How do you go about sampling wavetables?

[chk071]

Woah, easy with me there... i am merely mortal.

Seriously, i can't do any of the things you stated. I'm happy if i get the wavetables sampled.

[aciddose]

Lame.

You mean "re-sampled"

That's assuming the samples they used in the software were even the original raw data.

Nothing is cooler than dumping the 15-bit or 12-bit samples raw from the TR-xxx and playing them back with true fully anti-aliased nearest neighbor interpolation.

... if you're a huge nerd.

[fmr]

Yeah, but... when it comes to actual manual sampling, and editing, there's not much. Actually, i tend to sample the single waves now, and edit them all in Wavelab, because that seems like the most precise thing to do for me.

Microwave had 300 ROM waveforms (Largo is essentially a Microwave, in what concerns wavetables, AFAIK). Besides, many wavetables were created with just a few waveforms, having the other slots interpolating. You will have a really hard job in front of you to recreate all the wavetables if you are going to sample each waveform.

I downloaded the Microwave wavetables from the net a while ago (more than a year, if I remember well). I did a search now, and it seems they are no longer available.

[fmr]

Lame.

You mean "re-sampled"

That's assuming the samples they used in the software were even the original raw data.

Nothing is cooler than dumping the 15-bit or 12-bit samples raw from the TR-xxx and playing them back with true fully anti-aliased nearest neighbor interpolation.

... if you're a huge nerd.

Waldorf didn't use samples. The waveforms were generated though additive synthesis.

[aciddose]

Yes actually a significant chunk of the M1 ROM at its "huge 4 mbit" is total BS: lots of the samples are simply filtered copies of others since the M1 had no real interpolation (this was pre-FIR "polyphase" filters) since it was way too expensive or even impossible to implement in the low density chips they could create.

Some of the samples in the expansion carts/chips are pretty cool "m1 plus" but most of the basic samples are super boring. You'll actually find Korg and Roland samples from these early units re-used in things like SNES carts, arcade machines and heaps of other sample-based devices.

That's one of the reasons that mid-90s period was so boring: everyone was recycling the same old same old garbage M1 samples which gave us that "really god awful crap sampler sound".

[chk071]

You will have a really hard job in front of you to recreate all the wavetables if you are going to sample each waveform.

I don't plan to. Just the Q wavetables (Alt1 and Alt2). I like those the best anyway.

[aciddose]

That's assuming the samples they used in the software were even the original raw data.

Waldorf didn't use samples. The waveforms were generated though additive synthesis.

Yes but they were stored in some format even if not by PCM. Technically additive is waaay more easy to reproduce than raw PCM since the data is extremely predictable (low entropy) and can be described with very simple formulae.

It would make more sense to reproduce the bit-exact additive synthesis process and re-synthesize the exact waveforms. You could then add extra parameters like any bit resolution (or floating point =~inf) and sampling rate.

If you understand the discrete Fourier transform you know that any block of N samples can be described exactly by N amplitude+phase pairs. If the spectra isn't complex (such as 1/N or 1/N^p) it makes it even more simple and can often be expressed in a single equation like for the "saw wave": saw = (-1/N, 1)

[fmr]

You will have a really hard job in front of you to recreate all the wavetables if you are going to sample each waveform.

I don't plan to. Just the Q wavetables (Alt1 and Alt2). I like those the best anyway.

Q Alt 1 is avaible in the Wave Edit repository: https://waveeditonline.com/index-2.html

[fmr]

That's assuming the samples they used in the software were even the original raw data.

Waldorf didn't use samples. The waveforms were generated though additive synthesis.

Yes but they were stored in some format even if not by PCM. Technically additive is waaay more easy to reproduce than raw PCM since the data is extremely predictable (low entropy) and can be described with very simple formulae.

It would make more sense to reproduce the bit-exact additive synthesis process and re-synthesize the exact waveforms. You could then add extra parameters like any bit resolution (or floating point =~inf) and sampling rate.

