Quantcast
cubase 5 summing why not 64 bit floating ? - Gearspace.com
The No.1 Website for Pro Audio
cubase 5 summing why not 64 bit floating ?
Old 25th January 2009
  #1
Here for the gear
 
🎧 10 years
cubase 5 summing why not 64 bit floating ?

Hello!

I got some informations:

Cubase 5 64 bit, but summing 32 bit floating!


What do you think Why?

Sonar,Ableton audio engine 64 bit float!
Old 25th January 2009
  #2
Lives for gear
 
jupiter8's Avatar
 
🎧 15 years
Because it's an absolutely useless feature put in for marketing reasons only.
Old 25th January 2009 | Show parent
  #3
Lives for gear
 
Robert Randolph's Avatar
 
🎧 15 years
Quote:
Originally Posted by jupiter8 ➡️
Because it's an absolutely useless feature put in for marketing reasons only.
Old 25th January 2009 | Show parent
  #4
Gear Guru
 
theblue1's Avatar
 
🎧 15 years
You guys are basing your answers on what rationale, precisely?


I'm always fascinated by people who are 100% sure of something, yet offer no insight into the thinking behind that certainty.

Old 25th January 2009 | Show parent
  #5
Lives for gear
 
Robert Randolph's Avatar
 
🎧 15 years
Quote:
Originally Posted by theblue1 ➡️
You guys are basing your answers on what rationale, precisely?


I'm always fascinated by people who are 100% sure of something, yet offer no insight into the thinking behind that certainty.



Excellent troll, would respond again. A+++++
Old 25th January 2009 | Show parent
  #6
Lives for gear
 
🎧 15 years
I would assume that it´s already in use where ever it makes sense.
Old 26th January 2009 | Show parent
  #7
Lives for gear
 
DeadPoet's Avatar
 
1 Review written
🎧 15 years
If 32bit gives you unlimited headroom, then 64bit will give you ... errr ... how much is two times infinity?


What converters should I buy to go with my 64 bit mixer?


Herwig
Old 26th January 2009 | Show parent
  #8
Moderator
 
psycho_monkey's Avatar
 
🎧 15 years
Quote:
Originally Posted by Robert Randolph ➡️
Excellent troll, would respond again. A+++++
Who's accusing who of trolling here?! Surely you've both made enough posts for that to be a daft comment?!
Old 26th January 2009 | Show parent
  #9
stop it or the thread will be destroyed in ...

10, 9,8 ,7 ,6, 5....
Old 26th January 2009 | Show parent
  #10
Lives for gear
 
synthoid's Avatar
 
🎧 15 years
The '64 bit' refers to the address space of the application -- it can address an amount of RAM that's practically unlimited (limited only by how much you can install). The summing precision is something altogether different, it refers to the type that's used to implement the summing bus.

I don't know why everyone would laugh off the question, when PTHD sports a 48-bit summing bus, and that's about the same amount of precision as a 64-bit floating point number (a lot less dynamic range, but so what? 64 bit float is a native type for Intel processors, just as 48 bit fixed point is natural for the Motorola 56K family that's used to implement PTHD systems).

-synthoid
Old 26th January 2009 | Show parent
  #11
Lives for gear
 
Fredo's Avatar
 
🎧 15 years
A 64-bit engine is technically not better than a 32-bit engine for summing.

A 32-bit engine does all calculations on the fpu, thereby using registers that can be as large as 80 bit. (The accumulator even extends this)
A 64-bit engine has no other choice than using SSE instructions, of which the registers are limited to 64 bit strings. Some people will argue that SSE strings are 128 bit, but that's incorrect. The 128 bit string is split up in two 64 bit strings.

So, if well written, a 32-bit audio engine can be more precise than a 64 bit engine, since it can hold larger strings of numbers.
That is when the numbers are kept as long as possibly can in the registers and are brought back to 32-bit as less as possible.
On the other hand, 64-bit is more precise than 32-bit thereby avoiding any rounding errors. But these rounding errors are neglectable when it comes to adding and subtracting numbers, which is what a summing engine basically does.
Rounding and additive (cumulative) errors are important for any process where DSP is concerned, but all DAW and plugin manufacturers are already using double precision for most -if not all- of their critical calculations.

In practice 32 bits vs. 64 bit will make no difference at all, unless you are really abusing the audio engine and are burning every gain stage by dozens of dB's and/or in scientific tests set up to expose the problem. In every day use (read: proper use) there shouldn't be any difference.

