Topic: 24,32,56,64bit what the heck???

Since this seems to be always a cause of confusion I am going to explain which bits do have which meaning.

1. Binary Format
In principal there are two different binary formats (the format of the executable code running on your host CPU): either 32 or 64bit. While the 32bit format supports 32bit processor registers and a 32bit address space the 64bit binary format supports 64bit processor registers and a 64bit address space. The entire kernel of your OS (including all drivers) is running in one of these binary formats. Applications (including plug-ins) may run in one of the binary formats, even a different one from the kernel. 32bit Windows only supports 32bit Applications though. 

2. Audio Format
On the other hand there are two audio processing data exchange formats used by the interface to and from the DAWs: 32 and 64bit floating point. PowerCore supports 32bit floating point precision.

Internally the PowerCore DSPs accessing the audio data in either 24 or 48bit fixed point precision and do all digital signal processing with a 56bit precision. All conversion is handled by PowerCore seamlessly.

3. 32 vs. 64bit Driver Binaries
OSX and Windows come into two general flavors: 32 and 64bit kernel.
3.1 OSX
OSX (10.6+) includes both a 32 and 64bit kernel. You decide which kernel version you are running. By default the 32bit kernel will be used currently. In any case you will need the correct drivers. However regardless of the kernel used the application space is 32 and 64bit compatible. All necessary mappings between the two models are transparently handles in the OSX kernel.
OSX drivers, like all other OSX software, can be provided in universal binaries, that is the executable will be delivered in more flavors. It will then appear seamless to the user for all platforms supported by this binary. The PowerCore driver 4.0.4 will be delivered in a PPC and Intel 32/64bit universal binary.

3.2 Windows
Under Windows you choose either the 32 or the 64bit version and install the OS with one of the kernels. You will then need the drivers accordingly. PowerCore 4.0 is delivered in 32 and 64bit.

4. 32 vs. 64bit Plug-in Binaries
32bit DAWs can load 32bit plug-in binaries. All DAWs support a 32bit floating point plug-in interface; some of them also a 64bit floating point plug-in interface.
64bit DAWs can load 64bit plug-in binaries. Some of them bridge to 32bit plug-in binaries. All DAWs support a 32bit floating point plug-in interface; some of them also a 64bit floating point plug-in interface.
PowerCore 4.0 delivers 32bit plug-in binaries. A later release will additionally support the full 64bit binary path.

5. 32 vs. 64bit Floating Point Audio Format
32bit floating point contain a significant 24bit mantissa that delivers a dynamic range of 144dB. The signal is required by the DAW to not exceed 0dB or it will clip at one or the other position in the audio path. PowerCore transforms the 32bit floating point to 24bit fixed point without loosing precision, processes the audio internally with 56bit precisions and transforms the result back into 32bit floating point.

While it is crucial for the audio quality to be as precise as possible during the algorithmic processing due to the summation of processing errors in loops, the transport between such algorithms needs only to be as precise as required for the output medium.

PowerCore 4.0.4 for OSX:
contains universal 32/64bit driver binaries and 32bit plug-in binaries with a 32bit floating point processing interface to the DAW.
PowerCore 4.0 for Windows 64bit:
contains 64bit driver binaries and 32bit plug-in binaries with a 32bit floating point processing interface to the DAW.
PowerCore 4.0 for Windows 32bit:
contains 32bit driver binaries and 32bit plug-in binaries with a 32bit floating point processing interface to the DAW.

Please put your setup in your signature!

This will help us to relate the issue you are talking about to a certain configuration.

Re: 24,32,56,64bit what the heck???

hagen@TC wrote:

PowerCore 4.0 delivers 32bit plug-in binaries. A later release will additionally support the full 64bit binary path.

the sound quality will change/ get better after that right?
thanks

Re: 24,32,56,64bit what the heck???

firubbi wrote:
hagen@TC wrote:

PowerCore 4.0 delivers 32bit plug-in binaries. A later release will additionally support the full 64bit binary path.

the sound quality will change/ get better after that right?
thanks

32 vs 64 bit as it relates to operating systems and software has nothing to do with sound quality.

http://www.whitecat.tv

PowerCore 6000 (4.3.3 RCIV) - Macbook Pro Core 2 Duo 3.06 - OSX 10.6.8 - Nuendo 5.5

Re: 24,32,56,64bit what the heck???

firubbi wrote:
hagen@TC wrote:

PowerCore 4.0 delivers 32bit plug-in binaries. A later release will additionally support the full 64bit binary path.

the sound quality will change/ get better after that right?
thanks

How would that be? According to the original post: In principal there are two different binary formats (the format of the executable code running on your host CPU): either 32 or 64bit.
We are talking about the executable format. The code containing all GUI staff of the plug-in. Remember: all processing is done in the DSPs and even then as long as you don't change the audio format no change to the sound is to be expected.

