oversampling questions on Mac

[cfmsp]

Chris,

(note: I've made edits based on more education since posting this earlier)

I have a couple of questions about use of upsampling. I've been a straight RBCD listener up to now.

 

The available software/hardware:

 

MBP: iTunes / AudioMidi;

Metric Halo ULN-2 Dac (with on-board upsampling and internal clock);

Wave Editor 1.4.4 (with iZotope 64-bit SRC)

 

 

scenario no. 1:

 

RBCD played by iTunes via Firewire to ULN-2

 

I'm assuming here that I want to keep AudioMidi set to 44.1k, and I'm assuming that if the ULN-2 is set for 88.2k that any upsampling will then only occur within the DAC, and not by iTunes / CoreAUdio, is that correct?

 

scenario no. 2:

 

RBCD ripped, converted to 88.2k, and played back by Wave Editor 1.4.x - via Firewire to ULN-2 which is set at 88.2k.

 

It looks like Audio Midi Setup changes settings in the ULN-2. I"m assuming that I should always try to set Audio Midi to match the material. If there is a mismatch - I have the same question as in Scenario 1 - will sampling rate conversions occur in/by the DAC and not within iTunes/CoreAudio?

 

 

scenario no. 3:

 

same file as in no. 2 is now saved as AIFF file and the played back via iTunes (latest version).

 

I'm assuming there is a easy possibility to screw this up (i.e. unwanted SRC by CoreAudio or iTunes).

What is best practice for iTunes and non-RBCD sample rates?

 

I've read on the Benchmark wiki site that iTunes SRC is high quality and therefore it is not a problem to set Audio Midi at 96k and leave it, although I have a hard time believing that would be a good thing (given that the same wiki says that it is NOT a good idea to upsample - and this from a manufacturer who upsamples everything to 110khz as I understand it).

 

I'm also assuming that if upsampling IS good, then why wouldn't I want that to done by Wave Editor (with their well known SRC software) or by the ULN-2 rather than iTunes?

 

Question regarding the appropriate place for upsampling - there seem to be three (four if you count CoreAudio) alternatives. 1) Rip & convert before saving file. 2) Upsample within iTunes (CoreAudio, if you're unlucky) or 3) Upsample within the ULN-2. Are any of there preferable with regards to upsampling RBCD - or are the 'benefits' of any of them negligible/questionable?

 

 

this is my first foray into oversampling and I'm asking about 88.2k as it seems to be theoretical best rate to use with RBCD AND the most difficult to accomplish on the Mac.

 

Thanks much,

Clay

 

 

 

 

[Ashley James]

Clay I'm certain that leaving the Audio Midi at 24/96 and leaving it will be fine, it's what I do. I spent hours trying all the different permutations and heard no difference and than Martin tells me I won't and there's no technical reason why anyone should. But they do and they'll be cross with me now.

Sample Rate conversion changes don't affect sound quality in themselves but there may some other factor that can in some equipment. We can't hear them in Audio Midi or our own kit and we did loads of testing and also had various visitors try as well.

 

Ash

 

 

 

[cfmsp]

 

 

 

thanks Ashley...for sharing your experience testing the permutations.

 

Chris, et al, I'm still curious about upsampling, even if it may only be from a theoretical standpoint.

 

I.e., apparently a portion of Amarra's proposition is playback / sample rate conversion performed outside of CoreAudio/iTunes with dedicated hardware (304).

 

I'm curious how to take a peek at this given that Amarra is not yet available and will be overly expensive in any event.

 

I'm also wondering if a part of the Weiss' success is due to it's own world-class SRC - Saracon.

 

thoughts?

 

Clay

 

 

 

 

[BobH]

Hi Clay,

 

Been thinking and doing the same things recently, for the same reasons! I'm using XP, rather than Mac, but I'm sure the principle is the same.

 

My findings so far are, upsampling (using R8Brain Pro) to 24/88 is superior to 24/96 and both are superior to the 16/44 original. R8Brain approaches the problem of getting to 24/96 by doing a software DA conversion first, then it resamples the result, then does a DA conversion back again. In my tests I found that this resulted in a tad better detail separation and possibly a bit weightier bass. It was outweighed, however, by being markedly less involving, it all sounded a bit flat.

 

In comparison to the original CDP playing the RBCD, the ripped computer version is miles better, every time. I prefer the sound of 24/88 to 24/96 and, in all cases, I prefer the sound of the pre-converted file to the sound of either a media player, or a dac, doing the SRC for me. There is a penalty to pay in hard drive space but I am now of the opinion that software SRC and a nos dac may well be worth a look-see for some folks. It certainly does it for me!

 

 

 

[weiss2496]

Whether upsampling (or oversampling, which is the same) makes sense is up to the taste of the listener and/or the quality of the upsampling done in the DAC chip. Modern DAC chips employ some upsampling, e.g. one type of DAC chip upsamples all input rates to a constant conversion rate of 352.8 kHz (at a 44.1kHz base rate). So depending on the input rate the DAC chip sees, it engages factor 8 or 4 or 2 upsampling. With an external upsampler it is possible to bypass that internal (DAC chip) upsampling to some extent, especially for the most critical upsampling stage, 44.1 to 88.2, it can be advantageous to have an external upsampler - provided that the external upsampler is better.

Another benefit of upsampling all files to a common rate can be that the player software and DAC do not have to switch sampling rates all the time and thus there is some consistency in playback quality. Disadvantage is that the files eat more disk space.

Our Saracon (and the hardware unit SFC2) we built for our pro audio customers (the Mastering Engineers), but both can find their uses in the audiophile world as well. With the hires downloads it could be attractive to have the ripped CDs at a higher rate as well.

Daniel