The Maillard Reaction
Max Output Level: 0 dBFS
- Total Posts : 31918
- Joined: 2004/07/09 20:02:20
- Status: offline
Mix engine bit depth comparison?
I know that SONAR uses a 64bit floating point mix engine. I love it.
I'm wondering what Pro Tools 11 native uses (the one with out the HD or HDX hardware) and I can't seem to Google up a clear answer.
Many of the older PT tutorials I am watching speak about clipping on aux inputs and a work around of inserting master faders to use as proxy input meters. I'm reflecting on how that sort of concern is pretty much a non issue with SONAR and its 64bit floating point engine.
Learning that people do this work around highlights my appreciation of the fact that SONAR can let you select either pre or post fader metering.
Is Pro Tools native up to par now or does 64bit floating point mixing require the more expensive HD option?
Thanks.
best regards,
mike
|
bitflipper
01100010 01101001 01110100 01100110 01101100 01101
- Total Posts : 26036
- Joined: 2006/09/17 11:23:23
- Location: Everett, WA USA
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 17:39:49
(permalink)
It's pretty much a non-issue with native Pro Tools, too, since it's 32-bit float like SONAR. HD is 48-bit integer, but uses a clever sliding window that gives an effective range equal to floating-point.
Now, whether or not 64-bit precision does anything for you or not, that's another question. True, there are times within a calculation where 64-bit precision is necessary. But whether it's needed from start to finish, I'm dubious.
 All else is in doubt, so this is the truth I cling to. My Stuff
|
drewfx1
Max Output Level: -9.5 dBFS
- Total Posts : 6585
- Joined: 2008/08/04 16:19:11
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 18:17:43
(permalink)
bitflipper
Now, whether or not 64-bit precision does anything for you or not, that's another question. True, there are times within a calculation where 64-bit precision is necessary. But whether it's needed from start to finish, I'm dubious.
I used to think it might be beneficial. Then I happened to actually think about it one day. Then I did the calculations. Then I tested it to confirm things. Oh, and of all the people who claim they can hear the difference between 32bit and 64bit, I've yet to find a single person who can, you know, actually accurately describe what the artifacts from doing 32bit calculations sound like. But that's not their fault, because how are they supposed to imagine they hear something specific if they aren't told specifically what to imagine ahead of time?
 In order, then, to discover the limit of deepest tones, it is necessary not only to produce very violent agitations in the air but to give these the form of simple pendular vibrations. - Hermann von Helmholtz, predicting the role of the electric bassist in 1877.
|
dmbaer
Max Output Level: -49.5 dBFS
- Total Posts : 2585
- Joined: 2008/08/04 20:10:22
- Location: Concord CA
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 18:22:41
(permalink)
I'm not particularly trying to "blow my own horn" here, but I just recently wrote a tutorial that explains why (IMO) this is largely a non-issue. Cubase, for example, is still 32-bit only and there doesn't seem to be a swelling rebellion against that. But certainly some plug-ins will do their stuff internally in 64-bit and downcovert before passing the data to the output port. So, even if using SONAR in single precision mode, you're likely to have some double-precision action going on under the covers and totally unbeknownst to you. You can read the full story here, if interested: http://soundbyte.arsov.net/Wordpress/2013/07/15/ofdigitalbitsanddecibels/
|
yorolpal
Max Output Level: 0 dBFS
- Total Posts : 13829
- Joined: 2003/11/20 11:50:37
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 18:24:03
(permalink)
@Drew...Here here! Or should that be hear hear?
|
SuperG
Max Output Level: -63 dBFS
- Total Posts : 1371
- Joined: 2012/10/19 16:09:18
- Location: Edgewood, NM
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 18:58:43
(permalink)
|
SuperG
Max Output Level: -63 dBFS
- Total Posts : 1371
- Joined: 2012/10/19 16:09:18
- Location: Edgewood, NM
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 18:58:43
(permalink)
|
SuperG
Max Output Level: -63 dBFS
- Total Posts : 1371
- Joined: 2012/10/19 16:09:18
- Location: Edgewood, NM
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 18:58:43
(permalink)
|
The Maillard Reaction
Max Output Level: 0 dBFS
- Total Posts : 31918
- Joined: 2004/07/09 20:02:20
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 21:07:31
(permalink)
Hi Drew,
I enjoyed learning about your calculations... that's why I think of it as a non issue... I love it because I don't have to think about it. I still mock best practices as that just seems normal. It's just nice to know for sure there is nothing that can clip behind the scenes.
