Removing digital noise from old samples?

Apologies if this is the wrong venue. If so, perhaps someone with more experience can point me in the right direction.

I’ve imported about 700 old samples from an 35 year old hardware sampler in order to create a custom sample library. They’re usuable, -except- for some digital artifacts–which are OK until the sound decays. It’s 90% ‘noise’ but also contains all these tiny flakes of ‘digital’. And then there are the attacks, which often have an ugly digital ‘click’.

Can WL do these kinds of fixes in batch, without having to tweak each individually? Or are there more appropriate (perhaps ‘AI’) tools to handle this en masse?

TIA. Happy Holidays

You can try this Steinberg plugin:

This is usually a sign that the samples were originally recorded at low resolution, e.g. 8-bit. If that’s the case, there’s not much you can do, as information would never have been captured in the first place, i.e. data loss.

1 Like

Is it possible to mask with dither?

regards S-EH

I have some limited personal experience in the context of some old Ensoniq ASR 10 samples off a set of Ensoniq DVDs.

The digital artifacts are almost certainly ‘embedded’ and way too strong for something like a properly implemented dither to mask.

As PG suggests, sound repair solutions or simply accepting them as they are maybe the only practical options.

There is no substitute to treating them individually … although if it is a personal project you might be able to just do the ones needed as they are required.

1 Like

As each sound decays there is this subtle ‘DX7’ style FM distortion. Like tiny ring modulators off in the distance. :smiley:

So far the best luck I’ve had is running them one by one through Spectralayers.

I’m sure at some point there will be a way for Spectralayers to ‘learn’ this kind of noise footprint, then use WL’s batch facility to ‘pipe’ each sample through SL and automate the process.

Until then…

That’s most likely aliasing then … the original sound source that was sampled had frequency content higher than half the sampling clock frequency. Older hardware had to use a lower sampling frequency in order to reduce the amount of storage to something that would fit in a bank of EPROMS. Unfortunately, when reproduced, this aliasing “folds” these false frequencies back into the audible band, where you hear them. It can’t be filtered out using conventional techniques after the fact … anti-aliasing filtering needs to be done on the analog signal before A/D conversion.

The possibly good news is that these will have some kind of characteristic pattern, which may explain why SL might work better. On the other hand, I see people paying good money to get this kind of “character” (a.k.a. noise) back into their recordings … :smiling_imp:

1 Like

But what did you do in SpectraLayers? How did you manage to reduce the noise?

As you mentioned, Wavelab batch processor would be a better option than SpectraLayers since you have 700 samples. For the batch process, it will probably a matter of finding the right specialist plugin(s) and assigning these in the custom plugin chain. Did you try the Acon Digital restoration plugins?

If you could upload one of the samples maybe others here would have a better idea of the nature of the digital noise and how best to remove it.