24- or 32-bit integer samples: how to?

[notorand]

While there's very little support to (32 bit) floating point samples, 24- and 32-bit integer samples
are very common.
Moreover, the use of floating point representation leads to approximation with in turn partially kills
the advantage of using 32-bits.
Is there any plan for 24- or 32-bit integer samples or any idea on how to get these sample sizes?

[binandadas]

When You Refer to 24 bit or 32 bit....Its the sample information that is being recorded at this bit rate.....Increasing bit rate increases dynamic range....Ex: For 16 bit, the amount is approx 16*6 = 96dB (Max)..You can't store information beyond that.....It will result in clipping...CD quality is 16 Bit.....So, For Normal recording even 16 bit can be of good choice!!!

[Sti-Jay]

24- and 32-bit integer samples are very common.

Oh, never heard about such samples... Wait, are you meaning float. What, doesn't integer mean that sample would use a logarithmic representation of amplitudes?

[tresf]

@binandadas,

Thanks for the detailed information. Since this is an old thread (4 years old), it's probably more constructive to bring conversation regarding internal sample rates over to our GitHub issue tracker: https://github.com/LMMS/lmms/issues

-Tres

[diiz]

While there's very little support to (32 bit) floating point samples, 24- and 32-bit integer samples
are very common.
Moreover, the use of floating point representation leads to approximation with in turn partially kills
the advantage of using 32-bits.

That's actually completely untrue. Floating point is much more accurate in representing audio due to the exponential dynamic range they provide as opposed to the linear range of integer samples.

Furthermore, if we're talking about recorded samples, there's no advantage whatsoever in using anything above 24-bit. In fact, there's no recording equipment in existence that could actually record even at 24-bit accuracy. Any real circuit produces thermal noise which is louder than the smallest bit in a 24-bit sample. So even recording equipment that is labeled as "24-bit" actually only records at about 20-bit accuracy or so.

Is there any plan for 24- or 32-bit integer samples or any idea on how to get these sample sizes?

You can already load 24-bit or 32-bit integer samples into LMMS, they're just internally converted to 32-bit float as all processing is done in floating point.

And I only just now noticed the OP is 4 years old. Oh well, already wrote the post so might as well post it, maybe it helps someone...