24bit 96khz - Will Things Sound Better With That?

I now have ability to use 24bit bit depth and 96khz sample rate. It’s a bit too much to use it for playing, since it eats way more CPU, but do i use it for rendering songs to disk?
Will it really sound better?

Also, after it’s done and i mastered the song in such sample rate and resolution, how do i then alias, dither and what parameters do i set for that?

Thanks for answers in advance.

actually, if your song has not been conceived from the beginning as a 24/96 song, it may sound worse.

for example, not every VST/AU plugin has been designed for such quality, so some of them could output no sound when used at 24/96, or sound horribly.

also, if you are using 44khz samples, an upsampling is needed when rendering. this process would be transparent if you are using 48khz sampples, but may create artifacts when using 44khz samples.

in general, however, rendering at 24/96 should give you better results especially when you are using lots of tracks, since the higher bit resolution and sampling frequency will result into better mixing.

A point worth noting is that 32bit and 32bit floating point are available nowadays.

The bit depth is the ‘dynamic range’ a recording is capable of reproducing. Musical of a ‘classical’ style can benefit from a high dynamic range.
The frequency of the recording affects the highest frequency that can be reproduced (just less than half the sample rate is the limit.) At 96khz although you cannot hear all the frequencies, they do affect the recording - as said by it-alien, this could be good or bad.
There is little point in mastering to a higher frequency that what you are playing/writing with.

Compressed pop music often doesn’t warrant the 16 bits offered by a cd.
Messing between mathematically incompatible sample rates is something i particularly avoid.
Im reasonably certain most peoples playback equipment arent of the quality to benefit from 24bit/96k.

For archiving on my computer i use 24bits at 44.1khz.

Foo?'s method:

  • Record all sources at 96khz 32bitfloat.
  • If a used sample or synth can’t up convert to these rates then ditch it and get something better.
  • Compose and mix at the same rates.
  • Render mixdown at the same rates.
  • Build digital mastering chain using the same rates to work on mixdown.
  • Using a real time dither output the mastering at 24bit to match hardware performance (this makes an improvement).
  • Output to tape (in my case Tascam BR-20T with Ampex 456 tape calibrated).
  • Record tape output back into the computer at 44.1khz 16bit. Minor edits. Done.

Your converter quality is very very very important. The difference between good and bad ones can be quite remarkable. Latest generation stuff is the way to go, they’re still evolving the quality of digital and in my opinion still have a long way to go.

Working at 96kHz (or higher) will give you even shorter latency (when using ASIO). Of course system load will be higher because of 24/96 :D

I’ve noticed a couple of funny things when going up in sample rate. Sometimes it just sound worse, when I’ve tried it’s sounds more harsch in the highs, could be my lousy card (AP24/96) but I’m going for a fireface soon.
Another thing is that some synths speed up, e.g. like you do with a sample and then press a key an octave higher.
Probably what Foo refers to in the second line.

Question, how do you select the bit depth ?
As far as I know I haven’t seen a selection anywhere, to use 16 or 24.

afaik playback is 32 bit with linear interpolation.

It is. If you’re dithering down to another bit-rate for a rendered file use a program other than Renoise that allows you some ‘shaping options’. Dithering in real-time is only needed if you are outputting the playback to an external device (like a tape deck).

If sampling up to 96khz sounds worse do inspect your source material. If you’re not ditching stuff then you’ll most certainly have to review your EQs and Filters because they’ll sound different at the higher rate.

You can still get a good sound at 44.1khz - we did the last Hunz album mix at that rate and it turned out really well. But if you’re after that delicate warm characteristic that you get with analog audio then best to try and up your sample rate as much as you can.

Also, one more question:
i see the 32 bit and 32 bit float options in the render song menu, which one do i choose and what are the differences?

My friend once conducted a small experiment, to see if there really is a difference between 44100 and 96000. The participants were shown pairs of short samples from recording and then choose the better sounding one from the pair. I won’t get into detail about the methodology of the experiment. Some important facts are:

  • the records were all originally made in 96000. The 44100 examples were downsampled from the original
  • the records were made on a MotU 828 mkII interface. The same one was then used for playback during the experiment
  • the participants listened to the examples using rather decent Beyerdynamic headphones (don’t remember the exact model)

There were 11 participants, all trained sound producers (students of sound production departement on Frederick Chopin Music Academy in Warsaw), with me being one of them. According to the results, there was 53% of chance that a participant will choose a 96000 as the better one. This means there was no correlation between the sampling frequency of playback and sound quality as percieved by the participants. The highest score (mine :) ) was a bit above 70%, which is still statistically weak.

Of course the experiment had several weak points. While the Beyerdynamic headphones had a huge headroom in terms of frequency reponse, it’s unsure what are the true parameters of MotU’s phone-outs. Might be, that all signals are downsampled or processed in some other way for the headphone outputs. Also, the recordings were all originally made in 96000kHz. It’s possible that the difference woul be more significant if the 44100 samples were not downsampled, but made originally in CD-quality.

It’s also important to note that the samples were from classical recordings, made using few microphones, mixed by means of analogue Studer console. This means the experiment did not test the potential benefits of using higher sampling frequencies for mixing digitally.

The experiment did not test the differences between resolutions. All samples were played back in 24-bit.

Although the experiment was nothing big and nothing definitive, you may still take a lesson from it. Whatever happens, don’t expect that changing from 44100 to 96000 or even 192000 kHz will make your jaw drop, your brain blow up and give you a boner. We’re talking about slight details here. Details most people might never know about. However. If your technology ever allows you to use higher sampling frequencies and/or resolutions, there’s no reason not to do that. HDD space is cheap nowadays. Even if you’re going to downsample it to 44100, 16-bit in the end, your records might benefit from being processed at higher parameters during the production. It has been a long tradition that the parameters of studio recorders is higher than that of consumer equipement. Back in the analogue days studios used awesome tape recorders whose recording parameters were definitely better than the abilities of vinyl or, hehe, tape cassette, but that didn’t mean there was no sense in using all the power for production, before eventually crippling the material to make the end product. Home-recording were still sounding worse back then, and they still are today, if you listen closely.

That’s my opinion at least.