Up-sampling to higher DSD rates - what is this wobbling?


Hi @tedsmith

I was just reading this article:


And this part caught my interest (more than the rest):

Going from PCM > DSD64 > DSD128 the writer notes (and shows) this wobbling - what is it and why does it look worse going to higher sample rates?

He does say it’s not audible but I’m still interested why going to higher DSD sample rates creates more wobbling.



As the caption states it’s the ultrasonic noise. If you want it to be smooth then filter it with the lower cutoff filter than is used for PCM then you’ll see that DSD can be more accurate than PCM over the audio band - of course it won’t sound as good as using the gentler filter that’s possible in DSD.


Thanks Ted!

When you say “that’s possible in DSD” in this part quoted, you mean the DirectStream DAC here?

Where the gentler filter can be used because of the DS DAC’s very very high up-sampling rate?


In general the purpose of upsampling is that you allow a simpler gentler slope analog filter to be used to get rid of any aliasing. For Redbook with no upsampling you need a filter that goes from full level at, say, 20k to next to 0 at 22.1k. That’s a very steep filter and it not really possible in analog without lots of compromises. If you upsample even just by 2, then you can go from full level at, say, 20k to 0 at 68.2kHz (88.2k - 20k). 4 times give you a final cutoff at 156.4k (4 * 44.1k - 20k)… There was nothing specific to the DS in my first comments. DSD allows a much gentler anti aliasing filter than PCM simply because it’s oversampled to a higher rate. There is a rise in ultrasonics if the filter has a high frequency cutoff or a shallow slope.

The high internal oversampling in the DS isn’t directly related to that - it just allows upsampling, say, 352.8k by a small integer factor. Going to quad rate DSD allows the ultrasonic energy to be spread over a wider frequency range and hence also lowering the amount of ultrasonic energy that’s not being filtered out.