In other words, if you can hear (and measure) the difference between a 32-but engine and a 64-bit engine, then something is/was wrong (badly designed) with the 32-bit engine to start with.

Fredo
Old 26th January 2009 | Show parent
  #12
Lives for gear
 
hugol's Avatar
 
🎧 10 years
Quote:
Originally Posted by Fredo ➡️
A 32-bit engine does all calculations on the fpu, thereby using registers that can be as large as 80 bit. (The accumulator even extends this)
A 64-bit engine has no other choice than using SSE instructions, of which the registers are limited to 64 bit strings.
What? This doesn't make sense - I really don't get your logic.

If you're adding two numbers and storing the result as 32-bits then why on earth does it matter how long the registers are for that operand - you've just thrown away the additional accuracy anyway!


Equally how can you possibly argue adding 32-bit numbers together gives you more precision than adding 64-bit numbers?


By the way I agree with your conclusion, just not some of your other logic.
Old 26th January 2009 | Show parent
  #13
Lives for gear
 
Fredo's Avatar
 
🎧 15 years
Quote:
Originally Posted by HugoL ➡️
What? This doesn't make sense - I really don't get your logic.

If you're adding two numbers and storing the result as 32-bits then why on earth does it matter how long the registers are for that operand - you've just thrown away the additional accuracy anyway!
Where did I say that?
In a properly build DAW, the numbers are not stored in 32-bit strings, but kept in the FPU registers as long as possibly can before they are dumped (and rounded) to a 32-bit string. Thereby the intermediate roundings (80 bits =>32 bits) are much less frequent, than when using SSE strings which need to be dumped as soon as the 64bit strings are "full".

Of course, depending on the internal structure of each an every DAW -and the way things are tested- there are arguments in favor of both techniques. All I said is that in normal, proper, non-abusive use, the (rounding) error in a 32-bit mixer will remain in the LSB's, exactly as in a 64bit mixer.

Fredo
Old 26th January 2009 | Show parent
  #14
Lives for gear
 
hugol's Avatar
 
🎧 10 years
Fair enough - if you know you're keeping the running total at higher resolution then yes that makes sense - you're not loosing accuracy due to rounding.

This is all assuming someone's gone to the effort of validating the (probably C++ generated assembler) actually does things that way. But sure I guess that's not such a big deal - even if that bit is hand-cranked.
Old 26th January 2009
  #15
Here for the gear
 
🎧 10 years
Thank you Fredo!!!

You right!!!
Old 26th January 2009 | Show parent
  #16
Deleted User
Guest
I would take a STABLE 32 bit Cubendo over any 64 bit version of it anyday

None of that matters if its going down all the time
Old 26th January 2009 | Show parent
  #17
Lives for gear
 
tomdarude's Avatar
 
🎧 15 years
and none of that matters if native daws are still flawed with their implemention of ADC

analog subgroup inserts, parallel processing, etc...with varying cpu load...I don´t get anything near phase accurate processing
Old 26th January 2009 | Show parent
  #18
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by tomdarude ➡️
and none of that matters if native daws are still flawed with their implemention of ADC

analog subgroup inserts, parallel processing, etc...with varying cpu load...I don´t get anything near phase accurate processing
well said
Old 26th January 2009 | Show parent
  #19
Gear Guru
 
theblue1's Avatar
 
🎧 15 years
Quote:
Originally Posted by tomdarude ➡️
and none of that matters if native daws are still flawed with their implemention of ADC

analog subgroup inserts, parallel processing, etc...with varying cpu load...I don´t get anything near phase accurate processing
Here's an internal routing test I just performed on my (64 bit) DAW...
[starting with a 1000 Hz test tone generated in Sound Forge 8 and imported to my DAW; one side straight out; the other routed through a compressor plug via track insert, then through two separate busses in series -- each with a compressor on a buss insert -- exported as a stereo file and brought back into SF8]
  • left (upper): -> straight to main output
  • right (lower): -> compressor plug-in via track insert-> buss1-compressor/insert -> buss2 compressor/insert -> main output
    [total: 2 buss sends and 3 fx inserts]


They line up to the sample.

