Mix to 44khz/16bit or Render at higher Samle & Bit Rate?

[Studio de' Lara]

Hello,
I am curious as to how everyone is handling the final mix down resolutions lately.
Is it better to stay at a higher rate to the last rendering (if you use a build mix type of rendering), or to mix down (real time to either another computer / recorder) 16bit/44khz?
I have heard that adding dither or SRC is not as good as the real time transfer. Especially if the mixdown recorder/computer has good converters.
Any input is appreciated. I am doing a little test and will post my results once the mixes are done.
Regards,
Rich

[Bob L]

This has always been an area of heated debate.

Since the new SRC hi-res routines have gone into SAWStudio, the idea of samplerate converting at mixdown time has opened more options.

The routines are very good indeed, and can now generate excellent results without having to go through external hardware conversion.

I have always suggested hardware conversion of hi rate sessions, played out through expensive converters to be re-recorded to a mix machine. Now, that may not be necessary.

One thing to be careful of, though, is to make sure to select the conversion in the Mixdown dialog, not on the MT itself. The live MT conversion is samplerate converting every track independently, then summing the results. This is a lot of individual conversions and any artifacts will be ganged up in the final mix. By leaving the MT at the original hi rate, and selecting the rate change in the Mix dialog, the engine sums each track in its original rate and then converts only the final sum... one level of artifacts.

Bit Res conversion and Dither, is another topic altogether. There are those that have their favorite plug-ins and fancy algorithms to do this conversion, but when I have bit res to change, I still favor running the signal through good converters and re-recording.

Bob L

[Naturally Digital]

One thing to be careful of, though, is to make sure to select the conversion in the Mixdown dialog, not on the MT itself. The live MT conversion is samplerate converting every track independently, then summing the results. This is a lot of individual conversions and any artifacts will be ganged up in the final mix. By leaving the MT at the original hi rate, and selecting the rate change in the Mix dialog, the engine sums each track in its original rate and then converts only the final sum... one level of artifacts.

Bob L

Hi Bob,

Good point. I hadn't thought of that. :o

Is there any chance of adding an option to SRC at the output channel of the console? Just wondering if the possibility exists. Many of us are forced to do our mixdowns/mastering in realtime to incorporate external hardware or DSP (mixtreme, in my case). Having your wonderful SRC available at the output would get around the above issue.

Dave.

[bertie]

Hello,
I am curious as to how everyone is handling the final mix down resolutions lately.
Is it better to stay at a higher rate to the last rendering (if you use a build mix type of rendering), or to mix down (real time to either another computer / recorder) 16bit/44khz?
I have heard that adding dither or SRC is not as good as the real time transfer. Especially if the mixdown recorder/computer has good converters.
Any input is appreciated. I am doing a little test and will post my results once the mixes are done.
Regards,
Rich

hi?

there is one thing you can do but you need to try to believe!!!
i do this in my studio without using expensive converters and other DAT recorders and things....

After you finish your session do a mixdown at highest resolution your soundcard can handle.
Then open the mixdown file in e.g. WMP windows media player and also if you have Cool Edit Pro or similar software for recording set it up to record from WMP to xxx Software (NOTE: you need to set windows mixer to record internal sounds within windows). And what is going to happen here is you are playing e.g.192000khz file in WMP and at the same time recording it on 44100 on another software and all this with one machine and the results are 192000khz file will be recorded to 44100khz but the quality WILL NOT CHANGE!, then you can burn your song into supported 44100khz audio CD while still having 192000khz quality sounding. :)

It Works (but i don't know how :) , maybe Bob or people at Microsoft can explain why)

And remember if you want your soundcard driver to support maximum sample rate you MUST HAVE WinXP Professional!

Bertie, (lifetime SAWStudio user)

[AudioAstronomer]

It just uses linear interpolation when resampling and chops off the LSB. REally not the best way to do it Bertie :) That's technically though, if it sounds good by all means go for it.

Im going to do some tests right now with different ways of mixing down and see what sounds best to me. I havent explored this yt and today is a good day!

[bertie]

It just uses linear interpolation when resampling and chops off the LSB. REally not the best way to do it Bertie :) That's technically though, if it sounds good by all means go for it.

Im going to do some tests right now with different ways of mixing down and see what sounds best to me. I havent explored this yt and today is a good day!


have you tried it. give it a try :)
First time when this thing came up in my head i thought technically it will not work but you know sometimes things may go to another direction :) beyond our imagination :)

Bertie (lifetime SAWStudio user)

[AudioAstronomer]

I just tried it :) I personally did not like it, but it did change the sound for sure.

What Ive found, which is contrary to what I thought would happen.....

I mixed down a session from 24/44.1 to 24/96. I build mix to a hot track. now I set the SRC to highest quality. Now the 24/96 mix is being resampled to 44.1

Assign the mix to output 2. This is so I can use the solo-hottrack for A/B.

When I a/b, with my eye's closed (rapidly moving between hot tracks till im confused) I picked the mix that was mixed to 96 and being resampled live.

