Sample rate

[MackyParticipant]

I use the mixer for recording with Logic x, sometime I had to Work with project at 44.100. I’d like to cha te the siamole rate

[[XAP]BobParticipant]

Re sampling is at least lossless, even if it’s not cpu cheap.

Unfortunately it would be cup expensive in the QU as well…. I know we got stereo playback in various sample rates and bit depths, but I’m not convinced there will be enough spare cpu to re sample 64 channels (in and out) on the fly.

I might be wrong, but I’d suggest that an offline re sample in logic is the way to go.

[AndreasModerator]

The Qu is running at fixed 48kHz, this is a design decision to reach the price tag of that unit.
I totally understand the request for 44.1kHz since I did CD productions as well, so some sort of re-sampling will be required.
Personally I’d place the resampling right at the end of the production chain when exporting the downmix, hoping that Logic does a good job there. Checking sound of resampled cymbals easily show up issues there.
Changing the samplingrate of the recorded files first and operate the project on 44.1kHz may be an option as well, but if you later recognize issues from the early downsampling, there’s no way back.
Downsampling from 48 to 44.1 khz isn’t easy, since it has to be done correctly to avoid audible artifacts to pop in. It can not be lossless, since you have to cut frequencies at the end of the audio range which requires really good and steep filters to accomplish that. Upsampling is a lot easier…

[MikeShandParticipant]

Voxengo’s R8Brain (free) usually does a pretty good job.

[GCumbeeParticipant]

We have been fighting this battle for 20+ years now. I was pushed in the 90’s to record at 48K. Then in 2000’s to 96 and 192. I fought it hard. We always had to come back down to 44.1/16bit for CD and conversion was always and still is an issue.

I had studio clients in mid 2000’s that made the decision to just stick with tracking and mixing at 44.1/16. Some 44.1/24. It seemed bit conversion was less damaging than sample rate. It would not have bothered me if AH had chosen 44.1 but then some would not have been happy with that. You can’t win with this.

[AndreasModerator]

Since I’m (currently) focusing on video, I’m very happy with the 48khz, but sometime I’ll come back to CDs and then I wish to have 44.1 too…
Operating a project on higher samplingrates may make sense, since virtual instruments can easier operate inside the safe-area in terms of Nyquist and filters have a somewhat higher precision. Of course you need to come down to 44.1/48kHz finally.
Downsampling from 96/192 to 48 resp. 88.2/176.4 to 44.1 isn’t that big deal as well, since this basically requires a brickwall filter and dropping of unneeded samples. Odd conversions (i.e. 96 to 44.1) require additional math (interpolation) to generate samples at time locations not contained in the original material.
Regarding bit-depth its indeed easy to drop bits at the end (downmix), but for recording I really recommend to pick at least 24 Bits since it allows to record with some headroom and still allowing larger gain manipulation (compressors) without sacrificing quality.
The correlation between dB and Bits is simply 6dBfs/Bit, since each additional bit doubles resolution (or output voltage when keeping same resolution). That’s where the 96dB SNR from CD is coming from (16*6=96). So 8 Bits more provide about 48dB more flexibility.

[[XAP]BobParticipant]

Those interpolations are what I meant by lossless. There are obviously frequencies which get lost, but since they are inaudible frequencies the whole operation can be considered lossless (ok, you can get interesting artefacts around those very high frequencies, but both 44.1 and 48kHz have quite some margin over even the best human ears).

The digital waveform isn’t a stepped wave, it’s a lollipop graph. It is so easy to forget that there is only one analogue wave which fits the lollipops, here isn’t any (audible) information that isn’t encoded on the digital waveform…

However in the QU we work at a fixed 48kHz – this makes the internal operation nice and simple, which brings the price point down to the price we paid…
There is a cost to resampling, and we didn’t pay it. We have had a bonus in terms of stereo USB playback from different sample frequencies and bit depths – but I would suspect that we simply don’t have enough cpu in the QU to cope with resampling all the inputs…

[AndreasModerator]

Sorry, Bob, that I need to correct you (dealt too many years in synth design and all the Nyquist stuff).
Sure, frequencies above 16-20kHz can’t be heard, depending on age and music volume preferences… But the math behind sampling-therory does not care about this and also will handle frequencies above that margin.
If you do a quality recording of some (natural) instrument at 48kHz, the sample stream likely contains frequencies up to 24kHz (half of sampling frequency). Downsampling to 44.1kHz without previous brickwall filtering at 22kHz will fold back any higher frequency parts back into the allowed range. An unfiltered 23kHz signal will then not only occur at 21kHz but also at 1kHz, which will be audible by hopefully anyone of us.
Even if the energies from these bands are normally very low, they’re easily recognized as wrong, since they’re strongly non-harmonics compared to the remaining stuff.
The sampling theorem isn’t very tolerant in that respect, I’ll tell you… 😉

[MackyParticipant]

The question was only because my friend Andrea most of the studio that Work with me use 44.100 and send me files at This sample

[[XAP]BobParticipant]

That filter is why there is a significant cpu cost – I’ve never quite got my head around where (in the frequency spectrum) things get folded, but I do know that a decent re sampler will kill them dead (I.e filter them out before the resampling).
If you up sample (as per the OP) then you don’t have the folded audio, because the signal is already bandpass filtered, then the subsequent down sampling (because they’ll presumably want 44.1kHz back) should also be perfect (I think).

I think we’re probably singing from a very similar hymn sheet. The main point is that resampling isn’t “guesswork”, it’s an exact science. Loss of bit depth is slightly less accurate, but we’re so far above the point where it makes a difference (assuming a good dither) that no one will hear that either. But we keep the accuracy whilst mixing – because we’re affecting all sorts of interesting maths on the sound first.

I.e. Don’t be scared of using a decent re-sampler. Do check for band pass filters and dither settings…

[AndreasModerator]

Perfectly stated, Bob!
After my post I’ve tried to find some application to generate a sound example for bad resampling, but actually any I’ve tested were fine (so various filter rolloff, but not significant). So my warning seems to be a little outdated, lesson learned… 😉

[TonyParticipant]

Hi guys i recenntly got a qu 32 for my studio was using presonus 44vsl but now after connecting qu 32 my
Some tracks in my old projects dont play in the correct place in the project. Could it be a sample rate problem

[AndreasModerator]

Same question, same answer:
Possible. Depends how your DAW handles the difference in samplingrate between the interface and your project. Maybe you just need to switch your project to 48kHz to get this fixed, maybe there’s more work…

[Author]

Posts