Exporting 32bit

standardly when I finished mastering {still 32 bit file} I save file in Wavelab as 24 bit, without dithering. Than import track to S5 and do dither from 24 to 16. Works well. Or you can import 32 file to S and make dither.

But wait if you export 32 bit and save it as 24 bit dont you lose 8 bits without dithering

I think if you export 32 bit file to 24 from Sonar, it doesent matter if is dither on or off. I dont see any 32 bit dither, only 8, 16 and 24. I think dither dont affect 32 to 24 or I am wrong???

I think if you export 32 bit file to 24 from Sonar, it doesent matter if is dither on or off. I dont see any 32 bit dither, only 8, 16 and 24. I think dither dont affect 32 to 24 or I am wrong???

Well - now we then have to ask ourselves why there is such a thing as a 24 bit dithernoise. When we finalize a production for CD, the very last step is add a 16 bit dither noise to a file with a larger bitdepth. This noise (as probably most of us know by now) is masking the artifacts created when DECREASING a wordlength (and for that matter - a files size).

So, what about larger files in 32 bit floating format then?

To start with - the 32 bit floating file is larger than the 24 bit fixed file. Regardless the "Mantissa"-aspect the 32 file is larger than the 24. It does contain more digital information which makes it more precise. This increased size is most likely subject to the "floating numbers", and the fact that decimals can be quite endless, when needed, in the floating format. Decimals = precision.

A recently exported project from Sonar 5 gave me the following sizes:

24 bit fixed PCM - 95,5 MB
32 bit IEEE - 127 MB

The way I look at this is - It takes XXX MB of digital data to represent how my project is sounding to my ears, when I'm playing, from start to end, it in Sonar. When I export it to 32 bits, the file shold, in fact, sound the same to my ears as when playing it in Sonar. And it kind of does. So, let's say that 127 MB of data represent what I hear and what my ears detect.

If we then export the audio to 24 bits @ fixed PCM format, the audio is represented by 95,5 MB of data. So what happened to the remaining 31,5 MB's? Well, decimals were rounded. Did this rounding cause any artifacts? Maybe, but at very low levels.

When we apply the 16 bit dither we do it because the artifacts comes up in the lower range of the 16 bit dynamic register, a register which our ears can detect from top to bottom. Hence the 16 bit dither raises the noise floor a few dB, but manages to mask the artifacts (which by the way are not pleasant at all to hear). So if we apply 24 bits of dither we mask artifacts with an even lower level noise - noise and artifacts which is inaudible to our ears, due to the fact that it's so far down in the bottom of the 24 bit dynamic range?

So what you say. What the hell should I apply 24 bit dither for?

Here's my opinion. A 24 bit file might be subject to later treatment. Let's say mastering. People might eq, compress and limit the file further. This treatments commonly makes mixes sound louder and maybe better (loud should not be compared to good all the time) and ensures that most of the later-on 16 bits will be active. So when we tweak our 24 bit file with eq, compression, limiter and WHATEVER we also - in nearly al cases increase the noisefloor. We "lift" up the bottom so it's closing in on the top (the holy 0 dB-limit which should never be exceeded).

So as we lift up the bottom and the noise floor it is better to have a flat common noise (stretching the spectrum from 0 Hz to - say - 22000 Hz for 44.1 kHz audio which is constant in level throughout the whole range (24 bit dithernoise). If we instead have not dithered the file as it changed from a greater (32 bit) environment (Sonar) or a 32 bit IEEE floating file, the artifacts from rounding errors are thrown all over the bottom of the dynamic range. You can not know where they are. There might be a bunch at one frequency and less at another. Imagine a 1.5 dB boost of frequencies at maybe 2 Khz and it happen to lay a bunch of artifacts just there. Subtle, non-msical distorsion. As you eq, compress and stuff these subtle artifacts might interfere with what will audible be heard later on in a 16 bit file. If we instead have a flat noise in our 24 bit file, we also have a better guarantee that the music ishas been more conserved throughout the frequencyspectrum, as we mask artifacts throughout the very same spectrum.

