The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Good dither practices, what are yours? Dynamics Plugins
Old 19th January 2017
  #151
Lives for gear
 
Hyder boy's Avatar
 

Quote:
Originally Posted by DeyeYguy View Post
Lynx AES16e mixer does a nice job offering dither LAST on every DA out for us guys feeding consoles, i only employ dither when i make a change ITB, even if it's a .01 fader movement, otherwise if the "BOX" serves as a tape machine only untouched, dither remains off. I'm not positive i have my head wrapped completely around the proper implementation of dither in a hybrid context, but the previous method makes sense to my small brain, the earlier posts seem to support the idea as legit as well

Bob O has emerged as the dither conscience of Pro Audio Universe, and for good reason!

thanx Bob O
I'd love to hear more on this, as I am on Lynx AES16e as well. How do you implement it? I'm using hardware inserts in pro tool.

Thanks
Old 19th January 2017
  #152
Lives for gear
 
JayTee4303's Avatar
Quote:
Originally Posted by Hyder boy View Post
Hey thanks
I get most of it, I always dither when bouncing final to 16 bit. This business of dithering when using hardware inserts (specifically pro tools here) has me confused.
A hardware insert is simply a D/A>outboard>A/D circuit.

D/A there, or for monitoring, "polls" the bit registers, and uses the value obtained to generat an analog signal. This is subject to the same truncation error potential, as recalculating bit register values from on sample rate to another.

Put another way...D/A is also a sample rate change, from say 48k, to infinity k, except...there are only 2**24 possible input values. A/D assigns infinite input levels to nearest estimated matches in a 24 bit system. D/A reverses that.
Old 19th January 2017
  #153
Lives for gear
 
JayTee4303's Avatar
Quote:
Originally Posted by ZenMaster View Post
Atm I'm under the conclusion that you should dither before D/A in any case. That's HW inserts and the master buss. I understood the reason being that the software can process the audio in multiple bit depths. For example REAPER's mixing bit depth is 64 bits. If I understood correctly one should dither because most D/A's are 24 bit.

Please correct me if I'm wrong.
Capture a track. 24 bit file. Make no changes. Save it. Copy the file to a thumb drive. Copy from thumb drive to second PC.

Open it. Bit for bit copy of original captured track.

No dither necessary.

Ok?

Now, capture 2 tracks, 24 bit files. Mix them with zero changes. Yes...they will be processed in FP...EVERY TIME you hit play. Live, in near realtime, thru the CPU, each time.

The original two files are bit for bit unchanged from original capture.

No dither necessary.

You DO dither the master buss, to avoid hearing trunc error in the analog monitors, but the files are unchanged.

Ok?

Now, capture one track. Insert a plugin. The data stream is converted to 32 bit FP. Then processed. Then converted back to 24 bit fixed. The re-conversion to 24 bit induces trunc errors, which cannot be undone by the master buss dither later on, so you dither at the end of the plugin chain, so the rest of the channel strip works with dither noise, which is better than working with trunc errors.

Again...the original capture is unchanged after saving and shutdown. Next session, that original file is re converted to 32 FP before the plugin chain, processed, dithered again, and again, the rest of the channel strip gets dither nouse, better than trunc error.

In its simplest form, this is about commitment.

Critical point: TWO DIFFERENT KINDS OF COMMITMENT.

Type 1: Changing a capture file's sample rate, for export or D/A... irrevocable committment, chose one...dither noise or trunc errors, FOREVER.

Type 2. Changing a capture file's sample rate EARLY IN THE AUDIO CHAIN, such that you are commited to either trunc error or dither noise, the rest of the chain, ONCE, TODAY, but ALSO, EVERY SUCCESSIVE PASS thru the CPU, because once done, it can't be undone later in the chain. The master buss dither won't touch it...too late.

Yes...in the case of type 2 commitment, you HAVE NOT changed the original file, not till you D/A or SR change via a render, but...you can't USE the original file...you have commited to the 32 bit FP version, this pass, next pass, every additional pass...until you remove the plugin.

"But..." you say..."mixing two channels forces the datastream to 32 FP, too!"

True.

But it STAYS 32 FP, till the master buss dither, which lessens the problem.

The plugin section of the chain, converts to 32FP, but, importantly, CONVERTS BACK TO 24 fixed.

THATS where you need the additional dither layer, right before the re-conversion at the end of the plugin chain.

Now... I begin to get why nobody answered my question... DSP processing in the audio interface.

Yes... same problem, most likely... if the interface processing converts to 2FP then back to 24 fixed.

Not only can't you dither by choice there... but there's no place else to dither either. SOL.

And THAT...if I have this right...opens up a whole FORUM worth of worm cans.

Like... which interface DSPs process at 32FP? Which ones don't? Which ones dither? What kind of dither?