The low end was much much tighter. A growling synth rumbling near 40hz
instantly became VERY tight, but analyzer shows the low end was still very much there. A bass guitar previously masked by the synth became slightly more audible. And the high end was much much more tamed. A hi-hat I felt was a bit peircing in the mix was much nicer, and the snare drum sat very nicely in the mix now. I was very surprised. The whole thing seems to round out better.

Then I tried putting a compressor across the bus. My favorite 2-bus comp, blockfish, in stereo mode with a moderatly high setting, and set to slow. No saturation, no Air ettc... The difference becomes IMMEDIATLY noticable. My wife even came from the other room and asked ow i fixed the lowend so well....

explanations? perhaps others try it?

[bertie]

I just tried it :) I personally did not like it, but it did change the sound for sure.

What Ive found, which is contrary to what I thought would happen.....

I mixed down a session from 24/44.1 to 24/96. I build mix to a hot track. now I set the SRC to highest quality. Now the 24/96 mix is being resampled to 44.1

Assign the mix to output 2. This is so I can use the solo-hottrack for A/B.

When I a/b, with my eye's closed (rapidly moving between hot tracks till im confused) I picked the mix that was mixed to 96 and being resampled live.

The low end was much much tighter. A growling synth rumbling near 40hz
instantly became VERY tight, but analyzer shows the low end was still very much there. A bass guitar previously masked by the synth became slightly more audible. And the high end was much much more tamed. A hi-hat I felt was a bit peircing in the mix was much nicer, and the snare drum sat very nicely in the mix now. I was very surprised. The whole thing seems to round out better.

Then I tried putting a compressor across the bus. My favorite 2-bus comp, blockfish, in stereo mode with a moderatly high setting, and set to slow. No saturation, no Air ettc... The difference becomes IMMEDIATLY noticable. My wife even came from the other room and asked ow i fixed the lowend so well....

explanations? perhaps others try it?

and what is you final conclusion? :)

[AudioAstronomer]

Mixing to higher sample rate then using the high quality SRC to come back down indeed sounds much better than straight to my ears..... strange as it is

[Yura]

I have heard so much funny stories in my life!
so much snobbishness of famous sound engeneeres, so much words
of musicants about digital sound that was too funny! so much self-fraud
in impressions of sound quality!
guys! come down to the man's ears! or go to dolphinpool. you will then
take a unique state or get to Guiness book.
I was in a very delicate situations when sound-expert was as red as cancer
after his defineing of "what is the samplerate of this soundtrak"
I played back the file of 32000 Hz and asked him: well, is it 44100 or 48000?
"it's sounds like 48000" hi said. and hi didnt guess that I may have in my stuff
even possible to use 32000, and that I canbe so mean to play with him so.
well, I can and you can distinguish 32000 and 441000 if there is much weight
of hiest frqns, but you NEVER distinguish 44100 and 48000 or with everything
hier!!!!!!!!!!!!!!!!!!!!!!!!!!! can you hear 20000 tone signal from the tone generator???? if you can, fly to me and I tutn you into a millionaire.
all those from such a fild that one snobe say to another that hi doesnt love the Whirlwind cables and prefares only Apogi Cables, because they sound much better in his stodio.
dont fool yourselfs. forget about own persona, close all doors in your studio
and make some correct tests with the help of 1 unconcerned person.
you may not speak to any about the true results.

[Bob L]

David,

The Hi Res SRC is available to the output stage of a buildmix operation... as I said, leave the MT samplerate as recorded, then set the samplerate change in the BuildMix dialog... this routine places the SRC after the mix on the outputs.

There is a separate quality setting in the Process/Mixdown menu for setting the quality of this SRC algorithm.

Bob L

[Bruce Callaway]

Hi Bob,

If I select the highest quality in the build mix menu as you suggested, is dither in the build mix stage still relevant or does the new SRC (at the highest quality) negate the use for dither. Is dither a totally separate debate?

FYI, I am experimenting with using the leveliser to maximising volume for mastering (rather than Waves). I am still not sure Is the selection for dither in this a personal choice or not necessary?

I have stuck to 24/44.1 to minimise SRC impact. Your response implies it is now not a problem to go to maximum quality on the sound card (24/96 in my case).

Thanks
Bruce

[Bob L]

Bruce,

Dither has nothing to do with SampleRate Conversion.

Dither is used when bit resolution conversion is required... for instance from 24 bits to 16 bits... dither is arguably suppossed to make things sound better when throwing away the extra 8 bits.

Bob L

[Bruce Callaway]

OK........Just checking
Cheers
Bruce

[TotalSonic]

Hi Bob,

If I select the highest quality in the build mix menu as you suggested, is dither in the build mix stage still relevant or does the new SRC (at the highest quality) negate the use for dither. Is dither a totally separate debate?

