HydrogenAudio

Hydrogenaudio Forum => General Audio => Topic started by: NateHigs on 2021-04-22 11:17:36

Title: Measuring Jitter
Post by: NateHigs on 2021-04-22 11:17:36
Can anyone point me towards a guide for measuring jitter in an audio interface? Is a scope necessary? I've got a handheld one, but I think it lacks the needed features.

I have an audio interface that can run natively at difference clock speeds, I'm interested to see what one (if any) offers the least jitter.

Thanks
Title: Re: Measuring Jitter
Post by: NateHigs on 2021-04-22 12:10:08
So, I've done some very basic testing. I think these noise peaks are showing jitter (at least some of them are). Interestingly, the noise floor is different depending on sample rate.

These tests were performed by generating a 440Hz sine wave at all 4 sampling rates, creating a recording loop back into the device, and recording the result. Unfortunately, the loop contains some hardware (a hardware compressor that was bypassed), so that could add some noise I would image. Connection are balanced.

I think the result at 44.1 looks really good, but I would love some feedback from people. Maybe I'm off base now, I don't know what a modern audio interface is capable of.

edit. added silent recording at 44.1. Some psi noise (50hz and harmonics)
Title: Re: Measuring Jitter
Post by: saratoga on 2021-04-24 00:45:55
I think you probably cannot measure jitter in loop back since both the A/D and D/A are likely to be using the same clock and so any jitter introduced by one will be cancelled out by the other.
Title: Re: Measuring Jitter
Post by: Roseval on 2021-04-24 10:08:59
I recommend using the  J-test if you want to measure jitter: https://www.nanophon.com/audio/diagnose.pdf
Have a look here http://archimago.blogspot.com/2018/08/demo-musings-lets-listen-to-some-jitter.html and other post for how to measure jitter.

Title: Re: Measuring Jitter
Post by: NateHigs on 2021-04-24 10:56:14
I think you probably cannot measure jitter in loop back since both the A/D and D/A are likely to be using the same clock and so any jitter introduced by one will be cancelled out by the other.

My assumption would that it would be doubled... I don't know tho

I recommend using the  J-test if you want to measure jitter: https://www.nanophon.com/audio/diagnose.pdf
Have a look here http://archimago.blogspot.com/2018/08/demo-musings-lets-listen-to-some-jitter.html and other post for how to measure jitter.

Thank you very much, I will investigate!
Title: Re: Measuring Jitter
Post by: saratoga on 2021-04-24 17:42:41
^^^Jitter is a change in clock frequency.  You measure it by playing  a tone with one clock and recording  it with a second more accurate clock and then looking  at how much the frequency changes. If there is only one clock you would  be subtracting that clock from itself which will give zero difference.
Title: Re: Measuring Jitter
Post by: NateHigs on 2021-04-27 19:53:09
^^^Jitter is a change in clock frequency.  You measure it by playing  a tone with one clock and recording  it with a second more accurate clock and then looking  at how much the frequency changes. If there is only one clock you would  be subtracting that clock from itself which will give zero difference.

I forgot to thank you for this comment. Explanation makes perfect sense. I do have another interface in which runs a separate clock. Maybe I’ll mess around with that.
Title: Re: Measuring Jitter
Post by: splice on 2021-04-28 00:28:43
The use of the same clock when looping D-A to A-D will only cancel if the same clock tick does the D-A and A-D of each specific sample. In practice there are latencies due to different digital path lengths in the D-A and A-D. So I would expect to see some jitter, but it would be a mixture of the clock jitter at the two different times.
Title: Re: Measuring Jitter
Post by: saratoga on 2021-04-29 20:55:23
The use of the same clock when looping D-A to A-D will only cancel if the same clock tick does the D-A and A-D of each specific sample.

No, that isn't necessary so long as they derive their sampling rates from the same oscillator, which almost all practical devices are going to do since it wouldn't make sense to have multiple crystals on a typical device. 

In practice there are latencies due to different digital path lengths in the D-A and A-D. So I would expect to see some jitter, but it would be a mixture of the clock jitter at the two different times.

Absolute latency doesn't affect jitter, so you don't need to worry about that. 

You will of course always measure nonzero jitter since getting a true zero would imply a noiseless system which is not possible.
SimplePortal 1.0.0 RC1 © 2008-2021