How to do loopback test of DAC/ADC on macOS?

[Music1969]

I might have to re-do IMD plot.

Even though there's a hump (visually not good) the actual values at various levels don't look too bad

ADC plot below for UA X16, RME, MOTU, Lynx Hilo (I was doing loopback)

 

[Music1969]

Also my x-axis has made the hump look worse because I took more measurements at 2.5dBFS intervals in the hump region and 5dBFS for the rest, but haven't scaled it evenly in my axis axis. I'll re-do that. It was a very quick and dirty measurement just to see if there was a hump.

The hump will look less bad with a proper x-axis scale.

Matrix Element i has same DAC chip and has slight hint of hump visually, extremely slight.

 

[Music1969]

Taken at 2.5dBFS through full range this time. Instead of just in the problem region and 5dBFS in the rest.

I've got my scale out of whack and it's driving me nuts.

Still, I'm confident of 2 things:

1. The actual IMD_DIN values of mine are wrong

2. The shape of my curve (remember this is loopback) , especially shape/size of "hump" is correct.

As expected, the hump is still there.

But with accurate intervals and x-axis scale (even if absolute values are suspect) the hump is not as visually bad.

I feel a bit better ! LOL

 

[Music1969]

So I got the digital multimeter out to actually set to 4 Vrms analogue output and properly set the calibration/scale in REW.

I'm happier with these results now (in terms of numbers being much more accurate than earlier). Whether an APx555 would show a huge difference from my results, very possible still of course LOL.

I am more confident in these IMD_DIN numbers now and while the hump is still there (as expected) the values aren't that bad, compared to interfaces amir measured that don't have the ESS hump (including non ESS interfaces).

Definitely not state of the art but not broken (I think)

Not great but not broken (all these are DAC + ADC together)

 

[Music1969]

I saw your multitone @pkane for the UA interface and realised my input level was too high.

At lower level it looks better. This is going to 48kHz, 96k sampling rate

In audible band

 

[Blumlein 88]

I saw your multitone @pkane for the UA interface and realised my input level was too high.

At lower level it looks better. This is going to 48kHz, 96k sampling rate

View attachment 113239


In audible band

View attachment 113241

You are getting the right results now. I wondered about your first multi-tone results.

 

[Music1969]

For lol's

135 tones, 96kHz sampling rate, no filter, full span 0-48kHz

Still getting distortion free for roughly ~16 bits?

 

[Music1969]

No audible concerns for jitter but I expected cleaner.

If I had an APx555 it would probably be cleaner on the D to A but have to settle with loopback results I guess

 

[Blumlein 88]

No audible concerns for jitter but I expected cleaner.

If I had an APx555 it would probably be cleaner on the D to A but have to settle with loopback results I guess

View attachment 113268

I'm a little surprised. In loopback you aren't measuring jitter because presumably the same clock is used for both adc and dac. So even though jitter is present it will be the same in both and the result is artificially too good. Usually looking perfect around the central spike though it might have a few idle or constant noise tones in the FFT. Yours is showing the central jitter test tone with a little variation around the base. It really shouldn't be that way in loopback. You should expect that if connecting two devices with different clocks.

 

[Music1969]

Any idea what that those spikes at 3kHz, 6kHz, 12kHz, 15khz, 18kHz could be @pkane ?

48kHz jitter test.

 

[Music1969]

I'm a little surprised. In loopback you aren't measuring jitter because presumably the same clock is used for both adc and dac. So even though jitter is present it will be the same in both and the result is artificially too good. Usually looking perfect around the central spike though it might have a few idle or constant noise tones in the FFT. Yours is showing the central jitter test tone with a little variation around the base. It really shouldn't be that way in loopback. You should expect that if connecting two devices with different clocks.

Hmmm good points

Better view

 

[Blumlein 88]

In loopback would expect something more like this. 130k fft.

 

[Music1969]

In loopback would expect something more like this. 130k fft.

View attachment 113275

Yes the same as what I got, minus 3/6/9/15/18kHz spikes.

There's something oddly repetitive/consistent about those spikes I got. They're not random spikes (I don't think).

I don't know what happens at 3kHz multiples though. 48kHz sample rate.

It's like I'm running a 20/21bit test.

 

[Music1969]

So with 48kHz sample rate I get spikes at multiples of 3kHz

And with 44.1kHZ sample rate, I get the spikes at multiples of 2.75 kHz LOL

Below is using a 32 bit floating file I generated and then use Audirvana to playback at 24bit/44.1 kHz

 

[Music1969]

Interestingly I see @pkane 's loopback J-test has ticks at 3kHz, 6kHz, 9kHz also (low level) - 48kHz sample rate also. Different interface.

Do you know what the significance of these J-test spikes are @JohnPM ?

For me, with 48kHz sample rate I get spikes at multiples of 3kHz

And with 44.1kHZ sample rate, I get the spikes at multiples of 2.75 kHz

https://www.audiosciencereview.com/...nterface-quick-measurements.7652/#post-179444

 

[JohnPM]

Julian Dunn's paper describes the jitter test signal.

 

[Music1969]

Interestingly I see @pkane 's loopback J-test has ticks at 3kHz, 6kHz, 9kHz also (low level) - 48kHz sample rate also. Different interface.

Hi @pkane do you know what these (low level) spikes at 3kHz, 6kHz, 9k Hz are for 48k sample rate J-test?

And with 44.1k sample rate, are they then at 2.75kHz multiples also like I saw?

Do they correlate with something in particular? Or random?

 

[pkane]

Hi @pkane do you know what these (low level) spikes at 3kHz, 6kHz, 9k Hz are for 48k sample rate J-test?

And with 44.1k sample rate, are they then at 2.75kHz multiples also like I saw?

Do they correlate with something in particular? Or random?

Not sure. The frequency interval seems to be Fs/16. This equates to 3k for Fs=48k and 2.75k for Fs=44.1k. Could be the result of some quantization/truncation error in the ADC or the DAC. Might be worth investigating turning on dither in software or on the device.

Here's a digital loop-back recording of REW J-Test showing some side effects of computation, albeit at a much lower level. There's no ADC or DAC in the path, so no jitter possible. Just the loopback driver passing through 32-bit samples:

 

[Music1969]

Might be worth investigating turning on dither in software or on the device.

Can you do either of these in REW or your UA interface ?

For J-Test in REW i don't see this option.

 

[pkane]

Can you do either of these in REW or your UA interface ?

For J-Test in REW i don't see this option.

I don't think it's appropriate to dither J-test test signal It's supposed to be a perfect square wave. Can't test the UAD interface for now, but I thought turning dither on or off might help decide if it's a quantization problem. But it might be some internal clock or other signal causing it.