Sample Rate Conversion and adding dither are two totally different things. Adding dither (which is a fancy name for low level random noise) is a way to smooth out the sound by reducing quantization errors when you reduce the word depth (aka "bit rate") of a sound file. In other words - you only need it if you are going from 24bit to 16bit (or other similar bit rate conversion). The effects of dither are extremely subtle. Usually you can only hear the difference when listening to fades or reverb tails (and often only after cranking them up in headphones). Usually truncated files (going from 24bit to 16bit without adding dither) will have their fades or tails kind of fizzle out as they disappear to nothing because in 16bit audio the least significant bit is still above the noise floor - adding dither basically smooths out the sound of these. Also truncated files can sometimes have their stereo field collapse a little bit more when going from 24bit than if you add dither.


FYI, I am experimenting with using the leveliser to maximising volume for mastering (rather than Waves). I am still not sure Is the selection for dither in this a personal choice or not necessary?

To change word lengths after using the Levelizer set "normalize" to at the most 99% (if you choose 100% and have the peak level to go to digital 0 then the added noise of the dither (usually about a 1/10th of a db) can push the level over digital zero - so you need to back off the level a litte to accomodate for it). You then either choose the SAW native dither you wish to use on the final mix from the "Multitrack" menu or place the dither plugin of your choice after the Levelizer as the final plugin in the chain on the output channel. Build your mix to a new 16bit file - and your done.

It was pointed out by Chris Smith (aka "The Real Computer Guy" - aka "Dark Phader") a number of years ago that's SAW's native dither added the same noise to both L&R channels and wasn't producing noise as random as some other algorithms do. I don't believe that this has changed since then - and it might explain why sometimes it sounds to me as if the stereo field collapses a bit more when using it then when using other choices. There are a number of dither plugin options readily available though that can give you an alternative though if you want it.

Waves IDR is a fine choice. So is the DX Dither plugin that comes with Sony's CD Architect 5. The Megabitmax algorithm that is part of Izotope Ozone is probably a better one. And for those on a budget - a freeware dither VST plugin that I actually like is RDR from http://www.ismusic.ne.jp/sweetboy/RDR/

Interestingly enough - there was a recent thread on Brad Blackwood's "Mastering Demystified" forum where a number of name mastering engineers
basically felt that dither choice is a really over hyped issue - and that as long as you are actually using dither when doing word length conversions then you're just as well off since the differences between using different dithers are so minimal and are usually only perceivable on very high end monitoring systems.


I have stuck to 24/44.1 to minimise SRC impact. Your response implies it is now not a problem to go to maximum quality on the sound card (24/96 in my case).

I also to date have stuck with 24/44.1 when tracking for the very same reason.

Best thing to do in this case is make a comparison and see which sounds better to you - either building a mix at the original sample rate that you recorded at, sending out via the best quality da you have and recording back in through your best quality ad at the highest rate supported, or doing a software src upsampling using the various different apps that you have.

After you've upsampled remember that ultimately your files need to be brought to whatever rate your delivery format is (i.e. 44.1 for CD / 48kHz for video / 96kHz for DVD). A mastering engineer receiving 24bit/96kHz files for CD usually will do one of two things - either use a hardware digital src box like the Weiss or Lavry which have a large amount of dsp devoted just to such a task - or loop out via dac and record back via adc to 44.1. So - the question that is debateable is how much of what is gained by going to 24/96 is retained when going back down to 16/44.1

I have not yet drawn definite conclusions to this myself - but one thing I can say is that there seems to be a bad trend where people just take others words that they read on internet forums as truth without running their own tests and deciding what subjectively sounds better to them. But when running tests - beware the placebo effect, beware missing what is the true variables, beware comparing apples with oranges, and most of all beware not trusting yourself!
:)

Best regards,
Steve Berson

[Bruce Callaway]

My renewed interest in this area started when I tried using the levelizer with dither type1 and type 3 in SAW to master a song and I found that I could not pick the difference (with a pop/rock song with no fade out). I know I need to do more tests to be sure.

Thanks for all the information, I do remember the dither debate from some years ago however just wanted to check where this was all at in respect of the SRC update (and yes my brain had faded on the actual details). Your comments about dither problems not being audible on all but the best monitors seems most pertinent to me. Thanks also for the info on normalising when mastering. I wasn't sure how all of this fitted together in the levelizer. I always check the master files on the PAS Spectrum Analyser software against suitable commercial CDs tracks.

I was just interested in using SAW products as against the Waves IDR to see if I could notice a difference (better or worse). To date I hard hardly noticed any difference without a lot more testing.

I was considering trying 96Khz sampling rate, perhaps I should stay at 44.1Khz (which still sounds fine to me). I was thinking it may give me more depth in the signal.

Cheers
Bruce

[Pedro Itriago]

Maybe not much, but if the work you do now needs to go later to something like dvd-a or sacd or whatever comes next, your source is already @24/96 and don't need to be telling yourself "gee, wish I've done it in 24/96 rather than 16/44.1".

