Bit-perfect and PCM vs DSD: which digital format claims actually matter

A reader emailed me last month asking whether he should re-buy his entire Miles Davis collection in DSD because a forum post convinced him his FiiO K11 was "wasting" his SACD rips. He'd already spent four evenings converting files. The honest answer took me about ninety seconds to type back: no, and here's why, and also stop trusting forum posts written by people who've never run a null test.

This is the conversation I have more than any other when I'm not talking about earpads. Format claims are where digital audio gets religious, and where the gap between what's technically true and what's audibly true opens widest. So let's go through the actual claims — bit-perfect playback, PCM versus DSD, hi-res sample rates, MQA's legacy — and sort what changes your sound from what changes your bank balance.

What "bit-perfect" actually means (and why it's not optional)

Bit-perfect just means the digital data leaving your source and arriving at your DAC's input is identical, bit for bit, to what's on the file. No resampling, no volume-control attenuation in the digital domain, no format conversion the player decided to do on your behalf. This one is not a maybe. If your playback chain isn't bit-perfect, you are, by definition, not hearing the recording — you're hearing a processed version of it, even if the processing is inaudible.

The usual culprit is an operating system's shared audio mixer grabbing the stream to apply system sounds or sample-rate conversion. On Windows this shows up as WASAPI exclusive mode versus the shared mixer; on macOS it's Core Audio's mixer doing the same thing unless you're running something that bypasses it. Most decent playback software — Roon, Audirvana, the manufacturer apps for streamers like the WiiM Ultra (check price) — has a bit-perfect or exclusive mode toggle, and you should have it switched on. This is genuinely free, it costs you nothing, and it removes a real variable. I'd put it in the "no-brainer" bucket alongside using a proper USB cable rather than blaming the cable for everything else.

Where it gets murkier is when people extend "bit-perfect" into "therefore every bit-perfect source sounds the same." That's not automatically true — jitter, clocking, and the DAC's own implementation still matter, which is a topic I've gone through in more detail when I wrote about DACs and network streamers. Bit-perfect is a necessary condition, not a sufficient one.

PCM versus DSD: two philosophies, one destination

PCM (pulse-code modulation) samples amplitude at regular intervals — 44.1kHz/16-bit for CD, up through 192kHz/24-bit and beyond for hi-res files. DSD (Direct Stream Digital) is the format Sony and Philips built for SACD; instead of multi-bit samples it uses a single bit at a very high sample rate — 2.8MHz for DSD64 — relying on noise-shaping to push quantisation noise up out of the audible band.

The marketing pitch for DSD has always been that it's "closer to analogue" because it dispenses with the ladder of discrete steps. I'll be blunt: I don't think this holds up in practice for most listening. Every DSD-capable DAC I've measured or read measurements of ultimately converts DSD to PCM internally, or filters it in ways that make the "purity" argument largely theoretical. Chord's DACs famously do this differently again, with their own custom filtering rather than an off-the-shelf chip, and even among people who love that approach, nobody's claiming DSD itself is the reason.

What I will say, because I think it's the more useful and slightly contrarian point: the difference you hear between a PCM remaster and a DSD/SACD reissue of the same album is overwhelmingly down to the mastering, not the container. Older SACD reissues frequently used better tape sources, more careful engineers, and less compression than the CD pressed a decade earlier. Compare the two blind and the DSD version often does sound better — but the format isn't why.

Hi-res sample rates: where returns actually diminish

The Bluetooth SIG and various streaming services love a big number, and 192kHz/24-bit sounds twice as impressive as 96kHz/24-bit on a spec sheet. The engineering reality, per the Nyquist–Shannon sampling theorem, is that 44.1kHz already captures everything up to 22.05kHz — well beyond confirmed adult hearing, which the AES and most peer-reviewed hearing research puts at a ceiling nowhere near that for most people over 25. The extra headroom in 96kHz and 192kHz files buys you gentler anti-aliasing filter slopes near the audible band, which is a real, if subtle, engineering benefit. It does not buy you information your ears can use directly above 22kHz.

Where hi-res genuinely helps is bit depth, not sample rate. Going from 16-bit to 24-bit lowers the noise floor by roughly 48dB in theory, which matters far more for mastering headroom and dynamic range than anything above 20kHz does for playback. If you're choosing between a 96kHz/24-bit file and a 192kHz/24-bit file of the same recording, I wouldn't pay extra for the second. If you're choosing between 16-bit and 24-bit of a well-mastered recording, 24-bit is the one worth having, mostly because it usually indicates the label didn't slam the master through a brickwall limiter to make the CD version loud.

I tested this properly a few years back with a small level-matched ABX session in my Adelaide studio, using a RME ADI-2 DAC FS (check price) for its measured transparency and its built-in loopback for level matching. Across a handful of tracks and several listeners including two mastering engineers, nobody reliably picked 96kHz from 192kHz of the same file. Everybody picked the 24-bit version over a heavily limited 16-bit one, instantly.

MQA, in the past tense now

I'll keep this brief because it's largely a historical note rather than a live buying decision. MQA promised "origami" folding of hi-res data into a smaller file with authenticated provenance, using a specific temporal-smearing correction filter as the headline audible claim. Whatever your view on whether that filter did anything you could hear — and Archimago's and other measurement-focused blogs spent years picking this apart — the commercial reality is Tidal has moved on and MQA's business went through insolvency proceedings. I mention it because readers still ask, and because it's a useful cautionary tale about proprietary formats generally: don't build a collection around a codec a single company controls unless you're comfortable with that company's fortunes becoming your fortunes.

Where the DAC and amp stage actually earn their keep

Format arguments distract from the parts of the chain that reliably do change what you hear: the DAC's output stage, its behaviour into your headphones' actual impedance, and whether you're running balanced or single-ended. I've written before about how a DAC's output impedance interacts with your headphones more than the DAC chip itself does, and that's a far bigger lever than PCM versus DSD will ever be. If you're assembling a desktop rig and trying to decide where the money should go, I'd put source format dead last on the priority list, behind headphone fit, amplifier matching, and room noise floor if you're at all close-listening.

For a practical desktop chain, something like the iFi Audio ZEN DAC 3 (check price) or the FiiO K11 (check price) will play back DSD, MQA-derived hi-res, and standard 16-bit files with no meaningful difference in the DAC chip's contribution to the sound — the differences you'll actually hear come from the amp section and how it drives your specific headphones. If you're weighing up a full desktop rig from scratch, I've laid out the buy order in my piece on the best desktop DAC/amps, and it might save you buying the wrong thing first.

A practical checklist that isn't a checklist

If I had to boil this down for the reader with the Miles Davis SACDs: keep your player in bit-perfect or exclusive mode, don't re-buy an album purely to get it in DSD, do pay attention to whether a reissue is remastered from better tape, and stop worrying about 192kHz unless storage space is free and you like big numbers on a screen. Spend the money instead on getting your headphone-amp pairing right — that's the piece of this chain that's actually audible, every time, to everyone in the room.

My colleague Theo covers the deeper engineering side of jitter and clocking if you want the electronics-first version of this argument; I'm coming at it from the listening chair, and from that chair the format war has always been the least interesting fight in the room.

What I'd actually tell you to do this weekend

Pull up your existing library, check your playback software is set to bit-perfect, and stop there. Don't convert anything. If you want to spend money on sound quality this month, spend it on a proper listen-and-return trial of headphones that actually fit your head, because that variable dwarfs every format debate we've just been through.