SOOOOOooo. SO.

If you have ever asked yourself why your music never sounds the same (colder and less brilliant) when you have exported a project to a 24 bit file, you should dig into the 24 bit dither issue. If you can not spot any sonical difference between playing a project in Sonar and listening to the 24 bit file, you don't have to bother.

This whole subject of 24 bit dither is really a little of "overkill", but if you care about your babies, please give some time to evaluate it.

This is my point of view. I'll print files at 32 bits if I'm going to process them further and if the host can read 32 bit. If I need to have a 24 bit fixed PCM-file, for some reason, I will surely enable 24 bit dihering when doing the conversion from 32 to 24 bits. If I want to print a 16 bit file directly from Cakewalk I'll enable one of the POW-R-fellows.

As a summary - there should be no need to print 24 bit files if you can mix in hosts that can read 32 bit floating points. Just put 16 bit dither at the very last and final stage, just when converting the whole **** to 16 bits of wonderful wave. The only need for a 24 bit file is if the music is sent to a mastering house who prefer a fixed 24 bit PCM format or if you use a host which can not read 32 floating bits correctly.

Pattor well man that is a great explaination thats for your time in explaining it. And to all of you to for giving your info. Well i then well export in 32 bit but first i run it through my digital Roland v mixer7200 to get its summoing then i will go to wave lab and master it then ill dither it to 16 would that be right? I do want to send it to a mastering house. But first need the cash.

Ok dude here i go i have a Rme digi 9652 card. And it is connect to my Roland VmC7200 mixer it is a mixer that was made in the year 2002.

It has 24 channels and has 32 bit processing. I have fiber optic cables going to the Mixer through adats coming from my Rme card.

So what i do is send the tracks from Sonar to the mixer giving each intrument its own track. i have 4 band eqs on each track and i love the way they sound.

It saves computer power. Some people will say that digital mixers don't have there own sound i think they do. I love the way the Roland vm sounds very pleasing. There are people will disagree with me, I think that mixing on a desk is better even a digital one becasue the tracks are being seperated on hardware, Mixed

and send back, using its mixing bus. Anyway i record 24bit and send it back to Sonar the way i do it is play the song though and recored a wave file while doing it.

It goes back using adat and i export the file also all in the digital domain. I think my board sounds pretty analog even though it is digital. That is why to my ears it is pleasing. I prefer for some stupid reason Sonar 4s sound over Sonar 5 working with my board I might change my mind later. Just a first impression. Hope that answered your question.



http://www.carusomusic.com/roland/vm7000_series.htm





Comp 2.4 Amd fx53
2 gig Ram
500 gig raid 0 hard
geforce 6800 gt 256 ram
windows xp sp2
Sonar 4
Sonar 5

Ok dude here i go i have a Rme digi 9652 card. And it is connect to my Roland VmC7200 mixer it is a mixer that was made in the year 2002.

It has 24 channels and has 32 bit processing. I have fiber optic cables going to the Mixer through adats coming from my Rme card.

So what i do is send the tracks from Sonar to the mixer giving each intrument its own track. i have 4 band eqs on each track and i love the way they sound.

It saves computer power. Some people will say that digital mixers don't have there own sound i think they do. I love the way the Roland vm sounds very pleasing. There are people will disagree with me, I think that mixing on a desk is better even a digital one becasue the tracks are being seperated on hardware, Mixed

and send back, using its mixing bus. Anyway i record 24bit and send it back to Sonar the way i do it is play the song though and recored a wave file while doing it.

It goes back using adat and i export the file also all in the digital domain. I think my board sounds pretty analog even though it is digital. That is why to my ears it is pleasing. I prefer for some stupid reason Sonar 4s sound over Sonar 5 working with my board I might change my mind later. Just a first impression. Hope that answered your question.

