Is there any benefit of converting tracks that were recorded already as 24/44 to 24/48 before mixing?


No. Think of it this way… Printing a black and white picture in color will still produce a black and white picture. In this case… nothing is being added or can be added by going to 48 khz.

I would never go to 48kHz >>unless<< you had gear such as a Soundblaster or ADAT, etc. that works natively at 48k.

That bounce back down to 44.1kHz when you eventually make a CD could cause you more problems than it’s worth to use that 4kHz higher sample rate. And you’ll only be packing zeros on the front of the data anyway when you upsample it.


I would never go to 48kHz >>unless<< you had gear such as a Soundblaster or ADAT, etc. that works natively at 48k.

I don’t know how I feel about that really. I know the sample conversion is not a clean conversion like from 88.2 to 44.1 and antialiasing etc are factored in. I have been going 24/48 for years now and have been quite happy with the results. Of course, this is 24/48 in the raw tracks and the ONLY sample rate conversion I do is at the very very end of the mastering phase. Anyone else have thoughts?

Thinking about it, I always worked in 48khz because I had an old EMU APS card based on the SBLive which had the whole 48/44.1 screw ball thing going on. I will have to try 44.1 with my new EMU 1820M to see if there really is any advantage. The EMUs have such great convertors and clocks, I would hope the high end filters would be cutting off cleanly above the audible range. I KNOW I can’t hear crap above about 16.5 or 17 khz…

Hmmm, high frequency audio perception is a tricky subject. Test have been done (I recall Hi-Fi News doing this in the late 70’s) that show how we can detect the presence of a low pass filter that is above our tested HF hearing limit. I.E., people whose hearing is measured as deaf above 15kHz could hear an 18kHz LP filter being switched in and out, statistically reliably, in double-blind tests.

We don’t hear like a measuring instrument, any more than we see like a camera. Our ears and eyes collect an imperfect assortment of cues and clues and our brain maps out some interpretation of reality which is good enough to ensure our survival, sensing predators and our own prey etc. I’ve always thought our ears can hear into things in ways that measurements don’t reveal. For example, I still think good vinyl discs sound better than CDs, in spite of the background noise. And valve amps seem to sound better than transistors, although maybe that’s because we like a bit of 2nd harmonic distortion, in the same way that we like too much salt, sugar and fat in our food.

Anyway, I guess my point is that we should not rely too heavily on measurements; what we percieve as being better is what matters. We can train our sense of hearing to be discerning and to appreciate well-made recordings but that don’t mean what we hear is natural or in any way accurate or correct.

Merry yuletide wishes to all.

I usually use 48 KHz to record because I want the quality I can get through my ADAT, and 96 ends up taxing the computer too much with high track counts. If I didn’t use ADAT I’d probably use 88.2 when I could.

I used to use 16/48 due to the old “soundblaster” thing. For the last couple of years I’ve been recording at 24/44.1. Not really found the need to go any higher than that.

Bubba post #1: I agree.
Tim: I agree
Bubba post #2: I have no idea what you’re trying to say. ???
Tuster: your argument applies more to recording at 48k or higher than to recording at 44.1k and converting to 48k for mixing.
Guitars69 and Mark A: Agree.

There is one case when it’s worthwhile to convert up to higher rates, but you’d want to at least double the rate for it to be worthwhile. Some chorus and pitch-shift plugins use a variable delay line. In this case, the more samples there are per second, the better the results – regardless of the fact that there’s no extra information in the extra samples when they were created by upsampling. It’s due to the fact that these plugins are taking nifty little shortcuts that are sensitive to samples per second and has nothing to do with sampling theory per se. Note that most pitch-shift plugs use Fourier transforms, not variable delay lines. However, delay-line PS plugs have the advantage that they don’t mangle the phase information and can be great for use in building stereo images from complex mono signals. But understand that this is a strange corner case rather than a contradiction of the general concept about upsampling: it’s not just a waste of time, it does add artifacts because the math as implemented on a computer is an approximation. The difference between high-quality resampling algorithms and low quality ones is minimizing the impact of round-off errors.