yes i had the same result as you. the question is what's the reason, is it waves or pro tools tdm.3*s wrote:I corrected the timing difference of the l2 files, thats the result I posted.
waves TDM vs. Native wankery!
-
- KVRist
- 494 posts since 18 Jul, 2004
-
- KVRian
- 1243 posts since 24 Oct, 2003 from Maine
The transients sound kind of like bit crushing. I'm guessing the TDM and RTAS use different size data types, which makes sense since Pro-tools hardware uses a different memory architechture than the PC.
♫♪♫♫♪♫
-
- KVRist
- 494 posts since 18 Jul, 2004
forgot to say thank you to bmanic for the files! i just have a little request: could i have the original 24bit-file and could you give the exact settings or presets you used. i will try it with an other native host(samplitude), without dithering. could you do the same, once again only with tdm? and then we compare again.
-
- KVRAF
- 12235 posts since 18 Aug, 2003
Not sure about better, but doesn't it stand to reason that RTAS and TDM sound different? Isn't that kind of the point of TDM, that having the dedicated DSP allows for extra calculations, which would mean that code could be written differently?
-
- KVRAF
- 2208 posts since 13 May, 2005
TDM sounds slightly crispier. However, depending on the host used, native plugins sound different, too. There's a very audible difference between the old Logic PC and Samplitude for example...
-
- KVRAF
- Topic Starter
- 11386 posts since 3 Feb, 2003 from Finland, Espoo
yes, I'll provide a new file for testing. Also, I'll re-do the test and post exact screen shots of what settings were used. I have no memory of the settings used. Looks like I even screwed up the L2 test by having the files wrongly aligned. Sorry bout that, but use your ears guys (provided that you have accurate monitoring). The difference is there to spot, it ain't no voodoo. Listen to the transients.defjamm wrote:forgot to say thank you to bmanic for the files! i just have a little request: could i have the original 24bit-file and could you give the exact settings or presets you used. i will try it with an other native host(samplitude), without dithering. could you do the same, once again only with tdm? and then we compare again.
Cheers!
bManic
"Wisdom is wisdom, regardless of the idiot who said it." -an idiot
"They don't ban hate speech; they ban speech they hate." -an oracle
"They don't ban hate speech; they ban speech they hate." -an oracle
-
- KVRist
- 401 posts since 4 May, 2004
ProTools signal path is 48-bit fixed point both TDM and RTAS, with the Waves doing calculations in 56-bit fixed point for TDM plugins (as those are fixed-point DSPs), and 64bit floating point calculations on the CPU for the RTAS version. I've read it on their site or a brochure or something, it's their info.
The TDM must be fixed-point (due to fixed point DSPs it uses), which means smaller resolution (as it doesn't scale down to low values as floating point one does). Waves plugins are designed to clip softly on overloads. Obviously, with the less precise fixed-point math this will probably sound a bit harsher. It's even worse for EQs, I'm pretty sure they tweaked to match, but it must have been a nightmare (IIR coeffs on fixed-point is not a pretty thing to deal with).
Hearing it as better is classical effect of "better is better" approach where you percieve a Futterman amp as solid gold while looking at the logo, but in a blind test it's "strikingly similiar" to a cheapo Pioneer.
The TDM must be fixed-point (due to fixed point DSPs it uses), which means smaller resolution (as it doesn't scale down to low values as floating point one does). Waves plugins are designed to clip softly on overloads. Obviously, with the less precise fixed-point math this will probably sound a bit harsher. It's even worse for EQs, I'm pretty sure they tweaked to match, but it must have been a nightmare (IIR coeffs on fixed-point is not a pretty thing to deal with).
Hearing it as better is classical effect of "better is better" approach where you percieve a Futterman amp as solid gold while looking at the logo, but in a blind test it's "strikingly similiar" to a cheapo Pioneer.
