ClearBits^TM Audio Demo

David Ching, DC Software
March 19, 2026 (originally written September 25, 2009)

Welcome

Since the late 1990's, PC's have been used as digital audio players. I noticed that various audio applications seemed to "sound different" even when playing the same file on the same hardware. Attempts to explain this did not always resonate with my peers, many of whom reported little or no audible difference.

This question has persisted over time. For example, Audirvana is often praised as one of today's best sounding audio applications, while Roon also deserves special mention. This raises a natural question: can playback software itself influence perceived sound, even when identical digital data is used?

ClearBits Thesis

This paper describes a Windows audio player which varies the sizes of the audio buffers passed to the system for playback. The sizes of the audio buffers are randomized using different algorithms.

The goal is to explore whether this variation can affect perceived sound quality. Depending on the nature of the randomness, listeners may perceive differences, although the magnitude and consistency of such effects can vary.

ClearBits randomizes the size of each audio buffer streamed to the device’s playback endpoint.

An audio player typically performs the following steps to play a music (e.g. .wav) file:

1. Open the .wav file
2. Read the format of the file (e.g. 44 KHz/Stereo/16-bit)
3. Open the wave output device, setting the format to match #2
4. Seek in the file to the start of the audio samples
5. Read a fixed number of audio samples (bytes) from the file into a buffer
6. Pass this buffer to Windows to be played, using e.g. the waveOutWrite() API
7. Go to #5 and fill the next available buffer. Keep filling the buffers that Windows has finished playing, until the entire file has been played.

The process of using randomly sized audio buffers is called ClearBits. Hence, the enhancement is:

5. Read a random number of audio samples (bytes) from the file into a buffer

Depending on the nature of the randomness, listeners may perceive differences in sound quality, although the magnitude and consistency of such effects can vary.

ClearBits Music Player

The ClearBits Music Player (ClearBits.exe) demonstrates the ClearBits effect on Windows 10/11 by playing a list of WAVE / MP3 files while allowing the Audio Buffer Size algorithm to be changed on-the-fly.

Click to download ClearBits.exe (4.3 MB)

❶ Alter the algorithm used to determine buffer size. Choices include:

• [F1] Fixed : all buffers have the same size.
• [F2] Random – VC rand() : the size is determined by the VC++ rand() function, notorious for its poor randomness.
• [F3] Random – CAPI : the size is determined by the CryptGenRandom() function, known for good randomness.
• [F4] Random – random.org : the size is determined by real-world occurring random numbers from random.org. (probably the best randomness of all these choices)

When the combobox is changed, you should hear any perceptual differences --if present-- immediately.

NOTE: Due to some versions of Windows’ tendency to glitch when presented with small buffers, all buffers are at least big enough to store 2 seconds of music samples, at the current bitrate. The buffer size is then increased by the random amount.

❷ Displays the current format of the wave out device

❸ VCR Controls: Previous File, Play or Pause current file, Next File

Additional Seek Controls:

• F5 : Seek back 1 minute
• F6: Seek back 10 seconds
• F7: Seek forward 10 seconds
• F8: Seek forward 1 minute

❹ Searches for files containing the specified string

❺ Display file paths that can be played

• Select desired file, then Play/Pause.

❻ Load or Clear the file paths in the list.

• You can load individual wave (.wav) and MP3 (.mp3) files
• You can also load playlists (.m3u) files.
• Instead of clicking Load, you can also drag and drop the files from Explorer into the list.

❼ Check to play a random file after the current file has finished. Uncheck to play the next file in the list.

• Hold the Control key while clicking << Back or Next >> to jump to a random file.

❽ Drag window corner to resize

Suggested Use

Rip a few of your best-known songs from CD to .wav files (but ClearBits can also be used with lossy compression formats).

Drag and drop them from Explorer into the ClearBits Music Player.

Click the Play button, listen for a few minutes to get initiated, and then alter the Audio Buffer Size combobox to pick another randomness algorithm. Any perceptual differences --if present-- may be heard immediately.

