A 24-bit file plays as 16-bit when any link between the file and the converter can only carry 16 bits. The player, the Windows mixer, the USB sound device, the cable and the receiver each have a format they accept, and if one of them offers 16 bits the lowest 8 are simply dropped. Nothing breaks and nothing sounds wrong, so usually nothing warns you. To check, set the device’s Default Format in Windows to 24-bit, use a player with an exclusive (bit-perfect) mode that reports the format the device actually accepted, and pick hardware that genuinely takes 24-bit. For most listening, the lost bits are below anything you can hear. For archives and transfers, you want them intact.

How the link works

The USB sound card below is linked to Amazon.com.au with an affiliate link, marked “(paid link)”. As an Amazon Associate, Digital Greenhouse earns from qualifying purchases. You pay the same price either way. It’s the card we bought with our own money and use ourselves.

What happened on our bench

We’re building Ironbark, our own FLAC player for Windows. It isn’t publicly released yet. One of its jobs is bit-perfect playback: handing your sound device exactly the numbers that are in the file, then proving it. Ironbark fingerprints every block of samples it sends and compares it with a second, independent reading of the file.

On 12/10/2026 John played a 24-bit, 96 kHz FLAC through a DSD TECH SH-AU02A, a small USB sound card with optical and coaxial digital outputs, feeding a Marantz AV receiver. Ironbark put up a warning: the device was taking 16-bit samples, so this 24-bit track was losing its lowest bits on the way, and the playback was not bit-perfect.

The warning was right. Here’s what was going on.

A 24-bit sample is 3 bytes. Computers prefer to move numbers around in 4-byte boxes, so the usual way to send 24-bit audio is in a 32-bit container: the 24 bits of sound, plus one empty byte. Ironbark asked Windows for exactly that. The card is built on C-Media’s CM6542, a chip that follows the older USB Audio Class 1 standard, and it said no. It accepts 24-bit audio only packed, three bytes to a sample, with no empty byte.

When a device refuses the 32-bit box, the open-source sound library Ironbark uses (PortAudio) works down a list of fallback formats. On that list, 16-bit comes before packed 24-bit. The card accepted 16-bit, so it was opened at 16 bits, every sample was cut to its top 16, and the music played on as normal.

John measured it directly, asking the card what it had agreed to after each kind of request, at both 96 and 44.1 kHz:

Ironbark asked forThe card ran at
24-bit in a 32-bit container16 bits (refused, fell back)
24-bit packed in 3 bytes24 bits
16-bit16 bits

The fix, in Ironbark 0.46.1 the same day, was to stop trusting the format it asked for and check the one it got. If a device lands below 24 bits, Ironbark now tries packed 24-bit as well and keeps whichever carries more bits. Its Details window names the result (“96 kHz 24-bit exclusive”). With that version installed, the DSD TECH line read 24-bit · bit-perfect ✓. A 16-bit CD rip stays bit-perfect this way too: its 16 bits go out with 8 zeros underneath, which is exact.

And the far end agreed: the Marantz receiver’s 96 kHz indicator lit up. Optical and coaxial carry the same S/PDIF signal, so it makes no difference to any of this which of the card’s two digital outputs you use.

One 24-bit sample three waysThe same 24-bit sample drawn as bytes. First in a 4-byte, 32-bit container: three bytes of sound and one empty byte, which the sound card refused. Then packed into 3 bytes, which the card accepts. Then cut to 16 bits: the top two bytes kept and the lowest 8 bits lost.most significant → least significantAsked for: 24 bits in a 4-byte (32-bit) containertop 8middle 8lowest 8empty bytethe card refusedit fell back to 16Packed: 24 bits in 3 bytestop 8middle 8lowest 8what the card takes24 bits ✓Cut to 16 bits: what the player gottop 8middle 8lowest 8 lostlowest 8 bits dropped16 bits, no warningThe top 16 bits are the same in all three; the cut drops only the lowest 8.
Same sample, three packings. The card refused the 4-byte box, so the player's fallback cut every sample to its top 16 bits. Packed into 3 bytes, all 24 arrive. Bytes drawn from most to least significant.

