Dithering opinions needed

[zip]

Hey! Long time no post... (been swamped at school)

I'm pretty up to date on dithering (dither once after all digital manipulations) but I'm doing something different here and wanting to check my assumption.

I'm transferring cassette media to a digital format using a Motu and preparing to pre-master using Logic Platinum.

I want to save the original .wav files and burn them to a disc...so if the HD crashes I have a backup. Here's the assumption...

I should not dither the transferred unaltered material when bouncing in Logic and ONLY dither the final pre-mastered file.

In other words...when I bounce the file to .wav I should NOT dither...in case I need to import the material from the CDR back to the hard drive - where after manipulation it will be dithered ONCE.

Is this correct??

Thanks! zip >>

 

[M.Brane]

The only time you need to dither is when you go to a lower bit/sample rate. If you just want to back up your files just save 'em as data i.e. files/folders. No need to burn an audio CD to save audio files for backup.

 

[zip]

Hmmmmm...

According to what I have researched you also need to dither whenever performing manipulations to the audio...EQ, compression etc...as a last step so the bits are not truncated after the algorhythm is performed.

The only reason I am saving this to CDR is to protect against a hard drive crash. Once I transfer this rare bootleg I have to give the tapes back and won't have access to them again. It's a brand new HD so call me over-protective...

Sounds like I'm on the right track however.

Thanks,

zip >>

 

[MONTE]

What M.Brane is saying is that if you are just backing up the files, back them up as .wav files and make a data disk. You do not need to dither in this case. Also...even if you were making an audio disk you only need to worry about dithering if you are (as M.Brane stated) lowering the bit rate or doing a sample conversion. There are several threads on dithering here if you'd like to do a search.

 

[pglewis]

The only time you need to dither is when you go from a higher bit-rate to a lower one.

If you're recording at 24-bits, then keep everything at 24-bits up to the point that you convert to 16-bits in preparation for burning a redbook audio CD. To archive them, save everything in the unaltered 24-bit format as data files and not as an audio CD.

If you're at 16-bits, you shouldn't have any dithering to do.

According to what I have researched you also need to dither whenever performing manipulations to the audio...EQ, compression etc...as a last step so the bits are not truncated after the algorhythm is performed.

With DSP, any necessary dithering should be done for you. In the case of 16-bit audio, the DSP will bump up the bit rate for its calcuations and then dither back down when it's done. The audio is only at a higher bitrate temporarily, while being used internally by the DSP, so there's nothing needed on your part.

I think it's mostly academic, given the application. But no reason not to do your best with the audio anyway.

 

[zip]

Interesting...

I'm recalling a post from Sonusman stating dithering should also be used with other manipulations. (other than conversion from 24 to 16 bit)

I was unaware the DSP algorhythms do not truncate the bit depth.

Thanks guys...

zip >>

 

[littledog]

Interesting...

I'm recalling a post from Sonusman stating dithering should also be used with other manipulations. (other than conversion from 24 to 16 bit)

Unless you are able to specifically reference what Sonusman wrote, chances are you may be remembering something out of context. Rounding errors can accrue when performing manipulations at any bit depth, but they are much less noticeable at higher bit depths. That is why many plug-ins perform their internal calculations at 32 bits or 48 bits, and only round down to 24 (or 16) when they are done. Perhaps this is what you are referring to.

But dithering is a seperate issue. The information everyone else has provided you on this thread is correct. Dither only when going from 24 (or 20) to 16 bit, and only as the very last step in the mastering process.

 

[sonusman]

Actually, he read it right and assumed right.

Now, I cannot attest to the most up to date plugin's if in fact they do some dithering on their own. Some have an option to dither at the plugin's output, like the L1/L2 and a couple others I have seen. BUT, unless you can find that option on your plugin, or somewhere in it's documentation, you should assume that it DOES NOT do ANY dithering, but rather truncates back to the original bit depth.

If the original source file is 24 bit, don't worry a bit (ha ha..a pun...) about dithering. Any truncation taking place will be well out of your equipments usable dynamic range. But if the original source is 16 bit, it is a good idea to apply dithering before it goes to CD. If the plugin doesn't do the dithering for you, then you should do it before saving the final file.

WARNING!!! DO NOT DITHER MORE THAN ONCE!!! Dither ONLY once, and as the final step before the audio is burned to CD.

Ed

 

[pglewis]

But if the original source is 16 bit, it is a good idea to apply dithering before it goes to CD.

I don't follow you, Ed... if the source is already at 16-bits, what's dithering going to do? Even if the DSP is truncating, the extra bits are tossed to the bit bucket and not available for the dithering algo. What am I missin' here?

 

[sonusman]

Okay, let's assume that your audio is really clean all the way down to -96dB, the dynamic range of a decent D/A converter.

When you apply DSP to the audio, and it does so at let's say 24 bit (fixed, 32 bit floating point...most argue that they are about the same in the real world...). 24 bit is something like -110dB of real world dynamic range, but about -144dB internally (meaning when it doesn't have to do a A/D/A conversion of any type). So, the audio was processed and an extented bit depth was created to do so (digital works that way). The resulting file could be using more than 21 or 22 bits to represent it. You will not hear truncation at 20-24 bits. The first 16 bits will be very clean and free mostly of quantization errors. But when that DSP truncates the audio, what you can effectively have is audio that as it is fading out will seem to drop off the face of this earth instead of fading out smoothly. Adding that bit of shaped noise at -93dB (which is what dithering does on at 16 bits) will make that last 3dB of the dynamic range seem to fade out smoothly.

