I posted a similar question in the mastering forum but received no response. I’ll tell you what I finally ended up doing and the results have been good. I set up a montage with 96/24 files (the project I am working on consists of 24 songs). My first few attempts at getting to 44/16 did not sound all that good. My best results happened when I went from the montage and processed the files with a compressor and limiter in the master section and some eq on some of the songs. I rendered this file to a new montage at 96/24. My thoughts were to maintain the higher bit rate files for the processing. I used the newly created montage and then using Crystal Resampler and UV22, rendered a new montage at 44/16. From there I created the cds and DDP files. A few of my failed attempts had the all the processing done in one run - Crystal, comp, limit, dither then I also tried the same with the Crystal at the end of the chain. Neither sounded as clean and open as the final way. My recording consisted mainly of piano and vocals (up to six vocalists on a song) so the mixes were pretty straighforward.
Its safe to say nothing will sound worse working at 96K and SRC’ing later, but some plugs/processors will sound different and generally better. Even simple things like EQ shape can be affected in the 10th octave, when the a-a filters are out of the way. So there’s definitely no harm working at 96/24, and possibly some benefit.
I too thought it made sense to stay in the same SR/Bit depth as whatever files I had coming in from the mix engineer.
The Tischmeyer DVD (available on the uaudio site) offers a pretty specific workflow with Wavelab that I found
useful. He creates a new set of files that remove DC offset, normalize to -3 db and that are 32 bit. So when
you create a new montage, you have files that are somewhat the same volume. Helpful.
jbird: Because the result of the two methods above should be technically identical (if all is done “In the Box”), it seems that if the results sound different, then something is wrong, and that something should possibly be looked at for a fix. If the two methods result in identical rendered files as confirmed with the Wavelab File Comparator, then it doesn’t matter which method you use - the 1 step or the 2 step. If you’re using Waves plugins that integrate Waves dither or noise shaping (which many of their compressors do), then an apparent “random” function is introduced making bit comparison impossible (thanks Waves, not). But I’ve not seen that in any of the Steinberg dithers or plugins (thanks Steinberg). From what I can tell, results using the Steinberg dithers and plugins always result in bit indentical renders as long as everything has the same settings and is in the same order in the processing. It shouldn’t, and apparently doesn’t matter whether this is done in 2 steps or 1 step. Unless I’m missing something obvious. But as I said, this can be easily confirmed with the File Comparator tool.
EDIT: I just tried this with our current Waves (which is still version 7, but newer than the last time I tried this), and the “random” thing appears to be gone. It used to be you could toggle the dither or noise shaping and the random function would come or go. Now there’s nothing random, and multiple renders should be bit identical. Thanks Waves.
Why do you disagree? What Dave says is: If you start with a 96K file it will never be a disadvantage to do the processing at 96K and SRC later rather than SRC first and process at 44,1K, which was the OP’s question.
Maybe you think he said: “Its safe to say nothing will sound worse THAN working at 96K and SRC’ing later”?? (which he obviously didn’t say and didn’t mean).
For a good comparison of SRC software have a look here. R8brain pro is not alone at the top:
Appreciate the comments, but the way the quote is written states quite clearly that there is nothing worse than working at 96k and SRC to 44.1k and it is this I disagree with.
As for R8Brain Pro, on the site you mention there are comparisons galore, and you cannot get better than perfect , which is what their graph shows for R8Brain Pro in Linear Phase mode. There may be others almost as good, but none better which was my point. Another gripe with these comparisons are the obvious biases - there are so many different options on the Reaper SRC, the iZotope SRC etc - yet none of the tweaks in R8Brain Pro are trialled & posted here.
I stand by my original comment - in years and years I still have not found anything better than this.
However, graphs do not tell the whole story, and this graph does not show the ringing artifacts which can and do appear in LP SRC, which is why R8Brain also includes a Minimum Phase option. This gives a very different type of conversion that is almost like (note “like” and not “the same as”) going through a high end reference analogue device.
I’d like to see the SRC comparison site updated to reflect these additional options, as well as the Ultra Steep Slope options (not always a good idea as too steep a slope can also cause artifacts) and the “no clipping” options, as one mistake often made when performing SRC is to forget that this can alter the peak values (which I guess might be another argument against excessive limiting/compression for volume, which is far worse than SRC in the first place)
Well, FWIW, I agree with LutzR in the way I read what Daved said. To me, “nothing will sound worse working at 96K and SRC’ing later” means “nothing will sound worse WHEN working at 96K and SRC’ing later”, not “nothing will sound worse THAN working at 96K and SRC’ing later”.
Odd - and a little disconcerting - to discover how different people can get such different meaning from the same words.