Hi Bit,
I wonder if I have misunderstood the 64 Bit Double Precision option. I always have it checked, but I assumed it was an above and beyond type of process. I just assumed that SONAR's basic mix engine was 64 point float. Are you saying it's 32bit float until you select 64 bit Double Precision?
Hi David,
I'm off to enjoy your article.
I'm still hoping to learn what the Pro Tools native version mix engine is? I think the latest hardware is running 64bit float point but I can't find the info about the native version.
The point made by the tutorials that caused me to ask this question is that clipping does occur on the internal "aux input" tracks.
so I'd like to learn if that changed in the past 2 versions.
Thanks.
best regards,
mike
|
The Maillard Reaction
Max Output Level: 0 dBFS
- Total Posts : 31918
- Joined: 2004/07/09 20:02:20
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 21:11:25
(permalink)
Here is an interesting quote from the link that Super G posted:
"The instruction sets and architectures of the Freescale and TI families of chips are completely different, and this means that no existing TDM plug-ins will run on HDX hardware. One of the major differences is that whereas the 56k chips used fixed-point processing, the new TI chips use floating-point arithmetic. In terms of resolution, where HD used 24-bit word lengths for plug-in processing and 48-bit for mixing, HDX uses a 32-bit floating-point resolution for plug-ins and a 64-bit floating-point resolution for mixing. This has two important consequences: firstly, it means increased dynamic range for the whole system, making it pretty hard to overload the mixer even when working with a large number of tracks. And secondly, both native and DSP plug-ins are now processed with the same resolution, meaning there can be better uniformity between the sound of native and DSP plug-ins."
|
Jeff Evans
Max Output Level: -24 dBFS
- Total Posts : 5139
- Joined: 2009/04/13 18:20:16
- Location: Ballarat, Australia
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 21:16:35
(permalink)
Specs i5-2500K 3.5 Ghz - 8 Gb RAM - Win 7 64 bit - ATI Radeon HD6900 Series - RME PCI HDSP9632 - Steinberg Midex 8 Midi interface - Faderport 8- Studio One V4 - iMac 2.5Ghz Core i5 - Sierra 10.12.6 - Focusrite Clarett thunderbolt interface Poor minds talk about people, average minds talk about events, great minds talk about ideas -Eleanor Roosevelt
|
drewfx1
Max Output Level: -9.5 dBFS
- Total Posts : 6585
- Joined: 2008/08/04 16:19:11
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 21:27:35
(permalink)
Unless you set Sonar's engine to use 64 bit double precision, it uses 32 bit single precision. mike_mccue
Here is an interesting quote from the link that Super G posted:
"The instruction sets and architectures of the Freescale and TI families of chips are completely different, and this means that no existing TDM plug-ins will run on HDX hardware. One of the major differences is that whereas the 56k chips used fixed-point processing, the new TI chips use floating-point arithmetic. In terms of resolution, where HD used 24-bit word lengths for plug-in processing and 48-bit for mixing, HDX uses a 32-bit floating-point resolution for plug-ins and a 64-bit floating-point resolution for mixing. This has two important consequences: firstly, it means increased dynamic range for the whole system, making it pretty hard to overload the mixer even when working with a large number of tracks. And secondly, both native and DSP plug-ins are now processed with the same resolution, meaning there can be better uniformity between the sound of native and DSP plug-ins."
I actually almost posted part of that quote here too, as I think it illustrates the fact that audio writers often don't have a good technical understanding of digital audio:
"HDX uses a 32-bit floating-point resolution for plug-ins and a 64-bit floating-point resolution for mixing. This has two important consequences: firstly, it means increased dynamic range for the whole system, making it pretty hard to overload the mixer even when working with a large number of tracks."
If they're really using 64bit floating point and we're not talking about bugs or plug-ins that are deliberately trying to clip, I'd say it's "pretty hard to overload the mixer" in the same way it's "pretty hard" for someone to accidentally blow out the Sun like a birthday candle.
In order, then, to discover the limit of deepest tones, it is necessary not only to produce very violent agitations in the air but to give these the form of simple pendular vibrations. - Hermann von Helmholtz, predicting the role of the electric bassist in 1877.
|
The Maillard Reaction
Max Output Level: 0 dBFS
- Total Posts : 31918
- Joined: 2004/07/09 20:02:20
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 21:34:10
(permalink)
|
backwoods
Max Output Level: -49.5 dBFS
- Total Posts : 2571
- Joined: 2011/03/23 17:24:50
- Location: South Pacific
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 21:35:04
(permalink)
"If they're really using 64bit floating point and we're not talking about bugs or plug-ins that are deliberately trying to clip, I'd say it's "pretty hard to overload the mixer" in the same way it's "pretty hard" for someone to accidentally blow out the Sun like a birthday candle."  Love it! But we can still blow out our monitors through negligence...