Other than that, you can stick with 16/44.1


. So - the question that is debateable is how much of what is gained by going to 24/96 is retained when going back down to 16/44.1

:)

Best regards,
Steve Berson

[Carlos Mills]

Hi Steve,

Your comments are very interesting. Thanks for sharing them with us...


To change word lengths after using the Levelizer set "normalize" to at the most 99% (if you choose 100% and have the peak level to go to digital 0 then the added noise of the dither (usually about a 1/10th of a db) can push the level over digital zero - so you need to back off the level a litte to accomodate for it). You then either choose the SAW native dither you wish to use on the final mix from the "Multitrack" menu or place the dither plugin of your choice after the Levelizer as the final plugin in the chain on the output channel. Build your mix to a new 16bit file - and your done.

I use L2 for limiting (one setting for each song, spread throw the Multitrack). Since they have a "built in" IDR dither algorithm, I was using them for Dither. But then I realized that IDR (or any other dither algorithm plug in that I know) has an accumulative effect, meaning that using 12 of them in the same EDL is pretty CPU intensive (they do not bypass even if there is no MT entry in the track). Since the algorithm is always the same, there is no problems in patching it only in the output track, specially if you are using a PIII for this... :rolleyes:


It was pointed out by Chris Smith (aka "The Real Computer Guy" - aka "Dark Phader") a number of years ago that's SAW's native dither added the same noise to both L&R channels and wasn't producing noise as random as some other algorithms do. I don't believe that this has changed since then - and it might explain why sometimes it sounds to me as if the stereo field collapses a bit more when using it then when using other choices. There are a number of dither plugin options readily available though that can give you an alternative though if you want it.

You know Steve, I had the opportunity to test the famous Apogee UV-22 Dither algorithm. Do you know what I discovered? That it has exactly the same noise to both L&R channels. My procedure to test this was to create a dither only file (applied on digital silence) and them visually compare L & R channels in sound file view. I think that if we are dealing with stereo files, L & R channels will obviously never be the same. So if we apply the same very low level noise in quite different channels (L & R), the interactions between each channel will always be different ;)


Interestingly enough - there was a recent thread on Brad Blackwood's "Mastering Demystified" forum where a number of name mastering engineers
basically felt that dither choice is a really over hyped issue - and that as long as you are actually using dither when doing word length conversions then you're just as well off since the differences between using different dithers are so minimal and are usually only perceivable on very high end monitoring systems.

I couldn't agree more with this.


I also to date have stuck with 24/44.1 when tracking for the very same reason.
Best thing to do in this case is make a comparison and see which sounds better to you - either building a mix at the original sample rate that you recorded at, sending out via the best quality da you have and recording back in through your best quality ad at the highest rate supported, or doing a software src upsampling using the various different apps that you have.

Today, I prefer to record and mix 24/48. I have the subjective feeling that some of my 48 kHz tracks sound better than the 44.1 ones. Since SS will make all the math for all the processing in a higher samplerate, it seems to me that the error margin would be smaller for all these calculations. After the mix is ready, I use SS high resolution SRC to convert it to 44.1 (as Bob showed before in this thread) and I apply dither as the last step of the mastering process. I am not trying to prove anything, just adding some thoughts for food...


I have not yet drawn definite conclusions to this myself - but one thing I can say is that there seems to be a bad trend where people just take others words that they read on internet forums as truth without running their own tests and deciding what subjectively sounds better to them. But when running tests - beware the placebo effect, beware missing what is the true variables, beware comparing apples with oranges, and most of all beware not trusting yourself!
:)

This is perfect! ;)

Best regards,

[AudioAstronomer]

The apogee uv-22 is not random I might add as well. They do even try not to call it dither. If you look, it is normally referred to as a "process". I participated in a test (since I had an apogee unit wht uv22) to see how it worked and indeed it was nothing more than a "looped" sound, identical in both channels. A software tester found it was almost identical to the hardware version with the only anomolies being attributed to conversion and transfer inadequacies.

My personal favorite dither is that available on the voxengo plug-ins. It is truly random in both channels and in the research I've done into it, it retains much more information from the source file. I've been experimenting with the SAW dither and it seems to make the mix sound a little papery or "closed off". Just an observation, seems to be similiar to Steve's experience as well.

[Bob L]

Now try using the mix with no dither... and see how transparent it is...

Realize, this is when keeping things in the integer math as much as possible... no Waves in the loop... just the Levelizer and my eq and my gates and compressors and a select group of excellent third party plugins.

You may find that the need for dither is once again way overated and primarily due to the nasty effects of other processing and summing engines.

You mention about reverb trails and fades sounding fuzzy... I sure do not notice that problem on any of my favorite projects, for example the Steven Lee CD... all done at 44100 and 16 bits with no dither needed.

Also, try using my softedge method for the final trails in a mix... each track region using a different length softedge... I doubt you will find any fuzziness in the fade trails, even in the best headphones cranked up loud.

