Doepfer

I would welcome your help to understand wavetable synths.

I thought I had them sorted in my mind but now I am not sure

I have a 12 bit Akai S900 , max bandwidth 12Khz.

I sampled a 6Khz triangle wave and truncated it to one cycle and looped it, 

It sounded like a square wave buzz.

I sampled a 6Khz sine wave and truncated to one cycle , looped and ,

It sounded like a square wave buzz.

I went on to do the same with more complex wave forms sampling 1Khz wave forms and......

You-get the idea they all sounded like a square wave buzz.

I wondered if the looping frequency was dominating the sound so I lengthened the looped wave form a little to no improved effect.

I tried a low pass filter to cut the looping frequency and it sounded like a more dull buzz.

Why?

How do wave table synths manage to sound as they are described.

Thanks for you help

Thanks Zoe, as always your explainations are succint but I am not very bright so I am afraid I do need more detail.

The akai sample rate is 24 Khz bandwidth 12 KHz.

I have read the sample rate should be at least twice the fequency of the  wave form that is to be sampled.

Well, the sample rate was 24Khz for a 1Khz wave form.

Still sounded like a buzz.

I am missing something.

I have  actually sampled at 128Khz f or a 3 khz waveform to no avail on my Mac.

I am doing something wrong I am sure,

I would like the Doepfer wavetable module but would like to understand my short commings 1st.
--- In Doepfer_a100@yahoogroups.com, Zoë Blade <zoe@...> wrote:

>
> Hi!
> 
> If you're sampling something 12,000 times a second, and that thing is moving up and down 6,000 times a second, then you're going to get a sample of it being up, then a sample of it being down, and so on.
> 
> You'll have to sample it playing at a much lower frequency than 6kHz.
> 
> Hope that helps,
> Zoë.
>

Hi!

If you're sampling something 12,000 times a second, and that thing is moving up and down 6,000 times a second, then you're going to get a sample of it being up, then a sample of it being down, and so on.

You'll have to sample it playing at a much lower frequency than 6kHz.

Hope that helps,
Zoë.

Zoe! You are a genius, it falls into place, sample rate at least twice the maximum harmonic!!! Not merely twice  the fundamental!! 

No wonder it sounded rubbish.

I never got the inference of the reference to the component  harmonics, I feel a right plonker!

Thanks Zoe.

Means that my 25 year old Akai with 750 kilobytes of ram is woefully inadequate

Regards Anoop


--- In Doepfer_a100@yahoogroups.com, Zoë Blade <zoe@...> wrote:

>
> Hiya!
> 
> OK, so when you sample something, you're basically storing a very blocky picture of it.  The higher the sample rate, the smoother the picture, the lower the rate, the blockier the picture.
> 
> As a separate thing entirely, bear in mind that any given waveform can be said to be made up of lots of sine waves at different frequencies.  So, for instance, a 6kHz triangle wave consists of a 6kHz sine wave, a 18kHz sine wave, a 30kHz sine wave, and so on.  The higher the sine wave frequency, the quieter it is, but they're there.
> 
> The sample rate you're recording at, how blocky or smooth the representation of the sound is, should be double the speed of the *fastest* sine wave you want to hear.  The more of them you capture, the more accurate the sound.
> 
> But you don't really need to know all that, I'm just trying to explain what the "should be at least twice the frequency" bit means.  It's NOT twice the frequency of the note, but twice the frequency of the fastest harmonic in that sound that you want to hear.
> 
> If you want to know all about sampling, check out these articles:
> 
> http://www.soundonsound.com/sos/aug00/articles/synthsec.htm
> http://www.soundonsound.com/sos/sep00/articles/synthsec.htm
> 
> Basically, in order to get the best sound possible, you want to record a really low frequency sound at a really high frequency sample rate.
> 
> As far as wavetable synthesis goes, however (such as the other feature of the A-112 module), you don't need to manually sample this from another synth.  It's far easier (to certain geeky types of people, anyway, heh) to make these sounds from scratch on a computer, without them existing as sound until you've copied it across to the A-112 and played it.
> 
> I happen to have bought an A-112 module the other month, so I'll look into making some handy wavetable .syx files for everyone in a little while.
> 
> With wavetable synthesis, if you were to record an oscillator, it would have to be at an exact frequency that matches the frequency the A-112 is recording at, as you need to use exactly 256 samples to store each and every cycle of the waveform.  This would be nigh on impossible to get just right, hence making the waveforms digitally on a computer and copying them across.
> 
> But really, as a musician, you don't need to know all this.  All you need to know is that if you play a lower note into the sampler, or you sample at a higher sample frequency, the result will sound better.
> 
> Does that help at all?
> 
> Thanks,
> Zoë.
>

Hiya!

