Manaberry
Senior Member
Well, as soon as you use plugins that don't oversample, having headroom to manage the aliasing with saturation for instance is quite useful.
What sample rates and bit depths do you work with?
- Thread starter dman007
- Start date
Well, as soon as you use plugins that don't oversample, having headroom to manage the aliasing with saturation for instance is quite useful.
AllanH
Senior Member
I use 48k/32 bit float. I chose 32 bit float, primarily because it's a more "native" format. My audio interface supports 32 bit floats, and Cubase calculates in 32/64 bit floats. There is (imo) really no reason to up/down scale from 24 bit int. My guess is that memory usage is identical based on alignment.
d.healey
Libre Wave
Are you working with your own audio recordings or with sample libraries? I'm don't think any sample library is using 32bit samples.
Last edited:
Sunny Schramm
Senior Member
48khz, 24 bit
Rossy
Senior Member
Just started yesterday and as this is my very first attempt at scoring to picture, I'm hooked. This is so much fun and I wish there were more opportunities like this, it's great to have no music and only effects, I can clearly hear how my music has impact. It's a great learning experience and I can't wait to download nucleus tomorrow and start implementing it.
AllanH
Senior Member
Primarily sample libraries. My thought process was that since reading from disk is already much slower than accessing memory, I should take any (hopefully one-time) conversion hit during sample loading. All processing by Cubase would then be in 32 bit float. In practice, I could not tell any difference between 24 bit int and 32 bit float. [I'm not sure I know what I'm talking about, but my software engineering brain arrived at this as most-likely being the best]
Last edited:
d.healey
Libre Wave
Wrong thread?
Rossy
Senior Member
So sorry
Geomir
Senior Member
I have a simple question about sample rates and bit depths! I hope it fits into this topic! Let's say I am starting a new song project from scratch, and I am seeing this:
In the right column I can obviously select the "sample rate" and the "resolution" of my song project. I understand most of you would choose the "standard" 48 KHz / 24 bit for high quality sound.
But my question comes here: Let's say that in this particular project all the libraries that I am going to use are 44 KHz / 16 Bit (Hollywood Strings Gold, Hollywood Brass Gold, Symphonic Choirs Gold, VSL Synchronized Special Edition, etc).
Do I have any reason to select 48 KHz / 24 bit instead of my libraries' default 44 KHz / 16 Bit? Will I gain anything in sound quality, since my libraries already hit a lower limit?
In the right column I can obviously select the "sample rate" and the "resolution" of my song project. I understand most of you would choose the "standard" 48 KHz / 24 bit for high quality sound.
But my question comes here: Let's say that in this particular project all the libraries that I am going to use are 44 KHz / 16 Bit (Hollywood Strings Gold, Hollywood Brass Gold, Symphonic Choirs Gold, VSL Synchronized Special Edition, etc).
Do I have any reason to select 48 KHz / 24 bit instead of my libraries' default 44 KHz / 16 Bit? Will I gain anything in sound quality, since my libraries already hit a lower limit?
AllanH
Senior Member
I would say "yes": 16 bits give your DAW a total of 2^16 = 65,536 distinct values to represent -1 to +1 (the raw audio). 24bits gives it 2^24 = 16,777,216. In other words, you gain an additional 8 bits of resolution corresponding to 256 levels between "each of the 16-bit bits".
This allows effects (compression, etc.) to introduce less-noticeable noise. Many inserts do actually up-sample to higher internal resolution to avoid the "rounding" problem. However, your DAW may not, and so I see no reason not to take the higher precision.
Here's a good link from Wikipedia, that explains it in more detail and also discusses how this relates to noise and dynamic range.
Audio Interface: I would want to ensure that my audio interface runs in the same bit-depth as my DAW as that reduces the amount of processing required to hand off the audio from the DAW to the audio interface.
If you need to deliver in 16 bits (e.g. CD), you should (imo) down-convert as the very last step after mastering.
Geomir
Senior Member
Thanks for the detailed explanation and the useful link! I can understand all the math behind it (my first university degree is in Mathematics after all!), and the advice to work my project as high as possible and downgrade only at the last step is really smart!
My main question is if all of this would have any effect since all my sound source will be coming from 24 KHz / 16 Bit music libraries.
Last edited:
AllanH
Senior Member
I'm not sure there is a universal answer to your question. Cubase, for instance calculates in 32 or 64 bit floating point internally, which eliminates the concern about accumulation of rounding errors.
I have chosen settings that keep the number of conversions to a minimum, as the intuitively seems the best. I cannot hear any difference between 24 bit integer and 32 bit float but I do believe I can hear some hissing at low volumes for 16 bit with some effects processing. I'm not sure if you save any memory with 16 bits over 24. That could be, as 16 bits align nicely.
Another mathematician, I see
I have chosen settings that keep the number of conversions to a minimum, as the intuitively seems the best. I cannot hear any difference between 24 bit integer and 32 bit float but I do believe I can hear some hissing at low volumes for 16 bit with some effects processing. I'm not sure if you save any memory with 16 bits over 24. That could be, as 16 bits align nicely.
Another mathematician, I see
Geomir
Senior Member
Wow did you also study Math? You know, music composition is connected with Math (in a way)!I'm not sure there is a universal answer to your question. Cubase, for instance calculates in 32 or 64 bit floating point internally, which eliminates the concern about accumulation of rounding errors.
I have chosen settings that keep the number of conversions to a minimum, as the intuitively seems the best. I cannot hear any difference between 24 bit integer and 32 bit float but I do believe I can hear some hissing at low volumes for 16 bit with some effects processing. I'm not sure if you save any memory with 16 bits over 24. That could be, as 16 bits align nicely.
Another mathematician, I see
OK so just to feel "safer", from now on all my music projects will be 48 KHz / 24 bit, just to make it sure that I get more detail and less noise! After all not all my libraries are 16 Bit!