You all might be amazed how much resolution is present in the 16 bit 44100 format... as long as all kinds of plugins and tube mic-pre algorithms and the rest are not added to the mix. :)

Once again I suggest, if you want really random noise for dither... then simply mix in a track of cassette hiss or tape hiss from your favorite analog deck... record 5 mins and then loop that for longer projects. You may suprise yourself... but then again, I thought we got into digital because we were tired of dealing with the tape hiss. :)

I never could understand why people want to add this noise floor back in... of course it muttles the mix.

Bob L

[TotalSonic]

Now try using the mix with no dither... and see how transparent it is...

You mention about reverb trails and fades sounding fuzzy... I sure do not notice that problem on projects like the Steven Lee CD... all done at 44100 and 16 bits with no dither needed.

Even though from what I've heard on the promo videos I think he's technically a great musician I'm not personally a fan of Mr. Lee's brand of jazz fusion (more of a Miles/Sun Ra/Coltrane man myself) - but I think I obviously need to pickup this CD so I can judge for myself what you are saying here. Do you have a link to where I can purchase this?

Anyway my confusion comes from this: you state you feel there is no need to dither when tracking at 16bit - but aren't all those individual 16bit files in the multitrack being summed to a 24bit sound file for SAW's output??


Also, try using my softedge method for the final trails in a mix... each track region using a different length softedge... I doubt you will find any fuzziness in the fade trails, even in the best headphones cranked up loud.

Bob - I'm not having a problem with SAW's fades at all - I think the controllable slope can make them incredibly smooth. The problem results in any DAW I've heard when taking 24bit files and changing to 16bit - and its effects are most noticeable to me when working on material where acoustic instruments are tracked in real ambient spaces.


You all might be amazed how much resolution is present in the 16 bit 44100 format... as long as all kinds of plugins and tube mic pre algorithms and the rest are not added to the mix. :)

I agree that a lot of converters combined with the noise floor of the room, and self noise of the mic and pre our recorded files actually never really take advantage of sometimes even 16bits. However - in my direct experience for most modern converters paired with great pres I can definitely hear the benefits of recording to 24bits (even if real world performance is only 18-20 - the last 4 bits being the marketing ones!)


Once again I suggest, if you want really random noise for dither... then simply mix in a track of cassette hiss or tape hiss from your favorite analog deck... record 5 mins and then loop that for longer projects. You may suprise yourself... but then again, I thought we got into digital because we were tired of dealing with the tape hiss. :)

Actually - this is in fact something similar to this exact idea offered on CD from Crane Song - makers of some amazing eq's, comps, pres and converters - http://www.cranesong.com/analogdither.html


I never could understand why people want to add this noise floor back in... of course it muttles the mix.

There's been hundreds of white papers done on the benefits of dither for the past 20 years - just checking out the AES archives will pull up lots of them. The quantization errors introduced from simple truncation are pretty visible - and in some cases in my esperience also audible.

A nice intro for those reading this who don't want to read the real technical stuff (which is frankly above my head and puts me to sleep also) is Izotope's Dither Guide available at http://www.izotope.com/products/audio/ozone/OzoneDitheringGuide.pdf

You can also download this with wav file examples so you can judge for yourself at http://www.izotope.com/products/audio/ozone/dither_files.zip

Mastering engineer Bob Katz also has a great article on dither at
http://www.digido.com/portal/pmodule_id=11/pmdmode=fullscreen/pageadder_page_id=27/?PHPSESSID=f5d6dcaf0561e479673e8271b9430e89

Having said that - I find there is indeed a ton of material where dithering makes very little difference to my ear. One example is some hip-hop where the material originates primarily as 16bit samples, and often the final level is peak limited to all get out and rarely goes below maximum. But for acoustic music I've found that proper dithering can make a big difference is preserving the "you are in the room with the musician" vibe. OMMV.

Again - like everything in audio this debate is subjective and is best judged by the listener!!

Best regards,
Steve Berson

[TotalSonic]

Maybe not much, but if the work you do now needs to go later to something like dvd-a or sacd or whatever comes next, your source is already @24/96 and don't need to be telling yourself "gee, wish I've done it in 24/96 rather than 16/44.1".

Other than that, you can stick with 16/44.1

Very good point Pedro! I've been thinking of picking up Minnetonka's Discwelder Bronze which has basic DVD-A authoring capabilities for only $99 just for this reason - so that I can listen to my 24bit mixes even when away from my DAW.

Best regards,
Steve Berson

[AudioAstronomer]

Ive been studying dither for the last few days (I do pick something new to go over every few days, even if it's review)....

I find the in Saw it sounds remarkably better without dither of any sort when mixing. I remember when I did various listening and scientific tests in samplitude, it REQUIRED dither to sound decent when mixing. Nuendo as well. They both exhibited spikes of quantitization noise up into -90dbFS which is entirely audible. This was measured while over sampling... and primarily an effect of some sort of aliasing occuring within the quanitization noise itself.

Now, I must say there is certainly something to say about mixing with noise. I personally LOVE noise when I mix. I use digital because of the editing and arranging powers. Especially in saw. Ive gon eback to analog in this studio constantly. I miss noise!! I lust for that soft almost inaudible hiss when you bring a fader up to max. Or the blissful, somnolent hum of your favorite preamp... the comforting drone of tape hiss. I love it. I have one track dedicated in all my digital projects to noise if I feel I need it. I let all sorts of noise enter my tracks purposefully to help emulate somehow the comforting noise of the analog world. Heck, one day I turned a vacuum on in the control room and left the doors open, sounded nnnniiiiiiicccceee. You cant "hear" it, but the feel changes drammatically.

Anyways, just one mans twisted view. I dont wish to go into some long diatribe about the technical merits of dither, because Im sure most of the people here are well versed on this. Anything I have to offer is a matter of opinion and many many hours of testing and listening and analyzing, and di I mention listening? You guys can do that yourself :)

[Bob L]

Steve,

You are correct... Dither is a debate that will last many more years to come... debates like this kind of bore me. :)

They never seem to resolve anything... most opinions expressed seem to be what people have read about somewhere, rather than their own opinions from detailed testing and responsible decision making.



There's been hundreds of white papers done on the benefits of dither for the past 20 years

There have also been hundreds of papers written on the need and benefits of using Floating Point math for audio also... now, we are finally seeing some of that being questioned by many more articles and also by just observing some of the best digital audio companies coming out of the closet about using integer math.

Meanwhile... while everyone is debating... I say get busy making music.

Yes Steve, SAWStudio is summing even 16 bit files at 24 bits... and yes, if you mix back to 16 bits without dither, the bottom bits are dropped... does it hurt... you have to find out for yourself... do what I suggest and setup a session with just the built-in eq and compression and so forth and mix using just the Levelizer or nothing... you compare the digital fade or reverb trails when playing back at 24 bits in the MT or after the final 16 bit mix with no dither... I'd say you would be hard pressed to really be able to hear a difference in a blindfold test, with no phychology entering into the equation.

By the way, you or anyone can order the 'From The Ground Up' CD directly from their website at www.StevenLeeGroup.com (http://www.stevenleegroup.com/)

Whether their style of music is to your liking or not... there is no question about the expertise of each of the players in the band... and the CD is definitely a testomonial to the sonic quality capable with SAWStudio and a completely virtual mix at 44100, 16 bits.

Bob L

[Yura]

you compare the digital fade or reverb trails when playing back at 24 bits in the MT or after the final 16 bit mix with no dither... I'd say you would be hard pressed to really be able to hear a difference in a blindfold test, with no phychology entering into the equation.
Bob L

blindfold test, with the assistanse of second man - very soberness thing for
our brains.

[Naturally Digital]

David,

The Hi Res SRC is available to the output stage of a buildmix operation... as I said, leave the MT samplerate as recorded, then set the samplerate change in the BuildMix dialog... this routine places the SRC after the mix on the outputs.

There is a separate quality setting in the Process/Mixdown menu for setting the quality of this SRC algorithm.

Bob L

Huh? :confused: Sorry Bob, I was referring to a situation where I'm doing a 'loopback' to a new track in realtime, or recording to a new instance of the program, another program etc.

The issue you raised is that for realtime work (let's just say I'm outputting a mix to an external CD burner to do a 44.1/16bit trial CD for the client) in SAWStudio the SRC would be performed on each of the tracks, consuming vast resources.

What I'm suggesting is a REALTIME option to 'place the SRC after the mix, on the outputs'.

Sorry if I wasn't clear about that.

Seems logical to me. Keep the highest resolution during the live mix and process the SRC once only. Or, perhaps once for each active output track. ;) OK, maybe a native SRC plugin then...

Dave.

P.s., this is of course for downsampling. If I were upsampling files then I would want that to happen as it currently does. Upsample > process > mix > etc.

[Bob L]

If I was outputting to an external CD Burner and wanted to use the 1 stage final SRC, I would simply build the mix first, like I said, using the Build Mix output stage SRC. Then playback the mix file to the external CD Burner.

A separate plug-in would be nice... but I am not sure its that useful. The buildmix operation does not take that much time, and I would always want a mix file of the final tune anyway to layout the final cd burns, along with all the rest of the tunes.

Bob L

[Carl G.]

Bob,
That's my main application... to have a final mixed file to burn to CD.

I record at 24bit/44.1. (I do use various pluggins & lots of volume changes/etc)
Mix down at same.
On real critical stuff in the past I reduced bit rate/dithered thru Sound Forge...but now on most stuff I use SS to reduce bit rate/dither (3) simply because, time being of the essence... it is much better doing it all inside SS!
(quicker and more efficient).
I used to tell the diff in quality but got tired of straining at the nat's eyelash:)

Of course I always keep the 24bit as well as the 16bitmix.

Carl

[Bob L]

Carl,

Now you are getting into the flow. :)

Although let me suggest you try building the same mix without the dither and do a critical listening test one more time... I would almost bet you will find the No Dither mix more transparent and having an overall clearer sonic quality about it... if you can hear any difference at all. :)

Put the two mixes up side by side as stated in the comparing mixes thread... and see what kind of results you come up with.