If you understand the discrete Fourier transform you know that any block of N samples can be described exactly by N amplitude+phase pairs. If the spectra isn't complex (such as 1/N or 1/N^p) it makes it even more simple and can often be expressed in a single equation like for the "saw wave": saw = (-1/N, 1)

Yes, and Waldorf waves were even simpler, because they were ALL forcelly symmetric, therefore they just computed the first half of the waveform, and simply inverted it to get the second half.

[fmr]

You will have a really hard job in front of you to recreate all the wavetables if you are going to sample each waveform.

I don't plan to. Just the Q wavetables (Alt1 and Alt2). I like those the best anyway.

Q Alt 1 is avaible in the Wave Edit repository: https://waveeditonline.com/index-2.html

OK, I think I found a solution for you:

1. Download the Komplexer wavetables from here: http://synth.stromeko.net/Downloads.html
2. Download Audio-Term from here: https://www.facebook.com/Audioterm/?__x ... tn__=HHH-R
3. Unzip and copy the Komplexer wavetables to the folder data\04 KTERM - Wavetables (Komplexer, WT) inside the Audio-Term folder
4. Open Audio-Term and convert any wavetable you like to the Serum format. The two Q/microQ/Komplexer wavetables (they are the same) are there, as are 9 wavetables containing ALL the ROM waveforms of the Microwave.

I could post the Q wavetables converted, but I'm not sure I can legally do that. and I don't want to open a can of worms.

[chk071]

Thanks but... as they say "The journey is the destination". I really wanted to try this out manually for a long time now.

I already set up a project including automation, and already recorded the Alt1 wavetable. The only thing i'm confused about now (probably miss something obvious) is that the waveforms, when i import them to Dune 3, are all a octave higher than they should be. Any idea why that is happening? I did everything exactly according to the hints in this thread, and in the Expanse tutorial i linked to. Played the notes in F0 minus 24 cents, recorded in mono, 44.1 kHz, 16-bit wav.... there's no tuning going on in Largo either. Strange.

Edit: Ok, my bad. Seems like it works fine when i import 1024 samples long wav files.

[aciddose]

That's something to ask the software manual or forum. It's 100% about the software you're using.

RIFF WAVE does include tuning information inside the "smpl" chunk. I've written a tool to display this information... it may be useful to you although most good sample editors will also show this information somewhere... sometimes they don't offer that at all though especially if RIFF WAVE is "just another format" they import.

http://xhip.net/temp/rifftool.zip

One "dumb" thing about this tool is I load the file completely into memory before parsing it. Normally it would be possible to load only smaller chunk headers at a time (8 bytes?) and decide whether to load the entirety of the chunk or not (so "data" could be skipped) but that would make the code a bit more complicated and I'm lazy. It works fine for small samples just don't go loading any 90 gig files.

You just launch this and drag & drop, it should be capable of printing at least the most basic info about the file such as the "fmt " and "smpl" chunks along with some "INFO" chunks which may tell you what the tuning parameters are for the file.

To understand how Dune 3 interprets that data you'll need to read the manual. In order to strip that data... well... my tool can't do that currently but it is reasonably easy to accomplish. Normally I just use a hex editor and fix up the chunk sizes by hand. It isn't too much work on a single file but if you have 100s of samples it would be awful without a tool to batch process them.

[fmr]

Edit: Ok, my bad. Seems like it works fine when i import 1024 samples long wav files.

Yes, waveforms must be 1024 samples.

[AnX]

Thanks but... as they say "The journey is the destination". I really wanted to try this out manually for a long time now.

I already set up a project including automation, and already recorded the Alt1 wavetable. The only thing i'm confused about now (probably miss something obvious) is that the waveforms, when i import them to Dune 3, are all a octave higher than they should be. Any idea why that is happening? I did everything exactly according to the hints in this thread, and in the Expanse tutorial i linked to. Played the notes in F0 minus 24 cents, recorded in mono, 44.1 kHz, 16-bit wav.... there's no tuning going on in Largo either. Strange.

Edit: Ok, my bad. Seems like it works fine when i import 1024 samples long wav files.

at 2048, should be a single cycle waveform, are you getting that?

[Shabdahbriah]

... in other news.

viewtopic.php?f=33&t=517146