Think about studios who mix DAW tracks and livesynth tracks to 2 track tape. Hybrid summing. DAW farms. Plugin farms. "Live" softsynths.

Think about the poor bastard who has 12 PCs and 12 interfaces, scattered across 4 rooms, connected via both ADAT and copper, with a 240 channel ADAT multiplex in the middle? With 100 pre-programmed routings, that each involve around 19.5 million variables.

Yeah, that would be me.

It's ok. The concept is in place. I know what I need to know, and I know what I need to find out.

I will, then I'll work thru thise variables too.

When I get to it.

**** sounds great now, and clients ain't bitchin yet.

I have time to work with, and really, this IS what we do. We use tech to make pretty noises, we identify shortcomings, then learn new tech to minimize them, for noises that are prettier still.

Nother day, nother dollar!

:-)
Old 19th January 2017
  #154
Lives for gear
 
Hyder boy's Avatar
 

Quote:
Originally Posted by JayTee4303 View Post
A hardware insert is simply a D/A>outboard>A/D circuit.

D/A there, or for monitoring, "polls" the bit registers, and uses the value obtained to generat an analog signal. This is subject to the same truncation error potential, as recalculating bit register values from on sample rate to another.

Put another way...D/A is also a sample rate change, from say 48k, to infinity k, except...there are only 2**24 possible input values. A/D assigns infinite input levels to nearest estimated matches in a 24 bit system. D/A reverses that.
As always, thanks for responding. I get all that, and I understand what dither is and does. I just haven't heard of anyone using dither in conjunction with hardware inserts specifically. More specifically, how are you implementing it?
Old 19th January 2017
  #155
Motown legend
 
Bob Olhsson's Avatar
 

Quote:
Originally Posted by Hyder boy View Post
As always, thanks for responding. I get all that, and I understand what dither is and does. I just haven't heard of anyone using dither in conjunction with hardware inserts specifically. More specifically, how are you implementing it?
Tens of thousands of people did it every day for over a decade using Pro Tools TDM systems with Avid's dithered mixer plug-in. Those of us using native systems have been dithering hardware inserts.
Old 20th January 2017
  #156
Lives for gear
 
Hyder boy's Avatar
 

Quote:
Originally Posted by Bob Olhsson View Post
Tens of thousands of people did it every day for over a decade using Pro Tools TDM systems with Avid's dithered mixer plug-in. Those of us using native systems have been dithering hardware inserts.
Well thank you for that. If you'd be so kind as to answer one last question (for me at least), exactly where in the chain should one insert a dither plugin, before or after the hardware insert?

I've been on pro tools since 9, before that, mostly tape. Computers are not my strong point.

Thanks
Old 20th January 2017
  #157
Gear Nut
 

Quote:
Originally Posted by JayTee4303 View Post
Now, capture one track. Insert a plugin. The data stream is converted to 32 bit FP. Then processed. Then converted back to 24 bit fixed. The re-conversion to 24 bit induces trunc errors, which cannot be undone by the master buss dither later on, so you dither at the end of the plugin chain, so the rest of the channel strip works with dither noise, which is better than working with trunc errors.
^could you explain this^

if what you wrote is true, then wouldn't "you dither at the end of the plugin chain" being pointless?
In this case the plugin itself should apply dither, in fact I have no chance to sniff the data stream from the plugin to the daw to apply dither and if I do it as a second plugin then truncation errors would have already been introduced.

The point is: who does the conversion? Where in the chain? If I stack two plugins in series would the data stream stay at 32fp between the two?
Old 20th January 2017
  #158
Gear Guru
 
Karloff70's Avatar
 

Quote:
Originally Posted by Hyder boy View Post
Well thank you for that. If you'd be so kind as to answer one last question (for me at least), exactly where in the chain should one insert a dither plugin, before or after the hardware insert?

I've been on pro tools since 9, before that, mostly tape. Computers are not my strong point.

Thanks
Before. As to output the audio it has to make a 24bit file, and you are dithering to make sure that file doesn't have any truncation happening.
Old 20th January 2017
  #159
Lives for gear
 
JP__'s Avatar
 

Quote:
Originally Posted by Karloff70 View Post
Before. As to output the audio it has to make a 24bit file, and you are dithering to make sure that file doesn't have any truncation happening.
Dithering do not prevent truncation, it prevents distortion...
Old 20th January 2017
  #160
Motown legend
 
Bob Olhsson's Avatar
 

Dither belongs between processing or level changes and the feed to a converter. The converter dithers internally from its processing to the DAW input.