Please put your setup in your signature!

This will help us to relate the issue you are talking about to a certain configuration.

Re: 24,32,56,64bit what the heck???

firubbi wrote:
hagen@TC wrote:

PowerCore 4.0 delivers 32bit plug-in binaries. A later release will additionally support the full 64bit binary path.

the sound quality will change/ get better after that right?

Ok think of it like this - presumably you have digital photos on your computer somewhere - if you buy a new computer, do your photos look any better?

Why not?

Re: 24,32,56,64bit what the heck???

Fitz wrote:

Why not?

Double-precision processing isn't possible in 32 bits thus more errors result, which is why double precision sounds better at least to my ears.

Re: 24,32,56,64bit what the heck???

biggot wrote:
Fitz wrote:

Why not?

Double-precision processing isn't possible in 32 bits thus more errors result, which is why double precision sounds better at least to my ears.

I see there is really a need for my article. Have you read the above?

The executable binary type has nothing to do with the audio data format.
Despite I don't think you will be able to hear the difference of 32bit vs. 64bit floating point in the DAW audio path between the plug-ins. As explained we do not talk about internal algorithm precision, as this is entirely defined by the algorithm of the plug-in. You may read up a little here:
http://en.wikipedia.org/wiki/Dynamic_range
http://en.wikipedia.org/wiki/Absolute_t … of_hearing

Please put your setup in your signature!

This will help us to relate the issue you are talking about to a certain configuration.

Re: 24,32,56,64bit what the heck???

hagen@TC wrote:

I see there is really a need for my article. Have you read the above?

Hi Hagen,

I did read and thank you.

As you say it's about error correction algorithms, with more bits allowing greater accuracy.

Of course the by-product of that is more allowable memory registers and not the other way round but I understand why people disagree and I am not wanting to cast any doubt on you, on the contrary what you have written brings even more clarity to the issue for me personally and I'm sure for anyone who reads particularly the last paragraph, even in a general way.

Thanks for all the work you have done for the platform. I really hope it can continue towards the full 64 bit computational datapath whether implemented as 32 bit ganged or otherwise, since 56 bits is enough to run DSP algorithms I'm sure.

Hope I'm making sense!

All the best to poco users the world over

Re: 24,32,56,64bit what the heck???

biggot wrote:

Hope I'm making sense!

Not even slightly.

Re: 24,32,56,64bit what the heck???

Okay: Madonna's first hits were recorded 16 bit........so if you can't make good music 16 bit, you sure not can make good music 24,32,or 64 bit (floating where ever possible ;o)
Let's make music instead of talking bits & bytes  ;o)

Re: 24,32,56,64bit what the heck???

robbie wrote:

Okay: Madonna's first hits were recorded 16 bit........so if you can't make good music 16 bit, you sure not can make good music 24,32,or 64 bit (floating where ever possible ;o)
Let's make music instead of talking bits & bytes  ;o)

Basically if anyone doesn't really grasp the concepts of bit depth with regard to CPUs and audio, just be advised that the type of CPU does not change your audio in any way whatsoever (so long as it's fast enough to run your DAW of course...).  A 24bit 44.1KHz file recorded on one machine is not going to 'sound better' on another just because the CPU is different.

Bit depth in audio is a bit like the number of colours in a photograph.  At the end of the day the only thing that is going to make a difference to the image is how good your printer is.

Re: 24,32,56,64bit what the heck???

And even black and white pictures can be fabulous......(ask Madonna)

Re: 24,32,56,64bit what the heck???

Fitz wrote:

Basically if anyone doesn't really grasp the concepts of bit depth with regard to CPUs and audio, just be advised that the type of CPU does not change your audio in any way whatsoever (so long as it's fast enough to run your DAW of course...).  A 24bit 44.1KHz file recorded on one machine is not going to 'sound better' on another just because the CPU is different.

It also depends on processor scheduling, since the OS can proritize bit depth dynamically thus resulting in poorer quality.

Fitz wrote:

Bit depth in audio is a bit like the number of colours in a photograph.  At the end of the day the only thing that is going to make a difference to the image is how good your printer is.

Or monitor.

Precision also effects the end result as Hagen has implied error corrrection algorithms are all important and subjective analysis does not always produce and adequate result.

Re: 24,32,56,64bit what the heck???

biggot wrote:

It also depends on processor scheduling, since the OS can proritize bit depth dynamically thus resulting in poorer quality.

Please explain how processor scheduling would cause changes to digital data in a file or data stream.

Now explain how it would affect the internal processing that happens with an external DSP such as within a Powercore, bearing in mind all the computer CPU is doing is executing instructions that move data into it to be processed, and collecting the results.

biggot wrote:

Or monitor.

Precision also effects the end result as Hagen has implied error corrrection algorithms are all important and subjective analysis does not always produce and adequate result.

And as we all know, displays are affected by your CPU...