|
The Maillard Reaction
Max Output Level: 0 dBFS
- Total Posts : 31918
- Joined: 2004/07/09 20:02:20
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 21:38:57
(permalink)
|
The Maillard Reaction
Max Output Level: 0 dBFS
- Total Posts : 31918
- Joined: 2004/07/09 20:02:20
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 21:42:57
(permalink)
Jeff Evans
Mike you might find this interesting too:
http://www.soundonsound.c...0/articles/pt_0610.htm
Thanks Jeff, That is very similar to the tutorials that have caused me to ask the original question, which is; Does this still apply? I find it interesting that the article explains the the old LE versions had more headroom than the TDM systems. I'm wondering how this evolved from Pro Tools 9 onwards. best regards, mike
|
The Maillard Reaction
Max Output Level: 0 dBFS
- Total Posts : 31918
- Joined: 2004/07/09 20:02:20
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 22:17:15
(permalink)
Found it: page 199
Clip Indication
Because Pro Tools uses 64-bit floating point calculations for the mixer, there will never be any internal clipping in the mixer. However, it is possible to
clip at the converters on your audio interface. It is
also possible to clip when writing audio to disk in
fixed point audio files (16-bit or 24-bit files only,
32-bit floating point files will not clip).
page 200
Auxiliary Input and Instrument Track,
and Sends Clip Indicators
Auxiliary Input and Instrument tracks, as well as
Sends clip indicators display yellow when the signal exceeds 0 dBFS (even though there is no internal clipping). This provides a warning that these
audio streams can clip at the output converters or
when writing fixed-point file formats to disk.
It seems like it finally caught up with the times.
|
Jeff Evans
Max Output Level: -24 dBFS
- Total Posts : 5139
- Joined: 2009/04/13 18:20:16
- Location: Ballarat, Australia
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 22:36:47
(permalink)
It is all interesting stuff but for me personally I don't worry about it because there are far more important things to worry about in terms of getting a great mix than the summing procedure.
And also if you adopt a K System approach to signal flow and have VU meters effectively telling you what the rms levels are everywhere in your system and the headroom built into that will also take care of any clipping or peaks then you will never have any issues with clipping on any track or buss or your masterbuss.
The Masterfader concept in PT is also interesting in that you can have more than one and place them where ever you like too.
Specs i5-2500K 3.5 Ghz - 8 Gb RAM - Win 7 64 bit - ATI Radeon HD6900 Series - RME PCI HDSP9632 - Steinberg Midex 8 Midi interface - Faderport 8- Studio One V4 - iMac 2.5Ghz Core i5 - Sierra 10.12.6 - Focusrite Clarett thunderbolt interface Poor minds talk about people, average minds talk about events, great minds talk about ideas -Eleanor Roosevelt
|
cliffr
Max Output Level: -80 dBFS
- Total Posts : 539
- Joined: 2010/02/19 21:44:43
- Location: Wellington, New Zealand
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/25 23:52:09
(permalink)
dmbaer
... Cubase, for example, is still 32-bit only and there doesn't seem to be a swelling rebellion against that.
Na, Cubase has been 64 bit for a while now. I only know beacuse I was thinking of adding it to my tool kit a while back so checked that first. I never did get it, and right now I feel no burning need to. Sonar is working quite well for me :-) Cheers - Cliff
i7-950 24 GB, GTX 580, W7/64 Ultimate, Sonar Platinum, Alesis MasterControl, KRK Rokit RP8g2s Some Real piano, basses, and guitars, Komplete 8Ultimate, Ibanez guitars, MusicLab RG/Strat/LPC, Trilian, Omnisphere, RMX, EWQL SO Platinum, Pianos, Choirs, VOP, Gypsy, Goliath, SD2, MOR, Ra, HS, HB, too many plugs, Midi controllers, and all kinds of weird gadgets My Soundclick Page
|
SuperG
Max Output Level: -63 dBFS
- Total Posts : 1371
- Joined: 2012/10/19 16:09:18
- Location: Edgewood, NM
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/26 00:32:17
(permalink)
|
The Maillard Reaction
Max Output Level: 0 dBFS
- Total Posts : 31918
- Joined: 2004/07/09 20:02:20
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/26 07:42:56
(permalink)
"It is all interesting stuff but for me personally I don't worry about it because there are far more important things to worry about in terms of getting a great mix than the summing procedure."
This all seems well and good, but when you are watching a tutorial and the presenter demonstrates that there is a stage that IS GETTING CLIPPED but you don't know it because you have to do a work around to have a meter run on a proxy of the signal I felt it was worth learning when and where to be aware that it may happen. I think it must have been referring to the TDM systems.
