Jitter of DAC. Or is it jitter?

[JohnYang1997]

As amirm has measured many DACs, some have good jitter some have poor jitter performance. Some shows a pattern of modulation distortion. What's going on?
44.1khz vs 48khz
The retro issue still exists.
Here I specifically use three DACs to demonstrate this issue.
Topping DX3 Pro
Topping D30
SMSL M3

The jitter test condition was J test which uses 1/4 of the sampling frequency, in this case 48khz. So basically amirm runs all DACs in 48khz mode and test 12khz signal and see if there's unwanted noises in the frequency spectrum. Theoretically the unwanted noises only happens with 12khz signal. If I play 12003hz there shouldn't be unwanted noise, or even 6666hz 16666hz etc.

Problem rises as D30 has great jitter performance with 44.1khz and 88.2khz and slightly better with 44.1khz AND DX3Pro has good performance with 48khz and 96khz and slightly better with 96khz.

 

[JohnYang1997]

Spoiler: D30 44.1khz j test

Spoiler: D30 44.1khz 6666hz

Spoiler: D30 96khz j test

Spoiler: D30 96khz 6666hz

 

[JohnYang1997]

Spoiler: DX3Pro 44.1khz j test

Spoiler: DX3Pro 44.1khz 6666hz

Spoiler: DX3Pro 96khz j test

Spoiler: DX3Pro 96khz 6666hz

 

[JohnYang1997]

Spoiler: M3 44.1khz j test

Spoiler: M3 44.1khz 6666hz

Spoiler: M3 96khz j test

Spoiler: M3 96khz 6666hz

 

[JohnYang1997]

Funny thing I noticed is that, m3 performed best here and it's the cheapest one here. Idk what's going on. Sure I modded the unit but that's only at the output stage. Shouldn't have anything to do with digital stuff.

 

[peterq]

Willing to see what you find, maybe some mod of DX3? From what you described, I think jitter here is different with what jitter I know in telecommunication.

 

[amirm]

What shows up in the output of running J-test or any other test tone is not necessarily jitter. That is why I often say there are spurious tones, noise, etc. For example, if the reference voltage is modulated, it too shows up as sidebands just like jitter does. Similarly if there is leakage of other periodic tones/timers and noise, they too bleed into the output.

To know if something is truly jitter, a) the sidebands need to be of equal amplitude and b) increase in level when the test tone is increased in frequency.

In other words, the "jitter" test is a catch-all test that while sensitive to jitter due to high frequency of test tone, also shows up any and all other sources of interference, modulation, etc.

Finally, what makes j-test good is that it doesn't have dither in it so it has a very low noise floor. That is because j-test is really a square wave so it is created with absolute fidelity (it is just a binary value). Sine waves on the other hand are floating point values converted to integer and then dithered so the noise floor is higher which may mask some distortions, spurious tones, etc.