Stupid question about dithering

How to do this, that and the other. Share, learn, teach. How did X do that? How can I sound like Y?
Post Reply New Topic
RELATED
PRODUCTS

Post

Ok so basically this is what I want to know.

Say I decide to run a project @ 192khz, 32bit quality with a bunch of synthesizers for a sound design project. I would use this sample rate to get the best quality i can and then dither each of the tracks down to 48khz 24bit.

Now my question is can I load all of those dithered files into a 48khz 24bit project and then dither it again down to 44.1khz and 16bits when i am ready to export the final track?

For some reason I'm thinking it will probably not work so good and it might smudge the sound by dithering twice.
:borg:

Post

You shouldn't dither from 32bit to 24bit. There's no need for, makes no sense because the 16bit dither is much louder. Only dither one time when you going 16bit as last step.
║▌║█║▌│║▌║▌█

Post

Dithering isn't sample rate conversion.
Dithering/Noise Shaping is only for bit rate reduction.
[ Mix Challenge ] | [ Studio Page / Twitter ] | [ KVRmarks (see: metering tools) ]

Post

Also, you'll gain nothing by recording at 192/32, though you'll put quite a bit of "stress" on your system, depending upon how many tracks or VSTs you use. Use "oversampling" if your VSTs support it, but recording audio or rendering all your tracks at 192 kHz wastes a LOT of CPU cycles in processing all that data.

Valerian and Compyfox were both on the money in their responses. You might want to go back to what your DAW's manual (or manufacturer's Website, or Wikipedia, or something) says about dithering and study it again. Dithering should only be used at the last stage, and really only when you're going down to 16 bits or lower.

This wasn't a "stupid question" though--many people are confused about dithering, noise shaping and many factors involved in digital recording. You may have helped someone else by asking this!

Steve
Here's some of my stuff: https://soundcloud.com/shadowsoflife. If you hear something you like, I'm looking for collaborators.

Post

Thanks for the responses everyone! I still have so much to learn with mixing and mastering my own stuff :oops:
:borg:

Post

V0RT3X wrote:Thanks for the responses everyone! I still have so much to learn with mixing and mastering my own stuff :oops:
We all do! It's an ongoing education! I learn something new about my DAW or VSTs each week.

Steve
Here's some of my stuff: https://soundcloud.com/shadowsoflife. If you hear something you like, I'm looking for collaborators.

Post

planetearth wrote:Dithering should only be used at the last stage, and really only when you're going down to 16 bits or lower.
Actually, dithering is also important if you want to go from 32bit float to 24bit integer. If you don't dither, you introduce all kinds of things - most notably "bit truncation noise/distortion". Which is not a good thing.



I can't find the video comparison anymore. Was it on the YT RealHomeRecording blog or not. All I know it was a two parter that clearly showed the up's/down's of using higher sampling rates, or better said "oversampling". 192kHz might make sense for both filters and compressors (which distort the signal), they might also work for synths. Especially with content that is hovering around the 22kHz range. But working at a higher sampling rate constantly only makes sense for a couple of applications.

i.e. Mastering.
Or... if you create HD audio content in surround. But then you have the suitable rig.


Here is a video I found that clearly shows you the difference between recordings at either plain 48kHz or 96kHz.
http://youtu.be/j9_yPDzdt60

The 96kHz simply records "more", definitely no doubt about that. You also feel "more" if you listen to the content. But if you don't produce HD audio content, then recording at higher SRC definitely gives you a tad more "resolution" in the upper frequency range, while SRC'd down to 48kHz gives you more room to work with.


Rule of thumb:
Working in the HD Audio realm: do whatever the funk you want

If the end product is either CD (Redbook), MP3 or DVD-Audio:
- you can record at 96kHz and then SRC (downsample) to 48kHz or 44kHz


192kHz is really excessive and barely any Blu-Ray audio release offers that setting currently. And definitely not for surround either. So it's "still" save stick with lower sampling rates.


One thing that you shouldn't do however:
Don't record lower than 24bit.

And if your ADC is "only" 24bit (which determines the dynamic range and the noise floor), then 32bit is a waste of resources if you'll never ever going to clip your signals. I'd only use 32bit during mastering and if you're not sure if your signal will ever clip during editing.

Else, "internal" 32/64bit is taking care of that prior to reaching the summing bus.


YMMV.
[ Mix Challenge ] | [ Studio Page / Twitter ] | [ KVRmarks (see: metering tools) ]

Post

It's so easy to get sucked into the marketing about numbers and we all have to resist the urge to think more=better. I do it too, you're not alone. :)

Short answer? 48kHz/24-bit on modern audio interfaces is pretty awesome. I have a couple of links here for you should you want to learn more why:
http://people.xiph.org/~xiphmont/demo/neil-young.html
http://xiph.org/video/vid1.shtml
http://xiph.org/video/vid2.shtml
http://wiki.xiph.org/Videos/Digital_Show_and_Tell
http://www.soundonsound.com/sos/jun98/a ... ital2.html
http://www.youtube.com/watch?v=BYTlN6wjcvQ

However, that's not to say that you shouldn't enable oversampling on digital plugns: that's a different matter. Trust your ears when setting the oversampling. If you can't hear a difference leave it on 2x instead of 4. It'll save that CPU for you later when you add more tracks or FX. :)

Regarding dithering: you should only dither once when exporting to a different sample/bit-rate. Dithering actually adds noise to mask audible problems when changing sample rates. See the xiph vid2 above for some hands-on example of what I mean.

Good luck!

Post

Compyfox wrote:Actually, dithering is also important if you want to go from 32bit float to 24bit integer. If you don't dither, you introduce all kinds of things - most notably "bit truncation noise/distortion".
Yeah, that truncating noise will be sitting at -144dB level. Which is about 40 dB below the noise floor of any typical DA converter. So that's not really relevant I'd say.

32bits float is composed of 24bits mantissa & 8 bits exponent. By converting to 24bit integer you don't lose anything. Maybe one single bit, but that can't be significant in real life.
We are the KVR collective. Resistance is futile. You will be assimilated. Image
My MusicCalc is served over https!!

Post

Don't forget to check that your synths and plugins support what ever samplerate you choose.

It was only a year ago that all Waves plugins in the Gold bundle went to support 96k.
Not all support 192k yet as I remember.

But the thinking is good - to render a midi track with a synth in as high rate as supported.
I heard some mention they create less artifacts doing that.
Then move the audio track to your project and work in a rate that more relevant for your system - not walking on its knees.

Post Reply

Return to “Production Techniques”