Toneboosters and bit-depth question

[badass_billy]

Got a technical question on a topic I know very little about. After doing some more in-depth technical reading on bit-depth, I started taking a closer look at my projects in Reaper, and came across a tool, Bit Meter, that comes stock.

And something looked a little off when I placed it immediately following a Toneboosters plugin in the effect chain.

Normal (other plugins):



Toneboosters:



Any idea if this is something that has any real-world consequences?

[aberration123]

Might be that Toneboosters only use 32bit floating point, where Reaper uses 64 I think. Some do some don't. Pretty sure Bitwig only uses 32bit floating point.

Got a technical question on a topic I know very little about. After doing some more in-depth technical reading on bit-depth, I started taking a closer look at my projects in Reaper, and came across a tool, Bit Meter, that comes stock.

And something looked a little off when I placed it immediately following a Toneboosters plugin in the effect chain.

Normal (other plugins):



Toneboosters:



Any idea if this is something that has any real-world consequences?

[aberration123]

Barricade by any chance? does it say no dither in the bottom right hand corner? or are you applying 24 bit dither?

[badass_billy]

Barricade by any chance? does it say no dither in the bottom right hand corner? or are you applying 24 bit dither?

Same deal with every plugin I tried, v3 and v4. No dither. Not really worried. Always get great results from their tools, so I doubt it has any real impact on sound quality.

[MogwaiBoy]

Try some other non-Toneboosters plugins too if you have any.

As someone said above, this is because they operate at 32-bit floating point internally.

Unfortunately, 64-bit audio is still not really a thing. Reaper is ahead of its time in this regard, as were Voxengo many years ago, but most 3rd party plugins run at a 32-bit audio rate internally, including most of the biggest players in the game, ie Waves, Native Instruments, Izotope, Fabfilter, etc. They argue that they are doing your CPU and harddrive space a favour.

[Ploki]

What could possibly be gained by 64bit fp?
-250dB noise floor?

[Aleksey Vaneev]

They argue that they are doing your CPU and harddrive space a favour.

That's some very outdated opinion. Doubles are practically not slower than floats on modern x64 CPUs.

[Aleksey Vaneev]

What could possibly be gained by 64bit fp?
-250dB noise floor?

If convolution and FFT are concerned, 64-bit FP is much better as 32-bit FP peak error is very large.

[Trancit]

What could possibly be gained by 64bit fp?
-250dB noise floor?

If convolution and FFT are concerned, 64-bit FP is much better as 32-bit FP peak error is very large.

This means to me 64bitFP makes perhaps sense in speciallized plugins but what shall a 64bitFP mixing engine do in a DAW (compared to 32bitFP)?? ... In practice = hearable result... not in pure theory...
You have a very deep DSP understanding so most likely you are the best one to ask...

[Aleksey Vaneev]

This means to me 64bitFP makes perhaps sense in speciallized plugins but what shall a 64bitFP mixing engine do in a DAW (compared to 32bitFP)?? ... In practice = hearable result... not in pure theory...

Pure theory usually does not consider peak error, or distribution of error. In average, 32-fp is adequate, but its peak error can be huge. That's what one hears when comparing 64-fp and 32-fp.

[Trancit]

This means to me 64bitFP makes perhaps sense in speciallized plugins but what shall a 64bitFP mixing engine do in a DAW (compared to 32bitFP)?? ... In practice = hearable result... not in pure theory...

Pure theory usually does not consider peak error, or distribution of error. In average, 32-fp is adequate, but its peak error can be huge. That's what one hears when comparing 64-fp and 32-fp.

Tbh, I have no clue what "peak errors" are and I am personally not able to hear any difference between switching in a DAW 64bit engine on/off... but perhaps I don´t know what to listen for...
Otoh...if I don´t do with decent monitors/headphones... how shall any normal consumer??

[MogwaiBoy]

I'm all for the highest quality of digital audio signal with the lowest mathematical noisefloor to give it the best chance to survive heavy processing better without accumulating rounding errors/degredation etc - and 64-bit audio through-and-through is a clear upgrade in that sense..... at least in the numbers. In the real world most people won't notice or care - but I respect devs like Voxengo who care for us

I know Reaper can even export 64-bitFP wave files - that is total overkill (and disk space).... for now.

Go back to the 90's when 16-bit audio was essentially considered lossless

[badass_billy]

Try some other non-Toneboosters plugins too if you have any.

Yea, I found a few more. Waves and Fabfilter (oddly). Most of my collection is 64.

[Aleksey Vaneev]

Tbh, I have no clue what "peak errors" are and I am personally not able to hear any difference between switching in a DAW 64bit engine on/off...

E.g. you could make two renders - in 64-fp and 32-fp. Then subtraction of one from the other produces a delta signal whose RMS is considered an average error. But this delta signal also has its peak values. The thing usually overlooked is that peak error spectrum is not just some sort of white-noise sound - it may be a spectrally-rich noise like the one produced by bit-depth truncation. That's why one may hear it, even though on average it's almost non-existent.

[Trancit]

Thank you... will try!