Bob L

[Car 24]

OK guys this thread makes my brain hurt. Now, 18 months after this thread initiated, which method gives the best sound quality on a CD?

I'm using an RME Multiface interface. What would the recommendations be for the recording rates and then the Buildmix SRC. Dither or not; if yes, what setting?

The goal is the best sounding CD I can produce.

[Craig Allen]

If you are doing it all in the box, record at 44.1kHz 24 bit. Dither with the random and non-correlated L and R. Or there is the Sonoris dither which has had positive reviews (I haven't heard it yet).

[trock]

HELP!

I read thru this whole post and am confused. i should NEVER rad the technical stuff, sometimes it can derail me for months trying to go wiht what everyone is talking about without knowing what the hell i am doing

after reading this am i or am not correct in my way of using SAW

my projects are set to 44.1/24 (i have no idea why)

i record all my tracks

i add what i need to with the built in EQ, comps, bobs reverb, and maybe the levelizer, automation etc, basically i build my whole song from start to finish.

and thats all i have been using/doing

then i go to "buildmix" highest quality and build it to a file and BAM i have my song. i thought iwas "dithering" cause it has to be 44.1/16 to play on a CD????

am i doing something wrong here? my tracks sound great to me

[Dave Labrecque]

Yeah, I'd expect your CD burning program to only accept 16-bit files, but who knows? Check it.

Also check your build mix dialog. If it's set to 16-bit, then you ARE getting 16-bit files at the output. However, if you don't have some kind of dither turned on in the dialog (or added via a plug-in), you are truncating the high-resolution information, rather than adding dither to get to 16-bit output files. Many feel this doesn't sound as good.

[trock]

what is my build mix dialog?

[Dave Labrecque]

what is my build mix dialog?

When you go to build the mix to an output file, you are presented with a build mix dialog box, or build mix dialog, for short. This dialog box allows you to set the sample rate, bit depth, dither options, filename and number of channels in the output file.

[Bob L]

Only you can decide which method creates the best sounding CD for YOU... each person is different.

There are many options for creating the final CD mixes... all in all... no matter what... you must get down to 44.1k and 16 bits for a standard CD... so...

Bob L

[trock]

thanks dave

i only double click my file and make sure of the name and where it is going

man who knew there was all this other stuff to decide??

i literally am doing my whole song in SAW now, using mostly all SAW components or Bobs levelizer and reverb. i do love the sonoris eq for bass.

anyway for some reason i have set my projects for 44.1/24 which i figure means i am recording with these settings on each track, then i build mix with the highest quality (although i have no idea what the highest vs the least quality is) to my file and thats it.

what should i be learning or worrying about with the fither and all the other things SAW does?

[Tree Leopard]

thanks dave

i only double click my file and make sure of the name and where it is going

man who knew there was all this other stuff to decide??

i literally am doing my whole song in SAW now, using mostly all SAW components or Bobs levelizer and reverb. i do love the sonoris eq for bass.

anyway for some reason i have set my projects for 44.1/24 which i figure means i am recording with these settings on each track, then i build mix with the highest quality (although i have no idea what the highest vs the least quality is) to my file and thats it.

what should i be learning or worrying about with the fither and all the other things SAW does?Learning about dither? Yes, indeed! :) Its a little bit difficult at first to get a picture of what dither does.

In basic terms, a dithering algorithm hides digital noise. When a sound file is converted to 44.1 kHz / 16 bit from a higher sampling / bit rate, the "dither" filters the noise and moves it to another frequency zone where its difficult to hear. The noise occurs as a result of the conversion process - more info being chopped, diced and sliced down to less info.

A similar thing happens with digital images. Reducing a hi-rez photo down to a jpeg for a website, you need to filter the image a bit so it doesn't look like crap.

OK, here's something you can try in the safety of your own home.

- Choose a quiet song recorded @ 44.1 kHz / 24 bit. Maybe the last minute of the song. Good too, if it has a long fade at the end.

- Pull the main output slider down to - 12 dB, which will mean the mix is going to be very soft. This will help you listen for any noise.

- In Process / Mixdown menu, make sure "highest" for Buildmix quality is selected.

- In the Build Mix option box select the same sample rate you are using - in this case @ 44.1 kHz / 24 bit - with dither OFF.


Now we're changing gears here...

- In the Build Mix option box select 44.1 kHz / 16 bit with dither OFF.

- Do the same again @ 44.1 kHz / 16 bit but with dither ON and make another 3 mixes using the 3 different types of dither offered in SAW.

You'll have 5 mixes:
- one at the original sample / bit rate (your "master")
- four CD ready mixes but with different dither settings (your CD mixdowns)


Then, with a good pair of headphones, just listen and see if you notice any slight differences.