Two lessons came out of it. The card wasn’t faulty, and Windows wasn’t misbehaving: every part did what it was designed to do. And the only reason we knew was that the player checked the format it had actually been given, which most players don’t do.

Every place the bits can go

The path from a FLAC file to your ears has more steps than it looks, and any one of them can be the narrow point.

The path from a 24-bit file to the receiver, and where bits can be lostA 24-bit FLAC file goes to the player, then through Windows either in shared mode (the mixer and the Default Format) or in exclusive mode (skipping the mixer), then through the USB driver and sound card, along the S/PDIF cable and into the receiver or DAC. Numbered markers show each place the audio can be cut to 16 bits, and the receiver, which often shows the sample rate but not the bit depth.24-bit FLAC fileall 24 bitsPlayerdecodes, picks a formatWindowsShared modemixer → Default FormatExclusive modeskips the mixerUSB driver + sound card24-bit in 3- or 4-byte slotsS/PDIF cableoptical or coaxialReceiver or DACturns it into sound1234561The playermay output 16-bit by default, orquietly accept a 16-bit fallback2Windows mixer (shared mode)mixes everything, then convertsit to the Default Format: maybe 163Exclusive modeskips the mixer, but has to use aformat the device lists4Driver and USB format24-bit in 3- or 4-byte slots: askingfor only one can miss the other5S/PDIF cableroom for 24 bits, but a chip thatonly does 16 fills just 166Receiver or DACoften shows the rate, not the depth:a 96 kHz light isn't proof of 24
Six links, any of which can be the narrow one. Red markers are the places 24-bit audio can be cut to 16 without a warning. The amber one is the receiver, which usually can't tell you either way.

1. The player

A player decodes the FLAC into samples and chooses a format to hand to Windows. Many players work internally in 32-bit floating point, which holds 24 bits comfortably. The risk is a player that outputs 16-bit by default, or one that asks for a format the device refuses and quietly accepts whatever it falls back to, which is what caught us.

2. Windows’ mixer (shared mode)

Normally, Windows mixes the sound of every program together, so a notification ping and your music can play at once. Microsoft calls this shared mode. The audio engine converts everything to floating point, mixes it, and converts the result to the device’s Default Format, which you choose in the Sound control panel. Microsoft’s own documentation uses the example of a device set to “2 channel, 16 bit, 44100 Hz (CD Quality)”: the engine converts its mix to 16-bit integers before playing it through the device.

So if a device’s Default Format is 16-bit, every program’s sound reaches it as 16-bit, your 24-bit album included. If the Default Format’s sample rate differs from the file’s, Windows resamples it too. Neither shows up anywhere in the player.

3. Exclusive mode, and the device’s own list

Exclusive mode (WASAPI exclusive, in Windows terms) skips the mixer. One program has the device to itself, and other sounds go quiet while it plays. In exclusive mode, Microsoft says, the program must use a format the hardware explicitly supports, and the data goes to the device “without intervention by the audio engine”.

That’s why players use it for bit-perfect playback. But it moves the question rather than answering it: now the program and the device have to agree on a format, and the device’s list of accepted formats may be narrower than its box suggests. Some take 24-bit only at certain sample rates. Some, like ours, take it only in one packing. What the program ends up with depends on how carefully it negotiates.

4. The driver and the USB standard

USB sound devices describe their formats in a standard way. The older USB Audio Class 1 lets a device say “24 bits of sound, in 3-byte slots” or “24 bits, in 4-byte slots” (the class’s format specification allows slots of 1 to 4 bytes). Both are legitimate. A program or library that only ever asks for one of them will miss the other. Class 1 devices like ours run at USB’s “full speed”, which is plenty for stereo at 24-bit and 96 kHz, but it’s one reason many older or cheaper devices stop at 48 or 96 kHz.

5. The S/PDIF cable

Optical (TOSLINK) and coaxial digital outputs both carry S/PDIF, a link from the CD era (more on the plugs). Each sample travels in a slot with room for 24 bits of audio: 20 bits as standard in the consumer version, plus 4 “auxiliary” bits that can extend it to 24 (IEC 60958-3, the consumer standard). So the cable can carry 24-bit. Whether a given device fills those slots is down to its chip. Some cheap USB-to-optical dongles are built on chips that are 16-bit only: Texas Instruments’ PCM2903C, for example, is a 16-bit, 48 kHz-maximum USB audio chip with an S/PDIF output, and it’s the chip in DSD TECH’s own cheaper optical adapter, the SH-AU01A.