Here's a screenshot of the routing from the DAW: http://bluetrip.com/images/bb_img/routingDAW.jpg
[please excuse the dorky colors; I've been experimenting]

PS... though I follow these debates, I'll leave it to others to argue for or against the benefits of 64 float vs 32 float, etc. I suspect my modest mixes don't push the envelope too hard. I do use my DAW's ability to use 32 float for realtime playback and 64 float for rendering/export. That's a nice feature, I think, that works out nicely for those of us with lesser hardware.

Old 27th January 2009 | Show parent
  #20
Lives for gear
 
synthoid's Avatar
 
🎧 15 years
Quote:
Originally Posted by Fredo ➡️
Where did I say that?
In a properly build DAW, the numbers are not stored in 32-bit strings, but kept in the FPU registers as long as possibly can before they are dumped (and rounded) to a 32-bit string. Thereby the intermediate roundings (80 bits =>32 bits) are much less frequent, than when using SSE strings which need to be dumped as soon as the 64bit strings are "full".
Good grief.

The question was why Cubase has a 32-bit summing bus, and your answer is that it has a 80-bit summing bus. Do you really have any idea whether the summing in Cubase is done using the x87 FP stack, or using SSE, or whatever? As far as I know the only "fact" concerning Cubase that has been cited is that its user documents say that it as a 32-bit summing bus.

It is by no means impossible to implement a true 32-bit summing on an Intel processor. It is also by no means the case that all "proper" or "correct" floating point code uses the (defunct and non-standard) x87 floating point stack. It is also not true that 64-bits is inadequate precision and that 80 is need to prevent the registers from getting "full" and needing to be "dumped" (haha). It is also not true that SSE has less preceision than the x87 floating point usit -- SSE provides 128-bit extended precision types in case you want to go beserk and allow the DAW to sum all the audio signals in the universe before the registers have to cross their legs and go to the potty.

-synthoid
Old 27th January 2009 | Show parent
  #21
Lives for gear
 
taturana's Avatar
 
12 Reviews written
🎧 10 years
64 bit is very useful for mixing... to say it doesn't work is simply not true.

i have used both and like 64 bit mixing better.

i enjoy it and enjoy having the option not to ( useful for some buggy plugins)
Old 27th January 2009 | Show parent
  #22
Lives for gear
 
hugol's Avatar
 
🎧 10 years
Quote:
Originally Posted by taturana ➡️
64 bit is very useful for mixing... to say it doesn't work is simply not true.

i have used both and like 64 bit mixing better.

i enjoy it and enjoy having the option not to ( useful for some buggy plugins)


1) Why is it useful?
2) What do you like more about it?
3) What difference did you notice?
4) You like having the option not to for buggy plugins, what do you mean?

We're talking about 64 bit mix buses, not 64-bit memory addressing, you're not getting these confused by any chance are you?
Old 27th January 2009 | Show parent
  #23
Lives for gear
 
🎧 15 years
dont think they will rewrite the audio engine until the future of VST3
is not clear
what if VST3 fails?
Old 27th January 2009 | Show parent
  #24
Lives for gear
 
Fredo's Avatar
 
🎧 15 years
Quote:
Originally Posted by lllubi ➡️
dont think they will rewrite the audio engine until the future of VST3
is not clear
what if VST3 fails?
The engine is 64bit "ready" so to speak. No rewrite is needed. VST3 is also 64bit ready, while VST2.x is still 32-bit per definition.

Fredo
Old 27th January 2009 | Show parent
  #25
Lives for gear
 
taturana's Avatar
 
12 Reviews written
🎧 10 years
Quote:
Originally Posted by HugoL ➡️


1) Why is it useful?
2) What do you like more about it?
3) What difference did you notice?
4) You like having the option not to for buggy plugins, what do you mean?

We're talking about 64 bit mix buses, not 64-bit memory addressing, you're not getting these confused by any chance are you?
1. It's useful because it sounds better. also I can have an higher resolution mix file. i normally mix to a 32bit wav file and 64bit IM is able to do that better than a 32bit engine. the difference in sound is quite palpable, especially if you master to 24bit ITB (for dvd, dvd-a, etc..). I have compared 32 bit mixing and 64 bit mixing in sonar, the files do not cancel out and the result is better. check the opinions of those who use it.
2. I think 64bit IM sounds more open and clear, no shrinking of image and better reverb tails. whether the end result is 24 bit or 32.
3. see 2
4. URS strip pro, for one, crashes sonar at any attempt of 64 bit mixing... also being able to select 32 or 64 bit to save cpu while listening and then mix in 64bitmode is pretty cool too.