Remember, your ears are capable of hearing audio well beyond the real world dynamic range of most 16 bit converters, so it is important that 16 bit audio is very smooth and clean and doesn't sound like the fade out drop off the face of the earth. Dithering helps with that at the expense of a little noise that your ear will not hear so well.

Bob Katz wrote a nice piece on all this over at http://www.digido.com . It is worth checking out.

Ed

 

[zip]

I read that article...

Which is why I assumed my assumption was indeed correct. I also DID read your post correctly about the effects of DSP on 16 bit audio...

...I was not questioning it - rather seeing if it also applies to this situation.

I am transferring very clean audio tape at 16 bit / 44.1 KHz then "mastering" in the digital realm using a combination of EQ, stereo enhancement, Multi-band compression and limiting. The Logic AD Limiter (which sounds best on THIS audio) does not have a dithering option as the L1 / L2 do...

Therefore I assumed it would be wise to use Logics' Powr dithering option when bouncing the final work to .wav stereo. I did NOT dither when saving the original audio straight to CDR as a back-up. So am still following the dither once mandate.

No offense to the others who offered their opinions but I have followed Ed's advice in the past and he has never steered me wrong.

Thanks for the follow up Ed...

zip >>

 

[littledog]

Sounds to me like everyone gave you the same advice: don't dither except during the final mastering process.

And sounds like you followed it.

 

[zip]

ummmmmm...

I did follow it but there was one important distinction...

Some said no need to dither 16 bit audio AT ALL...even during the final transfer. Sonusmans advice is to always apply some form of dither when using DSP - which was my original question....

Maybe I read them incorrectly...

zip >>

 

[pglewis]

Some said no need to dither 16 bit audio AT ALL...even during the final transfer

Yeah, and I still don't know what Ed's going on about there . I've read the Katz stuff several times.

For dithering to have any effect, you'd need more bits of information available to you than the target bit-width. If you only have 16 bits, and your target is 16 bits then there's nothing dither can do for you.

Adding that bit of shaped noise at -93dB (which is what dithering does on at 16 bits) will make that last 3dB of the dynamic range seem to fade out smoothly.

But, you need to have information from the 24-bit audio to feed the dither algo. If the DSP already truncated, then those extra 8 bits are gone. Bye-bye. Can't use them to your advantage for dithering.

 

[sonusman]

pglewis, you are missing the point big time buddy.

It is MOST IMPORTANT to dither audio that was originally 16 bit where DSP was applied to it.

You need that noise at the bottom of the dynamic range to smooth out quantization errors and for the fade outs to sound right.

When you apply DSP to 16 bit audio, a higher bit depth is created. Usually, the extra bits are just truncated at the output of the DSP, so at 16 bit, the audio will instead of fading out smoothly, will seem to just stop before a smooth fade out happens. Applying dither at 16 bit helps assure that there is at least some noise down there to smooth out the audio just dropping out to the ear.

When dither is applied to a higher bit depth, there is a pretty good chance that you can add some more percieved bit depth to the audio, because you are actually starting out with with a bigger dynamic range and the dithering noise keeps that last bit "on" so you can actually hear audio that is normally truncated and lost. That audio is "mixed" with the dithering noise.

Anyway, you are indeed VERY wrong to say that you don't need to dither 16 bit audio that has had DSP applied to it!!! THAT is originally why dither was created friend!!! This whole new 24 bit realm is a very new thing really.

Ed

 

[pglewis]

Ed, head back over to http://www.digido.com/ditheressay.html and look at the section on "How to Dither". You'll see from his description of how dither works that it requires more bits than your target bit-width. His example shows how adding a random number (ZZZZ ZZZZ) to the lower 8-bits of a 24-bit sample (YYYY YYYY) modulates the LSB (W) of the resulting 16-bit sample.

... In essence, the random number sequence combines with the original lower bit information, modulating the LSB. Therefore, the LSB, from moment to moment, turns on and off at the rate of the original low level musical information...

(emphasis mine)

Note that you still need the lower bit information for this to be of any use.

 

[Neve]

Check this out: http://www.cranesong.com/analogdither.html

The dither from the HEDD 192.

I have one on it`s way to me now.

Amund

 

[pglewis]

Dither CD?!

That's it! Where's Skippy? I'm getting to the bottom of this.

 

[zip]

I like...

...the healthy discussion here.

Maybe a "test" is in order?? I can bounce the DSP applied 16 bit files with and without dither and post short .wav snippets to my website. Just a thought.

The last time I "mastered" a CD bootleg (which was a direct transfer from tape to CD using Soundforge and no dither) I tried the comparison...

Personally I found the non-dithered, DSP applied files to be grainier sounding. Maybe that's my ears however...

So - why in the hell did I ask this question?? Because I didn't know the correct answer...



zip >>

 

[pglewis]

How dare you suggest using our ears, zip?!

I just wanna make sure I'm not misunderstanding the ones and zeros here. I thought I had a good grasp, and if I don't I wanna get the tech side straightened out in my head.

Where's skippy hiding?