6. The receiver or DAC

At the far end, an AV receiver or DAC takes the S/PDIF stream and converts it to sound. Most will happily play whatever arrives. Many show the sample rate, but bit depth often isn’t shown at all. The Marantz in our setup has a 96 kHz light on its front panel and no bit-depth light, and even Marantz’s current flagship processor lists the sample rate, not the bit depth, on its information screen. A “96 kHz” light tells you the rate made it through. It says nothing about whether 24 bits or 16 did.

Why nothing warns you

Because nothing has gone wrong, as far as each part knows. Cutting a 24-bit sample to 16 bits doesn’t make an error, a gap or a click. Each step accepts a valid format and passes on a valid signal. The only thing that can notice is something that compares the start with the end, and almost nothing in a normal setup does.

How to check your own setup

In Windows

  1. Open the classic Sound control panel: right-click the speaker icon in the taskbar and choose Sound settings › More sound settings, or press Win+R and type mmsys.cpl.
  2. On the Playback tab, select your device and click Properties, then Advanced.
  3. Default Format shows what Windows sends the device in shared mode. Open the list: if it offers a 24-bit option at your music’s sample rate, choose it. If every option is 16-bit, the device (or its driver) only offers 16-bit in shared mode.
  4. On the same tab, make sure Allow applications to take exclusive control of this device is ticked if you want to use a player’s exclusive mode.

In Windows 11 you can also get there from the Settings app: System › Sound, then More sound settings under Advanced (Microsoft’s steps).

We haven’t yet written down which formats our own DSD TECH card lists here; we’ll add them to this page when we have.

In the player

Use a player with an exclusive or bit-perfect mode that tells you the format it actually got from the device, not just the format of the file. Ironbark shows it in Now Playing’s Details (“96 kHz 24-bit exclusive”) and says plainly when a device takes fewer bits than the track has. Other players show it in their own ways; look for an output or “device format” line, and be wary of one that only repeats the file’s format back to you.

One thing to know about exclusive mode: any digital volume change alters the samples, so turn the player’s volume to 100% and set the level on your amplifier. Start with the amplifier turned well down, because in exclusive mode Windows’ volume slider works only on a device with a hardware volume control of its own (Microsoft explains why). Our card has no volume control of its own (Windows reports none), so the receiver’s knob is the only volume, which is why Ironbark 0.47.0 added a “Volume on my amplifier” setting for outputs like it.

On the far end

  • The receiver’s display confirms the sample rate, as above. Treat it as half a check.
  • A DTS music CD or an HDCD-encoded CD (or an exact rip of one) is the classic home check over S/PDIF. A DTS music CD stores its surround sound disguised as ordinary 16-bit, 44.1 kHz audio. If every bit arrives unchanged, a DTS-capable receiver recognises it and plays surround. If anything has been altered, even by a volume change, you hear loud white noise, so start with the volume right down. A DAC with an HDCD decoder works the same way: its HDCD light comes on only if the hidden code in the lowest bit survives. Both are good tests that the path is untouched, but both are 16-bit tests: a path that cuts everything to 16 bits passes them perfectly, so they can’t show whether 24 bits get through.
  • The rigorous way is to record the digital output with a second device that has an S/PDIF input, and compare the recording with the original file sample by sample. If they match, the path is bit-perfect at that format. It needs extra hardware, but it’s the only home method that genuinely tests 24-bit end to end. Comparing two copies of data like this is the same idea as a checksum, which we explain here.

Does it matter?

Honestly, for listening, usually not.

Each bit adds about 6 dB of range. 16 bits gives about 96 dB between the loudest sound a file can hold and its noise floor, more than the gap between a quiet room and a jackhammer, as we put it in WAV vs FLAC vs MP3. 24 bits stretches that to about 144 dB, which is beyond what any converter or ear can use.

The real value of 24-bit is in recording and editing. It gives engineers headroom: levels can be set safely low and processing doesn’t pile up rounding errors. Once a song is finished, 16 bits holds it with room to spare.

