Does 64 bit Sonar really sound better?

[Bill Lorentzen]

Anyone done any proper tests?

[redddog]

here we go.

[taturana]

my (non-scientific) real world tests have shown a definitive sound improvement... and bouncing /mixing with no loss of detail at all... that alone worth the feature...

IMO clearer mixes, reverb tails and better results...

i know we've been through this a couple of times already.

[PaRaNoId]

I can tell the difference easily with larger projects-
It helps to gain separation b/w instruments and makes the plugs sound smoother. More complete math = more resolution. More like analog summing as far as "glue" is concerned. my 2 cents...

[Ashofapheonix]

64 bit as in Sonar running on 64 bit windows? or 64bit Double precision engine that works on both 32bit and 64bit. All 64 bit, windows allows is primarly Sonar can access more ram. 64bit engine, allows more precise editing of audio data, but this is the programs internal mathematics is more precise (not referring to the user) and the engine works on 32 bit windows. Those getting better results with 64bit, sound quality wise its just all in there heads, or better yet their mixing skills improved.

[NoWo]

Reaper is working with 64 bits too.

Sound is good and much better than 32 bit float.

NoWo

[tamasdragon]

Tested or think? True is that 64bit can leads to higher precision, but alone it's a blanket statement. Blind test could prove something...but till then there are just rumours.
TD

[Jesse Skeens]

Originally Posted by PaRaNoId

I can tell the difference easily with larger projects-
It helps to gain separation b/w instruments and makes the plugs sound smoother. More complete math = more resolution. More like analog summing as far as "glue" is concerned. my 2 cents...

If you've read up on Paul Frindles posts you'd know that "resolution" and bits don't belong together.

The extra accuracy of 64 bit is so low in level it's insignificant.

[Jesse Skeens]

Originally Posted by tamasdragon

Tested or think? True is that 64bit can leads to higher precision, but alone it's a blanket statement. Blind test could prove something...but till then there are just rumours.
TD

There were tests on the cakewalk forum when it first came out. From what I remember some people actually like sonar 5 better. It's all smoke and mirrors though, just another spec to try and sell to you.

[Jesse Skeens]

Originally Posted by Ashofapheonix

64 bit as in Sonar running on 64 bit windows? or 64bit Double precision engine that works on both 32bit and 64bit. All 64 bit, windows allows is primarly Sonar can access more ram. 64bit engine, allows more precise editing of audio data, but this is the programs internal mathematics is more precise (not referring to the user) and the engine works on 32 bit windows. Those getting better results with 64bit, sound quality wise its just all in there heads, or better yet their mixing skills improved.

Exactly, if people would just learn to track and mix better they wouldn't be worried about bits so much.

[Ashofapheonix]

Originally Posted by Jesse Skeens

Exactly, if people would just learn to track and mix better they wouldn't be worried about bits so much.

I agree, even if their was a minor quality difference, a good engineer should be able to get the same results from either. That aside in general, people worry about ver minor differences around here for some reason. It doesn't really matter if you have lynx or apogee, in the big picture their about the same. Ditto for Onyx Pres v.s Mackie XDR. A good engineer would get about the same results from either or be just as limited by either.

[theblue1]

Originally Posted by Ashofapheonix

64 bit as in Sonar running on 64 bit windows? or 64bit Double precision engine that works on both 32bit and 64bit. All 64 bit, windows allows is primarly Sonar can access more ram. 64bit engine, allows more precise editing of audio data, but this is the programs internal mathematics is more precise (not referring to the user) and the engine works on 32 bit windows. Those getting better results with 64bit, sound quality wise its just all in there heads, or better yet their mixing skills improved.

I've read this a couple times and still not sure exactly what you're saying... but I don't think an improvement in sound from a 64 bit audio engine is all that far-fetched (in complex mix situations).

There are two issues, as you seem to acknowledge. A 64 bit OS version (Sonar has both a version for 32 bit and a version compiled to be run in Win 64/Vista 64). That will potentially improve speed performance.

And Sonar also has the option of using its audio engine at 64 bit internal resolution (regardless of processor/OS native processing resolution).

Now... whether folks can truly hear the diff, as others have suggested, is a job for proper double blind testing.

But -- unless there's something in this discussion I'm not getting -- I really don't think it's entirely unreasonable to expect that -- particularly with a complex mix where there are a lot of mathmatical processes modifying individual audio streams and/or mixing large numbers of audio streams -- that improving the mathmatical precision of those processes might not, indeed, pay off in better sound.


