Bit depth of Logic processing

Hi,

 

What bit depth is Logic working on ? You can only record and render in 24 bit which is a shame :/ but Ive heard that Logic's engine is working on 32 float ...

 

so whats the point of limiting the output do 24 bit then ? :-/

 

Correct me if Im wrong.

if you think it's a shame you can "only" record in 24 bits it's because you fail to understand how it works.

 

24 bits is more than enough with about 16.8 million discrete steps (a theoretical whopping 144 dB of dynamics). In reality it's somewhat less but still more than ample.

 

Most programs record in 24 bits, e.g. Logic, Cubase, ProTools, Ableton Live, etc. So Logic is no different there.

 

Logic's audio engine operates with 32 bit float and 64 bit "where needed". This is similar to most other sequencers and better than ProTools LE. Logic's own plug-ins generally operate in 32 bit and output 32 bit.

 

The extra 8 bits can be used as headroom when adding signals together. This is also why you can exceed the master output 0 dBFS and bring it down again without signal loss (assuming you don't have any plug-in inserted on the master).

 

I don't want to sound condescending but you shouldn't worry too much about these matters, it will have almost no impact on the sound or music you make.

 

I can see from your signature that you make Drum & Bass.. in which case you probably want LESS bits to get that crunchy sound. Pretty ironic Cool tracks by the way.

Well its a bit weird that the system operating on 32 bit float can't output 32 bit float file, and therefore has to truncate to 24bit (so output should be dithered then) ... afaik dithering should be done only once thats one thing and most of digital mastering is done on 32 bit float anyway. I ask because I switched from cbs sx3 to logic recently (and TBH Logic kicks Cbs butt when creativity and sound design is concerned).

 

Well Im not an expert in any way, thats just what Ive heard and wanted to ask

I don't want to sound condescending but you shouldn't worry too much about these matters, it will have almost no impact on the sound or music you make.

 

I agree 100%! blz, you acknowledge yourself you're not an expert. If you were an expert in those matters, you would understand that there is zero need to export a mix in 32 bit floating point, even if you intend to reimport it in a new Logic project file to master it. Basically put, 144 dB of dynamic should be all you need.

 

http://www.logicprohelp.com/viewtopic.php?t=20312

 

PS: I agree again, your tracks are really cool!

Careful with that phrase, David. Remember that once upon a time, 640K of RAM was all the memory anyone could possibly need!

Yeah but there's a difference between memory, which newer, hungrier software can always require more of, and the human ear, which for our purposes will never (at a human scale) change, and will never need even the full 140 dB 24 bit files offer.

Ok, I get the point. So last question remains :

 

Should I use dithering while exporting mixdowns for further mastering in a professional plant ?

Ok, I get the point. So last question remains :

 

Should I use dithering while exporting mixdowns for further mastering in a professional plant ?

 

No, use dithering when you bounce a 16 bit file to burn on a CD yourself.

heh sure in theory 144dB IS all you would need, but what about SNR?

but hey, then we'd figure are equipment is too crappy to handle more dynamic range and we'd be stuck on the same.

 

i mean, whats 144dB good for if SNR of a compressor through which you send the singal is 101dB?

i mean.. that 144dB is digital right? so the noise you get in is actually the one you get from the equipment, boosting the "digital" dynamic range would make no sense, cause you dont add noise in digital domain, the noise is added before?

heh sure in theory 144dB IS all you would need, but what about SNR?

but hey, then we'd figure are equipment is too crappy to handle more dynamic range and we'd be stuck on the same.

 

i mean, whats 144dB good for if SNR of a compressor through which you send the singal is 101dB?

i mean.. that 144dB is digital right? so the noise you get in is actually the one you get from the equipment, boosting the "digital" dynamic range would make no sense, cause you dont add noise in digital domain, the noise is added before?

 

Exactly, so what's the point of storing more bits for each sample, which would give you more than 144 dB dynamic, when you are not even using the available 144 dB of dynamic in the first place?