When 24 bits are cut to 16 in playback, the lost part sits around 96 dB below full volume. Done properly, with dither (a tiny amount of added noise that turns the cut’s distortion into a steady, very faint hiss), it’s essentially inaudible. Done crudely, by simply truncating, as happens in our story, it adds a trace of distortion at that same very low level. That’s measurable, and in quiet fade-outs at high volume it’s theoretically possible to notice, but in normal listening it sits far below the sound of the room and the music itself.

Where it does matter:

  • When you’re paying for, or keeping, hi-res. If you bought 24-bit files, it’s reasonable to want 24 bits to arrive. Knowing where they stop is useful even if you decide not to care.
  • When you’re checking files. If you’re testing whether a “hi-res” download is genuine, or comparing two masters, you want to know the path isn’t changing what you’re measuring.
  • When you’re recording, not playing. Archival transfers are recorded at 24-bit so the level can be set once, safely low, and the capture kept untouched as a master. Our cassette transfer guide explains why. A recording path that quietly drops to 16 bits takes away exactly the headroom you set it up for.

How to fix it

In order of effort:

  1. Set the Default Format to 24-bit at the sample rate of most of your music (often 44.1 kHz for CD rips, 96 kHz for hi-res), if the device offers it. This fixes shared mode for every program.
  2. Use exclusive mode in a player that reports what it got. That takes Windows’ mixer out of the path and tells you the truth about the device.
  3. Pick hardware that genuinely does 24-bit, at the rates you need, over the output you use. Check the chip’s own datasheet where you can find it, not just the listing.
  4. Or decide it doesn’t matter for this setup. For background music, a party playlist or a TV, 16-bit is fine and always has been.

The card we used

USB sound card (optical and coaxial out)

DSD TECH SH-AU02A

What we use it for: sending music from a Windows PC to an AV receiver over a digital cable, and testing Ironbark’s bit-perfect mode on a USB Audio Class 1 device.

Chip
C-Media CM6542 (named in DSD TECH’s own listing): USB Audio Class 1, S/PDIF out at up to 24-bit, 96 kHz
Connects by
USB, which also powers it
Outputs
Optical (TOSLINK) and coaxial S/PDIF, plus a 3.5 mm analogue output (per the listing)
Listed as
Up to 24-bit, 96 kHz
What we measured
In exclusive mode, at 96 and 44.1 kHz: 24-bit only as packed 3-byte samples. Asked for 24-bit in a 32-bit container, it fell back to 16-bit.
Volume control
None of its own: Windows reports no hardware volume, so set the level on your amplifier
Driver
Windows’ built-in USB Audio driver; nothing to install on our Windows 11 PC

Know before you buy: it does 24-bit, but a player has to ask for it the right way. In shared mode, check the Default Format in Windows. In exclusive mode, use a player that tries packed 24-bit and reports what it got; one that asks only for 32-bit containers will get 16-bit, possibly without telling you. Because it has no volume control of its own, turn your amplifier right down before the first play in exclusive mode.

See it on Amazon AU (paid link)

Common questions

Why does my 24-bit file play as 16-bit?
Because some link between the file and the converter only offers 16 bits: the player’s output format, Windows’ Default Format for the device, the device’s own list of accepted formats, or the chip that sends the digital signal. The lowest 8 bits are dropped and the music plays on as normal, so nothing warns you.
How do I check what bit depth Windows is sending?
Open the Sound control panel (Win+R, mmsys.cpl), select your playback device, click Properties, then Advanced. Default Format shows what Windows sends in shared mode. In exclusive mode Windows’ mixer is skipped, so check the format your player reports it actually opened the device with.
What is the difference between shared and exclusive mode?
In shared mode Windows mixes every program’s sound, converting it to the device’s Default Format. In exclusive mode one program uses the device on its own, in a format the device supports, without Windows changing the samples. Other programs go quiet while it plays.
Can optical (TOSLINK) carry 24-bit audio?
Yes. The S/PDIF signal on optical and coaxial cables has room for 24 bits per sample. Whether you get 24 bits depends on the device sending it and the one receiving it; some cheap USB-to-optical dongles are 16-bit only.
My receiver says 96 kHz. Does that mean I’m getting 24-bit?
No. It means the sample rate arrived. Many receivers don’t display bit depth, so a 96 kHz signal could still be carrying 16 bits.
Can I hear the difference between 24-bit and 16-bit?
For finished music at normal listening levels, almost never. 16 bits gives about 96 dB of range, more than a quiet room allows. 24-bit’s real advantage is headroom when recording and editing. It’s still worth keeping 24-bit intact for archives, transfers and for checking files.
Is a DTS test file a good bit-perfect check?
It’s a good check that a 16-bit, 44.1 kHz path over S/PDIF leaves the data untouched: a DTS receiver plays surround if every bit arrives and white noise if anything changed, so start with the volume down. It can’t tell you whether 24 bits get through.
Do you get paid if I buy the card through the link?
Yes. It is an Amazon affiliate link, marked “(paid link)”. As an Amazon Associate, Digital Greenhouse earns from qualifying purchases. The price you pay is the same either way. Our privacy policy explains what Amazon does and doesn’t tell us.