5. no, i am not confusing 64bit IM with 64bit OS. i work with 32bit win xp and won't change that for a while.

now what's with the hammer thinghy?
Old 1st May 2010 | Show parent
  #26
Here for the gear
 
🎧 10 years
mix engine reply/test cubase nuendo

Right now I am using nuendo 3 going to cubase 5.I also have protools 7 and 8
which I use sometimes but I am faster and have more plugins on the vst stuff so
I am using nuendo 3 until I finsish this project.Anyway i have seen alot of theory
so I won't go into that the question is:try this
take a conga or bass sample,anything solo instrument or vocal that is 16 bit
44.1k and open a track in your daw then do an import file
in nuendo or cubase it will ask you
Do you want to convert this audio file to 32 bit float?I click yes the software
re-renders the file as 32bit 44.1k.
Now playback the file.
The file will sound fantastic compared to the orig file and will be twice the storage space on your hard drive.It is increasing the quality of the orig file using floating point math.Bit depth is the future of recording.
In the future the larger the bit depth and length of files the greater the DAW

any DAW will be.Of course all converters now days are 24 bit.The hardware
has to increase to bit depths much ie greater 64bit audio card etc.Not a 64 bit driver I mean an actual converter that will spit out 64bit.

That is why I set my system for 32bit when recording mixing etc.
Now I have tried out cubase 5.1 and an import of a file at 16 bit 44.1 does not prompt to "do you want 64bit float conversion?" yet.So the audo engine as of yet will not do 64 bit files.It would quadruple the size of a 16 bit 44.k
file and increase the quality of detail 4 fold.That is the reason for 64 bit.
The files should be 4 times the size and the cpu/ram /drives would have to handle this in a mix.
I have not changed the sample rate 44.1 I have changed the bit depth.
The os of course would have to handle this.
Right now you cannot do this in any daw I think except sonar which I have not tried.
I found this out by getting a protools mix at 44.1 16 bit ,they sent me 40 tracks which I imported and converted to 32 bit float.They could not believe the quality change
I have tried 48k,96k,192k,and it is really not the sample rate that makes a difference it is the bit depth.
try it yourself if your daw allows it and you will see what I mean.
the futrue of daw is 44.1(or any sample rate) at 64 bit.I think that is why steinberg fans are clamouring for a true 64 bit engine.Your audio files will be 4 times their normal size and the quality will be awesome.
Old 1st May 2010 | Show parent
  #27
Gear Addict
 
🎧 10 years
Quote:
Originally Posted by blanco ➡️
try this
take a conga or bass sample,anything solo instrument or vocal that is 16 bit
44.1k and open a track in your daw then do an import file
in nuendo or cubase it will ask you
Do you want to convert this audio file to 32 bit float?I click yes the software
re-renders the file as 32bit 44.1k.
Now playback the file.
The file will sound fantastic compared to the orig file and will be twice the storage space on your hard drive.It is increasing the quality of the orig file using floating point math.Bit depth is the future of recording.
[..]
I think that is why steinberg fans are clamouring for a true 64 bit engine.Your audio files will be 4 times their normal size and the quality will be awesome.
That is not correct. The quality of the audio file is not in any way changed by simply converting it from 16 or 24 Bit to 32 Bit float!
The audio information's bit depth will still be "only" 16 Bit even though it indeed takes up more space on your hd.
Old 1st May 2010 | Show parent
  #28
Lives for gear
 
🎧 10 years
Lightbulb

Quote:
Originally Posted by DeadPoet ➡️
If 32bit gives you unlimited headroom, then 64bit will give you ... errr ... how much is two times infinity?


What converters should I buy to go with my 64 bit mixer?


Herwig
Clock, more accurate Clock.
ultra low jitter, with super clean power, double shielded power cables, and gold plated silver fuses.
Drawmer M-Clock or Grimm Audio cc1 0.5ps. 500fs.

also more Hard Drive speed/bandwith, a RAID10 with 4x Hitachi Ultrastar A7K2000 minimum.

http://en.wikipedia.org/wiki/Dynamic_range#Audio

the new generation DA chips allow 32-Bit,
Teac Esoteric was the first DA converter to use 32-Bit chips...