Hagen was talking about the precision that the audio data is processed at internally within the Powercore.  Bear in mind you could do the same calculations on a pocket calculator and get the same results.  The DSP is simply optimised for, and therefore very very fast at these types of calculations.

The computer's CPU word size has nothing to do with audio data apart from how quickly and efficiently calculations on said data can be performed.

If you think otherwise, then may I suggest you write a scientific paper on the phenomenon and try to get it published - because you have obviously discovered a form of magic.

Last edited by Fitz (2010-06-20 18:24:55)

Re: 24,32,56,64bit what the heck???

Hi Fitz

I don't really understand how it works, all I can do is trust my hearing and what I am "seeing" is the sound from the virus is being modified depending in order to keep it stable which is as I expect and respect from a technical point of view.

Also, the sound coming from the V-station is very poor no matter what I do since it has no precision.

Re: 24,32,56,64bit what the heck???

BTW I have Duende and althought it performs some calculations in the host system it does not do it dynamically, ie it is fixed point.

HTH's

Re: 24,32,56,64bit what the heck???

biggot wrote:

I don't really understand how it works,

Evidently.

Re: 24,32,56,64bit what the heck???

Fitz wrote:

Evidently.

Well coming from an old-school background, big studios and so on we talk about experience, so let me share mine here.

I bought a powercore firewire with a $500 voucher since what initially attracted me here was what I refer to as the alleged Virus debacle, because I wanted to test claims of tc non-compliance with customer expectations.

By the time I got it (virus) all issues were "evidently" fixed.

I had heard there are mid-year sales but wanted to try the virus after testing out the 24C and it (leveler) meeting my expectations, as well as cabinet sim etc., so got the Virus even though I could have waited and bought a V-station as well but since the latter is only 32 bit single precision I passed.

I don't regret not waiting for the virus as it well just might take an eternity for Novation to update their plug in for powercore and I am happy using an X-station anyway since the native version is crud as well.

Re: 24,32,56,64bit what the heck???

biggot wrote:
Fitz wrote:

Evidently.

Well coming from an old-school background, big studios and so on we talk about experience, so let me share mine here.

I bought a powercore firewire with a $500 voucher since what initially attracted me here was what I refer to as the alleged Virus debacle, because I wanted to test claims of tc non-compliance with customer expectations.

By the time I got it (virus) all issues were "evidently" fixed.

I had heard there are mid-year sales but wanted to try the virus after testing out the 24C and it (leveler) meeting my expectations, as well as cabinet sim etc., so got the Virus even though I could have waited and bought a V-station as well but since the latter is only 32 bit single precision I passed.

I don't regret not waiting for the virus as it well just might take an eternity for Novation to update their plug in for powercore and I am happy using an X-station anyway since the native version is crud as well.

That's very interesting but what does it have even remotely to do with CPU word length?

Re: 24,32,56,64bit what the heck???

Think I'll get the popcorn out, observe and not get involved this time!

big_smile

Re: 24,32,56,64bit what the heck???

Fitz wrote:

That's very interesting but what does it have even remotely to do with CPU word length?

Hello Fitz

I don't know what experience you have in relation to using powercore but I believe some of the processing is actually done in the CPU, so in that sense I believe it's self-explanatory, and in the world of DSP based audio systems, not unique albeit likely the first to implement such architecture.

Many kind regards

Re: 24,32,56,64bit what the heck???

biggot wrote:
Fitz wrote:

That's very interesting but what does it have even remotely to do with CPU word length?

Hello Fitz

I don't know what experience you have in relation to using powercore but I believe some of the processing is actually done in the CPU, so in that sense I believe it's self-explanatory, and in the world of DSP based audio systems, not unique albeit likely the first to implement such architecture.

Many kind regards

I ask you again to please explain how the CPU word length would alter the sound.

If you can't explain this, then may I suggest you stop posting rubbish.

Re: 24,32,56,64bit what the heck???

Fitz wrote:

I ask you again to please explain how the CPU word length would alter the sound.

As I said, self-explanatory, unless a moderator requests clarification I suggest you read the manual.

Re: 24,32,56,64bit what the heck???

biggot wrote:

As I said, self-explanatory, unless a moderator requests clarification I suggest you read the manual.

Well in that case it should be rather simple for you to explain shouldn't it?  Except we are still waiting.

The only thing that is 'self-explanatory' is that you clearly do not understand the concepts involved (which you yourself have admitted) yet you continue to post incoherent rubbish without any evidence to back it up.

Until you are prepared to support your 'belief' with actual evidence, may I suggest you stop wasting everybody's bandwidth.

Re: 24,32,56,64bit what the heck???

Fitz, I think you had him at "not even slightly." 

Funny thread though.  This problem to me is like trying to sweep up the beach - too much sand.  hagen@TC provided a broom - for those who cannot figure out which end is up, not much to be done about it.