Upsampling PCM or DSD in 2022


The purpose of this post is to ask the collective about the best options for upsampling today, and whether it’s worth doing. I stumbled into this topic after recently considering Paul McGowan’s take on DSD, and after reflecting on the upsampling in my home theater system.

Paul believes that DSD is world’s better than PCM. That caught my attention, because, until now, I have been operating under John Darko’s view that high res (i.e. 24-bit and above) is nice but not worth chasing--CD quality is good enough. But audio does seem analogous video. And 1080p isn't good enough for my video streams. So, I now want to give DSD and upsampling a shot.

In my home theater system, I use an Nvidia Shield TV streamer for its AI upsampling, driven by its graphics processor. Plenty of video content is still being released or only available at 1080p. Although upsampled 1080p isn’t as good as 4k, it’s better than basic 1080p. That upsampling makes a big difference for me. I strongly dislike watching 1080p content on my 4k TV. As far as I’m aware, the Nvidia Shield streamer offers the best video upsampling on the market, and it only costs >$200. It occurred to me that I might get similar gratification by upsampling audio too.

Upsampling can be performed at the DAC, streamer, server, or somewhere in between. Here are the major options I’ve considered so far:

  1. PS Audio’s Direct Stream DAC costs 6k. As an FPGA unit, it has lots of extra power that this manufacturer directs towards upsampling, and upsampling PCM to DSD is a major selling point for this device. Unfortunately, you have to get to the Direct Stream in the lineup to experience that feature.
  2. Chord’s Hugo M Scaler costs 5k. Although Chord builds FPGA DACs too, Chord sells a separate component for upscaling. In the audio chain, the M Scaler sits between a streamer and DAC. Because Chord separates out upsampling functionality into its own component, Chord’s solution is likely more expensive than PS Audio’s.
  3. HQ Player software costs >$300. HQ Player is a software service that can be installed on a server, or on a computer that sits between the server and streamer. Besides being affordable, you can pair HQ Player with Roon. The author of AudioBacon reports HQ Player introducing up to 30 seconds of lag to an audio stream when set to the most demanding upsampling algorithm even when used with a powerful Windows computer. But another commenter mentioned that his M1 Apple device introduced zero lag when running HQ Player.

Because the retail price of the components in my stereo system came out to about 5k (when new), HQ Player is where I’m looking for now. Please comment if:

  • You know about some other upsampling options I ought to consider;
  • You have opinions about the value of high res audio or upsampling; or
  • If you have anything you’d like to add to the conversation.
128x128classdstreamer

When I get a chance to listen critically, I'll post my findings here. I know in the past years casting Qobuz that I notice the difference between 16-bit and 24-bit content. But I haven't picked up on SQ improvements based solely on sample rate. This might be because bit depth variable has always been easier to test for. This will be my first time intentionally isolating sample rate. 

This leads me to another question, how might I go about upscaling the bit depth? (Is "upscaling" the right word here?) I can't easily attach screen shots here to show, but Roon's "sample rate conversion" area only seems to affect the sample rate, not the bit depth. 

FI know in the past years casting Qobuz that I notice the difference between 16-bit and 24-bit content. But I haven’t picked up on SQ improvements based solely on sample rate.

Yup. I notice a notable difference between 16/44.1 and 24/96 — not so much between 24/96 and 24/192. Not a fan of SACD. Something just sounds “over processed” with a resulting thin veil compared to PCM and hi-res PCM that sound more transparent. But, I’ll also say that well-recorded 16/44.1 stuff sounds just as good as some 24/96 stuff, so go figure. For me, I’d NEVER convert PCM to DSD — it just seems like more unnecessary processing and original signal degradation.

Anyway, there are many new DACs out today that are multi-bit R2R and non-oversampling (NOS) precisely because it avoids all the extra processing of 1bit and upsampling/oversampling DACs. Many people find these multi-bit DACs to sound more natural and analog sounding and less “digital” sounding. I’m personally agnostic and think either technology can sound great, but I recently acquired an R2R DAC and absolutely love it. In short, while upscaling in video yields pretty universally praised benefits, in 2-channel audio upsampling/oversampling and the brick-wall filters involved with single-bit DACs versus NOS multi-bit DACs is a bit more controversial. Single-bit devotees will tell you they measure better, and multi-bit fans will tell you they sound better — not dissimilar to the ongoing tube vs. solid state debate. Pick your poison, but I just wanted to point out the relatively recent popularity of R2R and NOS DACs (some R2R are NOS and some are not) that fly in the face of the upsampling/oversampling tech. To me, there’s a reason there are more and more R2R/NOS DACs hitting the market. Sorry to maybe complicate this for you, but hope you find this helpful and best of luck in your search.

Don't  care for converting pcm to dsd with my setup and Roon. Staying with pcm for pcm recordings most revealing and tranparent, dsd conversion sounds more like hifi, smooths things over, micro dynamics lost, sense of live performers in room diminished. YMMV.

 

And yes, do the fiber conversion, you may find you don't need or care for extra processing after this.

 

My take is processing makes noise, the more processing the more noise. Many off the shelf servers don't have the computing power to run high level processing, one should ask themselves why. Is it because they've determined costs are greater than benefits?

In my many years of experience with DACs and digital formats DSD is indeed best. And PS Audio's Direct Stream is the best sounding DSD DAC I've heard. Almost always a noticeable improvement of PCM files using it compared to directly through a high quality PCM DAC (both with and without the fiber conversion trick, which I also recommend).

I also highly recommend their SACD transport using I2C to send the DSD layer directly to their DAC.