Sound engines in samplers? [solved]

Samu

Sampling at alternate sample-rates and bit-depths can also give the sound some character. I had a blast going thru the factory samples of the Electribe 2 Sampler and there the sample-rate was used creatively and was set to just about cover the frequency-range of the sounds thus optimizing storage. Some sounds were sampled at very low sample-rates and some were sampled a lot higher than what is possible with the unit it self

So In my opinion it should not matter what sample rate the source sample has, it's upto the playback engine to handle the up/down sampling and give the user the creative freedom to choose how it's done.

The 'output' sample-rate should however match the de-facto standards regarding sample-rate and bit-depth.

wim

@Samu said:
Sampling at alternate sample-rates and bit-depths can also give the sound some character. I had a blast going thru the factory samples of the Electribe 2 Sampler and there the sample-rate was used creatively and was set to just about cover the frequency-range of the sounds thus optimizing storage. Some sounds were sampled at very low sample-rates and some were sampled a lot higher than what is possible with the unit it self

Two words: Blamsoft Resampler

Samu

@wim said:
Two words: Blamsoft Resampler

Is one of my favourite AU plug-ins

[Deleted User]

@mschenkel.it said:
@soundklinik

Sample rate sets the Nyquist frequency which in other words is the max frequency for that sample. the maths are pretty easy: since to have some sound you need to have a oscillation from compressed state and in compressed(+1<X<-1) it means that a frequency is determined by the time passing by from a peak to the next(from+1 to -1) so it means you need two samples for any given frequency. So if you have a 16k samples per second the max frequench you can get is 8000hz which is just one octave lower of my hearing range, while with 44.1khz you can get up to 22050hz note/sound. Loads of vintage gear is way behind the 44.1 cd standard. All in all I would recommend to use 44.1 samples; the bit depth instead can give some character adding harmonics, which is why many old synthesizers sound crunchy and cut so well through mixes so that could be much more up to tastes, specially when doing creative sampling(but it is a no no when trying to achieve hi quality samples)

Thanks for the detailed info and explanation.
I had most samples in 16 and 24 bit @44.1 kHz and bulk converted to 16/16 stereo to make sure I didn't have to always convert in AS....because of some apps...and also for space on my old 16gb air1 but I didn't take in account the quality

Will

@samu is of course right about 'creative' sample rates but if you're just looking for accurate reproduction of your sample library, 16/44.1 is the way to go. Plus, as you saw with Modstep, 16/44.1 is generally going to be way more compatible across apps.

For storage reasons, would be sweet to see an automatic sample rate conversion tool that analyzed the spectral content of each sample and reaampled it based on Nyquist. Sox or something could probably do it.

Samu

@syrupcore said:

>

For storage reasons, would be sweet to see an automatic sample rate conversion tool that analyzed the spectral content of each sample and reaampled it based on Nyquist. Sox or something could probably do it.

This would be a nice addition to a future version of AudioShare.

I actually used something similar to sox when my Electribe ES-1 was still working to convert/re-sample everything to 16-Bit 32Khz and I actually preferred the 'aliased' pre 16Khz hi-cut filter conversion/re-sampling

Will

^ why I like you @samu.

u0421793

I know it’s none of my business, but while you’re checking the new sample handling arrangement, check the linearity — i.e., is the measured 50% level of the source sample the same as the 50% of the outgoing signal; the 25% in the same as 25% out; the 75% in, 75% out, etc. If they’re skewed, it might indicate something
i.e., if they’re all up, or down, by the same amount,
vs a situation where 100% and 50% are equal to original, but 25% is down and 75% is up,
vs 25% down a lot, 50% down some, 75% down a tiny bit, 100% equal to the original.

The reason I mention this is because of treatments to the least significant bits, to remove the noise floor, may leave (for example) a 16 bit sample with only 14 bits of signal, and depending what happens to that afterwards, may be allocated or handled unexpectedly. Just mentioning.

Samu

@u0421793 said:
The reason I mention this is because of treatments to the least significant bits, to remove the noise floor, may leave (for example) a 16 bit sample with only 14 bits of signal, and depending what happens to that afterwards, may be allocated or handled unexpectedly. Just mentioning.

This is a valid argument as this type of editing is missing in many sample editors.
Some of the earlier 8-bit samplers (DSS-8 for the Amiga as an example) had 'built-in noise' and the only way to remove it was to clear the least significant bits to make 'clean' samples

This also leads to other interesting discussions regarding bit up/down sampling with or without dithering-noise and with or without interpolation. This is 'geeky' stuff I know...