Hi, just wanted to share with you test results of Sampler resampler (Adjust):
Not bad although there’s space for improvement - some distortions exceed 16bit noise floor, it’s always nice for a DAW to keep things closer to at least 24bit.
Kudos for impulse response with no audible ringing and linear phase.
Live playback in Renoise typically uses Cubic interpolation – it depends on your individual sample settings, but it’s typically Cubic by default. (some even turn it off completely for that OG Amiga/Atari lofi sound)
When rendering to disk, the “Precise” option uses much cleaner Sinc interpolation instead, which in most cases solves the usual anti-aliasing issues… but you should be aware that Sinc can sometimes introduce its own weirdness to certain sounds… the “perfect” math sometimes creates imperfect results… waveforms that might clip due to the anti-aliased smoothed peaks going beyond 0dB, for example… so it’s not a guaranteed clean solution.
At the end of the day you just have to use your ears… Does what you hear in Renoise live playback sound good? If so then just accept that kickass sound!
Don’t rely on those SRC results too much. Just use your ears.
Would be curious about a test in “precise” rendering mode.
Also Renoise live-resamples, maybe I am wrong, but aren’t the usual audio tracks in DAWs precalculating instead?
So I would be also curious about a result of Kontakt, or Bitwig Sampler module, doing the same sweep test. So other “live-resamplers” in comparison.
EDIT: So here is a lossless 96kHz sweep sample thrown in and you played it at C-4? I think C-4 then is the exact samplerate playback you set the audio device to, e.g. 44,1kHz. You are mentioning 44kHz though. Maybe I am completely lost with the test setup regarding Renoise and mixing up things here…
That’s a really good point about comparing it to other ‘live’ samplers like Kontakt or Bitwig. It would be interesting to see if Renoise’s trade-off for CPU efficiency actually shows up in a side-by-side test. I’d definitely be curious to see a ‘Precise’ mode render test too if anyone has the time to run it!
@dblue I surely am the one who can’t listen to anything Amiga if it goes thru anything else than crude ZOH “interpolation”
@ffx What I tested above is Sampler’s resampler which you rarely use for one time conversion of imported sample if it isn’t already in a format you want (Adjust button). Test needs 44,1kHz and highest supported bit depth. Live resampler is something else and expected to have worse parameters but it needs to be fast with minimum delay so you can use it for realtime composing and playback (selectable in Sampler, including crude ones having that retro signature). VSTs use their own resamplers, you can check if they are tested. Then if device is set to other rate than Renoise, it may go through Windows resampler (not very bad in 10 & 11, test available). Render uses separate resampler, which I tested below.
@EsmeD03 I did take a hassle to put all the test suite samples into separate tracks as C-4, render in Precise HQ mode to 44,1kHz 32bit, and cut manually to their original length:
It does have audible issues.
Could use stronger lowpass at Nyquist to avoid that -24dB aliasing/imaging.
Intermodulation Harmonic Distortion at -40dB.
Impulse frequency response could be better too.
DAW doesn’t want any distortions above -96dB, preferably -144dB.
Gapless support would be great in post production, but no DAW supports it yet (although it’s very doable, with open sources).
Surely space for improvement. There are opensource implementations with all the great features needed for offline rendering.
@rutra80 Thanks for the new test. Also make sure that dithering in the audio preferences is disabled, and also soft clipping in the end of the master chain. I think I don’t understand you response
@dblue If test suite was loaded as samples, each set to AA & Sinc, played back at C-4 - did they get processed by Sampler interpolation or went only through Render Precise HQ resampler?