Upsampling 16/44.1 collection a good idea?

[tellyoulater]

What do you mean?

I think antcollinet inferred my FIR filters were anti-aliasing filters and they're not. I correct a resonance around 3 KHz measured in anechoic chamber, apply a slight EQ on one of the room modes and I play with the phase.

 

[antcollinet]

I think antcollinet inferred my FIR filters were anti-aliasing filters and they're not. I correct a resonance around 3 KHz measured in anechoic chamber, apply a slight EQ on one of the room modes and I play with the phase.

OK fair enough - equalisation can improve the sound.

And in this case, upsampling to at least 24 bit is worth doing to avoid additional (audible) noise/distortion from the second quantisation.

It won't (on its own) make any improvement though.

 

[MaxwellsEq]

I think antcollinet inferred my FIR filters were anti-aliasing filters and they're not. I correct a resonance around 3 KHz measured in anechoic chamber, apply a slight EQ on one of the room modes and I play with the phase.

OK that makes a lot more sense!

This thread is about people who upsample 44/16 to possibly "improve" the sound, by using "better" reconstruction filters than are available in their DAC hardware or convert to DSD to perhaps avoid part of the DAC hardware. The goal is. "better" output than the original published content. There is much debate about whether the benefits exist or not

This thread is not about people doing speaker or room correction.

 

[Julf]

There is much debate about whether the benefits exist or not

Much debate, but very little facts and evidence.

 

[Brent71]

I've been thinking of upsampling my 44.1kHz collection to 88.2, 176.4kHz, 352.8, or even 705.6kHz for playback...

I use SRC resampler for Foobar2000 and have it set to upsample 4x except for 176.4, 192, 352.8 & 384. It obviously doesn't take up any more space on your HDD, but you still get any benefit of upsampling with your DAC.

 

[antcollinet]

but you still get any benefit of upsampling with your DAC.

Of which, there is none.

 

[MRC01]

I use SRC resampler for Foobar2000 and have it set to upsample 4x except for 176.4, 192, 352.8 & 384. It obviously doesn't take up any more space on your HDD, but you still get any benefit of upsampling with your DAC.

Of which, there is none.

Theoretically, the PC can use a higher quality resampling method than the DAC chip because the PC has more processing power.

In practice, how much better is it really? It's hard to find any evidence that oversampling before the DAC makes any audible improvement. Even measurable improvements (too small to be audible) seem to be rare and marginal. Especially when oversampling at integer multiples which is less computationally expensive

Put differently, the safe bet is that the DAC chip's built-in oversampling is plenty good enough. That seems to summarize the 16 pages of this thread.

 

[GXAlan]

Would you do it with your digital photos?

Well, there IS the drizzle algorithm.

Linear Reconstruction of the Hubble Deep Field

https://www.researchgate.net/publication/1840344_Drizzle_A_Method_for_the_Linear_Reconstruction_of_Undersampled_Images

But for audio, I wouldn’t do it

 

[AnalogSteph]

Put differently, the safe bet is that the DAC chip's built-in oversampling is plenty good enough.

At least as far as "proper" DACs are concerned and assuming some digital headroom is kept to be on the safe side. There's the proverbial exception to every rule if you keep looking for it long enough.

 

[dualazmak]

It looks the debate/discussion re-started again!

Let me share, therefore, at least my "practical" stance and policy as well as the rationales thereof, again here on this thread, even though I have already shared it in my post #39 on this thread.
- Summary of rationales for "on-the-fly (real-time)" conversion of all music tracks (including 1 bit DSD tracks) into 88.2 kHz or 96 kHz PCM format for DSP (XO/EQ) processing: #532

 

[JIW]

I use SRC resampler for Foobar2000 and have it set to upsample 4x except for 176.4, 192, 352.8 & 384. It obviously doesn't take up any more space on your HDD, but you still get any benefit of upsampling with your DAC.

If you don't give the interpolator enough headroom, you will get distortion from intersample overs as pointed out by John Siau. Once the distortion is in the signal, it cannot be removed.

