SRC comparison: Your Ears, Not Your Eyes

[prismatic]

Up until now, I have been using TC Spark for conversion of my mixes from 48khz to 44.1khz. It sounds a million times better than the best Pro Tools conversion setting ("Tweakhead"). But, for the time being, I am switching to PC format.

Has anybody compared the conversion from TC Spark to the latest hot offerings, notably Saracon and R8Brain Pro? I don't trust graphs on this one...After all, Pro Tools conversion supposedly "looks good." How do they sound? Should I keep an old Mac around just for TC Spark, or is it time to move on?

Options I have considered include Saracon, R8Brain Pro, Wavelab, Soundforge. But I am open to suggestions.

Except: I plan to track at 48khz. I am skeptical about the "track at your target sample rate" philosophy.

[mixerguy]

Apogee Rosetta 200 SRC

[prismatic]

Thanks, mixerguy. I am sure the Rosetta 200 is outstanding.

However, I should have mentioned that I don't need real time conversion. I have been looking at ITB solutions.

[sleepwalker]

I use Adobe Audition.

Anyway, Peak was getting slagged for their SRC so their performed some tests.

http://www.bias-inc.com/products/peakPro5/resampling/

It's a very good read. Interestingly enough, you can use Audition to perform these same tests using the Spectral Display. They're using MATLAB I believe.

I compared the SRC of Cubase SX3 and Wavelab 5 to Audition's and the pre/post filters are MUCH better in Audition.

Originally Posted by prismatic

Thanks, mixerguy. I am sure the Rosetta 200 is outstanding.

However, I should have mentioned that I don't need real time conversion. I have been looking at ITB solutions.

[d-dmusic]

Using my ears, I find the SRC in SX3 to be just fine. In fact, excellent.

I tried the free version of R8Brain, due to the amount of talk regarding SRC, and besides an audible hiccup in the R8Brains conversion, I found no sonic difference at all. This, as I said, was just using my ears.

There may be Golden Ears out there who really do hear something that I don't though and I'd really like to know what that is. Is it a phase related problem ?...a difference in the bottom end ?....top end ?...mid range ?....

If there is a "best" SR-convertor, what is it providing that our convertors are not ?

[theblue1]

Originally Posted by prismatic

...

Except: I plan to track at 48khz. I am skeptical about the "track at your target sample rate" philosophy.

Actually, I think it would be better to characterize that philosophy as "track at your target sample rate or an even multiple of it."

Skepticism is a very good thing. But it implies that the skeptic will then use his brain to get at the truth of the matter. But -- clearly -- if whatever technique you decide to employ sounds better to you, then, there ya go.

If, at some point it doesn't, the alias error inevitably resulting from SRC from a rate that is not an even multiple of the target would be the first culprit I would recommend investigating.

Good luck in your endeavor.

[RainbowStorm]

I use L3 type1 ultra for the bit reduction process... By far the best I've heard... When it comes to SRC I use whatever that comes with Cubase SX 2... Most of the time I record at 44,1KHz and don't need to do any downsampling.

[radiant]

Sonar's POW-r . I can't think of anything better.

[RainbowStorm]

Originally Posted by radiant

Sonar's POW-r . I can't think of anything better.

Have you tried L2 or L3 type1 ultra dithering? When I made A/B testing of the two (L3 vs POW-r) I found the L3 type1 ultra to be superior. Try it if you haven't, if you have compared it and still use POW-r, no need to change anything...!

[RainbowStorm]

I have also tested R8Brain. I was NOT impressed...

[norman_nomad]

I´ve recently been using R8brain (free) in place of Nuendo´s built in SRC which has a relatively lazy alias filter. I compared both by ear and I preferred the R8brain... not a night and day difference but you could hear some of the high end roll off in the Nuendo files.

Then again... I never thought 'Tweakhead' sounded horrible in PTLE.

I´m not sure why you´re skeptical about tracking at the target sample rate. What do you expect to gain by going 48k? Are you working with crummy converters?

With 'good' converters (think Apogee, Lynx, Kurzweil... even the new RME stuff), I´d rather track at 44.1k than track at 48k and SRC down to 44.1k.

[radiant]

Originally Posted by RainbowStorm

Have you tried L2 or L3 type1 ultra dithering? When I made A/B testing of the two (L3 vs POW-r) I found the L3 type1 ultra to be superior. Try it if you haven't, if you have compared it and still use POW-r, no need to change anything...!

