Just my guess, but I suspect that the upsampling filter setting you mention is likely simply repeating each sample. For illustration, if the original sample stream values are 32571, then with the x2 'upsampling' it would simply be, 3322557711. The same values repeated. If so, it's not performing signal reconstruction. It will appear as a double-rate digital signal on a DAC, but without any performance advantage over the original non-upsampled signal.
It just occured to me that if fed to a DAC with it's own digital image rejection filter, such a psuedo-upsampled signal would trick that filter into not removing the lower half of the first image band, while still removing the image products above that. In other words, it would raise the filter's cut-off frequency. The subjective sound may or may not be prove desirable.
Last edited:
Exactly. It is just repeating numbers. The benefit of this SRC method is we can trick DAC into believing he is receiving higher sampling rate signal to move cheap DAC filter into higher frequency. We can choose our own filtering method at nyquist, no filter, high quality FIR/IIR or analog filter.
This is the originally 44.1K now upsampled to 88.2K without interpolation in Audirvana. We can see this on Equilibrium and how custom designed FIR/IIR actually cuts the mirror image. You will clearly see the leaking, if your cutoff filter setting is too gentle.
It is utterly subjective what type of the filter is preferred, including no filter. 🙂
Last edited:
So long as the DAC's internal digital filter is not bypassed, it will still provide some image filtering, so, there's no true NOS option in that case. In the psuedo-upsampled mode you've mentioned, the lower half of the first image band will remain. Those lowest image frequencies are the most difficult to remove without use of digital filtering that cuts off at Nyquist. So, if the goal is to be rid of the sound character of a digital filter, and you're not concerned about removing the lowest image frequencies, why not just go NOS in the first place?
Right. The output of the DAC will be exactly the same in each case, regardless of the data rate at the input.
I'm curious which part of iZotope upsampling you call "psuedo". Upsampling is a purely mathematical process different from interpolation.
Please show me your NOS DAC measurement. I have a feeling that the modern DAC chips should have far superior objective performance than any NOS.
Why would you want to see a NOS DAC measurement? Practically everyone knows NOS DACs don't have SOTA measurements, that's not why we DIYers build them. Those of us who build them, build them for the sound, not the numbers.
I think a picture here would be worth a 1000 posts. What exactly does a 10kHz sine wave sampled at 44.1K and upsampled to 96K with no interpolation look like?
And now, I would just repeat the OP's question.
Has anyone actually compared Oversampled DAC without digital filter vs NOS?
Has anyone actually compared Oversampled DAC without digital filter vs NOS?
As what, the samples don't line up? There is no possible up-sampling from 44.1 to 96 without interpolation.
Last edited:
Before and after is the same. Upsampling is just repeating numbers and playback 2x or 4x as fast.
The part where the samples are simply repeated, as you agreed is what is being done. So, there is no math at all taking place, yes?
Oh, I see. Your question is 44.1 to 96K. I don't use that numbers. I only do 44.1 to 88.2.
- Status
- Not open for further replies.