The simplest way I've found to think of it is that any time there is math, the results need to be dithered down in order for the results to be useful and not distorted. Digital audio is nothing but numbers.
Old 20th January 2017
  #161
Gear Guru
 
Karloff70's Avatar
 

Quote:
Originally Posted by JP__ View Post
Dithering do not prevent truncation, it prevents distortion...
Whoops, my bad, still finalising these late learnings.....lol
Old 21st January 2017
  #162
Gear Addict
Quote:
Originally Posted by Bob Olhsson View Post
Dither belongs between processing or level changes and the feed to a converter. The converter dithers internally from its processing to the DAW input.

The simplest way I've found to think of it is that any time there is math, the results need to be dithered down in order for the results to be useful and not distorted. Digital audio is nothing but numbers.
I read earlier in this thread where it was said that dithering in floating point was in a sense pointless ( that was my interpretation of it anyways)

so what I'm wondering is, is this true? And if so then would inserting a dithering plugin prior to the hardware insert plugin be the same as dithering the floating point. Otherwise is the DAW still operating in floating point between these two plugins?

Where my confusion lies is knowing exactly at what point the DAW outputs fixed point. For instance on the master fader would it be better to add dither post master fader?

Also in Studio One there is a second set of channels with faders and inserts pre and post fader dedicated to the outputs. My thought was inserting dithering here ( output channels) post fader. That would be the vary last place in the DAW to add dithering before going to the converters. Is this just over thinking things or a solution?
Old 21st January 2017
  #163
Motown legend
 
Bob Olhsson's Avatar
 

No, you are dithering the conversion from what is often 64 or even 80 float to fixed.
Old 21st January 2017
  #164
Gear Maniac
 
Yuri Korzunov's Avatar
 

Quote:
Originally Posted by Lowend Bump View Post
I read earlier in this thread where it was said that dithering in floating point was in a sense pointless ( that was my interpretation of it anyways)
Dithering is noise.

Additional noise in audio is bad.

But if you convert to integer format appear rounding (quantization) error.

This error give harmonics correlated with audio stuff.

If add low level noise and after it make rounding these harmonics are smoothed on spectrum.

Noise level of audio in target integer format will increased slightly.

I.e. dithering work together with rounding only.

For target 24-bit and more integer format dither don't applied because there quantization noise level about -145 dB, but DAC have output noise level about -120 dB.
Old 21st January 2017
  #165
Gear Addict
Okay that clears it up.
The finer workings of dither isnt a strong point for me.
I aprciate all your input.
Old 21st January 2017
  #166
Here for the gear
 

Interesting read. So are you guys saying that truncation takes place every single time I instantiate a plugin on a channel and therefore would require dither after every plugin instance, or only at the end of a chain? Excuse my ineptitude, I'm trying to wrap my head around this...very late in the game unfortunately
Old 21st January 2017
  #167
Gear Guru
 
Karloff70's Avatar
 

At this point I'd say, can we please have a little 'reading the thread from the beginning before asking the same questions all over again' type behaviour, chaps?

Only because otherwise the thread instead of a useful resource will become a mess of repetition and confusion. It is confusing enough for most already....

Old 21st January 2017
  #168
Lives for gear
 
apartment dog's Avatar
 

Quote:
Originally Posted by Karloff70 View Post
At this point I'd say, can we please have a little 'reading the thread from the beginning before asking the same questions all over again' type behaviour, chaps?
Makes me think of a contemporary play.
Every five minutes a new person enters the room (stage) and asks the same question. The conversation halts, starts all over again, and so on.

But literary fun aside
I love this thread!
Thanks for the great info.
Old 21st January 2017
  #169
Motown legend
 
Bob Olhsson's Avatar
 

Quote:
Originally Posted by waldster View Post
Interesting read. So are you guys saying that truncation takes place every single time I instantiate a plugin on a channel and therefore would require dither after every plugin instance, or only at the end of a chain? Excuse my ineptitude, I'm trying to wrap my head around this...very late in the game unfortunately
With a TDM system dither was required before and hopefully the developer provided dither back into Pro Tools. The dithered mixer did this.

In native systems, the connections are all 32/64 float except for connections to converters which should be dithered.

It's easier to put on your nerd hat and learn the theory about this than to try and come up with some standard workflow concept.
Old 23rd January 2017
  #170
Gear Addict
 

This dither chat has had me pondering and I was mid debate today with an old engineer who told me in no uncertain terms that it was a waste of time adding dither on every hardware insert and only added noise.

Never one to listen to old engineers I figured I'd come home and see just how much noise it adds. SOOOOOOOOO

I loaded up and instance of ableton. Recorded a spot of noisefloor from my lynx aurora and proceeded to start adding dither plugins (In this case I used the airwindows TPDF (Which shows up in my plugin window as TPDFDither64 and has no controls)

I added away until my ram ran out and I could not hear hide nor hair of this alleged noise. By my count I was at some 400 instances by the time my computer crashed (albeit on a single track). Am i missing something? I would expect 24bit dither to sit at a noisefloor of about -144 and I would expect each doubling of instances to raise the noisefloor by 3db? surely 400 instances would at least show up on the meter?