It has been difficult keeping track of which tutorials apply to the TDM systems, the HD, HDX, and native Software packages.
"And also if you adopt a K System approach to signal flow and have VU meters effectively telling you what the rms levels are everywhere in your system and the headroom built into that will also take care of any clipping or peaks then you will never have any issues with clipping on any track or buss or your masterbuss."
My OP question was a reaction to the fact that you don't have meters in the signal flow at the input of the summing busses and so you don't know what your head room is.
It is sort of a non issue as you can certainly place a meter as a first insert... but you still will not see it unless you go look for it.
The most important thing is to know if it is a possible issue. If it really is 32bit float now a days... then it is a non issue.
all the best,
mike
|
Jim Roseberry
Max Output Level: 0 dBFS
- Total Posts : 9871
- Joined: 2004/03/23 11:34:51
- Location: Ohio
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/26 10:24:54
(permalink)
Working with 16Bit audio... way back in the very early days of PC DAWs (using the original version of S.A.W.), you could clearly hear the results of several stages of destructive processing (caused by rounding error). For lack of a better description, the audio had a "buzzy" fuzzy like character. Not pleasant.
I won't argue the fact that I could hear the difference between 32Bit Float and 64Bit dual-precision Float in a blind test.
But if you give me the 64Bit option... and it comes at virtually no CPU hit (on a fast modern CPU)... and it makes any thought of rounding error completely a moot point, why not take advantage?
I agree with Jeff that there are MANY other factors that contribute more to the quality of a mix.
Same with recording at higher sample rates... (other factors have a greater impact on final results)
|
Jeff Evans
Max Output Level: -24 dBFS
- Total Posts : 5139
- Joined: 2009/04/13 18:20:16
- Location: Ballarat, Australia
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/26 10:45:15
(permalink)
If you have a meter present especially first insert in line on a buss then the meter is showing you the sum of all the signals present on that buss. By maintaining the correct rms level on a buss then it means there cannot be any clipping or overload on that buss. (just out of interest Studio One has K metering options on the buses which is pretty handy, ie it is all built in. They have put it there because it is very useful) I do it by not actually inserting a meter at all but by simply putting any buss into solo mode. My main VU meters are on the stereo buss but as I have all my busses sitting at unity then when a buss is in solo that exact signal is passed through to the stereo buss and hence my meter. The actual level of a buss is often lower than the reference level. Because several busses in my case may be used and will all sum together on the main stereo buss and that is where the final mix level reaches the reference level. The idea of putting masterfaders in PT on the buses is interesting in that they can act as a metering device showing you signal level at those points. There is no need to ever overload any buss in a multitrack digital recording system. If there is overload anywhere it means that something is simply wrong. All the tracks feeding that buss need to be grouped and dropped accordingly. The K system is a one way of keeping track of it all and it seems to work very well. If you use it there will be no overload anywhere and you will never hear distortion. The VU meters work well because even if the signal is just 3 dB over it looks way bad on the meter and very obvious. It alerts you very quickly. If the VU meters are working correctly then you are free to never worry about buss overload issues and hence you can focus on more important things. The other thing about VU metering too is the ballistics. When things are working right the needle only just reaches 0dB VU and rarely goes over. If the meter swings wildly and even occasionally then it too becomes very visible and you need to go off and find the offending track that is causing it. Once you stabilise that signal then the VU meter on the buss will settle down again and only just reach 0dB VU as it should be. You won't see any of this with peak metering anywhere near as well. It is good to know though that signals on busses can go over and there won't be an issue. I see that as a safety net, not the norm. My approach is to never let it happen in the first place.
post edited by Jeff Evans - 2013/07/26 18:08:28
Specs i5-2500K 3.5 Ghz - 8 Gb RAM - Win 7 64 bit - ATI Radeon HD6900 Series - RME PCI HDSP9632 - Steinberg Midex 8 Midi interface - Faderport 8- Studio One V4 - iMac 2.5Ghz Core i5 - Sierra 10.12.6 - Focusrite Clarett thunderbolt interface Poor minds talk about people, average minds talk about events, great minds talk about ideas -Eleanor Roosevelt
|
drewfx1
Max Output Level: -9.5 dBFS
- Total Posts : 6585
- Joined: 2008/08/04 16:19:11
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/26 12:14:40
(permalink)
Jim Roseberryut if you give me the 64Bit option... and it comes at virtually no CPU hit (on a fast modern CPU)... and it makes any thought of rounding error completely a moot point, why not take advantage?