I found POW-r better than other waves offerings but haven't tried the L3. You've got me real curious now.

[RainbowStorm]

Originally Posted by radiant

I found POW-r better than other waves offerings

Good for you!

Originally Posted by radiant

but haven't tried the L3. You've got me real curious now.

NOT good for you!

You should give L3 type1 ultra a chance, I couldn't believe my ears the first time I heard the difference between that and the POW-r... Up today the best dithering I've encountered...

[prismatic]

Originally Posted by theblue1

Skepticism is a very good thing. But it implies that the skeptic will then use his brain to get at the truth of the matter.

I don't doubt it.

[prismatic]

Originally Posted by norman_nomad

I´m not sure why you´re skeptical about tracking at the target sample rate. What do you expect to gain by going 48k? Are you working with crummy converters?

With 'good' converters (think Apogee, Lynx, Kurzweil... even the new RME stuff), I´d rather track at 44.1k than track at 48k and SRC down to 44.1k.

I am using 2 channels of Apogee Rosetta for 90% of the tracks. I do lots of overdubbing. For drums, additional tracks go through the venerable Digi001, clocked to the Apogee, which helps a lot. Less high end detail than the Apogee, but great otherwise, so I use it on tracks where less high end is required. Pres and mics make a much bigger difference to my ears, as does mic position.

Pro Tools Tweakhead, now THAT sounds bad to me. Made me doubt my mix. If you have compared Tweakhead 48khz > 44.1khz to tracking at 44.1khz, I have no doubt that avoiding Pro Tools SRC sounded much better than using it.

Seriously, Itunes' mp3 converter sounds better!

The bump up in quality by using TC Spark was dramatic. This is why I would like to know people's experience of how it compares to SRC's available for the PC.

[defjamm]

if testing r8brain, try the pro version with the ultra steep mode. you can download the demo:
'The demo version of r8brain PRO allows you to evaluate all features of the software, but its conversion is limited to the first minute of an audio file. The demo version does not support batch conversions.'

http://www.voxengo.com/product/r8brainpro/

[prismatic]

Thanks for all the replies! I don't seem to be as picky about dither as some of you. The ultra dither in L2 seems good.

Good point about trying R8Brain Pro. A minute of audio should be long enough for comparison.

[mixerguy]

Originally Posted by RainbowStorm

I use L3 type1 ultra for the bit reduction process... By far the best I've heard... When it comes to SRC I use whatever that comes with Cubase SX 2... Most of the time I record at 44,1KHz and don't need to do any downsampling.

well..... the Apogee Rosetta 200 SMOKES the ITB options I've compared it to - Waves or stock Digi.

[RainbowStorm]

Originally Posted by mixerguy

well..... the Apogee Rosetta 200 SMOKES the ITB options I've compared it to - Waves or stock Digi.

Interesting... I have only tried the software versions of Apogee uv-22 and Apogee uv-22hr and Waves L3 is on a different level. But I have actually never tried dithering with hardware. I think I should...!

[Max]

Originally Posted by RainbowStorm

Interesting... I have only tried the software versions of Apogee uv-22 and Apogee uv-22hr and Waves L3 is on a different level. But I have actually never tried dithering with hardware. I think I should...!

The deceased Mastertools (UV22) software plug-in really can't be compared to the hardware version of UV22HR.

[Doublehelix]

The R8Brain Pro uses different algorhythms according to Voxengo, so the 2 programs should not be compared directly.

I like R8Brain Pro, and now I am really liking the new Crystal Resampler that comes with Wavelab 6. Very nice indeed!

[prismatic]

Originally Posted by Doublehelix

The R8Brain Pro uses different algorhythms according to Voxengo, so the 2 programs should not be compared directly.

I like R8Brain Pro, and now I am really liking the new Crystal Resampler that comes with Wavelab 6. Very nice indeed!

Thanks! Just the kind of input I was looking for. Do you hear an audible difference between those two?

Also, those of you who have brought dither into the discussion: Does it make that big of a difference? So far, it has been my experience that SRC affects the quality much more.

[theblue1]

Originally Posted by prismatic

Up until now, I have been using TC Spark for conversion of my mixes from 48khz to 44.1khz. It sounds a million times better than the best Pro Tools conversion setting ("Tweakhead"). But, for the time being, I am switching to PC format.