If not then why the hell have I not been dithering everything all along?
Old 23rd January 2017
  #171
Motown legend
 
Bob Olhsson's Avatar
 

The bullsh!t about "double dithering" has been spread far and wide...
Old 23rd January 2017
  #172
Gear Addict
 

Quote:
Originally Posted by Bob Olhsson View Post
The bullsh!t about "double dithering" has been spread far and wide...
I'm not particularly bothered by the concept of double dithering but I do wonder why I can't hear this dither when I add several hundred instances? surely it should present as a slowly rising full spectrum hiss?
Old 23rd January 2017
  #173
Lives for gear
 

Quote:
Originally Posted by africantigercow View Post
surely it should present as a slowly rising full spectrum hiss?
The recorded noise from your converter is approx -110dBFS.

Dither at the 24bit level is down at approx -140dBFS.

Two uncorrelated signals at same level gives you an increase of 3dB.
If you add two instances of dither at -140dB you get -137dBFS.

If you add a third instance of dither you get less than +3dB since this third addition is at a lower level than the summed response of the two first instances.

So for every instance of added dither the relative increase gets smaller and smaller. That and the fact that you start out with a noise file of -110dBFS means you have to add many many instances of dither in order to get to an audible level.

While it is good practice to stay at 64bit float during post, you should never worry about adding 24 bit TPDF dither if you need to render or take a roundtrip to outboard FX a couple of times.
Old 24th January 2017
  #174
Quote:
Originally Posted by africantigercow View Post
I loaded up and instance of ableton. Recorded a spot of noisefloor from my lynx aurora and proceeded to start adding dither plugins (In this case I used the airwindows TPDF (Which shows up in my plugin window as TPDFDither64 and has no controls)

I added away until my ram ran out and I could not hear hide nor hair of this alleged noise. By my count I was at some 400 instances by the time my computer crashed (albeit on a single track). Am i missing something? I would expect 24bit dither to sit at a noisefloor of about -144 and I would expect each doubling of instances to raise the noisefloor by 3db? surely 400 instances would at least show up on the meter?
One instance of properly implemented TPDF dither yields a noise floor of -139 dBFS

10 log (400) = 26 dB. So the accumulated dither noise was still below -113 dBFS.

David L. Rick
Seventh String Recording
Old 24th January 2017
  #175
Lives for gear
 
Arksun's Avatar
Quote:
Originally Posted by Karloff70 View Post
Who dithers where, and why? Go......
When I output to 16bit audio file, its the only time its needed.
Old 24th January 2017
  #176
Well, you can lead a horse to water...
Old 24th January 2017
  #177
Lives for gear
 
Arksun's Avatar
Quote:
Originally Posted by David Rick View Post
Well, you can lead a horse to water...
He asked how each of us use it, so there's my answer

I work entirely ITB, the DAW works at 32-bit float, anytime I mixdown audio inside a project its always to 32-bit float, whats the point in dithering then?

Even if you just save the entire mix as 24-bit audio file without dithering, the distortion caused by truncation is well below the noise floor of the best converters on the planet, whos gonna hear it?

A one time truncation to 24bit for final mixdown (which lets face it is not gonna be a wave file that has a maximum peak of -90dbfs) is a complete non-issue as far as I'm concerned. Going down to 16bits is another matter however, then I'll dither.
Old 25th January 2017
  #178
Gear Maniac
 
Yuri Korzunov's Avatar
 

Quote:
Originally Posted by Audiop View Post
Two uncorrelated signals at same level gives you an increase of 3dB.
If you add two instances of dither at -140dB you get -137dBFS.
Quote:
Originally Posted by Audiop View Post
If you add two instances of dither at -140dB you get -137dBFS.
Up to 3 dB. It is fairly for 2 signals.
Noise is stochastic process. Accumulated summ of two noise sources give lesser 3 dB growth in most probably cases, I suppose.
Old 25th January 2017
  #179
I finally found a proper ABX test. There aren't many around!

Archimago's Musings: 24-Bit vs. 16-Bit Audio Test - Part II: RESULTS & CONCLUSIONS

In short: We get a solid 50/50, not really a surprise!

Last edited by FabienTDR; 25th January 2017 at 12:47 PM..
Old 25th January 2017
  #180
Gear Maniac
 

Quote:
Originally Posted by FabienTDR View Post
I finally found a proper ABX test. There aren't many around!

Archimago's Musings: 24-Bit vs. 16-Bit Audio Test - Part II: RESULTS & CONCLUSIONS

In short: We get a solid 50/50, not really a surprise!
And where is 16bit truncated in test?? This only says that people almost can't differentiate between 24bit (truncated/undithered output) and 16bit dithered.
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump