DP6--Improved Sample Rate Conversion

[XYZ]

Infinite wave just published sample rate conversion plots for DP6. You will see it is still not perfect, and not the best of the best, but it is now nearing the head of the pack. Clearly, MOTU has done some algorithm improvements in this area.

Just compare DP6 to DP 5.1 and you will see what I mean.

http://src.infinitewave.ca/

[Shooshie]

What an EXCELLENT resource! Thanks. It's very interesting to explore the various titles out there to see how they far with sample rate conversion anti-aliasing. Some, like Ableton Live (regular) appear to be almost unusable. Then there are Pro Tools 7.x, and Logic 8 (Leopard), which match Core Audio -- zero artifacts, over a slight noise floor. So which is more desirable? Virtually-inaudible artifacts that occur at the level of Core Audio's noise floor? Or a constant noise floor that masks all artifacts and makes for zero detectable artifacts? DP has the higher signal/noise ratio, with artifacts at, let's say, -xy dB, where -xy dB is virtually inaudible. Core Audio, Logic, and Pro Tools have zero artifacts, but their constant noise floor also = -xy dB. In both cases, -xy dB is inaudible to human ears on commercial playback equipment at normal volumes. So which is preferable? Is the fact that the artifacts are even present, even at -xy dB, indicative of potential audible problems after multiple processes? Or is it more likely that in compression or limiting, we'll bring up the noise floor in Logic or Pro Tools to audible levels?

Well, that's a very interesting page. I have a desire to put that in the DP tips sheet, though I don't know if it's beneficial that way. I think I'll do it just so the link will always be handy.

Shooshie

[Shooshie]

Check out "eac3to (SSRC)" in the graphs. Perfect. No noise floor to speak of, and no artifacts.

Also, "iZotope 64 bit SRC Steep, No Alias"

At the other extreme (bad), check out "Reaper 2 ("Ir2x2" Preset)" or "Sadie"

I was surprised to see the results for Waveburner 1.2. Absolutely terrible! I hope my version (1.5.2 fares better)

[kassonica]

Thanks for this.

I've just done a project at 88 and after bouncing I noticed (what I thought a big difference) in the sample rate conversion but put it down to my mind was playing tricks because this was 6 it must be better but this confirms what my ears told me.

I've always been very critical of DP's sample rate conversion and dither algorithms and have hoped they would work on them and it seems they have YAY.


I wonder if they have tweaked the dither as well ?

[jlaudon]

It was after checking out that web site that I ended up buying Sample Manager - it incorporates Izotope's SRC and also Mbit dithering. It makes a difference after using DP's SRC.

OT, but since someone mentioned Waveburner, it might be good to check out Wave Editor - a wave editor (duh) and mastering software - also the same company (Audiofile engineering), as it also has this Izotope SRC and dithering (and it also imports/exports DDP 2 files - which nothing else at this price can do - $79).

[XYZ]

Also, remember this challenge is a hard one, it is not from 88.2 to 44.1kHz, rather from 96 to 44.1, so this is a hard one to do. If you have a choice, you can not worry about it by recording at 88.2 if you are going to end up on a CD, or 96 [to go down to 48] if you are working on movies.

What I liked is that MOTU has improved the under the hood processing to pro levels, so no need to use another product, unless you are doing a lot of back and forth, or, as shoosh says, you are compressing the heck out of things, and have a lot of dead air. I mean those numbers are down to -150 or so dB. So if you are compressing by 20 dB [I'd consider that a lot] this would bring up this floor to -130. That is extremely low, even with a lot of compression.

Further, don't we all ultimately truncate to 16 bits which is, at very best!, -96 dB?

To further what shooshie was saying, anything below about -100 dB is getting a little crazy anyway. When was the last time you recorded a track where the microphone self noise+preamp self nose was even below -90? For me? I'd bet never. And now lets add in background room noise.

Finally, this ultra low level hash would have to be correlated to add up from track to track. This would only be the case if all the tracks had exactly the same frequency components. If that were the case, this would be a boring song to say the least. So, yes, I'd rather have uncorrelated artifacts vs. a constant noise floor [which will add with each track], but even 8 tracks of the "pro tools" example would be a 9 dB increase, worst case. Now you'd only be at a noise floor of -160 to -170 dB. These are still "massively small" values.

Certainly a low enough "clean" floor for anything but the most incredibly demanding situation, and probably even excellent for that. If I evern do recording that is requiring 140 dB of dynamic range, I'm going to be interested in what audience exists that can hear it.

Anyway, thumbs up to MOTU for improving the "under the hood" DP engine.

[bigcityrecording]

Or, you can do what I have been doing for years: I record at the highest sample and bit rate, then master in the analog domain. This way there is no SRC - just high resolution becoming analog and then shaped and molded before being resampled at 44.1/16bit.

I have always felt the result was better that way.

Eventually, it will be better to stay digital 100%, but perhaps not yet...

[Frodo]

There's just something about analog that cannot be escaped---- yet.