Explanations why you should record and use 32bit floating!

Explanations why you should record and use 32bit floating wave files, x64bit floating instead of 24bit in Cubase.
Also the benefits with x64bit floating double precision audio engine. I have two videos that explain the benefits.


32bit floating wave files, x64bit floating

Its faster for the computer to use floating-point instead 24bit file.
Todays computer both reads and write and work faster with 32bit and 64bit floating and calculations then example locked 24bit files. To record and work “true 24bit environment and recordings”, 24bit will not cut it even if you AD/DA are only in 24bit. You will loose quality especially if you continue work with the recorded files inside the computer DAW doing other processing. You need at least 32bit wave or 64bit floating points wave to perform “true 24bit” recordings and mixing.

I hope Steinberg add support of x64bit floating wave-files format too. Today you can use only 32bit floating waves in Cubase.




64bit double precision audio engine

I hope Steinberg upgrade their old 32bit floating audio-engine that currently Cubase are using to a newer one in x64bit floating double precision engine. Yes, Steinberg’s 32bit audio engine sounds great today already but if they upgrade it using 64bit double precision engine, it would sound even better. :wink:

You need 64bit double precision audio engine to mix true 24bit recordings.

Other internal softwares like plugins VSTI are already using 64bit double precision engine “internal” in Cubase and both SONAR, Reaper, Studio One, Bitwig, Reason, and new PRO TOOLS 11 has x64bit double precision floating audio engine already today. Its about time Steinberg upgrade theirs too!


So make a concrete example and understanding. All x64bit plugins use x64bit double precision floating, VSTI, EQs, filters, reverbs etc. 64bit double precision can be explain as it sounds warmer, more open musical “internal” but you can’t hear it because Cubase 32bit audio engine cut the audio quality in half. The Cubase 32bit audio engine are the weak link here. Also 32bit audio engine make “mixing” levels etc errors.


Here is the videos!

Extra…

Best Regards
Freddie

Hi Freddie,

I agree about the 64 bit double precision engine, but to record at anything higher than 24 bit is useless. Not even the most expensive converter in the world gives you full 24 bit of clean resolution, so why bother going higher? What benefits do you think that will give you?

Just to be clear, I’m referring to recording at higher than 24 bit. I do however understand the benefits of a 32 bit or even a 64 bit mixing engine.

Ehh… the guy’s argument for 64-bit mixing is fine, but the end result is still going to be 24-bit or less depending on the plug-ins and such used in the mix.

Until there’s some revolution in audio recording that makes microphones as they are now obsolete, a 24-bit file is biggest you’ll ever need for a recording to have pristine quality. So if your stems or mixdown are higher than 24-bit, you’re just wasting space on your hard drive.

I don’t thing better quality audio files have anything to do with the OP. And as stated, usually, your AD only works at 24 bit max. The gain reside here:

I am no expert in bit resolution or floating point calculation, but it would just make sense that computer would perform better at there native bit resolution.

Just my 2 cents worth!

Processing, yes, data, no. An audio file of 24-bits will always load faster than an audio file of 32-bits, even if the machine is 32-bit, 64-bit, or 1024-bit and greater.

The greater the bit depth of your data, the more space it requires in memory.

In terms of processing, a 64-bit system just won’t handle a native 32-bit application. A 32-bit environment has to be emulated for it, which is why it’s slower and often much less stable. Doesn’t matter for data, though.

He just deliberately opened a can of worms, stirring it all together in one pot :wink:

The second video says It all. If you change gain, sum or subtract you get around 3 dB to 6 dB noise depending on the correlation between the rounding noise.

But no one will ever hear if you add 6 dB noise at –138 dB on a heavily distorted guitar (although the guitarist will claim that it sounds better).

The funny part is that he properly would get the same figures if he had don the experiment in 64 bit floating-point. There will still be rounding noise even in 64 bit, although I admit that it will be added at around -306 dB.

I will rather have speed than bits. :wink:

If using 24bits, the computer calculations will be done using 32bit signed integers in its registers, which is faster than with floating point.

Of course coming in and going out will only be at the bit depth of the ADCs and DACs at the most.

However, use any FX and all its calculations will eat into the 24 bits with all the arithmetic rounding. Basic engineering dictates that ALL intermediate calculations MUST be done at higher precision than that required at the end. How much higher is dependent upon the type and number of calculations likely to be encountered.

Many hardware digital mixers internally calculate at 48 or 56 bits, but that could also be to make sure no internal summing points get overloaded, which 32bit floating would NEVER reach practically.

Since 32but floating still only has 24bits of significant binary digits, it provides NO guard bits to prevent calculation rounding affecting the resulting precision, so I support Cubase going to a 64bit internal floating format.

As we use all SSDs, although 64bit files would use twice as much space, there should be plenty of disk performance to spare, so I have no problem supporting the 64bit external floating point as well.

Of course, use of 64bit files in Cubase should just be another one of the file format options.