Obviously a computer still can’t throw a television out of a hotel window or get drunk and be sick on the carpet, so there is little danger of them replacing drummers for some while yet. -- Nick Mason
-
- KVRian
- 943 posts since 15 Mar, 2005
well i think in this instance, 'better' can be more objective. ie smoother, less distortion, and everything else that comes with higher resolution. of course these differences dont necesserily mean better subjectively.peejunk wrote:Hearing it as better is classical effect of "better is better" approach where you percieve a Futterman amp as solid gold while looking at the logo, but in a blind test it's "strikingly similiar" to a cheapo Pioneer.
although if you're suggesting the prefered TDM ones are actually of mathematically less resolution then
-
- KVRAF
- 4030 posts since 7 Sep, 2002
This comparison reveals Waves quality of converting their algorithms... In Waves L2 test it looks more like a whole variation to their algorithm - probably even some constants were changed.
But in my opinion it's pointless to say TDM sounds better than RTAS. At times I can hear TDM sounds duller, at other times I hear high-frequency transients are prononunced better in TDM (but it's probably the same as with UAD - simply, a soft distortion).
What about EQ comparisons? It's strange enough you've included dynamics plug-ins only.
But in my opinion it's pointless to say TDM sounds better than RTAS. At times I can hear TDM sounds duller, at other times I hear high-frequency transients are prononunced better in TDM (but it's probably the same as with UAD - simply, a soft distortion).
What about EQ comparisons? It's strange enough you've included dynamics plug-ins only.
Last edited by Aleksey Vaneev on Wed Feb 08, 2006 9:03 pm, edited 2 times in total.
-
- KVRAF
- 4030 posts since 7 Sep, 2002
peejunk, AFAIK Pro Tools use Motorola 56000 series of processors which are 24-bit DSP with only two 56-bit accumulators. 48-bit fixed point transfer is very different from the actual DSP calculations resolution.
-
- KVRist
- 401 posts since 4 May, 2004
I don't have expirience with large-integer calculations and how they're done. I do know one developer that did a some coding for 56k and other fixed point chips in general, and even some for TDM ProTools. He told me that fixed-point DSP is a black art. They use a lot of trickery to make use of both the fast DSP-specific math instructions fir high performance but also be able to achieve high calculation precision. It shouldn't be impossible, afterall, a 48-bit integer/fixed is just two 24bit ones stacked. I can see how it can get messy with multiply-accumulate, let alone more complex things but I'm sure there are textbook ways of dealing with it, as well as tricks of trade.
I read the info on the Waves site, and combined it with what my aquantance told me.
I read the info on the Waves site, and combined it with what my aquantance told me.
Obviously a computer still can’t throw a television out of a hotel window or get drunk and be sick on the carpet, so there is little danger of them replacing drummers for some while yet. -- Nick Mason
-
- KVRAF
- 4030 posts since 7 Sep, 2002
bmanic, can you please pass a simple 1 kHz 32-bit floating point sinewave at -6 dBFS through Waves L2 at 'no additional gain' setting? I'm curious about how it will be transformed (what kind of noise floor, etc).
-
- KVRist
- 494 posts since 18 Jul, 2004
ears are easy to fool(people could perceive a little distortion as better sounding).bmanic wrote:but use your ears guys (provided that you have accurate monitoring). The difference is there to spot, it ain't no voodoo. Listen to the transients.
which version of waves was used btw?
-
- KVRAF
- 4030 posts since 7 Sep, 2002
peejunk, '48-bit integer = 2 x 24-bit integers stacked' looks simple, but not as simple nor fast in implementation. That's why Motorola has 56-bit accumulators. But still constants (and probably some of the 'memory variables' have to be stored in 24-bit resolution), so that the general resolution is still 24-bit.
In native, internal processor calculations have 80-bit FP resolution (with a mighty 64-bit mantissa), but since memory variables are stored in 64-bit FP resolution, the overall precision is 64-bit FP (53-bit mantissa) as well.
In native, internal processor calculations have 80-bit FP resolution (with a mighty 64-bit mantissa), but since memory variables are stored in 64-bit FP resolution, the overall precision is 64-bit FP (53-bit mantissa) as well.
-
- KVRist
- 494 posts since 18 Jul, 2004
http://home.casema.nl/ajohnston/limiting/Aleksey Vaneev wrote:bmanic, can you please pass a simple 1 kHz 32-bit floating point sinewave at -6 dBFS through Waves L2 at 'no additional gain' setting? I'm curious about how it will be transformed (what kind of noise floor, etc).
not quite the setup you requested, but maybe you haven't seen it until now.