = (.0104/.0500)x 360 =75 degrees.

While we're at it, 75 degrees of 20kHz = (1130ft/20,000)x(75/360)= about 1/8 inch distance.

Did I make any sense at all ?

Mark, good discussion. I agree with your math. 1/8" seems about right.

Does that mean that we can't measure up to 20kHz phase response accurately? Or am I missing something with the REW discussion?

Can someone explain in more layman's terms how subsampling works? Is it simply using interpolation between samples to obtain the expected value?
Logged
I'm just a guy trying to do the next right thing.

This business is for people with too much energy for desk jobs and too much brain for labor jobs. - Scott Helmke

Mark Wilkinson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1104
Re: SQ/AR2412 Traces
« Reply #10 on: July 24, 2018, 11:00:07 AM »

I think I found what you are talking about.

http://www.rationalacoustics.com/forums/showthread.php?1094-ref-to-gen-delay-time-problems

"Transfer Function measurements can also be referenced internally to Smaart's signal generator by enabling "Allow Multi-Device Transfer Function"

Mark, good discussion. I agree with your math. 1/8" seems about right.

Does that mean that we can't measure up to 20kHz phase response accurately? Or am I missing something with the REW discussion?

Can someone explain in more layman's terms how subsampling works? Is it simply using interpolation between samples to obtain the expected value?

Nathan, excellent find.   I had no clue that's what Multi Device Transfer could do..
Seems I need to be hitting the manuals harder !

I think with Smaart, we can measure 20kHz accurately, to no worse than 1/2 sample off, at 48kHz (assuming we are using MTW transfer windowing with its 48k requirement.)
So this is the same 75 degrees from 96k sampling numbers above....which means to me, and I see every now and then based on random luck with mic position, that if the mic just happens that it's positioned halfway between inferred sample timing, the phase trace at 20kHz will tail upwards from flat by about 30 degrees, or downwards by 30 odd degrees, as we bump delay finder one sample.

Obviously a sample off is no big deal in the real world at all...heck, I was trying to locate a line array last week and couldn't even pin the damn thing to within 30 samples.   
I just love precision and knowing how things really work....

My take on sub-sampling is that it just re-samples the impulse response, keeping the mag and phase relationship through the spectrum the same, while forcing the very upper end of the freq spectrum (nyquist freq maybe) to zero degree phase.
But that take is just one hell of a layman's guess ....
Logged

Nathan Riddle

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2052
  • Niceville, FL
    • Nailed Productions
Re: SQ/AR2412 Traces
« Reply #11 on: December 04, 2018, 08:52:02 PM »

Brought upon by these developments:

https://forums.prosoundweb.com/index.php/topic,169366.msg1562589.html#msg1562589

Qu to SQ to QU measurements.
And
Qu to SQ measurements.

QU-IP1 > SQ-IP1 > SQ-MainL > QU-IP17 > QU-MainL
And
QU-IP1 > SQ-IP1 > SQ-MainL

Note the differences in impulse responses. There's a reason SQ sounds better...
« Last Edit: December 04, 2018, 08:55:33 PM by Nathan Riddle »
Logged
I'm just a guy trying to do the next right thing.

This business is for people with too much energy for desk jobs and too much brain for labor jobs. - Scott Helmke

Nathan Riddle

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2052
  • Niceville, FL
    • Nailed Productions
Re: SQ/AR2412 Traces
« Reply #12 on: December 04, 2018, 08:52:49 PM »

SQ at various levels of pre-amp gain. It handles being buried in the red (red constantly on) very well IMO.

Eh, my Focusrite Gen2 was clipping. Might disregard those other two images.

Still... looks great IMO.

...I really need Langston's suggestion for real testing: RME Babyface Pro (stupid name)
« Last Edit: December 04, 2018, 08:58:33 PM by Nathan Riddle »
Logged
I'm just a guy trying to do the next right thing.

This business is for people with too much energy for desk jobs and too much brain for labor jobs. - Scott Helmke

Frank Koenig

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1318
  • Palo Alto, CA USA
Re: SQ/AR2412 Traces
« Reply #13 on: December 04, 2018, 10:03:39 PM »

REW will do sub-sample alignment of impulse response

Indeed, sub-sample time adjustment can be performed by interpolation in the time domain or, alternatively, in the frequency domain by multiplying the spectrum by a linear-phase all-pass function representing an arbitrarily small time shift. The resolution of this time shift is limited only by finite register length (finite precision) effects. Why anyone would need a time shift of less than one sample, for anything other than prettying up a phase plot, is another question.

For purely electronic measurements where you're dealing with what is essentially a well behaved linear filter you call always fall back onto the old sine-wave generator and scope to sort things that are giving weird results. Fourier-based analyzers can get wonky in the top octave at fs = 48 kHz. Problems such a clipping, or other gross nonlinearities, will be obvious on a scope. A loop back test of the analyzer is always a good idea.

--Frank

PS: The cool thing is that shifting an impulse response by a fractional sample amount will make it look completely different, while shifting it by an integer number of samples will leave it unaltered except for the shift. The sampling theorem at work...
« Last Edit: December 04, 2018, 10:10:45 PM by Frank Koenig »
Logged
"Nature abhors a vacuum tube." -- John Pierce, Bell Labs

Mark Wilkinson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1104
Re: SQ/AR2412 Traces
« Reply #14 on: December 05, 2018, 01:38:21 PM »

Why anyone would need a time shift of less than one sample, for anything other than prettying up a phase plot, is another question.



Lol...and agreed! 

The anal, gotta have precise repeatable measurements part of me, wasn't happy till I figured out why phase at 20kHz using Smaart never seemed to wanna be at zero.   At least now I know why, and not to care...

But I do think fine timing is important...I believe most of the 96kHz sampling hullabaloo being discussed in conjunction with mixers, is probably more about being able to maintain tighter phase alignment throughout all the routing, than any other audio quality issue.

Oh, in addition to REW handling sub-sample timing on the measurement side, it appears that q-sys handles it too, on the signal side.
 

@Nathan,  sorry...I'm not getting what you're seeing / trying to show, with your impulse plots.... 

Logged

Nathan Riddle

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2052
  • Niceville, FL
    • Nailed Productions
Re: SQ/AR2412 Traces
« Reply #15 on: December 05, 2018, 02:50:57 PM »

Lol...and agreed! 

The anal, gotta have precise repeatable measurements part of me, wasn't happy till I figured out why phase at 20kHz using Smaart never seemed to wanna be at zero.   At least now I know why, and not to care...

But I do think fine timing is important...I believe most of the 96kHz sampling hullabaloo being discussed in conjunction with mixers, is probably more about being able to maintain tighter phase alignment throughout all the routing, than any other audio quality issue.

Oh, in addition to REW handling sub-sample timing on the measurement side, it appears that q-sys handles it too, on the signal side.
 

@Nathan,  sorry...I'm not getting what you're seeing / trying to show, with your impulse plots....

The length of the impulse response 'ringing' is significantly less for the SQ compared to QU.
Logged
I'm just a guy trying to do the next right thing.

This business is for people with too much energy for desk jobs and too much brain for labor jobs. - Scott Helmke

ProSoundWeb Community

Re: SQ/AR2412 Traces
« Reply #15 on: December 05, 2018, 02:50:57 PM »


Pages: 1 2 [All]   Go Up
 



Site Hosted By Ashdown Technologies, Inc.

Page created in 0.035 seconds with 24 queries.