Hi. I'm experiencing some issues with 96KHz resampling.

I'm using Foobar2k in kernel streaming mode with 24-bit audio on an M-Audio Audiophile 24/96 soundcard. I'm running an Athlon 1800+ with a VIA motherboard (which I suspect is the problem).

At first I was having some serious distortion/fuzz issues with 96KHz. Couldn't figure out why. 48KHz sounded fine. I tried upping the DMA latency buffer on my soundcard and that didn't do anything at all. Someone suggested it was due to the shitty PCI bus on VIA chipset motherboards, so I searched and found a "PCI latency" patch which I installed and which seemed to fix my problem. 96KHz is definitely better after this patch... but not perfect. I get a strange "fuzz" in the left channel (using headphones) on high-frequency parts.

I'm using digital out on the soundcard -> high quality Bolder coaxial digital cable -> ART DI/O digital/analog converter -> Headroom headphone amplifier.

I'm guessing the problem is due to my motherboard/chipset, but ... I'm hoping it's a problem in foob2k or something.

If anyone has any ideas, I'd be grateful.

- Chris

96khz resampling is totally useless in this case (it would only be of use if your source material was >96khz and you'd only be able to output <=96khz).
Your sound quality does not get improved (even without the compatibility issues).

Why? When would 96KHz be useful?

See modified post... (You were quicker than me)

rpp3po

Not necessarily.

I can hear audible improvements in upsampling to 96KHz.

Some very good DACs also have upsampling capabilities (the MSB Link III DAC has an upsampling upgrade card for instance) and it definitely does improve the sound quality.

But the difference is negligible, so I'll live without it. Still, I'm curious as to why that problem persists...

- Chris

As far as I know, from a technical point of view these upsampling hardware is of no logical use. If your source material has got a limited sample rate (e.g. 44.1khz) your sample rate converter would have to establish a magical connection to the mastering studio to request the missing bits from the master recordings. If it's not IN the recording, you can't extract enhanced resolution out of your media even using "really advanced mathematics"...
Nevertheless there is hardware out there which really sounds different (and maybe sometimes subjetively better) when upsampling from 44.1 to 96. But in these cases the sound material is modified using psycho-acoustic models - in this case not for compression but for so called 'resolution enhancement'.
Everybody using this kind of hardware should be aware that this modification really adds stuff to the audio signal which originally was not in the recording. This subjectivly might sound better in some cases but possibly worse in others.
Would you convert your CD's to the analog domain and back digital, the resulting files, when using these 'resolution enhanced' DAC's, will always have bigger statistical difference to the original source than a good DAC working at the same sample rate as the source.

Foobar resampling is just plain technical resampling, no esoteric. So in your case it should not improve sound quality.

Though it may result in a smoother signal from a digital amplifier. Possibly comparable to using a tube amplifier. Hah, whatever.

I can't see why.

Although, from a theoretical point of view, upsampling can't do anything to improve the signal, there's still the possibility that some DACs perform substantially better at 96 KHz than at 44.1 KHz. So in this case high quality upsampling would be good to overcome poorer performance of the DAC at 44.1 KHz. But with good DACs this shouldn't be necessary.

Edit: removed extra irony.

Apologies for reviving an old thread. But I have a question:

If digital brickwall filtering is used, could resampling possibly move the filter higher in frequency (in case of 96KHz, up to near 48 KHz) and out of the range of human hearing? Some have complained that steep digital brickwall filtering is responsible for poorer sound... maybe that's one reason why resampling sounds better to some people.

Just curious about this, if it sounds possible...

No, whether you oversample-upsample or not, the signal has to be filtered below 22.050 KHz in order to remove images present over this frequency (fs/2). This filtering can be done in software (SSRC & fb2K), or hardware: note that *all* moderns DACs oversample internally, in theory same thing that so-called upsamplers are supposed to do.



In addition to what I already said, all 44.1 KHz data has been already brickwall-filtered at the AD stage, and no "upsampler" can reverse this fact.

The only difference a properly working upsampler could make, is use a more "sharp" brickwall filter. However, the differences, compared to any modern DAC using a somewhat less "sharp" oversampling filter, should not be audible (we would be talking about around 0.1 dB better frequency response difference at 20 KHz, and cleaner signal over 21 KHz). Also, FB2K (SSRC) resampling should cause same effect, since it's filtering is very sharp.

I think that those "upsamplers" either do nothing and just push placebo effect, or do something to the signal that intentionally causes an audible difference, as some "high-grade" cables and tube amplifiers do.

If you measure what's happening between 20kHz and 24kHz at the output of a typical (oversampled) DAC running at 44.1kHz, and the output of (say) foobar resampling the material to a higher sample rate correctly, there is a huge difference. The filter in a modern DAC is nothing like a brick wall!

We've discussed how this could possibly be (but probably isn't) audible many many times. (Anyone who missed it: see all the nyquist is wrong and CD isn't good enough threads in the FAQ).


As for the original problem: are you sure the OS isn't resampling it back to 48kHz before sending it to the sound card? It shouldn't do (if you select the sound card directly), but if you route it through DirectSound all sorts of strange things can happen. waveOut in fb2k works for me.

Cheers,
David.

Ok, I edited my previous post and changed "slightly cleaner signal over 21 KHz" for just "cleaner signal over 21 KHz"