Well its a bit weird that the system operating on 32 bit float can't output 32 bit float file, and therefore has to truncate to 24bit (so output should be dithered then)

You're wrong there

 

Remember those extra 8 bits are basically used for the extra headroom when adding signals or processing internally.

 

Whether or not to dither when bouncing is a point of discussion - but only if we're talking flat TPDF dithering. Never ever use noise shaping when bouncing your 24 bit file for mastering.

 

Logic uses the much superior POW-r algorithms compared to Cubase's UV22 (bleugh), however be aware that POW-r#1 is not flat TPDF dithering. I wonder why Logic doesn't have flat TPDF but that's the way it is.

 

This is slightly misleading in the bounce dialog where it says that #2 and #3 are noise shaped and #1 is dithering, since this might lead you to believe that it's flat TPDF.

 

Finally, most applications won't load a 32 bit floating point file at all. They're usually only for internal use (such as a freezed file).

 

One last point of advice, when bouncing for mastering make sure you have about 3 to 6 dBs of headroom, i.e. peak at -6 to -3 dBFS maximum. This leaves some headroom for the mastering process without the need for digital attenuation, and still utilizes the dynamical range availble within the 24 bit format.

Remember those extra 8 bits are basically used for the extra headroom when adding signals or processing internally.

 

Since it is said to be floating point arithmetics I guess there have to be some bits (maybe the 8 bits not contained in the 24bit in/out format) for an exponent, i.e. the number by which the mantissa (the 24 bits) has to be shifted left or right. In this case the value of the 24 bits changes in a way that they represent a different step width, thus increasing the dynamic range to the expense of detail resolution. This makes sense only if these numbers are used in a chain of calculations at the end of which there is an attenuation that brings the results back into a range of practical use.

If you would like to export some interim results for further calculations, a floating point file format would in fact be useful. Indeed it would be ideal for mastering since it eliminates the need of any dithering.

But I admit freely I might be completely wrong.

 

One last point of advice, when bouncing for mastering make sure you have about 3 to 6 dBs of headroom, i.e. peak at -6 to -3 dBFS maximum. This leaves some headroom for the mastering process without the need for digital attenuation, and still utilizes the dynamical range availble within the 24 bit format.

 

Would you please repeat for me dummy what, if any, dithering type one should choose from those logic offers when a bounce is done for mastering?

Logic uses the much superior POW-r algorithms compared to Cubase's UV22 (bleugh), however be aware that POW-r#1 is not flat TPDF dithering. I wonder why Logic doesn't have flat TPDF but that's the way it is.

Hi lagerfeldt,

I'm sure that's actually what you're getting when you bounce to a 24 bit file, i.e. the mix engine will do this as part of its normal job.

 

This is slightly misleading in the bounce dialog where it says that #2 and #3 are noise shaped and #1 is dithering, since this might lead you to believe that it's flat TPDF.

Well, they are very close functionally. Pow-r 1 is essentially flat until just before nyquist. It's not really noise shaping the baseband, which is what the dialog is likely alluding to.

 

You're absolutely right about 32 bit files for delivery. If you can't make your mix dynamically fit nicely into a 24 bit file, then there's something wrong with your mix, eh?

In a floating point environment, the last 8 bits are an exponent. Logic does indeed use this for internal processing. You don't need to worry about dither from 32 to 24. Remember that your converter is only 24 bit, so you never hear 32-bit float. However 32 bit float makes a huge difference for internal processing. 32-bit simplifies summing and any other gain related actions where you might exceed the digital maximum of a 24-bit system. This is why you can run your mix hot into the main out and then just turn the main output down. Logic is summing the streams in 32-bit and the fader just attenuates the mix at 32-bit float back down to within the 24-bit realm. This causes no degradation in quality.

If you would like to export some interim results for further calculations, a floating point file format would in fact be useful. Indeed it would be ideal for mastering since it eliminates the need of any dithering.

 

Again, there's no need for dithering in the first place. Dithering is only when converting from 24 bits to 16 bits.

 

If you would like to export some interim results for further calculations, as long as you export it at a reasonable level (I don't know, let's say they're peaking somewhere between -24 dBFS and 0 dBFS) then there is no need for 32 bit floating point.