OK, so when you sample something, you're basically storing a very blocky picture of it.  The higher the sample rate, the smoother the picture, the lower the rate, the blockier the picture.

As a separate thing entirely, bear in mind that any given waveform can be said to be made up of lots of sine waves at different frequencies.  So, for instance, a 6kHz triangle wave consists of a 6kHz sine wave, a 18kHz sine wave, a 30kHz sine wave, and so on.  The higher the sine wave frequency, the quieter it is, but they're there.

The sample rate you're recording at, how blocky or smooth the representation of the sound is, should be double the speed of the *fastest* sine wave you want to hear.  The more of them you capture, the more accurate the sound.

But you don't really need to know all that, I'm just trying to explain what the "should be at least twice the frequency" bit means.  It's NOT twice the frequency of the note, but twice the frequency of the fastest harmonic in that sound that you want to hear.

If you want to know all about sampling, check out these articles:

http://www.soundonsound.com/sos/aug00/articles/synthsec.htm
http://www.soundonsound.com/sos/sep00/articles/synthsec.htm

Basically, in order to get the best sound possible, you want to record a really low frequency sound at a really high frequency sample rate.

As far as wavetable synthesis goes, however (such as the other feature of the A-112 module), you don't need to manually sample this from another synth.  It's far easier (to certain geeky types of people, anyway, heh) to make these sounds from scratch on a computer, without them existing as sound until you've copied it across to the A-112 and played it.

I happen to have bought an A-112 module the other month, so I'll look into making some handy wavetable .syx files for everyone in a little while.

With wavetable synthesis, if you were to record an oscillator, it would have to be at an exact frequency that matches the frequency the A-112 is recording at, as you need to use exactly 256 samples to store each and every cycle of the waveform.  This would be nigh on impossible to get just right, hence making the waveforms digitally on a computer and copying them across.

But really, as a musician, you don't need to know all this.  All you need to know is that if you play a lower note into the sampler, or you sample at a higher sample frequency, the result will sound better.

Does that help at all?

Thanks,
Zoë.

Hi Andy!

I do think you have an answer to my predicament. You seem to have explained what I am hearing, it does sound like a glitch caused by mismatching of sample rates and wave form frequency.

That number of samples error would easily be audible.

I've been dissapointed for 25 years about this!

How the heck do i get a sampler that samples like 100Khz?

Not really in Synclavier money range.

The point of all this is could I make descent wavetables for the Doepfer module.

The example  I have heard sounds like a ring modular applied to a tape recorder recording. 
--- In Doepfer_a100@yahoogroups.com, andy butler <akbutler@...> wrote:

>
> 
> 
> anoop.sahal wrote:
> 
> > Well, the sample rate was 24Khz for a 1Khz wave form.
> > 
> > Still sounded like a buzz.
> > 
> > I am missing something.
> 
> If frequency was absolutely spot on 1KHz that would have been exactly 24 samples
> for one cycle.
> That should have been ok, provided that the Akai can loop a sample that short (?),
> and it isn't trying to do some kind of crossfade (which under more normal
> use would help, usually you'd loop a much longer fragment).
> 
> ..but just a fraction off 1KHz, or if the sampling frequency isn't
> exactly 24KHz then the loop you're making is still 24 samples, but
> it won't fit precisely. There'll be a small glitch at the join
> ...and that will buzz like anything.
> 
> If you're sampling 6kHz then that's 4 samples and massive room for error.
> 
> Just the slightest error in how the loop joins up will strongly colour the sound,
> Whether the waveform doesn't match across the join, or the Akai does anything other
> than allow the raw digits to be sent to the DAC.
> I suspect that's what you're hearing. 
> 
> So, try sampling 100Hz.
> Try looping several cycles.
> 
> hope that helps,
> (my experience is with simulating wavetable sounds with digital editing).
> 
> andy butler
>

Hi again Anoop!

Yes, that's it exactly!

Sorry, I hope I wasn't too patronising earlier, I never know how much the person I'm talking to already knows... in this case, quite a lot by the sounds of things!

On an interesting and related note, you know how a wheel spinning really fast in one direction looks a bit like it's spinning slowly in the other direction?  You can get the same thing with sounds too, so (if I've got this right) if you've got a sample rate of 10kHz and one of the harmonics is 12kHz, it will sound like it's 8kHz instead, the sample rate *minus* two instead of plus two.  This is the reason why sampler inputs have a built-in low pass filter at half the sample rate.