Um, because in the real world it's already a moot point with 32 bits? The truth is the only reason people think 32 bit is an issue is because some marketing folks used carefully worded language to imply that it is. But I bet if you go back and carefully parse what they actually say, you'll see that they don't actually ever say that there's an audible difference - instead their very careful wording talks about errors in the abstract and leaves it to the reader to jump to the conclusion that those errors are a problem. Now here's the question for you: If this stuff was really a problem, then why would they use that very careful, manipulative wording?
In order, then, to discover the limit of deepest tones, it is necessary not only to produce very violent agitations in the air but to give these the form of simple pendular vibrations. - Hermann von Helmholtz, predicting the role of the electric bassist in 1877.
|
bapu
Max Output Level: 0 dBFS
- Total Posts : 86000
- Joined: 2006/11/25 21:23:28
- Location: Thousand Oaks, CA
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/26 12:29:03
(permalink)
bitflipper
I'm dubious.
Will you be creating a new login then? Can I have the bitflipper one?
|
bapu
Max Output Level: 0 dBFS
- Total Posts : 86000
- Joined: 2006/11/25 21:23:28
- Location: Thousand Oaks, CA
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/26 12:30:12
(permalink)
mike_mccue
I know that SONAR uses a 64bit floating point mix engine. I love it.
I'm wondering what Pro Tools 11 native uses (the one with out the HD or HDX hardware) and I can't seem to Google up a clear answer.
Many of the older PT tutorials I am watching speak about clipping on aux inputs and a work around of inserting master faders to use as proxy input meters. I'm reflecting on how that sort of concern is pretty much a non issue with SONAR and its 64bit floating point engine.
Learning that people do this work around highlights my appreciation of the fact that SONAR can let you select either pre or post fader metering.
Is Pro Tools native up to par now or does 64bit floating point mixing require the more expensive HD option?
Thanks.
best regards,
mike
I cudda sworn this was the SONAR forum.  I gotta get more rest.
|
SuperG
Max Output Level: -63 dBFS
- Total Posts : 1371
- Joined: 2012/10/19 16:09:18
- Location: Edgewood, NM
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/26 14:35:38
(permalink)
drewfx1
Jim Roseberryut if you give me the 64Bit option... and it comes at virtually no CPU hit (on a fast modern CPU)... and it makes any thought of rounding error completely a moot point, why not take advantage?
Um, because in the real world it's already a moot point with 32 bits?
Hmm... I assume there's no penalty for using 64-bit floats - so why not uses them internally for buses. It appears Protools uses 32-bit floats for plugins because their HDX boxes use 32-bit floats and they want to to use identical algorithms on both host and native plugs. The precise machine instructions for these plugs will differ since since the host and DSP differ, but results should be identical. Not a bad goal.
|
SuperG
Max Output Level: -63 dBFS
- Total Posts : 1371
- Joined: 2012/10/19 16:09:18
- Location: Edgewood, NM
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/26 14:35:38
(permalink)
drewfx1
Jim Roseberryut if you give me the 64Bit option... and it comes at virtually no CPU hit (on a fast modern CPU)... and it makes any thought of rounding error completely a moot point, why not take advantage?
Um, because in the real world it's already a moot point with 32 bits?
Hmm... I assume there's no penalty for using 64-bit floats - so why not uses them internally for buses. It appears Protools uses 32-bit floats for plugins because their HDX boxes use 32-bit floats and they want to to use identical algorithms on both host and native plugs. The precise machine instructions for these plugs will differ since since the host and DSP differ, but results should be identical. Not a bad goal.
|
drewfx1
Max Output Level: -9.5 dBFS
- Total Posts : 6585
- Joined: 2008/08/04 16:19:11
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/26 15:03:13
(permalink)
SuperGI assume there's no penalty for using 64-bit floats - so why not uses them internally for buses. There's no reason not to use 64 bit; there's also no reason to use it in situations where there is no benefit. It's a matter of keeping things in perspective - some things in the audio world are clearly audible and worth worrying about, while other things might be borderline audible at least some of the time. Other things will never be audible in the real world. And some people don't differentiate between the three while others of us do.
In order, then, to discover the limit of deepest tones, it is necessary not only to produce very violent agitations in the air but to give these the form of simple pendular vibrations. - Hermann von Helmholtz, predicting the role of the electric bassist in 1877.
|
cecelius2
Max Output Level: -61 dBFS
- Total Posts : 1494
- Joined: 2009/11/06 16:12:11
- Location: Pacific Northwest
- Status: offline
Re: Mix engine bit depth comparison?
2013/07/26 16:49:49
(permalink)
|