[Alu]

Ok, so I was a bit bored right now, and wanted to improve my soundquality with foobar. I use an M-audio Transit into a bel canto DAC2, meaning that I could at least take some inspiration of the following: http://www.empiricalaudio.com/frComputer_Audio.html (scroll down to the very bottom, then scroll up until you see "Recommended software configurations and other advice" in bold blue letters.)

So, I followed all the advice:

1. Set my transit to output 96000.

2. Enable the Secret Rabbit Code (SRC) Resampler in foobar.

3. Set the resample rate to 96000, and the other stuff to sinc best. (as indicated in the empaudio guide)

4. Use 24bit fixed-point.

5. Enable good old ASIO. (was enabled before anyways)

6. Time critical always did wonders so...ok.

7. The result: CPU usage around 30% and a darned choppy sound. (my processor is a 3500+ on 1gig ram.

Now, when I set my transit card to "Very High" in Latency, the result is a bit less choppy, but still not listenable of course. I even tried setting foobar to "Realtime" priority in the task manager and I tried ticking "Slow mode" as well. I changed my output from ASIO to wavout (Bufferlength = 4031ms) once, and the plaback was better, but there were still ticks an pops in it. I played with the ASIO settings (Buffersize and Thread Priority) but couldn't change the choppiness. When I remove the SRC resampler, and just use the standard SSRC one, it works when my transit is "Very High" once again in Latency, and playback is once again choppy when set to "Normal". The cpu usage is down to around 10% as well.

I don't really know what this SRC Resampler does more than the normal one, and the only thing that's different with it is a "Quality" scrolldown menu, from which you can choose the following: Sinc Best, Sinc Medium, Sinc Fastest, Zero Order Hold, Linear.
So...what else is there to do but to try em all out of course.

Results (with SRC to 96000 / 24bit fixed-point / ASIO Thread Priority = Time Critical and Buffer Size = 7 / Transit Latency = Very High / Task Manager Foobar Process Priority = Realtime):

Sinc Best = Playback choppy / CPU usage 30-35% (I got told that this mode would actually sound like the normal SSRC foobar resampler!)
Sinc Medium = Playback flawless / CPU usage 10-20% (I got told that the bass would begin to diminish in terms of quality with both Medium and Fastest modes)
Sinc Fastest = Plaback flawless / CPU usage 7-15%
Zero Order Hold = Playback flawless but Soundquality inferior, don't really know how to describe the sound...has a bit of a metallic aftertaste to my ears.
Linear = almost like Sinc Fastest (I really didn't listen to this a lot)

Linear and Zero Order Hold offer the lowest CPU usage results. (0-10%)
I don't really know what to think of these results, but I know that Sinc Best isn't working, and that's bad enough.
I guess I'll stick with the normal foobar SSRC for now, and leave the SRC to those who understand what they're working with.
Is SRC Sinc Medium better than Sinc Fastest? It probably is, but I couldn't hear that yet. Is SRC better than SSRC? I can't tell you as I don't really hear any difference here either (and so did another source).

I know that I really didn't find out a lot. And this is a lot to read. However, I'd like to know if anyone of you fellow computer-as-sourcers has tried something similar and got Sinc Best to work properly. And I'm hoping that maybe Steve could give us some input on this. Is it because I'm not using an USB Off-ramp f.i.? I doubt that though.

I used the album Samply Red - Perfect Rouge by Yoko Kanno for this, over B&W 705 speakers with Rotel amplification.

Also, don't ask me for the SRC, because it is now licensed by Empirical Audio!

EDIT: Someone told me that Sinc Medium and Sinc Fastest lack bass upon comparision. Sinc Best really is the best, but a difference between SRC (Sinc Best) and SSRC still can't be found.

There is a difference between the upsampling to 96000 and the standard 44100. I personally do prefer the sweeter sound of 96000, but somehow it tires me incredibly. Also, I can only hear a really obvious difference on my Japanese reference recordings. It's still hard to tell one from the other though.

 

[Alu]

Today I was messing around again, and it occured to me that the choppiness with the Secret Rabbit Code Resampler might have something to do with the various audio formats. It didn't really occur to me, it was more than a random thought. And guess what? I was right!

I tried with the following formats (SRC Resampler in Sinc Best):

FLAC - Choppy crackly playback all over. (and it was flacs that I was running the previous tests on)
APE - Perfect playback! (The plot thickens

)
WAV - Perfect playback!
MP3 - Perfect playback!

I didn't have any other formats to try, but I guess the result is clear. SRC works in Sinc Best, but it ONLY has issues with FLAC. So I went to download the most recent FLAC version that's out, installed it, and it was still choppy and all crackly.

I know noone cares, but this is interesting to me! And I have a place where I can write it down without it getting lost for now.

EDIT: This time I made the "tests" on my AKG K701.

Steve Nugent of Empirical Audio told me that when using FLAC, the CPU resources are taxed so much along with ASIO and SCR that there are no cycles left, so it drops data. This is called "underrun". Everything should be turned off and there are instructions for eliminating pops on the empirical audio computer audio webpage. If this does not fix it, he said, I'd need a faster CPU.

Since I already have an Athlon 64 3500+, the only thing I could think of was maybe using a dual core for future foobar tests. However I'll be sticking to SSRC, the internal foobar upsampler, as I don't want to spend money on a better CPU for this audio PC.