Ok. The mixer has a 32 bit processing engine. This is a normal thing, and has been for maaaany years in digital mixing. If the resolution was less, combining just a few seperate tracks of 16 or 24 bit recordings would make the output clip. This is why it's preferred to record at 24 bits, because faders can be lowered pretty much without degradation af audio quality due to loosing bits. And then these 24 bit recordings can be calculated summed in a 32 or 64 bit environment, giving fine results.

What I would like to make you aware of is the following, and please note that I'm no ADAT-expert and have never used the transfer protocol of ADAT:

When you use the ADAT connection, as far as I know, it only allows a transfer of 24 bit digital audio streams. Some ADAT chips might use less bits, only 20 so you better know that the soundcard and the mixer sends true 24 bit streams. And this implies - if you are recording a track in sonar at 24 bits and assign this tracks output to the ADAT-connection, you should be aware that the transfer of audio will be at 24 bits between your computer and your Roland mixer.

If you then insert effects in the effects-slot in Sonar or if you adjust a tracks fader in Sonar the 24 bits will be in a 32 (or 64 bit) floating environment. If you sum some tracks in Sonar with a dedicated Sonar-bus which is assigned an ADAT output, this bus will contain more than 24 bits of data but will be transfered in the ADAT connection at 24 bits. Having said this, it simply means that the artifacts created when the audio is re-adjusted to 24 bits from 32 (or64) will travel together with your audio to your mixer. And these artifacts are very subtle, but bad-sounding non-musical remains of calculations in a higher resolution. And the artifacts are in fact using the LEAST SIGNIFICANT BIT (LSB) of the 24 bit audio stream. The LSB is the very same thing as the most silent sound that the digital data have encoded. The artifacts are then taking this most silent sounds place when they are created, and not filtered out by using dither noise.

So what can be done in order to clean up the music?

1. Record audio at 24 bits in Sonar. Do nothing (no levelling, panning, effecting, bussing or anything at all) and assign the tracks output to one of the ADAT channels. If the fader in Sonar stays at 0 dB and the audio is not routed in Sonar, there will be no calculations of audio, and the 24 bit recording will pass out to the ADAT interface in a resolution of 24 bits.

2. Taking advantage of Sonars possibilities with effects, routing, summing, eq:ing and everything will mean that your original audio of 24 bits have increased in bits and will decrease back to 24 when just when it is entering the ADAT interface. And when the digital audio data is decreasing here, the artifacts of lost bits will join the 24 bit stream and will build up. Since artifacts are created at every ADAT channel, they will build up when you later combine everything in your Roland mixer.

3. Looking at #2 above, together with my former long posting about the 24 bit dithering can give us a sollution. And this sollution is to apply 24 bit dithering as a "last effect" in Sonar. Since you assign multiple channels to multiple different outputs you would need to dither each and every one to get rid of artifacts. The important thing is that the dithering has to come LAST and that a fader which is coming after the dither have to be set at 0 dB to let the audio pass through without creating new artifacts after the dithering. This means:

If you add one or more plug-ins to the effect-slot of a 24 bit recorded track in Sonar, you would benefit from using a 24 bit dithering plug-in as the last plug in in the slot. You would then have to leave the track fader at 0 dB to not destroy the effect of the dithering. If any gain changes have to be made with faders, you should use the (last in the chain) plug-ins gains-adjusting possibilities or route the track to a seperate bus with a dither-plug (and a fader set to 0 dB).

Having dithered the output of Sonar (if Sonar is increasing wordlength by FX, level adjustments or anything like that) will ensure that the artifacts will be filtered out as the digital data stream enters the ADAT interface and later on - your Roland mixer. You will simply have a lot cleaner audio. Not randomly spread distorsion which you have no control over.

Then you also have to be aware of the RME mixer. Make sure the faders used there are all set to 0 dB gain in order to not make unnecessary calculations in that one.

OKEY.

You mentioned that your Roland mixer sounded analogue, and you like its sound. Would you please try to check out if the mixer is using any kind of dithering? The impression you have from the mixer sounding good might be that the mixer use a dither-function when the internal audio (in the mixers 32 bit environment) is leaving for a 24 bit digital to audio-converter.

