What kind of dithering/noise shaping do you use?

When dithering down from 32 or 24 bit to 16 bit digital audio files, is there a consensus about which type of dithering and noise shaping is best or most accurate?

In Sound Forge there are five dithering options: rectangular (1 bit peak to peak), rectangular (2 bits peak to peak), triangular (2 bits peak to peak), high pass triangular and Gaussian (2 bits RMS to RMS).

For noise shaping in SF you can do either high pass contour or equal loudness contour.

I'm curious about the effect of each and if anyone has an opinion on what "sounds" better. I've been using the Gaussian dithering and high pass contour noise shaping so far and haven't had any problems.

Jason

 

bump

 

I typically prefer the POWr2 dither in SONAR to Triangular/Equal Loudness in SoundForge, but they both sound good (for 32-bit float to 16 bit).

 

Whatever option best fits the piece of music at hand. Although I oftentimes tend to gravitate towards Apogee's UV22HR.

 

Surface noise

 

Well, Jason, I have found Gaussian to be the worst dither for music. I rarely ever use SF for dithering, since I don't feel it offers enough customization for my needs.

I use Adobe Audition for my dithering. These days, I typically use the Shaped Triangular p.d.f. with a dither depth of around anywhere from .5 to .8 bits. I find it sounds smoother and cleaner than the Triangular p.d.f.. In some cases, I will use a dither depth of the default of 1 bit if I don't use any noise shaping at all. In AA, I favor the Noise Shaping 44.1KHz curve. If I am not too concerned with the bass, if there is an abundance of bass, or if I get an increase in noise, I use the C2 noise shape.

I experiment with not only by listening, but also by using the frequency analyzer to determine where that dither noise is going in relation to what I hear, and the effect it has on the music. With using noise shaping, I can get away with using lower amounts of dither, but not much lower than about .5 bits. On the other extreme, 1 bit us usually more than enough. Once, I accidentally used 7 bits of dither by accidant! (I forgot to add the decimal point). There was plenty noise, but the music didn't sound bad. I kept my work. I had no choice, since I didn't back up the 32-bit versions. That taught me to always make sure I set things how I want them before I dither without backups.

If I bump up a 16-bit file to 32-bit for processing, and that files has already been dithered once, I do not use dither again. It adds more noise and smears the sound a bit. In those cases, it's better to not use dither. But, it can't hurt too much, either. If I process a 16-bit file, it's going to get a dose of dither anyway at some point.

I set Audition to dither at .8 bits for processing in preferences.

If I do happen to use SF for some reason, I use the Triangular (2-bits Peak-to-peak) with a high-pass contour.

See, i'm all about tweaking to preserve the most and best sound of that 32-bit file for that 16-bit file.

 

Grant, thanks for that in-depth explanation, that's just what I was looking for. I had never used dither before but had always seen it in the drop down menus and thought I'd try experimenting with it. I guess my next move is to do some in-depth reading on the subject.

Jason

 

Bad news for you: there isn't much written on the subjuyct because there isn't much to tell. It's all about how something sounds. Bob Katz did write quite a bit on the subject. If it isn't on his website, it's in his book "The Art of Mastering".

If you go down from a high bit-depth, you should use dither. It is white noise added to the digital word to cancel out the effects of trunication. For the exchange of added white noise, you avoid digital distortion especially audible on fadeouts. The sound without dither is a grainy, harsh sound. It may not be that noticable on loud portions, but it no doubt affects the sound.

 

That has been my experience too but my use of dither is mostly to improve my RedBook playback of older recordings: [16 bit word length] -> [24 bit word length -> ambience recovery -> POW-R 3 dither algorithm] -> [DAC] with very good results.

 

There's a bunch of articles on mastering written by Bob Katz on his fine Digital Domain website, including this article on dither.

 

Regarding my mention of using Apogee dithering on post #4, here's what Katz says about it on the link that has been cited here:

"Apogee Electronics produced the UV-22 system, in response to complaints about the sound of earlier noise-shaping systems, declaring that 16-bit performance is just fine. They do not use the word "dither" (because their noise is periodic, they prefer to call it a "signal"), but it smells like dither to me. Instead of noise-shaping, UV-22 adds a carefully calculated noise at around 22 KHz, without altering the noise in the midband."

 

Thanks guys, that gives me something to start with, in addition to my ears.

In reference to this, on some DVDs, I've listened on headphones to some audio and noticed that there is a rather "garbled" sound in silent sections. Is this a result of dithering from a higher bit-rate, or am I hearing something else?

Jason

 

I didn't say there wasn't anything, just that there's not much. Bob Katz did well, but he is of the very few who have bothered to wrote anything. I have known about Bob Katz' articles for years.

But, there is no real way to know about dither without actually using it.

 

It could be anything. Without knowing what the sound went through, there's NO way of knowing why you are hearing what you hear.

 

What Grant said above.

 

Could be the audio compression used.

 

Hey Jason

Thread revival. Just wondering what dithering or noise shaping settings you have settled on. I've never used it when downsampling from 24 to 16, but thought I would at least do some testing.

You wouldn't happen to be 'anymajordude' over at Ex___ would you?

 

Well, since i've written that last post, i've settled on a good-sounding scheme for going from 96/32 down to redbook: adjust the sample rate quality to 350, and use a dither dpth of .8.

 

The newer Apogee units are very nice for dithering.

 

Is this for your own music or for comps that you're making? I've been doing my comps to DVD specifically to avoid dithering at all. I just leave everything at 24 bit. Of course you need some DVD authoring software.

best,
Jon

 

This is for CDs I am doing for others. For myself, I use a squeezebox for all my music, so I just keep the orginal 24 bit files on my server to use with that.

 

Yeah, but that's only for if you have the $$ to spend for hardware. Most of us are working in the box, and you can get some mighty fine results with a DAW these days. You just have to be willing to tweak until you get what you want.

 

Audio on DVD at high 24/96, or whatever is good, but you can't take it with you or make stuff for everyone. A plan for taking things down to redbook remains an option and a necessity for many of us.

 

As I mentioned before WaveLabd has two Apogee dithering plug-ins (or two options). Namely, the UV22 and UV22HR.

 

Are these plugins included with WaveLab? I think I have a copy of WaveLab 5 somewhere, maybe I will check it out.

Is the consensus then that it is better to use dithering/noise shaping than not use it when downsampling from 24/48 to 16/44?