Now the internal audio processing math part should be easy enough to test... math is not exactly subjective -- but that said I really don't think there's any real reason to doubt that the math is more precise... they would have to make a substantial error in the AE design for that to not be the case.

[AB3]

Are we talking about recording it at 64 bit or just the 64 bit mix engine (which is available on 32 bit XP).

As far as recording at 64 bit, there are no such sound cards that I know of. I do believe there is a 32 bit one out there.

So if it is just the 64 bit mix engine, switch it in and if you like the sound better, leave it.

Is this all that simple or am I missing something?

[George Necola]

I think we are talking again about clever marketing and magic happens between monitors and chair..

[theblue1]

Originally Posted by George Necola

I think we are talking again about clever marketing and magic happens between monitors and chair..

Well... mathematical precision is a real thing... there's no question that doubling the word length of memory address spaces the audio engine uses for internal processing will improve that precision.

Whether or not one can hear the diff is, I think, a good subject for a rigorous double blind testing session.

[dkatz42]

Originally Posted by AB3

As far as recording at 64 bit, there are no such sound cards that I know of. I do believe there is a 32 bit one out there.

If there is, it's just marketing hype. The best converters will give you 20-ish usable bits of data, and that's if your signal is peaking at 0dBFS. 12dB down, toss out two bits.

Doing 64 bit math will indeed provide more valid bits, but unless those extra bits percolate up to the bits that the DAC can actually reproduce, there's not going to be any effect (truncate the results to, say, 21 bits and much of the time you'll end up with *exactly* the same data stream at 64 bit computation than you would at 32.) The more math that gets done, the more likely the errors would accumulate to the point of showing up in the DAC output, all other things being equal (which they may or may not be.)

[sam c]

i do not believe your ear will hear. seriously, even the 16 bit to 24 bit has the same dispute going. if we could objectively listen, disregarding numbers, what would we hear?

i agree, marketing hype at its best/worst!

i had a few really good sounding tunes from my yamaha cassette 4 track recorder. remember those, with the magnetic tape you put on the window to read four tracks in one direction? back then we spent more time recording cause there was no internet to constantly debate this stuff. we just recorded and shared the tunes! ahhhh, a simpler time!

[George Necola]

if you read the WEISS article carefully.. you will stumble over 5 bit conversion (with a lot of oversampling), and internally 16/32/64 bit audioprocessing.

it's not the DAW that makes the diffrence in soundquality... it's the engineer. if someone can mix a whole "pro"-hiphoprecord inside the box in our most discussed system on gearslutz, 32bit internally is not that bad..

it's the engineer, not the gear.. but nice gear is cool to own.. so I will go now and get some more

cheerio

[Ashofapheonix]

Originally Posted by theblue1

I've read this a couple times and still not sure exactly what you're saying... but I don't think an improvement in sound from a 64 bit audio engine is all that far-fetched (in complex mix situations).

There are two issues, as you seem to acknowledge. A 64 bit OS version (Sonar has both a version for 32 bit and a version compiled to be run in Win 64/Vista 64). That will potentially improve speed performance.

And Sonar also has the option of using its audio engine at 64 bit internal resolution (regardless of processor/OS native processing resolution).

Now... whether folks can truly hear the diff, as others have suggested, is a job for proper double blind testing.

But -- unless there's something in this discussion I'm not getting -- I really don't think it's entirely unreasonable to expect that -- particularly with a complex mix where there are a lot of mathmatical processes modifying individual audio streams and/or mixing large numbers of audio streams -- that improving the mathmatical precision of those processes might not, indeed, pay off in better sound.


Now the internal audio processing math part should be easy enough to test... math is not exactly subjective -- but that said I really don't think there's any real reason to doubt that the math is more precise... they would have to make a substantial error in the AE design for that to not be the case.

You actually nailed it on the head. 64 bit windows only improves speed of sonar.

The 64 bit mix engine, and internal resolution improves mathematical processes of the processor. However, this is availible in 32bit Windows., I don't see why you'd set the internal resolution to something else to begin with.

My point is if there talking about running sonar 6 on vista or xp 64 and hearing a sound quality improvement, its in their head, the only thing you get is better performance.

If their talking about using the 64bit mix engine and getting better results its kinda "duh" the computer is more precise, and you could have been using this all along anyway.


On another note I don't see this forums love for double blind tests, which I've noticed 1. We're on different monitoring chains, 2. We don't know the context as to how the source was used. 3. We do not know the exogenous variables, the persons room, skill, if there testing microphones, we may know the rest of their chain ect.