 

So no, exporting 32 bit floating point is not a need.

Any particular school of thought on WHICH dithering to use? Embarrassed to confess i can not hear any difference

Any particular school of thought on WHICH dithering to use? Embarrassed to confess i can not hear any difference

 

No embarrassment: on most of the material I produce every day I can't hear a difference either. You may be able to hear the difference on a long fade in or fade out to silence, or on very dynamic material such as Jazz or Classical.

Any particular school of thought on WHICH dithering to use? Embarrassed to confess i can not hear any difference

 

Finally one who straightly writes what (as I reckon) many of us think...

That reminds me of an incident when a guy who condemned digital recording in general and audio CDs in particular came into a room with some electronic equipment and a record player in it. The record player was playing, and he was full of praise for the good old analog vinyl. He didn't know he actually was listening to the commodore amiga beside the record player that played back an 8bit recording of the vinyl record. After he was told the facts of course he heard the difference and put the blame on the low quality of the amplifier and the speakers. But there surely are much more important things than choosing the best method to bring the best of a 144dB range into 96dB. This is not the point that makes music to arouse attention.

You could just cut the 8 LSBs of a 24bit recording, and no one would know it except people with excellent gear and trained ears, and even in this case only where the sound allows it (I guess low volume passages with few discrete frequencies would undergo the strongest influence by different dithering/truncating algorithms).

Ok, there must be someone who keeps the standards as high as possible, but that's what mastering engineers are for. As we learned here, 24bits with some headroom and no dithering is the best choice for the handover to mastering. If you want to burn your own audio CD, do anything to get it into 16 bits, but even here leave some headroom since oversampling by the CD players' DACs might insert some overshooting samples.

Thank you for shedding some light on that topic and for all the advices

 

and for appreciation of my tracks on ms

 

Cheers

If you would like to export some interim results for further calculations, a floating point file format would in fact be useful. Indeed it would be ideal for mastering since it eliminates the need of any dithering.

 

Again, there's no need for dithering in the first place. Dithering is only when converting from 24 bits to 16 bits.

.

 

How about if you have 24bit Masters, and you need to make 320 bit mp3 files for promo downloads?

How about if you have 24bit Masters, and you need to make 320 bit mp3 files for promo downloads?

 

Now this is a different blob of paint on the palette... 320kbit is a resolution in time that cannot easily assigned to bit depth or sample rate. In fact an mp3 has an underlying sample rate that can be 44.1kHz or 48kHz, but this does not affect the mp3 file's bit rate, so a higher sample rate at a certain bit rate most likely reduces the accuracy somehow... But since mp3 is a lossy data reduction it's hard to decide what makes sense here. I think a good compression algorithm can work best with a high bit resolution, but higher sample rates don't do any good as they have to be converted to 44.1or 48 anyway.

BTW Logic converts everything higher than 48kHz to 48kHz mp3. If you want and mp3 to base on 44.1kHz, do the sample rate conversion first and have the result compressed.

When making mp3's (320 or not) you get better quality by using the 24 bit master instead of the 16 bit master.

 

Someone asked about what type of dithering/noise shaping to use.

 

A very short breakdown of Logic's would be:

 

·Use POW-r#1 when you're not too concerned about noise, i.e. pop/rock, etc. This will give you a fairly flat dither and just raises the noise floor slightly. The upside is you avoid the nasty truncation distortion from a 16 bit file made without dither.

 

·Never use POW-r#2 in Logic. It's broken compared to what it sounds like in Weiss hardware. Okay, I'm exaggerating slightly but it does sound worse than the same algorithm #2 in Weiss hardware. For some odd reason.

 

·Use POW-r#3 for material where you want very little apparent "noise" and if you're willing to use heavy noise shaping, i.e. add an extremely high pitched tone to your music. Most people can't hear it, but it can be bothersome for some people (including me), but only really audible in very low volume passages and long fade outs. At normal playback volumes and at 1 meter distance or more from the speakers it won't normally be audible.

 

So #1 or #3 depending on the situation.