How much advantage is there to 24 bit samples on a sampler like Kontakt or Halion over the 16 bit ones that Gigastudio uses?

It seems to me that streaming 24 bit samples with heavy polyphony, release samples etc, is going to be 50% more overhead on your CPU and hard disks with little to no audible difference in a mix. A number of sample content manufacturers are doing the 24 bit format though.

Has anyone done a comparison between a 16 bit and 24 bit version of the same library? Is there an audible difference? Also has anyone compared the overhead between 16 and 24 bit versions of the same sample library?

Laurence Kingston

Various experts seem to think that the 24-bit versions do sound better. I have a couple of very high-quality sample libraries where I have both the 16-bit and 24-bit versions, such as the newly released SampleTekk White Grand piano. To be honest, I have a bit of trouble hearing the difference -- for the libraries that I have compared, both versions sound off-the-charts good.

The 24-bit versions take up more disk space, of course. I don't believe that there is a significant -- if any -- CPU runtime performance difference since the integer values get converted to 32-bit floating point in both cases.

How is the Sample Tekk piano? I see that it's available in both 16 and 24 bit formats for Kontakt. The demos sound fantanstic, but then again demos usually do. How is it to play?

Laurence Kingston

I think it plays great; but, then, I'm no expert.

Like all the ultra-high-quality instruments, it takes a while to load. But it sounds fantastic!

i think the conversion to 32 bit is easier from 16 cause it just multiplies by 2, instead of by like 1.5 or something.

For a long time the people at Tascam refused to consider 24 bit samples because of the increased strain in resources, both CPU processing and band width for disk streaming. GS 3 will use 24 bit samples so I guess they are past that restriction now. The higher the processing rate the less you notice digital degradation, but even those effects in GS run at 24 bit rather than 16 bit. I would rather have 16 bit samples processed through a 24 bit reverb and D/A than have 24 bit samples processed through 16 bit reverb and D/A.

Robert

Silent223 wrote:

i think the conversion to 32 bit is easier from 16 cause it just multiplies by 2, instead of by like 1.5 or something.

ROTFL

Put it a different way. Take a 24-bit sample set. Slap a 16-bit truncating dither (use a good quality one) straight in front of it. Can you tell?

I'm guessing you can't, but then I'm lousy at hearing dither in the first place.

What frustrates me is that I seriously doubt that either I or anyone listening to my music will be able to tell any difference between whether I use 16 or 24 bit samples. At the same time my harddrive will be 50% more burdened at 24 bits than it would at 16, so what I'll be essentially getting is 50% less hard drive performance for negligable sound quality improvement.

SoundTrek's "White Piano" is available in 16 and 24 bit versions for Kontakt, but for most libraries you can either use the native 24 bit Kontakt version or import the 16 bit Gigasampler version. It seems like "24 bit" is a good salespitch, but I seriously doubt there is enough difference in the sound quality to justify the extra space and overhead it takes up.

I remember reading a topic on this somewhere once. The conclusion was that 24bit was most definately better in terms of sound quality and not only that but accuracy. There were also examples posted...

I can't remember where I read this, all I know is that since then I have done everything at 24bit, so it must be true.

That wasn't much help, was it.

Anyway, all that aside, if you are doing multiple editing to a soundfile, then it is always best to do it at 24bit so that the computer can process the data at the best possible way for itself. Probably.

Quote:

SoundTrek's "White Piano" is available in 16 and 24 bit versions for Kontakt, but for most libraries you can either use the native 24 bit Kontakt version or import the 16 bit Gigasampler version. It seems like "24 bit" is a good salespitch, but I seriously doubt there is enough difference in the sound quality to justify the extra space and overhead it takes up.

I tend to agree. Having said that, every time I have a choice between 16-bits and 24-bits for the same library, I end up going for the 24-bits (or sometimes both). I guess the "salespitch" worked on me!

I'm a 16 bit man for now, until 24 bits sell me over in real world usage.

Recently I bought a 24 bit drum loop sample CD. I listened to the loops through my 24 bit soundcard, and it sounded fine, but not good enough to knock my socks off, so I did some research.

I loaded some files into Sound Forge. Sure enough they were 24 bit.

As a test, I made a copy of some loops, chopped them to 16 bits (no dither), inverted the waveform, saved and reloaded the waveform (to clear out SF's 32 bit internal precision buffer), upped the bit resolution back to 24 bits, then pasted that new waveform on top of the 24 bit version.

In 1/2 the cases, the resulting file was digital silence. In other words, the source loop was 16 bits, and was simply upped to 24 bits.

In the other 1/2, I normalized the least significant 8 bits and listened to them. In all cases, it sounded like unwanted noise, stuff I'd want to clean out anyway!

So, I batch converted all the files to 16 bits and saved myself 50% hard disk storage.

I'm not saying 24 bits isn't better, but beware the hype - don't do it just because everyone else is. In some cases, you're being sold on a buzzword.

-Brian

Isn't the usual argument in favor of working in 24 bit that it preserves the sound quality through various manipulations? That is, 16 bits is probably fine for final rendering but feeding audio through several 16 bit processing stages will degrade the audio quality more than it will at 24 bit because the accumulated errors are greater?

Don't people doing image processing usually work at higher than final resolution for the same reason?

Quote:

So, I batch converted all the files to 16 bits and saved myself 50% hard disk storage.

What tool do you use to do a batch conversion from 24-bit to 16-bit, btw?

Quote:

Isn't the usual argument in favor of working in 24 bit that it preserves the sound quality through various manipulations? That is, 16 bits is probably fine for final rendering but feeding audio through several 16 bit processing stages will degrade the audio quality more than it will at 24 bit because the accumulated errors are greater?

Yep, I'd second that. 24bits gives you more headroom when fiddling with data. Try not to think of it in an audio sense, and in a computer sense. That lovely piece of guitar (or whatever!) you just recorded into your computer has now been converted to lots of 1's and 0's. The more bits you have, the more 1's and 0's you have that can carry data so the more accurate the soundfile is.

QED. Probably.

Of course 24 bits samples are more 'Accurate' than 16 bit ones.
But then again, beautiful music was made with the early 8-bit samplers like the Fairlight CMI and the Ensoniq Mirage. (not mentioning the AKAI's and Rolands on purpose here )

Being a purist with rather good ears I prefer 24 bits over 16 bits any time, having said that, I do prefer a great song with smaller bit-depth over a technically perfect 24-bit song that just doesn;t grab me....

my 2 cents

In the context of a final mix, I can understand going 24 bits and pretty much always render my mixes at 24 bits. That way I can EQ and compress away without accumulating multiple errors.

24 bits for an individual sample on a piano however is another situation entirely. If you were listening to just one note played softly with the volume cranked up past the point of headroom for the louder parts you might be able to tell a difference.

That is nowhere near a real world scenario though. Play two notes with 24 bit samples and the overall resolution is already well beyond 24 bits. Play a polyphonic piano part with 24 bit samples and you're getting closer to the 32 bit resolution mark. Add this piano to a mix with a bunch of other 24 bit samples and the 32 bit resolution of the DAW is going to be truncating extra bits all over the place. Theoretically it seems pointless, and yet that is what everyone is doing these days.

Having said that, I must say that I will probably opt for the highest resolution any time I buy something myself. I just ordered the Sonic Implants Orchestral Strings library in the 24 bit Kontakt format. This particular library uses release samples with room ambience and these might show a little difference at the tail end of soft passages. This kind of subtlty would be lost in a CD mix though.

Laurence Kingston