Help needed for Ensoniq's Digital Oscillator Chip (DOC)

[Spogg]

Thanks tulamide

I understand and agree with what you say completely.
If you want as accurate emulation of the original as possible, then you simply have to do what you say. There’s no other way, because then it would become an “inspired by” project, which is actually what I was kinda thinking about for myself; a bit like my QX7 and Q-50 projects. They had a similar structure and technique but were not much like the originals. That’s what I personally get interest and pleasure from doing. Plus, accurate emulation is a huge challenge

I did the interesting work, now you have to the hard work and I can’t wait to see where this all ends up.

Good luck my friend!

[tulamide]

No, no, my job is done, when I deliver the most accurate conversion of the waveforms, and some sort of scheme to combine them for the waves. For all the rest, a good starting point is Kevin's schematic, that features a way more accessible GUI than the original or the SQ8L plugin. I don't know how accurate the filters are, but maybe that's for the talented people that know how to do proper DSP code.

One last question, if I may?
The binaries are divided in two files of 32kB each. There are called "...low" and "...high". The addresses don't seem to make any call to low or high whatevers. Is it save to assume that I can add the two files together and then will get all the waveforms from them, because the addresses cover the whole 64kB range? (Note: I'm only interested in the ESQ1 and its EPROM, not the successor SQ80)

[Spogg]

Yes this bothered me for a while too.

I had to check the circuit, and chip U15B is an address decoder (74F139) which can enable 1 out of up to 4 ROMS (2 sockets are empty). Fortunately the ROMS (EPROMS actually) have 8 bit data bytes so the High/Low thing only refers to the ROM’s address space for the CPU address bus and not any data horror.

It may be a better bet to rip directly from ROM#1 image for the SQ80. The info says this is the same as the ESQ1 original. Or, maybe use it as a cross-check on a couple of waves. However it could be they modified some of the wave data for the newer SQ80. I read that folk think it sounds a bit better than the ESQ1. It depends on how authentic you want the ESQ1 emulation I guess.

When you get success with the rip please let me know the steps and I may be inclined to rip all the ROMs for the SQ80. I’d like those waves too! I’ve no idea of the work and time involved of course.

The filter chip is the CEM3379 which is an analogue 4 pole LPF. The output is constant with varying Q so my guess is it’s modelled on the Moog filter. Some folk are sensitive to the filter’s exact sound so maybe Martin would be able to contribute when the time comes.

Cheers

Spogg

[tulamide]

Thanks once again! I will check the SQ80 ROM #1, even if it should be identical, I at least save the step to combine the two files.

Of course, I will report. With my work done it should be a breeze to do the other 3 ROMs. But you never know...

Re: filter, I find it interesting, how to recreate it, because everything in the ESQ1 is digital, up until the sound is fed to the analog filter. There must be at least one D/A involved, long before the sound goes out to the jackets, maybe even a D/A for the filter and then A/D for other digital options? This may be crucial for the reproduction of the original sound!

But those are all questions for the more talented DSP guys among us.

[tulamide]

Soon.

Teaser

[Spogg]

Consider me successfully teased!

Regarding the signal flow:

The DOC obviously works on all the digital stuff and produces all the waveforms, LFOs, modulation control and envelopes which work on the “signals” in the digital domain.

The on-chip DAC in the DOC sequences all 8 audio signals which are then de-multiplexed into sample and holds for the filter chips’ audio inputs. By using a single inbuilt DAC any non-linearity will be the same for all 8 voices. The output from the filter chips then goes to the instrument audio output.

The DOC also generates multiplexed control values for an external D to A chip, and this is then distributed via de-multiplexing to sample and holds for the control inputs to the filter chips; Fc, Q, output volume (final envelope) and Pan position.

The DOC has a single A to D input and this input receives a multiplexed value from the analogue controls, like pitch bend and modwheel etc. but no audio. The multiplexing is controlled by the DUART chip and must therefore be mediated by the CPU (Operating System) to “inform” the DOC what parameter to adjust internally.

I had fun tracing this!

Cheers

Spogg

[adamszabo]

I wish I would understand all this discussion thats going on but it seems very exciting

[Spogg]

I wish I would understand all this discussion thats going on but it seems very exciting

You need to be of a "certain age" Adam and I don't think you qualify

For me it's been a blast, taking a trip down memory lane (literally). I was dealing with this stuff nearly 40 years ago and I never thought I'd need to do something like this again.

Great fun!

Cheers

Spogg

[tulamide]

The A to D input sounds a lot like standard CV fed into the digital domain. Damn, they were clever (given the time).

Meanwhile, I still have to solve minor issues and make the tool much more comfortable to use, but:

Teaser 2
(Caution: It's a cycle. Download and loop it in your preferrred audio renderer)

[Spogg]

The A to D input sounds a lot like standard CV fed into the digital domain. Damn, they were clever (given the time).

Hey you teased me again!

Sounds really "right". The bit depth is reported as 16 and the number of samples isn't a mutiple of 256 so what conversion process did you use on the raw wave data? I resampled it to 8 bits with anti-aliasing interpolation and this provided a slight background noise element, which is probably what the original sounded like.

Yes the DOC ADC input does exactly what you said. Everything happens in the digital domain except the filters, output VCAs and Panning which is all handled by the Curtis chips.
It is indeed a very clever design and it only uses 1 custom chip (the DOC). Everything else you could have purchased off the shelf at the time. I have to admire the Ensoniq devs for their ingenuity and achievement.

Cheers

Spogg