Sweeps report approx 3000ppm timing errors

[Tony Walters]

Hello,

When performing sweeps at 512k samples and 1m samples as examples, the measurment reports large timing errors and suggests verifying that sweeps and generated sweep file are at same rate (they are) and to look for dropouts.
This is the first time this error has surfaced in over 2 years of using REW, I have current product release on one laptop, and Beta4 installed on another. Same errors. have tried different volume levels for the sweeps. Same errors. I use the generator to create the measurement sweep .wav files with timing chirp, and then that gets copied over to Volumio folder and plays from there.

Also, the impulse reponse is very strange, maybe not surprising. Occasionaly, it also reports that distortion is abnormally high, but not all the time. Otherwise, the sweep response looks normal.

I replaced the USB cable for the UMIK microphone, same error.

My setup is RPI4 Volumio streamer, with FusionDSP, USB out to RME-ADI 2 DAC, feeding preamp to active crossover, feeding power amp. 2.1 system crossed at 50hz, low to sub, high to main speakers. Same error whether FusionDSP is enabled or not.

I tend to think the mic is ok (and the response file for UMIK is loaded). Maybe it could be the problem?

I posted on minidsp forum in the UMIK section and no responses.

I appreciate if anyone has seen this before, and how to resolve it.

Cheers,
Tonyw

 

[John Mulcahy]

You can check for dropouts by opening the Captured graph after making a measurement, it shows the raw sweep signal obtained from the mic. Look for flat spots or discontinuities. The graph is only for the last measurement made.

 

[Tony Walters]

You can check for dropouts by opening the Captured graph after making a measurement, it shows the raw sweep signal obtained from the mic. Look for flat spots or discontinuities. The graph is only for the last measurement made.

Hello John. Here is what the graph shows. I'm inferring that the signal level is really low but what could be causing it? I'm using all the same levels as before. At this level, I'm surprised it can generate a measurement result.

 

[John Mulcahy]

You can hide the Sweep trace and just look at the Captured trace. Zoom in on the Y axis to make it easier to see the signal and look through it to see if anything stands out. The level is OK. If you don't see anything attach the mdat file for a measurement that generates the error.

 

[Tony Walters]

Thank you John. In the meantime my sub plate amp failed and waiting for replacement. It did, however, give me a chance to re-run sweeps without sub. I took out the active crossover and ran the attached sweeps. No change, the timing errors seem to have stabilized around 1000-1100ppm.

 

[John Mulcahy]

the timing errors seem to have stabilized around 1000-1100ppm

Those do look like the corrections needed to get something resembling the one good measurement, number 5, which was 28.8 ppm. Seems like something has a clock that's a long way off, most of the time. I can widen the limits to allow adjustments that large to be applied, but that's a very big shift. What's the UMIK connected to? It gets its clock from the USB host.

 

[Tony Walters]

My laptop has the new USB connector while the UMIK has the full size USB so I have a USB hub Hyper which has a number of ports including full size USB. I am not using any of the other ports. However, this was failing on my other laptop with a full size USB thats why I dont suspect the hub.
The timing error is now always 1049ppm now. I do get the proper looking impulses on the overlays now,
I do notice that on the latest beta that the error message says that it could be reflections from objects in the path to the microphone. I havent seen issues before, but their is a coffee table. I can remove that and see if it changes anything. I suspect it wont because objects on the table are always changing yet the error is now the same each time - but I will do it and report back.

'best
Tony

 

[John Mulcahy]

If you have a stable difference it won't be due to reflections.

 

[johny_2000]

Hi John,

I have the same problem with brand new UMIK-2 and an unusually high timing error with acoustic reference in REW.

The idea that the delay is caused by the DAC seems to make some sense. I also used miniDSP 2x4 HD as an output device on this PC. The error was random from 2000 to 6000+ ppm with the exact same setup and microphone mounted on a stationary tripod. I even ran a series of automatic measurements with the same setting, and each one came back with the random high timing error. I then switched the output device mode to ASIO (with VB-Audio ASIO Bridge), reduced the buffer size to 8 samples in the miniDSP Device Console, and sampling rate to 48 KHz in Windows settings on both devices, which reduced errors to the 100-600 ppm range, and anything below 250 ppm completed without error messages.

I have a laptop on which I can run the test with the built-in sound card and UMIK-2 as a test for the cause of the error.
Is this a problem with the digitization delay of the UMIK microphone, or is it a problem with the conversion delay of the miniDSP DAC?
I remember REW warns in its documentation that the use of different playback and recording devices without a wired loopback reference leads to clock sync errors.

 

[John Mulcahy]

High timing errors are not caused by delays. They are usually caused by dropouts on either replay or capture, due to buffers too small or sometimes very long USB leads. For ASIO use the largest buffer setting available. There isn't usually a problem using the WASAPI Exclusive Java drivers (device names starting with EXCL in REW).

 

[johny_2000]

Thank you!

Tomorrow I'll try with the largest ASIO buffer size and EXCLusive WASAPI mode.

Regarding dropouts, could there be a problem with the frequency response of the speaker? They are definitely not perfectly flat across the measurement range.

 

[John Mulcahy]

That doesn't affect clock measurements.

 

[johny_2000]

High timing errors are not caused by delays.

That doesn't affect clock measurements.

John, if I understand you correctly, the ADC / DAC latency and Frequency Response of the speakers do not cause timing errors.
This can only be caused by dropouts during playback or capture due to a too small buffer, or an unstable USB connection.

I can definitely work on the USB connection since I have the option to connect it to the USB ports on the front (now) or rear of the PC. As well as a choice of USB2 / USB3 ports.

Regarding the buffer size, are you referring to the REW buffer size in the Soundcard Preferences or the ASIO driver buffer size (number of samples)?
I'm using the default 32k buffer settings in REW for both devices.

 

[John Mulcahy]

The ASIO buffer, you mentioned reducing it to 8 samples which is pretty much certain to cause problems. REW's buffers are huge, the 16k min setting is fine.

 

[johny_2000]

The ASIO buffer, you mentioned reducing it to 8 samples which is pretty much certain to cause problems. REW's buffers are huge, the 16k min setting is fine.

Oh, okay. The ASIO buffer has many options: from 8 to 2048 samples at 44.1 kHz, 4096 samples at 88.2 kHz, 8192 samples at 192.1 kHz.

In this case, what is the best sampling rate to choose for both devices (DAC and UMIK-2)?

I originally started with a sample rate of 96 kHz, but there were terrible timing errors. I then switched to 48kHz sampling rate along with ASIO mode and this seemed to make good progress in minimizing errors. However, the errors between measurements are still not very good (<250 ppm) and not stable.

 

[John Mulcahy]

48k is a good starting point.

 

[johny_2000]

Today I reconnected the UMIK USB cable to different ports on the PC. Firstly, it will not work on my USB3 ports with an "unknown device" under Device Management in Win 10. Then I plugged it to another USB2 port and it will be recognized fine. I set the sampling rate to 48 kHz for both devices. After setting everything up with REW to measure my speaker system, I ran a few sweep tone tests and they all passed fine! I then ran a few more decades of sweeps all morning and all the timing error problems went away. The timing error was very stable at 19.1–19.2 ppm in all measurements. The ASIO buffer size was set to the default value of 512 samples with "Safe Mode" enabled.

Thanks again for your help! Now I can get reliable measurements using the UMIK-2 microphone and an acoustic reference to work with.