Sources

Standards

  • Digital audio interface – Part 3: Consumer applications (IEC 60958-3:2021, edition 4.0): scope note and clause 4 (20-bit sample word, auxiliary bits as an optional extension to 24). IEC. link. Edition dated 02/09/2021, checked 12/10/2026.
  • Universal Serial Bus Device Class Definition for Audio Data Formats, Release 1.0: §2.2.2 and Table 2-1 (subframe size 1–4 bytes, bit resolution). USB Implementers Forum. link. Release dated 18/03/1998, checked 12/10/2026.

Microsoft

  • Device Formats (shared mode, the Default Format, exclusive mode). Microsoft Learn. link. Checked 12/10/2026.
  • Exclusive-Mode Streams. Microsoft Learn. link. Checked 12/10/2026.
  • Endpoint Volume Controls (hardware vs software volume in exclusive mode). Microsoft Learn. link. Checked 12/10/2026.
  • WAVEFORMATEXTENSIBLE structure (container size vs valid bits). Microsoft Learn. link. Checked 12/10/2026.
  • USB Audio 2.0 Drivers (also names usbaudio.sys, Windows’ USB Audio 1.0 class driver). Microsoft Learn. link. Checked 12/10/2026.
  • Fix sound or audio problems in Windows (More sound settings, Default Format). Microsoft Support. link. Checked 12/10/2026.

Manufacturers

  • CM6542 USB (Type-C) Ultra-low-power Audio Single Chip datasheet v1.93. C-Media Electronics. link. Checked 12/10/2026.
  • PCM2903C product page. Texas Instruments. link. Checked 12/10/2026.
  • DSD TECH SH-AU02A listing, with DSD TECH’s model comparison (main chip per model). Amazon AU. link (paid link). Checked 12/10/2026; title re-checked 13/10/2026.
  • DSD TECH SH-AU01A listing (TI PCM2903C, 16-bit). Amazon US. Checked 12/10/2026.
  • SR4300 User Guide (front-panel signal format indicators). Marantz. link. Checked 12/10/2026.
  • AV10 online manual, Information screen. Marantz. link. Checked 12/10/2026.

Software, patents and further reading

  • PortAudio WASAPI host API source, pa_win_wasapi.c (exclusive-mode fallback order). PortAudio project. link. Same order in release v19.7.0, checked 12/10/2026.
  • US 5,479,168, Compatible signal encode/decode system (HDCD). Pacific Microsonics. link. Checked 12/10/2026.
  • US 2010/0188947 A1, Automatic DTS CD versus CD-DA detection. link. Checked 12/10/2026.
  • 24/192 Music Downloads are Very Silly Indeed (Monty Montgomery). Xiph.org. link. Checked 12/10/2026.

Keep learning

Data & encoding WAV, FLAC and MP3 What bit depth and sample rate actually are, and whether hi-res is worth it. Data & encoding How a computer knows a file isn’t broken Checksums, and why “the same” can be proven. Reference guide Connector reference Optical TOSLINK, coaxial digital and the plugs at each end. Storage & media HDMI ARC, eARC and every HDMI version The S/PDIF ceiling, and what gets past it.