It isn't the bits, it's the hardware

From a purely technical standpoint oversampling can apply to DA conversion, not just ADC, so from that standpoint, you can use upsampling, oversampling or sample rate conversion to a higher frequency all interchangeably. Feel free to validate that with DAC data sheets that discuss oversampling.


But looking more at the (poorly) written paper linked attempting to compare a readily used term, over-sampling, to one practically made-up at least for this case (upsampling), and then to actually not really give any definition to upsampling except to define it pretty much exactly as asynchronous sample rate conversion, another well understood term, I am not surprised by the confusion.

erik_squires: "Actually that’s exactly how it works for upsampling, but different upsampling algorithms work differently. With the advent of cheap compute, Bezier curves are cheap and easy to do. "

I think you are missing a key element of how a typical asynchronous sample rate converter with inherent over-sampling works, namely that the first step would be an implementation of oversampling (typically fractional delay filters), which provides a smoother curve for the curve-fit which works over a smaller number of samples. Doing this keeps the spurious frequency components higher up allowing for easier final filtering.

Here’s an interesting article I ran across at Benchmark Media, I quoteth the relative part for this conversation:

An examination of converter IC data sheets will reveal that virtually all audio converter ICs deliver their peak performance near 96 kHz. The 4x (176.4 kHz and 192 kHz) mode delivers poorer performance in many respects.

The full article:

https://benchmarkmedia.com/blogs/application_notes/13127453-asynchronous-upsampling-to-110-khz

This again supports my hypothesis, that the converters themselves perform differently, it’s not just the data.

erik_squires

... with upsampling, you are not generating more data ...

Correct.

There’s no more clarity or resolution, or harmonics ...

Not necessarily, although if present, it would not be a consequence of more data, but more likely attributable to filtering, as others have noted.

Your mistake here is confusing correlation with causation, a common audiophile logical error.

But looking more at the (poorly) written paper linked attempting to compare a readily used term, over-sampling, to one practically made-up at least for this case (upsampling),

No, not at all.  This is not the only paper, and to claim it is is selective reading.

Upsampling and oversampling have long been quite clearly understood in the industry to mean two different approaches to the filtering problem. Only the poorly informed believe otherwise.

The former (upsampling) attempts to extrapolate new data points, whether by linear interpolation or by curve fitting. The latter replicates the data, to the rate at which data is received is now higher, but the amplitude is identical. That is, with 4x oversampling, you duplicate the same 16 bits.  With upsampling you do not.  Neither requires ASR.

And... I'm done. :) While you make good cases for the filter behavior being similar, and it is, this argument alone has already gotten us far from fact based, and my patience for that is now zero.

Buh bye.

In this limited industry somewhat exclusively does upsample mean resample by async sample rate converter (and it was pretty much a made-up term), which most people know, but what they don’t know is that the underlying technology which is pretty much exclusively some form of synchronous oversampling in the form of fractional delay filters, which provides an underlying shift upwards in the spectra from the original sample rate, coupled with a time compensated curve-fit which provides for the final sample rate and provides the jitter attenuation (something not needed with async streaming sources of course). So when claiming advantage of upsampling to a higher frequency, is it the inherent oversampling, the jitter reduction, or the pick of final sample rate, or some combination of?


As Cleeds pointed out, you can’t make a generalization to all cases based on one example.

That Benchmark found that "performance" based on data sheet and some simple performance metrics was better at 24/96 than 24/192 is not at all surprising. At lower speeds, you have less contribution to the output from the switching CMOS switches, less dynamic power (and less glitch energy) contributing to a quieter environment, and even simply more time to settle to the final value. Perhaps in Benchmark’s specific case, decoupling the output frequency from the input also removed sources of synchronous noise. Of course, they are running in 2x mode anyway, so technically the DAC is running close to 192KHz internally, so what exactly does that 96KHz even mean in their argument ... not to mention that it then runs into a much higher rate sigma-delta modulator. Does their product match the data, or does the data match the product ?


However, the items they cited w.r.t the data sheets and their tests, are not guarantees of excellent perceived sonic performance which would trade off high sample rate, system noise, with analog filtering. The article is also 10 years old, so what was best 10 years ago, may have shifted up 2x or more in terms of what was best.

I don’t think you have well made your point, simply because tests have been done with 24/96 native, and 24/96 down-sampled mathematically to 16/44.1 and then upsampled to 24/96 so that the playback path was identical, all that was changed was the information rate.