You could also A/ B the 44.1 / 16 mixes in SAW by bringing them back into a project and using the solo button & arrow key up/ down to quickly switch between the mixes.

If I remember correctly, Bob mixed the Lon Bronson without any dither at all - since he recorded @ 44.1 kHz / 16 bit in the first place there might not have been any need, in that case.

Andre

[Bob L]

There is no highest buildmix quality setting if your session is already at 44.1k... those settings are only for SampleRate Conversion when having to convert the session down to 44.1k from a higher rate.

The dither argument will rage on forever, whether or not people can really hear a difference or not... and which method is better or worse... that you will have to choose for yourself.

Try the various choices... listen intently (cause that's what it will take) and see which you think is better.

I use none... and if you listen intently to my projects you can judge for yourself whether there is any problem or not.

Bob L

[Tree Leopard]

There is no highest buildmix quality setting if your session is already at 44.1k... those settings are only for SampleRate Conversion when having to convert the session down to 44.1k from a higher rate.... oops... forgot that.

Andre

[trock]

WOW!

Andre

thanks for the info!

that was great reading and i now have a decent idea of the concept of this stuff.

i have to say, i will def try these and listen . but my stuff is sounding really good with dither off.

and also knowing now i don't need to buildmix to highest quality is great, i never knew "why" i was doing that in the first place, i just read "highest" quality and figured thats what i needed.

if i can ask this then. does using 24 bit on the project settings get me any better quality on my incoming tracks?

if bob did 16 bit and got those incredible sounds i am not sure i need that at all. however it is also because bob has way more experience too i am sure. so does 24 bit make up for some inexperience maybe? if you know what i mean, mic tech or whatever.

Bob, in the beginning i NEVER would have believed i didn't need outboard verbs, comps, eq's etc. and i NEVER would have believed i could be getting such good sounding recordings. you SWORE you could do all this "in the box" and i tell you now i really believe this is true (i still love my MCU though! :)

i am selling my API 550 EQ cause i never use it and will only keep the API pre amp cause it def helps my vocals and guitar. i use the channel eq's and comps, have never used the gate and don't even know what i would use it on. i then add a touch of verb from your reverb unit and maybe use the levelizer to bring volume up on a track if needed (and i use the sonoris eq on bass, default setting). i am using at most 9% CPU on all my songs (yes they are simple songs 20 track or less songs so i am sure this doesn;t apply to everyone) and i really LOVE the layout of SAW and how it makes mixing and recording a song fun.

with my other platforms i was always looking to buy more stuff (i stil love gear, don't get me wrong) but with SAW this is the first time i have ever downsized my setup.

interesting

[Car 24]

To recap what's been said:

1. Bob L records at 44/16, no sample rate conversion or dither

2. Everyone else has some feelings about which is best but there's no concensus

I've tried the different options except recording at 16 bit. I can't tell any difference. There's much more audible difference moving from playback system to system (studio monitors, computer speakers, home theater, car, etc...). Guess my ears are shot from too many years of racing.

[Pedro Itriago]

Seems I'm in a (most likely unnecessary) teaching mode today:

What the bit depth will do to your session is that gives you headroom. The higher the bit depth, the more dynamic range (this is excluding your actual soundcard's/converter's real performance) you'll be able to register. Also, a session with higher bit depth will let you get hotter tracks with less chance of clipping, although how the mixing engine is designed has also a lot to do with this, but I digress.

The more you know about sound/instruments/acoustics physics, you'll feel less inclined to need higher bit depth depending on what you're trying to accomplish.

A regular cd may benefit of not only higher bit depth but also higher sampling rate, but you'll always have to shrink them down to 44.1khz/16 bits if you want them to be played on a cd player. So, if you didn't go 44.1/16 all the way, you'll be on the mercy of the src and/or dithering algorithm you need to use to get your session down to 44.1/16

Then there's the view to the future, which means that you'll want to sell a cd today, but think that later you'll be famous or want to re-visit your works later and want it on say, DVD-A. This would mean that if all your work was made at 44.1/16, you'll need to upsample and zero fill you audio data to pump it up to DVD-A resolution. So, some people are already recording at high sampling/bit rates and dithering/src for a cd now, but are ready for a possible DVD-A or whatever comes in the future.

DVD video needs 48khz/24 bits, DVD-A goes 96khz/24 bits multichannel and can go up to 192khz/24 bits but you most likely stay in stereo if you use 192. SACD is just another thing.

Digressing a little more, there comes either HD-DVD oder blau ray, how will this impact the future of audio? most likely will mean that we'll be able to cram that 5000+ lp & cd collection into a single blue ray via mp3 :rolleyes:

[Jay Q]

if bob did 16 bit and got those incredible sounds i am not sure i need that at all.Trock, keep in mind that Bob uses very little FX processing. It's been mentioned in this thread, but the more processing you do, the more calculations occur and the more the signal can become degraded (because of rounding errors) in which case having the extra 8 bits helps. If you're doing pop/rock stuff, you probably won't notice, but anything subtler, and you might.

Jay