The Difference Between Bitcrushers And Sample Rate Reducers

bit

If you look around a few electronic music forums, you’ll see a lot of confusion over the difference between a bitcrusher – a filter that reduces the bit depth of an audio signal – and a sample rate reducer – a filter that does exactly what its name implies. With the popularization of 8-bit and retro synth music, this difference is obviously of grave import of concern to saints and kings alike. [Michael] is more than happy to walk us through the difference with real-time sample and bit rate adjustment with his audio hacker board.

The audio hacker board is an Arduino shield with a 12-bit DAC and a 12-bit ADC. With two 1/8″ jacks and a pair of pots, [Michael] was easily able to whip up a sketch that is able to adjust the sample rate and bit depth of an audio signal in real-time.

Contrary to nearly everyone’s opinion of what ‘8-bit’ music is, it’s actually the sample rate that makes music sound like a cassette deck jury rigged into a Nintendo Entertainment System. Reducing the bitrate just makes any audio source sound louder and worse.

Check out the excellent demo video of the effect of bitcrushers and sample rate reducers below.

[youtube=http://www.youtube.com/watch?v=8pt_p0DSpWc&w=580]

23 thoughts on “The Difference Between Bitcrushers And Sample Rate Reducers

      1. No, I think adcurtin in correct. There are single bit DACs for consumer audio. You would be hard pressed to tell the difference between 1bit PWM and usual 16 bit sound. See https://en.wikipedia.org/wiki/1-bit_DAC

        As for 8 bit audio/music, I think this somewhat of a misnomer. Thinking about a lot of the computers from that era, they didn’t have true 8 bit resolution for audio. The 8 bit refers to the processor, ie 8 bit processors such as 6502. A lot of the music coming out of these computers were single bit or had dedicated oscillators that were not 8 bit resolution.

        1. Another misnomer is that most people seem to think 8-bit music is only square waves, though several systems of the era had more advanced synthesizers (probably most famously the C64’s SID chip).

        2. 1-bit DACs aren’t usually PWM anyway. Instead the decimator usually outputs a square pulse or no pulse for each oversample clock. Doing it this way is easier at high oversample rates and makes it very easy to maintain DC bias by inverting the output every other clock.

        1. You can actually do a lot better than that via the magic of noise shaping. SACD can do about 120dB of dynamic range in the audible band while it’s just 64 times 44100 samples/sec. The trick is in concentrating most of the noise in the ultrasonic range.

  1. bit depth reduction adds correlated noise at overtones of the original frequency content and intermodulation. Sample rate reduction adds aliased copies, usually not at harmonics. The former makes it sounds crunchier and noisier; the latter (without a sufficiently good lowpass filter before resampling) makes it sound less tonal and more avant-garde.

  2. What I heard just now on that video is somewhat akin to the noise I heard on my Speccy all those years ago. I remember typing in two listings and a block of checksummed hex to get what even then was AWFUL sound. It was a Spectrum 48K sound sampler. One bit resolution of course.

    I’ve gotta build an amplifier and filter for my radio alarm clock project so maybe I ought to be taking notes about now.

    1. There were more than one PC-speaker .wav players (not to mention some games, capable of doing voice/music output to PC-speaker). Oh, and there is a driver for linux, making PC-speaker an ALSA device.

  3. For the record,
    (*) he obviously doesn’t do proper filtering in his sampling rate change demo. So, the sound is much worse (aliasing and mirroring) than what you could achieve. I guess he does not do a lowpass and I’m pretty sure that he simply uses a rectangle as “reconstruction filter” instead of a windowed sinc.
    (*) he also doesn’t do proper dithering in his bit depth reduction demo. So, the sound is much worse (noise modulation, harmonic distortions) than what you could achieve.

    If you do the sampling rate conversion properly, the sound should just get dull. If you do the bit depth reduction properly, the sound should just get noisier where the added noise is independent from the actual musical content.

    Then again, this all might just be intentional. One cannot deny that the “bad rate conversion” has at least some appeal to some ears. :-)

Leave a Reply to Grey LokiCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.