Has anybody compared the conversion from TC Spark to the latest hot offerings, notably Saracon and R8Brain Pro? I don't trust graphs on this one...After all, Pro Tools conversion supposedly "looks good." How do they sound? Should I keep an old Mac around just for TC Spark, or is it time to move on?

Options I have considered include Saracon, R8Brain Pro, Wavelab, Soundforge. But I am open to suggestions.

Except: I plan to track at 48khz. I am skeptical about the "track at your target sample rate" philosophy.

Since prismatic loves his TC Spark, maybe he'd be interested in hearing what one of their engineers, T. Lund, wrote in a white paper:

It should also be noted that the use of sample rate converters ought to be kept at a minimum.

http://www.broadcastpapers.com/white...essControl.pdf [sec. 2.2]

He goes on, in the same paper [section titled, "2.6 Alias Distortion"], to show how alias error can result from SRC and says -- pointedly -- "At the time of writing, the author has not seen a commercial sample rate converter chip not producing this type of result."

[emphasis added]

[sleepwalker]

Originally Posted by prismatic

I don't trust graphs on this one...After all, Pro Tools conversion supposedly "looks good."

By the way. I'd probably trust graphs if I was you. If you really like the sound of aliasing, I'd recommend getting it somewhere else other than your final SRC.

[prismatic]

Originally Posted by theblue1

Since prismatic loves his TC Spark, maybe he'd be interested in hearing what one of their engineers, T. Lund, wrote in a white paper:

Now theblue1, I never said I love TC Spark. Love would be a strong word for an SRC. All I said was that it did not mess up the sound the way Pro Tools' SRC did.

[theblue1]

You're right! My bad. I got wowed by that "million times better" thing...

Sorry!





BTW... I'm with you on the "ears over eyes" thing... if it sounds better -- that should call into question any previous assumptions. That said, I'm talking about general principles here, and it's my thinking that if one knows his target rate he should, all else being equal, use that rate or an even multiple -- assuming that the project will be staying in the digital domain the whole way.

[prismatic]

Originally Posted by theblue1

You're right! My bad. I got wowed by that "million times better" thing...

Sorry!





BTW... I'm with you on the "ears over eyes" thing... if it sounds better -- that should call into question any previous assumptions. That said, I'm talking about general principles here, and it's my thinking that if one knows his target rate he should, all else being equal, use that rate or an even multiple -- assuming that the project will be staying in the digital domain the whole way.

No worries.

I admit that graphs and measurements definitely have their place with audio. It's just that, being a musician first by training, the final test for me is always, "How does it sound?"

I have been encouraged to continue tracking at 48khz by comments of some very experienced people, such as this recent one by Bob Katz on another thread:

Originally Posted by bob katz

Usually 48K sounds more "open" to me. And 96K sounds "purer and warmer" while retaining the openness of 48K. I highly recommend 48K over 44.

In the case of aliasing distortion, of course I don't think that's something I'd want. Can anybody describe the sound of it or point out an audio example, so that I know what to listen for?

[smallbutfine]

Hmmm,
i just tried to remember a shootout of SRC's including saracon, r8brain (free and pro) barbabatch,...
From what i remember weiss smoked them all, but r8brain was on second place thereafter beeing r8brain free+barbabatch .
That said, i tested sonar6 with pow-r 3 dithering (which sounds a good portion better than sonar4 with pow-r 3 dithering to my ears) and r8brain free for both, src and dither.
Today i read in the voxengo forum that dither was completely removed from r8brain pro...
So, i will go for following chain for next tests - exporting 32 (or 64) bit from sonar (WITHOUT any dither!! Thats important...)
Then SRC with r8brain...with 32 bit float out format (so no dither here)
then dither with POW-R 3 (best dither i have at hand....)
Should give best results for now (that's my guess...)

But still - Sonar 6 sounds pretty good alone!

I do not really see a need to buy a damn expensive saracon - but i did not have a chance to compare really until now...and i still have the option to do dither in analogue domain having swissonic, mytek and studer converters that sound really good to my ears....

Just my few cents...

Kind regards,

Martin

[Ruudman]

How can one avoid using the eyes?
You can't claim there's no aliasing without proving it by visual measurement
(an example).

http://src.infinitewave.ca/


Ruudman

[theblue1]

Originally Poste