Click in the list to select a new song, and don’t forget the F5-F8 keys to seek within the current song.

Listen for any perceived differences, and if present, consider whether they are noticeable and consistent. You may also compare different randomness algorithms and note any differences in character or magnitude.

Boost the volume to 100% so that Windows does not reduce resolution with the digital volume control.

The temptation is to play test simple tones like sine wave, but more complex music may make any differences easier to perceive.

How does it sound?

Listener reports have been mixed. Some listeners report subtle differences, while others do not perceive a change. Those who do sometimes describe changes similar in character to different dither or noise-shaping algorithms, both in type and magnitude:

• Both change characteristics of digital hardness, sound staging, and low level detail.
• Both are subtle and may not be consistently detectable by all listeners or systems.

This is plausible since both use randomness to achieve their goals. ClearBits uses randomness to alter buffer size, and dither uses randomness to distribute truncation errors.

 

• David Ching, 11/19/2009 – 11/20/2009

The ordinary people who’ve tried this don’t hear a difference, but 2 other audiophiles and a sound card engineer could.

• Shannon Bailey, 1/5/2006

...I think the subtlety of difference and improvement, and the variability of difference from machine and sound card to machine and sound card, is going to be a large impediment to delivering consistent value....

• Frank Yerrace, 11/19/2009

I’m afraid I’m not hearing a difference.

• Graham Mayor, 8/18/2002

I have tried the app. with mp3 files on my hard disc. There is some improvement, but it does not overcome the essential blandness that is a characteristic of mp3 files....

 

Hypothesis

One possible explanation is that varying buffer sizes alters processing timing in a way that reduces cyclic timings at the hardware level, which could in turn influence timing-related effects such as jitter in the output stream. This hypothesis is speculative and has not been independently verified.

Other potential reasons for an audible difference

Random-sized buffers demonstrate that the audio player affects the sound, but they're not the only factors. Here is a sampling of what others have reported.

Do streaming services have a “sound”? | by Koscso Ferenc | Medium, 3/10/2024

 

• If you compare Tidal’s and Qobuz’s desktop apps, there is a chance (mainly on Windows) that the apps use the operating system’s audio subsystem in a slightly different way.
• In the case of macOS, apps use the so-called “Core Audio” sub-system. So, in the case of an Apple computer, there is a chance that there will be no meaningful difference between the apps’ output through your DAC as the Core Audio system provides the same path to the DAC for the applications.
• I did not research the local buffering mechanisms of Tidal/Qobuz apps. Theoretically, it can have an effect, as buffering requires quite a bit of power and power supplies can handle it differently. According to one of the best-known streamer developers, it can even induce extra jitter, which can be measured at the DAC — Roon buffers longer audio chunks locally than Tidal or Qobuz apps. PCs generally play audio streams from their memory, not directly from their network connection. Apps can load and buffer chunks in the memory differently, which can have a power sag effect.
• Very likely: different settings for the local audio sub-system
• Likely: apps have different ways of buffering the data while playing back, to the local RAM and this causes different power supply performance, and an indirect effect on the audio system
• Likely: your senses play a game to you

 

Gordon Keith, 6/12/2014

• I quickly noticed that different versions of memcpy sounded different
• Agree different buffer sizes sound different, too large and the sq is lumpy and without definition with poor treble, too small and the treble is too pronounced, so it seems to be a compromise. For 16/44 I get the best sound with 512 samples, 2048 bytes.

David Ching, 11/20/2009

I notice the perceived audio quality also changes when there is heavy system load:

1. there is a file copy occurring,
2. there are many processes running
3. the process and thread priorities of the music player are altered in task manager.

ClearBits makes a similar or larger difference as these.

Next

• To allow users to evaluate whether ClearBits produces perceptible differences under controlled conditions, the ClearBits Music Player will be enhanced with an ABX Testing feature.
• Discuss size of reference fixed buffer, from which a random size is appended. In general, the larger the buffer, the less noticeable treble. In general, larger buffers sound more natural. Smaller buffers sound more tinny.
• Experiment with WASAPI exclusive mode