Some DACs handle intersample overs well by having enough headroom at full volume while others can gain headroom by using the digital volume control of DAC chip if it is placed before any interpolator.

 

[antcollinet]

Theoretically, the PC can use a higher quality resampling method than the DAC chip because the PC has more processing power.

The theoretically part of that is obviously inaudible. Since the DAC without the PC upsampling, is. already putting out audible perfection.

 

[Julf]

Was this thread really worth resurrecting?

 

[Julf]

Theoretically, the PC can use a higher quality resampling method than the DAC chip because the PC has more processing power.

And theoretically the DAC might have to work harder to deal with the higher data rate, so the result might actually be worse. Again, purely theoretically.

 

[Julf]

I use SRC resampler for Foobar2000 and have it set to upsample 4x except for 176.4, 192, 352.8 & 384. It obviously doesn't take up any more space on your HDD, but you still get any benefit of upsampling with your DAC.

It doesn't take up any more space, but it takes up bandwidth and processing power of both the PC and the DAC and above all the connection between them. The DAC chips are optimized for signal processing - upsampling is what they do well, the bottleneck is often the handling of incoming data. Upsampling in the PC puts much more I/O load on the DAC.

 

[Mnyb]

The theoretically part of that is obviously inaudible. Since the DAC without the PC upsampling, is. already putting out audible perfection.

I think these PC up sampler and filters are used together with for example filterless R2R NOS DAC's .

You create a problem (filteterless NOS DAC's is a broken implementation , not up to "taste" it's just broken ) and then kludge around it with your PC you oversample high enough that the non existing filters in you broken DAC yield a very benign influence .

 

[boxerfan88]

PC side upsampling is really good, me thinks. It allows one to tinkle with one's own DIY filters. Or try those pre-programmed million tap filters like PGGB.

Secondly, most DAC upsample anyways as part of their conversion, and the DAC processing engine has limited horsepower. Which concerns me what sort of compromise they had to do with the filter design.

 

[Julf]

PC side upsampling is really good, me thinks. It allows one to tinkle with one's own DIY filters. Or try those pre-programmed million tap filters like PGGB.

It is good for tinkling. It doesn't improve audible sound quality (but can degrade it).

Secondly, most DAC upsample anyways as part of their conversion, and the DAC processing engine has limited horsepower. Which concerns me what sort of compromise they had to do with the filter design.

The DAC processing engine has enough *signal processing* horsepower. It is a DSP pipeline. What it struggles with is I/O processing. By upsampling in the PC you increase the load on the DAC.

 

[boxerfan88]

It is good for tinkling. It doesn't improve audible sound quality (but can degrade it).

Thus far the upsampled audio changes ranges from subtle to none. I have never encountered degraded audio quality due to PC upsampling.

The DAC processing engine has enough *signal processing* horsepower. It is a DSP pipeline. What it struggles with is I/O processing. By upsampling in the PC you increase the load on the DAC.

I disagree. Most DACs with good USB implementation can comfortably handle data rates at 384k some going up to 768k, no sweat at all. Most DACs cannot handle long FIR filters (except for the highest end ones) , lets say 64ktaps or more. I guess that they are in the 4-8ktap range, which limits LF resolution.

 

[Julf]

Thus far the upsampled audio changes ranges from subtle to none. I have never encountered degraded audio quality due to PC upsampling.

Let me guess - audio quality judged by sighted listening? It is funny how most tweaks and "improvements" lead to subjectively perceived improvements but no difference (or poorer performance) when compared properly in a double blind test.

I disagree. Most DACs with good USB implementation can comfortably handle data rates at 384k some going up to 768k, no sweat at all. Most DACs cannot handle long FIR filters (except for the highest end ones) , lets say 64ktaps or more. I guess that they are in the 4-8ktap range, which limits LF resolution.

How do you verify that they handle the data rates "comfortably"? Do you think the resolution limit from a 4k tap is audible?