My own tests with using 24 bit dither noise, is that it gives a more "musical" feeling and less of "digital feeling". For many years I didn't care that the music always was hit by this "digitaliness". But then I tried the greater resolutoion (24 bit) dither and said to myself: "Ahhh, whole lot of depth and dimension now". I have really good monitors so I can spot the difference between being careful and do things properly and not caring too much what is happening with the digital audio.

Puuhhhh.

This was a long post.

I need coffee now.

I believe that I get a perceivable difference in depth/smoothness...noticable even down to a 192K MP3... by saving the dithering until after the mixdown.

Yes. The more of musical information that is following different steps throughout the mixing, mastering and conversion operations, the more music should be able to punch through a crappy MP3.

If you exported your project as a 24 bit file (which was the only option before Sonar 5) without doing a 24 bit dithering, the artifacts would travel along to Samplitude and would give your audio a touch of this hard-to-describe "metallic-ness".

If Sonar would have had an optional 24 bit dithering function in versions prior to Sonar 5, and people would have enabled it at their export-audio-operations when saving 24 bit stereo files, we would not rave so much about that the 32 bit export sounds better. But still, I dream of that the Cakewalk-buddies will implement an optional true 24 bit dithering function in Sonar, which can be chosen just as easily as the ordinary 16 bit dithers.

ORIGINAL: ooblecaboodle

As with Cubase SX3, I observed my customary procedure of exporting the mixdown 48K/32-bit float wave to Samplitude 8.2, then using Samplitude 8.2 to dither to 44.1/16 (using the Pow-r psychoacoustic dithering in Samp.); comparing against a reference track (adding any further EQ that I might need); then Samp again to render an MP3.

Mind if I ask why you're using samplitude to bounce from 48-32 to 44-16?
Sonar has the POW-R algorithm built into the audio engine, including the much praised POW-R-3, so why would you do this final step in samplitude?

Sure,

When I make CD backups, and CD Demos, and MP3s for OMD sites, I use Samplitude. Organization-wise, it makes sense for me to put all the high-bit rate, high sample-rate files in a "Music Production" folder, in a sub-folder with the songs title...in the Samplitude work directory.

Since Samplitude works with virtual objects, I can have multiple VIPs that point to the same sets of master wave files. Then, no matter what I am doing, I can quickly make different demo CDs, backups, MP3s etc., at whatever quality I want, from the highest quality masters that I have. Samplitude also gives me more levels of dithering and other control that I have in SONAR 5 Producer...and it is much faster to work with, and does batch processing...so I can have presets that run, say, 3 different bit-level MP3s from a single wave in a batch, etc.

Plus, it ALSO sounds great.

Hope this helps.

Well, it is really good that you asked. I imagine myself when I first started, going through forums like this, picking up on stuff here and there...there are a lot of folks reading along...it is easy to forget that the reasons for working as we do are not always obvious, and even if you THINK you know why someone is doing something, your really don't, until you ask. Then you STILL might not understand it...LOL.

ORIGINAL: ooblecaboodle

Well, it is really good that you asked. I imagine myself when I first started, going through forums like this, picking up on stuff here and there...there are a lot of folks reading along...it is easy to forget that the reasons for working as we do are not always obvious, and even if you THINK you know why someone is doing something, your really don't, until you ask. Then you STILL might not understand it...LOL.

No problem, I was just wondering whether or not you'd realised that there was POW-R dithering in sonar!
I have no qualm at all about samplitude, in fact I'd love to try it out someday. I did actually suggest it in a studio meeting when we were discussing what to replace our old, knackered SADiE 4 with, but it seems that we're going the Pyramix route - which I'm equally happy about!!


Oh, and sorry for hijacking your thread, voyuers!!

Continuing OT for anothe post: I few weeks back I read a couple of write-ups on Pyramax, and it looks amazing...you really would have to go above Samplitude to Sequoia, its big brother, to get close, I think.