The DAC behavior is absolutely modeled. In the demo video he is playing around with pitch and decay all over the place so I would not view it as a 1:1 comparison. He raises the pitch on the hi-hats quite purposefully. Here's how the 707/727 actually work with regards to the modeling (from the 7X7-TR8 website).Shy wrote:Roland can't even make their TR-707 / TR-727 emulation sound remotely as good as the original.
Most of that demo on YouTube has a significantly raised pitch for the hi hats for a reason - they couldn't figure out and emulate the TR-707's DAC and analog processing which is responsible for most of the high frequency content of the sounds.
TR-707's and TR-727's internal samples have a low sample rate. The crash and ride cymbals are 25,000 Hz and other samples (including hi hats) are 22,500 Hz. So playing those original samples taken from the ROM, you could hear how extremely "lo-fi" they are compared to being played back on the hardware. It's a similar situation with many other "digital drum machines", including the lovely SCI Drumtraks and Oberheim DMX and DX. I don't know how exactly those drum machines work, but it's not just plain distortion or "exciter"-like processing that's responsible for the hugely significant addition of high frequency content. Plus all those drum machines modify the entire frequency balance as well and make the samples (stored in ROM) sound nicer and less harsh in the middle frequency range than they really are.
"CAPTURING THE ORIGINALS
Even though the TR-707 and TR-727 were primarily digital devices, it would not have been possible to perfectly replicate their sound using samples. Due to the low bit-rates of the original units, quantization noise became a problem, particularly during the sound’s decay. To reduce quantization noise, a clever design was employed where PCM sound was produced using non-decaying waveforms and decay was then introduced in the analog circuitry after being converted. Also, at the time these units were developed, deviations in the clock that triggered the PCM caused differences in pitch, and deviations in the analog circuitry downstream of the D/A converter caused variances in decay characteristics.
To replicate these sounds for the TR-8, we started with the original PCM wave data found on the classic machines. We then used our ACB process to completely model the PCM output stage, carefully including all of its quirks and instabilities. Modeling the analog envelope and amplifier stages that came after the D/A converter allowed us to implement the “Tune” and “Decay” parameters that are available on the TR-8, but not present on the original units."