PCM vs DSD?

@Logic Gates thank you for the clarifications, there is still a lot I have to learn on this topic. Still, my experience is that built-in interpolation filters for PCM perform worse than HQPlayer oversampling PCM/DSD. There is also one correction I'd like to make, which is:

This may be the case with some D/S DACs, however, with other DACs there are ways around the internal interpolation filters:

• Some R2R DACs can run in NOS mode and bypass the internal interpolation filters entirely. Both of my DACs have NOS modes that allow HQPlayer to do the oversampling and dithering.

• Some AKM DACs allow DSD to bypass the internal interpolation filters and play directly.

HQPlayer can upsample PCM up to 1.4112/1.536 MHz, although Holo Audio DACs are the only ones I'm aware of that can accept such sampling rates. My Gustard R26 takes up to 704.6/768 KHz, whereas the Bifrost 2/64 only accepts up to 176.4/192 kHz.

 

the ds dacs massively upsample. the earliest ones did so 64. the current ones x128 for the cheapo ones to x512. they are oversampling well into the mHz. Some pathetic upsampling in some meme media player will not mitigate this.

1-bit modulators are never used in proper converters. they do not work mathematically.

linear phase filters are almost always used! the only thing that is not minimum phase that is used is the first order low pass filter after the massive oversampling, decimation, and noise shaping. only horrible prosumer and heavily compromised equipment uses minimum phase anti-alias filters in 2024.

Stop using usb bullshit and get a real thunderbolt converter, pro tools HDX rig with the dad converters, or buy a small mixer or console

 

Perhaps I missed something, but I'm confused about this statement. What is wrong with USB?

 

It's not SPDIF.

 

High latency. Neither most windows computers nor purposefully configured macs are suited for low latency audio and video anyway.

 

There are some usb interfaces with a blend function and good sound (spl, Neve) and others with a digital mixer and bad to great sound (rme, UAD Apollo, apogee, Lynx) that you can use for low latency monitoring though

 

Does SPL still make recording interfaces? The Crimson is discontinued, and there's stlll the Marc One but that almost feels like an afterthought and you need a separate mic pre.

 

yeah it's gone but is sometimes cheap and will always work on mac.

 

Turns out there's actually a Crimson 3 at a pawn shop local to me, but they are asking $899 CAD which seems rather expensive for an old discontinued unit.

 

I didn't know there are DACs on the market that not only allow the user to switch filters but to also bypass them entirely.
I proposed that you could upsample the signal to "maybe 384kHz" because that's the highest sample rate the vast majority of DACs would still accept as an input. I imagine the Gustard and the Holo DACs are more of an exception and the Schiit DAC is the norm. An oversampling DAC would still use a much higher sample rate internally than what HQPlayer upsamples to. And they would have to oversample the digital input anyways unless you are allowed to bypass it as you noted.

I know they shouldn't be used. I noted that they typically use multi-bit modulators. I think oversampling DACs were still sold with 1-bit modulators back when this was still a novel concept.

I've measured the outputs of some audio interfaces in the past and they all had minimum phase interpolation filters. The caveat is that they are pretty much the kind that you mentioned: they are either aimed at the budget prosumer market or they are ~20 year old pro interfaces. This combined with the fact that a reasonably long symmetric FIR filter would add considerable amount of latency to the output makes me think that only minimum phase filters are used for interpolation as well, not just for reconstruction.

How sure are you about the linear phase interpolation filters used? Have you measured the output of some audio interfaces? Is there a brand that includes the type of interpolation filters used in their specifications? Do you happen to know the typical length of these filters?
The "audiophile approved" linear phase interpolation filters' favorite marketing talking point is their ~100k to million taps like the ones included in HQPlayer or the ones used in Chord products. Nothing of these sort of filters could make it into an audio interface. Are there are interfaces that employ sane linear phase filters? A good attenuation could be easily reached with just 1k taps without any fancy filter design tricks but that would still be 500 samples of latency added for a relatively small benefit.