With out knowing this how does double blind tests tell anyone anything besides the person performing the test, who can't really perform a double blind to begin with. Basically even if you guys conduct a double blind the test aint scientific.

[MoreDynamics]

I'm not sure if this would count as a "proper" test, but I've done A/B testing on a number of my mixes and can hear the difference on my more complex mixes that are summing lots of tracks and use a significant amount of FX processing. Probably the most noticable thing for me is the stereo separation doesn't fall apart with the 64-bit engine enabled. I've also patched Sonar 6.2's Bit meter, and A/Bed the 64-bit engine on and off, and at least to my eyes it appears that the 64-bit engine is doing something.

[dawhead]

Originally Posted by theblue1

Well... mathematical precision is a real thing... there's no question that doubling the word length of memory address spaces the audio engine uses for internal processing will improve that precision.

unfortunately, there's actual no question that doubling the address space will have absolutely ZERO effect on processing precision. doubling the precision of the floating point representatons definitely has a tautologically positive effect on processing precision, which may or may not translate into improved sound. but a 64 bit version of <your-daw-here> has nothing to do with the floating point precision.

thats because 64 bit address spaces have *nothing* do with floating point precision on *any* modern processor. on intel, floating point math is done by either the SSE/SSE2 vector unit or an even older FPU. on older macs, its done by either an FPU or the altivec vector unit. floating point comes in two variants: 32 bit or 80 bit. whether your processor's integer word size is 32 bit or 64 bit doesn't change this. it probably does change various other things, such as the memory bandwidth and/or the number of registers available, which will have beneficial effects on floating point math too, but only indirectly.

put bluntly, "64bit" versions of any DAW have nothing do with their audio processing pathways. what you should be asking is "does it do double precision (80 bit) floating point". DAWs running on a 32 bit processor can do this too (with about a 5% slowdown compared to single precision). all the 64-bit-ness does is allow a larger memory address space to be used, which includes easily handling very large files (larger than 2^32 bytes) (though this can be done in other ways on 32 bit systems).</your-daw-here>

[theblue1]

I usually qualify that with something like "properly designed, rigorously applied double blind testing methdoology."

But... point taken.


Still, I think that the more people learn about constructing good experimental methodologies, the more it clarifies their thinking and helps them sort things out in informal tests -- or even if there's any much value to the informal tests we might pursue.

I'm not so much trying to whip people into shape for presenting papers to peer-reviewed journals as try to promote more awareness of what goes into proper testing, with the notion that it will help inform their approach to evaluating both gear performance and some of the claims made for different products, techniques, etc.


But... yeah... there ARE a lot of variables. Acoustician Ethan Winer has (I believe) an informal write-up where he talks about the results of moving calibration mics by very small amounts in both treated and untreated rooms. While he found that treated rooms tended to have less nastiness, he also found that even in treated rooms a position change of a few inches could make for wide variations in pick up.

He analogized to the phenomenon of an informal ear test of something/anything... unless one was careful to position one's head in precisely the same spot (I can envision a plastic whiffle golf ball hanging on a string -- placing the center of one's forehead against it, say, as a position marker) for each pass, any difference heard might, more likely, be from room response anomalies.

It was a bit of an eye-opener.


Anyhow... that's one of the reasons that when talking about this particular issue I usually take the more-precision-is-better-but-who-knows-if-you'll-notice position... if you've got the CPU horsepower to spare, it presumably can't hurt...

[taturana]

i´ll keep using my 64bit IM DAW (and enjoy it a lot while doing it!) while you guys keep arguing if it´s good or not... (of course it is! )

and while i would love to see some test results i would not want to make them myself, as i can hear a good difference in my sonar 6.2 and am enjoying it... i spend my time with my ears and brain mixing music, not making useless blind tests... (in 64 bit that is...)

i´ve said it before and i´ll say again...

" 64 bit, one day you´ll be there and you´ll be loving it..."

i might just add that to my signature...

you know... this argument goes on and on and it just repeats.

how many people that say the difference does not exist have actual experience with a 64 bit IM daw? i´ve noticed that it is easier for people who actually use a 64 bit IM daw to recognize the difference... it´s subtle but it is there...... or rather, not there.... clean sound

but if your daw does not support 64 bit IM...
of course there´s no difference and the grapes are green....

[Geert van den Berg]

I thought it be interesting to ask this question to Justin Frankel creator of Reaper and guestmod at this moment.

Here's his take on this matter:

Reaper 64 bit engine...