Dither

[erhodes]

Is dither required or advisable when you are digitizing things like lp's or analogue tape?  I have software with a variety of dither algorithms but the instructions seem to be referring to recordings of live music.

If dither is to be used to dub analogue recordings, does anyone have advise on which algorithms would be best for that purpose?

[mjzee]

I'm intrigued by this topic, because I've never heard of "dithering."  I've digitized my fair share of LPs and tapes using AmadeusPro.  I've just searched the Amadeus user guide, and found no results for "dither."  What is it, exactly, and how would you do it?  I've found some web pages (here and here), but it doesn't help.

[erhodes]

Apparently it is various forms of white and pink noise that your software actually ads to the recording.  For a number of very obscure and highly technical reasons it's supposed to improve the dynamic range of digital recordings and, according to some claims, improve the overall sound.

I use Wave Lab and it has a number of dithering alogorithms but the instructions are sketchy.  If you are digitally recording a live band dither seems to be obligatory but I'm less clear about using it to digitize, say, my lp's.  Have been doing that for a long time, never used the dither function.  Just wanted to know if anyone else has experience with it or understands when it should be used.

[Scott Dolan]

  On 10/13/2016 at 4:06 PM, erhodes said:

Apparently it is various forms of white and pink noise that your software actually ads to the recording.  For a number of very obscure and highly technical reasons it's supposed to improve the dynamic range of digital recordings and, according to some claims, improve the overall sound.

 

Expand  

Perhaps if Kevin Bresnahan sees this he can weigh in with the facts, but as far as I know, digital has far more dynamic range than analog to begin with. So I have no idea why "improving" the dynamic range of the digital copy would be necessary or beneficial whatsoever. 

[JSngry]

mjzee included this link, but let me highlight it. The article and, especially, the comments go into more than enough detail for me: http://www.earlevel.com/main/1996/10/20/what-is-dither/

As I understand it, the point is not to increase the dynamic range, but to "blur" the quieter ranges to create just enough "fuzziness" to make them more easily processed by the ear. The auditory/visual equivalencies might not be exact, but the basic idea seems right to me. In both worlds, "fuzziness" is more difficult to ignore than "transparency".

I would like to hear Chuck's or Jim A's thoughts on the matter, because it (dither) seems to be something more relevant to higher end source recordings of a wide dynamic range than it is to any duplication of existing recordings.

 

[Scott Dolan]

Ok, that makes sense. I'd love to hear an A->B test so I could understand the difference better. 

My only thought is that if you're starting with an analog source, wouldn't the harmonic distortion already be there? 

[JSngry]

"Analog source" = the instruments, or the recording?

When I referenced ""source recording", I was referring to the in-studio performance. Perhaps dither is used during the actual recording? It would then carry on through subsequent stages of production/post-production, I think.

Which is to say, I don't know that it would be at all relevant to copying extant finished recordings, as the OP mentioned. Not unless you wanted to play games, and then that's something else altogether.

[mjzee]

I can understand a visual analogy - my guess is that it's similar to font smoothing.  A font can look jagged around curves; smoothing fills in the gaps and makes for a more pleasing appearance, although it's introducing material that's not in the original font.

[xybert]

I've done some audio engineering (although it's been a while and i was never an expert) and i don't recall dither being a thing when recording with Pro Tools etc. It was always a thing where you've already done your recording to Pro Tools in 24bit or whatever, so say you've recorded and mixed 30 tracks (vocals, drums, other instruments, overdubs etc) and you're bouncing your final mix down to a 16bit stereo wav file. If you don't add dither then you can get glitches/artifacts as a result.

I'm not familiar with modern methods of digitising LPs but i presume it'll effectively be the same as recording an acoustic instrument to digital. That's what i used to do: do a needle drop, put a mic in front of the speaker, set your level and hit record in Pro Tools. No dithering during the initial recording, but then if i'd decided to record it at a higher bit depth then i'd use dithering when bouncing it down to 16 bit.

In what step of the process is the program you're using asking you to select dither?

 

 

 

[Scott Dolan]

  On 10/13/2016 at 5:56 PM, JSngry said:

"Analog source" = the instruments, or the recording?

When I referenced ""source recording", I was referring to the in-studio performance. Perhaps dither is used during the actual recording? It would then carry on through subsequent stages of production/post-production, I think.

Which is to say, I don't know that it would be at all relevant to copying extant finished recordings, as the OP mentioned. Not unless you wanted to play games, and then that's something else altogether.