8-bit = 256 Dec.
100000000 Bin.
Dynamic Range = 48.16dB

16-bit = 65536 Dec.
10000000000000000 Bin.
Dynamic Range = 96.32dB

20-Bit = 1048576 Dec.
100000000000000000000 Bin.
Dynamic Range = 120.4dB

24-Bit = 16777216 Dec.
1000000000000000000000000 Bin.
Dynamic Range = 144.48dB

32-Bit = 4294967296 Dec.
100000000000000000000000000000000 Bin.
Dynamic Range = 192.64dB

64-Bit = 18446744073709551616 Dec.
10000000000000000000000000000000000000000000000000000000000000000 Bin.
Dynamic Range = 385.28dB
********************************

44100 x 64 x 2 = 5644800 = 5644.8kbps = 5.6448MB/s
48000 x 64 x 2 = 6144000 = 6144kbps = 6.144MB/s
88200 x 64 x 2 = 11289600 = 11289.6kbps =11.2896 MB/s
96000 x 64 x 2 = 12288000 = 12288kbps = 12.288 MB/s
176400 x 64 x 2 = 22579200 = 22579.2kbps = 22.5792 MB/s
192000 x 64 x 2 = 24576000 = 24576kbps = 24.576 MB/s

2" Tape Machines have 24 Tracks, Protools LE had 24 Tracks.
so...
96000 x 64 x 24 = 147456000 = 147456kbps = 147.456 MB/s = you need RAID10

Most 2TB HDDs Sata2 have sustained data read speed of 120MB/s to 60MB/s. (center to edge.)
download HD Tach by simplysoftware to measure your HDD performance.
147.456 MB/s x 60seconds = 8847.36 MB/minute = 8.8GB/minute (a complete DVD per minute.)
88473.6 MB for 10 Mintues. = 88.5 GB 10 Minutes.
885 GB for 100 minutes with 24 tracks at 96Khz.
Old 1st May 2010 | Show parent
  #29
Lives for gear
 
🎧 10 years
Lightbulb

Quote:
Originally Posted by Fredo ➡️
In other words, if you can hear (and measure) the difference between a 32-bit engine and a 64-bit engine, then something is/was wrong (badly designed) with the 32-bit engine to start with.

Fredo
Sonar x32 with 32-Bit engine (64-Bit engine dissabled)
sounds different than Sonar x32 with 64-Bit engine enabled.

because the 32-Bit engine was designed to be an Analog Summing bus emulation.

the 64-Bit engine was designed to be 100% Bit Transparent.

Logic also Bit transparent.
Cubase/Nuendo/Protools are analog summing bus emulation.
Mackie Tracktion with 64-Bit is not transparent, 50/50 emulation/transparency.
Reaper is on the Clean side.
Samplitude its on the Analog emulation side.
etc...
Protools TDM had 48-Bit summing engine.
Old 1st May 2010 | Show parent
  #30
Gear Addict
 
🎧 10 years
Quote:
Originally Posted by Dubai ➡️
Sonar x32 with 32-Bit engine (64-Bit engine dissabled)
sounds different than Sonar x32 with 64-Bit engine enabled.

because the 32-Bit engine was designed to be an Analog Summing bus emulation.

the 64-Bit engine was designed to be 100% Bit Transparent.

Logic also Bit transparent.
Cubase/Nuendo/Protools are analog summing bus emulation.
Mackie Tracktion with 64-Bit is not transparent, 50/50 emulation/transparency.
Reaper is on the Clean side.
Samplitude its on the Analog emulation side.
etc...
Protools TDM had 48-Bit summing engine.
On what facts do you base your statements that certain DAWs use "analog summing bus emulation"? I've never seen any claims from the the respective companies,
for example Steinberg, about such a feature.

If there was indeed some "analog summing bus emu" going on, I would think the companies implemented it because they'd think of it as useful for the user and/or their marketing,
and would thus mention it in their advertising or even their detailed feature lists for the respective software. But as mentioned above, I have never read such a claim,
especially concerning Cubase/Nuendo.
📝 Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 295 views: 74021
Avatar for anguswoodhead
anguswoodhead 26th March 2013
replies: 15929 views: 1532488
Avatar for Ragan
Ragan 11th January 2019
replies: 1296 views: 181391
Avatar for heraldo_jones
heraldo_jones 1st February 2016
Post Reply

Welcome to the Gearspace Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…

Forum Jump
Forum Jump