Sample rate - looping single cycle analog waveforms for blof

[insilico]

Im trying to make a multisample of some single cycle waveforms from analog synths. I recorded the raw waveforms of each note of my juno 106 and then cut out single cycles and looped them for my blofeld. Problem is, i recorded at 96k 24bit but it is recommended by waldorf to use 44.1k 16 bit. I tried this sample rate but it is almost impossible to make a perfect loop where the sample points hit the zero crossing. Often at higher freq's the zero crossing will land between two sample points. Is there a different technique i Should be using for sampling and looping? Is there a work around?

[Ashe37]

Im trying to make a multisample of some single cycle waveforms from analog synths. I recorded the raw waveforms of each note of my juno 106 and then cut out single cycles and looped them for my blofeld. Problem is, i recorded at 96k 24bit but it is recommended by waldorf to use 44.1k 16 bit. I tried this sample rate but it is almost impossible to make a perfect loop where the sample points hit the zero crossing. Often at higher freq's the zero crossing will land between two sample points. Is there a different technique i Should be using for sampling and looping? Is there a work around?

Multisampling every single note is going to eat the RAM and polyphony of your blofeld for breakfast. Have you tried using a single sample to see how much aliasing you get at the high registers?

[insilico]

Yeah i first tried sampling one note per octave but it sounded dull or had wierd artifacts if transposed more than a couple semitones. RAM isn't an issue because single cycles are so short and only take a couple of kb. Polyphony hasn't been an issue either so far. Im mainly concerned about making accurate loop points without affecting the sample.

[Ashe37]

Ok, try sampling at a multiple of 44.1 (say, 88.2) at 24 bit, make your loops, then resample them to 16 bit 44.1

[cartesia]

or just use 2 cycles instead of 1.. takes up more space.. but also, there might be changes in the sound output occuring over periods longer than a single cycle that give the synth it's character

[Shreddie]

I tried this sample rate but it is almost impossible to make a perfect loop where the sample points hit the zero crossing. Often at higher freq's the zero crossing will land between two sample points.

That is to be expected. Your sample rate is fixed at a specific frequency but the frequency of your waveforms is not, so there will be times when they simply won't match up on single cycles because they are effectively out of time with eachother. How often they don't match up will increase as the frequency of your sample increases, this is because it's approaching the nyquist limit and there are less samples used to capture each single cycle, so the chance of finding correct loop points for single cycle waves reduces exponentially as frequency increases. When sampling at 44.1kHz, at very high frequencies (15+kHz), correct zero crossings may be getting on for a second apart.

However, you may get perfect loop points at 2, 3, 5, 7, 11 etc cycles... I have some experience in this and it does seem to be mostly prime numbers (of cycles) on which the correct loop points/zero crossings will match up perfectly... There is a formula you can use to work it out for specific notes/frequencies at your chosen sample rate but I'm not at home so I don't have my book of secrets handy.

Is there a different technique i Should be using for sampling and looping? Is there a work around?

Not as such... You could use crossfades but auto crossfades don't usually work well for very short samples so you would have to do it manually which would mess up the sound a little... I avoid crossfades as all costs, especially with shorter samples. Either use short multi-cycle waves where appropriate or only use the single cycle waves that do land on the correct zero crossings and transpose them as required... Personally, I used the longer samples for the project I worked on a while back.

[Shreddie]

or just use 2 cycles instead of 1.. takes up more space.. but also, there might be changes in the sound output occuring over periods longer than a single cycle that give the synth it's character

That is very true as it happens... For this kind of thing, the longer the sample the better... I'm currently sampling the Andromeda for another project where size is less of a concern and I have noticed a couple of things...

Firstly, the oscillators themselves (filters are bypassed) have a bit of an envelope... Initially they have a higher amplitude and have a slight positive DC offset but settle down at about one second... After that first second, there is a continuous but very subtle variation in the waveforms... Trying to take single cycle waveforms out of these sucks them completely dry of life so I'm sticking with loops of several seconds.