Expand  

Right. And that's what I was getting at. If the original source is an already finished analog recording (which is what we're talking about), then I wouldn't think dithering would be necessary. 

Now, if this is a recording in a studio that hasn't been mastered yet, OK. 

  On 10/13/2016 at 6:15 PM, mjzee said:

I can understand a visual analogy - my guess is that it's similar to font smoothing.  A font can look jagged around curves; smoothing fills in the gaps and makes for a more pleasing appearance, although it's introducing material that's not in the original font.

Expand  

That's an excellent analogy. But again, we're talking about a source that's already mastered and pressed to whatever medium. Right? And if that's the case, the "fuzziness" (harmonic distortion) would already be in the source material. 

  On 10/13/2016 at 6:42 PM, xybert said:

I've done some audio engineering (although it's been a while and i was never an expert) and i don't recall dither being a thing when recording with Pro Tools etc. It was always a thing where you've already done your recording to Pro Tools in 24bit or whatever, so say you've recorded and mixed 30 tracks (vocals, drums, other instruments, overdubs etc) and you're bouncing your final mix down to a 16bit stereo wav file. If you don't add dither then you can get glitches/artifacts as a result.

I'm not familiar with modern methods of digitising LPs but i presume it'll effectively be the same as recording an acoustic instrument to digital. That's what i used to do: do a needle drop, put a mic in front of the speaker, set your level and hit record in Pro Tools. No dithering during the initial recording, but then if i'd decided to record it at a higher bit depth then i'd use dithering when bouncing it down to 16 bit.

In what step of the process is the program you're using asking you to select dither?

 

 

 

Expand  

Can you expand on that, xybert? Why would the bitrate change how the source material sounded? 

This is actually a very fascinating subject. 

Edited October 13, 2016 by Scott Dolan

[xybert]

  On 10/13/2016 at 6:44 PM, Scott Dolan said:

Can you expand on that, xybert? Why would the bitrate changed how the source material sounded? 

Expand  

I don't really fully understand the reason why, but whenever you're dealing with digital there's room for digital related glitches and i always just thought of it (adding dither) as a preventative thing. Like, you can choose not to add dither and get away with it, but you may as well use it (a bit like selecting the 'error correction' option when ripping a CD in iTunes).

I can't explain it properly, but it's like, you've got a digital recording of the sound at 24 bit, and your're then reducing it to 16 bit which reduces the 'quality' of... it's like if you've got a fence with 24 planks of wood and you're reducing it to 16 planks  but you still want it to look the same as when it was 24 (i know this makes no sense but imagine that it could be done)... you might do a great job of it but there's probably going to be a nail sticking out somewhere.

[JSngry]

  On 10/13/2016 at 6:15 PM, mjzee said:

I can understand a visual analogy - my guess is that it's similar to font smoothing.  A font can look jagged around curves; smoothing fills in the gaps and makes for a more pleasing appearance, although it's introducing material that's not in the original font.

Expand  

Yeah, there are some graphics programs where the word "dither" or "dithering" is actually used somewhere in the menu/settings. I don't do a lot of that stuff, but the word sounds familiar in that context. And yes, distorting the original for the purpose of ultimately clarifying it.

[xybert]

Here's an extreme but good example:

 

[Ted O'Reilly]

Sounds like corruption to me.  Is this a product of Dithers, Inc., Dagwood Bumstead sales rep?

[erhodes]

Thanks to those that responded and especially to mjzee for the links.  I found a few links of my own which are highly technical and which don't resolve the issue.  The best I can get from these sources and this Organissimo discussion is that dither is used to record live performances and when reducing a file's bit count, e.g., from 20 or 24 to 16 bits.  I can find no instance where dither is specifically recommended for digital dubbing of an analogue source and that appears to be consistent with the consensus in this thread.

Again, many thanks to all.

[Jim Alfredson]

The reason to use dither is to mask certain harmonics that may appear due to digital processing. Essentially what you are doing is raising the noise floor. The noise floor is the point at which the signal becomes indistinguishable from the background noise. The reason to do this is to mask the artifacts from converting a higher bit depth to a lower bit depth, like from 24bit to 16bit CD. Such conversions cause quantization errors which appear as harmonics. The dither masks those.

So if you're digitizing analog sources into your computer, you don't need dither. 

A good and easily understandable guide is here: http://downloads.izotope.com/guides/izotope-dithering-with-ozone.pdf