Incorrect. Calculations are still 32-bit, but result of the calculation is converted to 24-bit.
WARNING: SCIENCE CONTENT
Whenevre you process a 24-bit audio file, it introduces rounding error at most 0.5 bits. Now if you process it 16 times you (in worst case … and yes … the worst case happens on average with every 65 000th sample) have created an audio file with 16 bits of significnt data.
If you process a 32-bit floating-point audio file 16 times, you’ll get the same rounding error, but the result is like 24-bit floating point audio, which still has huge dynamic range, but increased distortion. It’s now 96dB below current signal level, which means … inaudible.
How true! Evolution of digital audio systems have brought us equipment/software with audio quality far exceeding human ear’s capabilities. And this applies to even semi-pro systems.
If you don’t believe:
Watch and download the examples from author’s website and find out yourself.
You will get exactly the same amount of rounding errors with 24-bit interger (fixed-point) and 32-bit floating-point files … but the DIFFERENCE is:
With 24-bit (fixed point) files rounding error is proportional to highest available value (0dBFS)
With 32-bit (floating-point) files rounding error is proportional to sample value
Now, with my example (16 stages of offline processing) distortion created by rounding error is about 0.0015% (completely inaudible … remeber … your speakers create at least 1% of distortion) for 0dBFS signal in both cases, but for -60dBFS signal rounding error distortion is:
With 24-bit: 1.5% … yes … this may be unacceptable in some sensitive material
With 32-bit … well … you guessed it … still 0.0015%
All your answers are here. This guy actually KNOW’s his stuff and very adeptly dispels myths, rumour and misinformation, and all using a very modest, dated consumer-grade e-magic converter!
What’s interesting about that video is that it sounds like he recorded it using a binaural mic set-up; as he moves from left to right, his position in the stereo soundfield moves correspondingly
Not really. Offline processing is something you do while working on audio file (instead of while playing back your mixdown). It can be destructive, but in case of many DAWs (like Cubase) offline processing is stored on EDIT-files instead of original audio file. You can always get back to original one by using Undo (disrecard/delete edit files). But that’s not important here…
What’s important is that result of the offline processing is stored to project audio format (16bit/24bit/32bit) while result of online (mixdown) processing is “stored” (temporarily in memory) always on audio engine format (32 bit float in Case of Cubase). That makes the difference: with online (mixdown) processing you can put many plugins in series without significan reduction of audio quality, because they talk to each other with audio engine format (32bit float), but if you do the same with offline processing result of every plugin is stored in project format and if this is not 32bit float, you will (sooner or later) introduce audible distortion created by rounding errors.
And BTW … all this thinking about bit-depths and decibels … etc … finally ringed a bell in my brains considering The Original Post and the distortion levels on that one. They are EXCACTLY what you would expect any complex 32bit floating-point DSP algorithm to create. Not better, not worse. So … if you think it sounds OK to put any VST plugin in your signal path, it’s just as OK to use Cubase’s SRC when it comes to distortion.
And now … if you are horrified how huge is the distrotion created by 32bit floating-point audio engine, watch the Monty’s video linked by “Sherz”: distortion created by the test equipment alone on the video is at -70dB … 1000 times more distortion than on Cubase’s SRC!!! And these equipment are/were used to measure distortion of analog audio gear!
CONCLUSION: As Bredo and I have tried to say: stop thinking about bit-depths, sampling rates and audio quality (it’s lot more than good enough). Just make music … well … unless you are a nerd and like to understand the basics behind your lovely tool provided to you by Steinberg.
CONCLUSION: As Bredo and I have tried to say: stop thinking about bit-depths, sampling rates and audio quality (it’s lot more than good enough). Just make music … well … unless you are a nerd and like to understand the basics behind your lovely tool provided to you by Steinberg.
this topic should be rated top secret…or not for clients entrance.
now we can’t blame equipment when our music and sound sucks…
I’ve read a number of posts on this subject over the years and that it’s still a matter of debate says something about the subject in and of itself, seems to me.
I never thought about it since I concluded that the differences - and I don’t doubt they exist - are very subtle indeed and, that being the case, not worth my hard drive space and processing power.
All your answers are here. This guy actually KNOW’s his stuff and very adeptly dispels myths, rumour and misinformation, and all using a very modest, dated consumer-grade e-magic converter!
Just a heads up. There were two videos made on this subject. Pt.1:
thanks for the explanation