Anyway, yes, samplers of that era will *not* sound as pristine as a CD quality sample.  If it's any consolation, I spent many years making music in trackers, starting out with 8363Hz samples, at 8-bit resolution as well.  It didn't sound great, but just the ability to use a sampler was amazing enough that I for one didn't really care. :)  Now it's a whole crunchy aesthetic in its own right!

All the best,
Zoë.

anoop.sahal wrote:

> Well, the sample rate was 24Khz for a 1Khz wave form.
> 
> Still sounded like a buzz.
> 
> I am missing something.

If frequency was absolutely spot on 1KHz that would have been exactly 24 samples
for one cycle.
That should have been ok, provided that the Akai can loop a sample that short (?),
and it isn't trying to do some kind of crossfade (which under more normal
use would help, usually you'd loop a much longer fragment).

..but just a fraction off 1KHz, or if the sampling frequency isn't
exactly 24KHz then the loop you're making is still 24 samples, but
it won't fit precisely. There'll be a small glitch at the join
...and that will buzz like anything.

If you're sampling 6kHz then that's 4 samples and massive room for error.

Just the slightest error in how the loop joins up will strongly colour the sound,
Whether the waveform doesn't match across the join, or the Akai does anything other
than allow the raw digits to be sent to the DAC.
I suspect that's what you're hearing. 

So, try sampling 100Hz.
Try looping several cycles.

hope that helps,
(my experience is with simulating wavetable sounds with digital editing).

andy butler

You don't need a sample rate as high as 100kHz.

OK, so A4 is 440Hz.  You could sample that at 112,640Hz to get 256 samples per oscillator cycle, sure.

But you could instead record an A1 (which plays at 55Hz) at 14,080Hz to achieve the exact same result.  This is one reason why most samplers let you specify what the original note of the sample is.

With a piano or violin, playing an A1 sample fast enough to get an A4 note will sound atrocious, but with a static waveform, it will sound the same at any note.

Bear in mind CD players are only 44.1kHz, and they sound just fine. :)

As far as being one or two samples out goes, in my experience it doesn't make that much difference.  If a sample loops badly, going from a very low point straight to a very high one or vice versa, it will make a click sound, and if it's only a single cycle, then the click will become an integral part of the overall sound.  That and a shape that's cropped or has space after it during each cycle will sound different, but it would need to be more than a few samples out to be really noticeable, I believe.

Hmm, thinking about it, it would be possible to write some software that reads in a waveform, finds the first 256 zero-crossings, and stretches or squishes each resulting cycle into 256 bytes each, which would completely automate the process of wavetable-ising a waveform for you, without you having to worry about the original pitch or sample rate (other than for improving the quality).  Tempting...

Hello

though Zoe explained the center of topic bravely. One hint to a very 
clearifying video I saw recently:

http://www.youtube.com/watch?v=cIQ9IXSUzuM

It explains a lot of things very clear, what is needed and what not 
(especially, why you don'tneed a 100kHz sampler).

Florian

anoop.sahal wrote:

> Hi Andy!
> 
> I do think you have an answer to my predicament. You seem to have 
> explained what I am hearing, it does sound like a glitch caused by 
> mismatching of sample rates and wave form frequency.
> 
> That number of samples error would easily be audible.
> 
> The point of all this is could I make descent wavetables for the Doepfer 
> module.

As I understand it, (and I'm by no means expert on the Doepfer module).

The Doepfer wavetable is really small, maybe 128 samples(??).

So by editing you would create a single cycle of your waveform that
fits exactly that number of sample.

Then the module will play back at whatever sample rate is needed to produce
the pitch you require.

 
Really what you need is software like Adobe Audition which lets you
edit at the sample level, and preview the result of a loop.

To get a triangle wave you could generate it in software at the precise frequency.
Or sample it and resample it (in software) to be the exact frequency.

If it sounds right when looped in software (and it probably takes a number
of tries to get it right) then it's ready to be loaded to the Doepfer module.

> 
> The example I have heard sounds like a ring modular applied to a tape 
> recorder recording.

If there's tones that aren't harmonically related then you're hearing aliasing,
which is what Zoe explained with the spinning wheel analogy.
google "digital aliasing", it's one of the big topics in any textbook on digital sound.


andy

> As I understand it, (and I'm by no means expert on the Doepfer module).
> 
> The Doepfer wavetable is really small, maybe 128 samples(??).
> 
> So by editing you would create a single cycle of your waveform that
> fits exactly that number of sample.
> 
> Then the module will play back at whatever sample rate is needed to produce
> the pitch you require.

Yes, that's right, only it's double that amount.  So in sampler mode, it stores a single sample that's 65536 samples long (if you see what I mean), and in wavetable mode, it stores 256 of them, each one 256 samples long.