I was wondering if it could be true that the Casio MT-100 really has “an internal sample rate” of 600kHz as I had found this information posted online. I quote:
The MT-100 and its cousins (such as the MT-68 and the MT-65) employ the NEC D931x chip to digitally synthesize sound. Two short waveforms, stored in a sort of “stairstep” format (instead of being stored as simple quantizedsound like PCM, the waveforms are stored in essence as a series of instructions meaning “amplitude goes n steps up/down”), are crossfaded with each other to produce each instrument sound. This was internally referred to as “consonant-vowel synthesis”, due to the usual form of each instrument patch — a sharp percussive “attack” waveform segues into a droning sustain phase.
The MT-100 has a wonderfully sharp and clear tone, with no distinguishable aliasing at all — perhaps due to the fact that its internal sample rate is approximately 600kHz. The output of the sound chip has 17-bit depth, with each of the 8 polyphonic voices internally processed at 14 bits.
This seemed crazy, though. CDs are around 45kHz. 600kHz is 12x that fast.. The highest audio resolution that is common on computers these days is I think 96kHz. That’s still just 1/6 as fast as 600kHz. How could a consumer keyboard made by Casio in 1982 and targeted at home musicians be capable of such an insanely high sample rate.
Twitter user @ahihih pointed out that the documentation seems correct, and that “running at a high sample rate is no problem when the DSP is computationally simple.” From the documentation:
The 931 produces a 17 bit audio word every 8 clock cycles - at a rate of about 600 KHz. This is an extremely high sample rate by digital audio standards and is of great significance to the bright clear Casio sound. In a sampled audio system, any harmonics of the signal which are above the Nyquist frequency (half the sample rate) are reproduced below the Nyquist frequency rather than above it. For instance in a Yamaha DX7, which has a sample frequency of about 50 KHz, if you try to create a sine wave X KHz above 25 KHz it will come out at X KHz below 25 KHz. If you tried to make a 40 KHz sine wave it would in fact be audible as a 10 KHz sound. This is a problem in the Yamaha DX7 because some of the more complex FM sounds have very high harmonics which can fold over the Nyquist frequency and become objectionably audible - for instance the Sitar voice played in the highest octave. Similarly when a recording is being made for a Compact Disk which samples at 44 KHz, the sound must be put through a really mean filter to remove all signals above 22 KHz which would fold over as "Aliasing Distortion". Since the Casio sounds are all made of step waveforms, they are very rich in upper harmonics which will extend way beyond the audio range - these waveforms would sound terrible from a DX7 because virtually every sound would have a significant harmonic content above 35 KHz and so would fold over unmusically to below 15 KHz. Since the 931 samples at about 600 KHz only those harmonics above 1185 KHz will fold down to below 15 KHz to become audible, and since the harmonics are getting rather low in energy at this frequency, there is very little "Aliasing Distortion".
I’m not sure, but I think this means that there must be a lot of supersonic overtonal frequencies interacting with the audible ones — like modulators, almost.