DSP Robotics Support

Interesting - OK, will do.

Though I don't suppose you have an old copy of SM to do it with, so it will be a .osm file, do you? I don't have Flowstone, and the wife would have a fit if I spent a c-note on synth software.

(This is becoming less appealing all the time isn't it?)

JLS]n~��������=��z|��������xnjb]OA=AJUdllfjp|����z���hS?52=Sx���ý���������ťx[UWOC@ACHLOOLE,###*Y���������������}�r6####

This is what one of the .bin opens as in Open Office. Make any sense to you? This file was named "esq1wavlo.bin" Can't figure out now how I got that first wav file.

.bin is short for "binary". You can't access binary data through a text editor. Instead, you need to know which application saved the data (so you know the format of it) and then use whatever tool is able to deal with the format. Since on his page they often talk about Linux, I'm afraid those are self-extracting Linux binaries (that can only be accessed in Linux). That's why I was so amazed you got something out of them.

However, my own search seems to be more successful. This link leads to a zip that contains a folder structure with .aif files (the Apple sound format very similar to .wav). The readme explains the structure in detail. It also means that I might be able to do the conversion now. But to do an accurate ESQ1 you have to recreate the osc behaviour exactly (3 osc fed by up to 6 waves each, depending on key played. Again, that's where I have to quit, but I do my very best to provide you with the raw audio data you need to recreate the ESQ1.

http://www.buchty.net/ensoniq/files/the_waves.zip

Yea, I can't duplicate what I did before; can't get anything to open to anything other than a .bin file. I can't figure why he didn't just post the wavs on his site, they're very small - only a fraction of a second each.

Thanks for digging up those aiffs; I think I stumbled across them long ago and had forgotten. i actually tried to trace them with the SM wave-draw oscillator - sounded like poo, as you can imagine; for reasons you have already mentioned.

Sent you a pm with instructions for an easy DIY way.

Ok, so after a long conversation we found out a lot of things. One of them: Kevin won't be able to do the full project. His Synthmaker version is too restrictive. I, on the other hand, don't have the expertise to do all the serious technical stuff, like proper filters, etc.

So I share what I found out so far, which should bring anyone, who is interested to do this ESQ1 recreation, a big step forward.

The structure of the ESQ1 is as follows.
The lowest layer are the waveforms. Those are one cycle of a wave, some of them synthetically created, some of them sampled from real instruments.
The next layer (and the first one, a user has access to) are the waves. The ESQ1 has 32 waves, but the definition is different to what you know as waves. A wave here means a collection of above mentioned waveforms (up to 6), with each waveform assigned to a certain midi key range.
The highest layer are the programs. A program manages up to three oscillators -each one associated with one of the waves-, the envelopes, the filter, etc.

It is advisable to build the ESQ1 recreation in this order as well. To get the original waveforms, you should download the ESQ1 WaveROMs #1 and #2 on this page: http://www.buchty.net/ensoniq/ (Scroll down a lot)
There are also aiff files of the waveforms, but the guy who converted them created Aliases for those waveforms that are used by several waves. Such Aliases are not visible/usable when you unzip the package, and so it is missing a lot of waveforms.

The WaveROM#1 and #2 are .bin files, which is a self-extracting format used in Linux. I couldn't find a way to convert them yet. Should they be just packed and not compressed, in theory all the sound data should be extractable with a binary viewer. Or maybe some one around here has Linux and can convert them to a more friendly Windows-readable file?

The format of those is 8-bit, 1MHz/26=38,462kHz. Which means, you need an audio editor that can deal with such an unusual format.

Just remember that the ESQ1 uses a multi-sample (read "ROMpler") type of synthesis to feed the OSCs, not wavetables. If you head for the original sound, you have to re-do it (maybe use the multi-sampler module in Flowstone, or build your own, tailor-made multi-sampler for this project).

All details about the waveforms can be found in this manual pdf, page 35ff (page 38 starts a list of all the names of the waves -those, that users can select-). http://www.synthmanuals.com/manuals/ensoniq/esq-1/musicians_manual/esq-1_musicians_manual.pdf

There's also these, which are .wav files.

Thanks for the forms! Do u know how many there should be in total?
I'm not in a hurry, but I can auto-chop these into a whole load of single cycles.
Herr Proff is our benefactor hehe

nix wrote:Thanks for the forms! Do u know how many there should be in total?
I'm not in a hurry, but I can auto-chop these into a whole load of single cycles.
Herr Proff is our benefactor hehe

That's the problem I currently struggle with. The two .bin files I referred to, are just an image of the Eprom's content. (The .wav files from Herr_Prof are exactly those images, just with a wav header)

The individual waves are of different sizes each and can vary somewhere between 256 and 32768 byte per multisample. I asked Rainer Buchty about this, but his answer is way over my head. We had our conversation in German, so hopefully my translation won't miss too much. [These sections are my comments]

"...[me explaining what I did]... 8-bit unsigned was a good start and assuming multi-samples wasn't basically wrong as well. But it looks like so: http://buchty.net/ensoniq/files/waves.xls
There's a key range of 128 keys. Each 8 keys get assigned one raw waveform [he means all seperate waveforms that make up one wave]. That's 16 multi-sample zones. This raw waveform as access parameter has the starting page(!), the WSR-value and the tuning (halftone and fine). As you can see from the indices, a lot of the waveforms are not only presented with a harmonic-reduced copy of themselves in higher registers [higher frequencies, higher midi keys], but also with even more reduced OCT+5, OCT and finally SINE.
Look for the WSR-value in the DOC data sheet, since it encodes the waveform length as well as the length dependent address resolution [He uses the word "Addressauflösung", not sure how to correctly translate it]: http://buchty.net/ensoniq/files/ics1261.pdf
Since the DOC has a fixed time window of 38461,54Hz, in the end the playback will always be a compromise between frequency resolution, effective playback speed and waveform length. [I don't understand this paragraph at all]
Also the granularity of the page [the Eprom is accessed through pages] is implicitly given by the waveform length, since the DOC, restricted by hardware, can only access and loop the waveform page by page.[Another paragraph I don't completely understand]"

But, Nix, maybe you can make sense of the linked documents and his descriptions? If so, don't use the wav files, because whatever address values you come up with, they are only valid in the original, unaltered binaries. I can access and chop them, I just don't have a clue where.

I'll look at this in the next couple of days,
and try to comprehend.
The place to chop them is at 0 crossings of full cycles? It's very troublesome that they vary in length.
It needs a keen eye to see what the cycle is, and this figure will need to be adjusted
when the form changes afaic tell.
hmm, if I pick this up, I would make folders of evolving single cycle sets-
and resample them- say to 2048. Could you work in this way?
They could be accessed by a mem array index.

It's a lot of work, I'm not in a hurry hmm
IMO- anything I can do u can Tula. I have SoundForge and an auto chopping schematic.