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 12th February 2017
  #301
Lives for gear
 

Quote:
Originally Posted by JP__ View Post
But a DC filter is audible too.
Blanket statements like that always trigger my interest. :-)

What kind of DC filter, in what situation and with what material?
Old 12th February 2017
  #302
Lives for gear
 
JP__'s Avatar
 

Wow, not even one hour...
Im quite impressed.
Old 12th February 2017
  #303
Gear Maniac
 
Yuri Korzunov's Avatar
 

Quote:
Originally Posted by JulenJVM View Post
Of course this distortion is not audible, but when I apply dither, I hear new things on my mixes, related mainly to spatial placement of instruments. Truncation distortion seems to take away small details from the music, that dither helps bringing back. That is also subjective, but my own experience.

Two questions, please. Are you able to hear DC Offset? Do you take measures to remove it?
I suppose, you applied for 16 bit conversion. It is other case than dithering of 24 bit, that effective for modern DAC.

DC offset inaudible (it is not oscillation), but it can cause overload and clicks.

In DSD stuff sometimes being non constant DC offset per album: zero in begin, further it grow, at the end it fall to zero.

When tracks of this album played in audio player without gapless ability, there possibly clicks at borders of the tracks due DC offset.

I suspect what such non-constant offset need remove manually with editing of waveform.
Old 12th February 2017
  #304
Lives for gear
 
edva's Avatar
Quote:
Originally Posted by JulenJVM View Post
The point I wanted to make was not about the effect of the process, but rather about dismissing it because the problem is not audible. I wanted to point out that we deal with issues outside the hearing range, because we know the theory behind it, so why treat dither differently?
Yes, I completely understood your point, and your technique of using one thing to cast doubt upon another is a valid argumentative technique, in that regard I agree.
My point is that the two sonic effects are too different to make a valid comparison.
But your point is certainly a valid one.
Old 12th February 2017
  #305
Motown legend
 
Bob Olhsson's Avatar
 

I just ran a test yesterday where dithering to 24 bit beat 32 float in the final sample rate converted, dithered 16 bit file. People forget that what might not be audible after a single process can become audible upon additional processing. It's important to listen with our ears and not our eyes and logical sounding unproven theories. In the case of masking, it's totally non-linear, depends on the specific frequency spectrum involved and varies from individual to individual.
Old 12th February 2017
  #306
Lives for gear
 
edva's Avatar
Quote:
Originally Posted by Bob Olhsson View Post
I just ran a test yesterday where dithering to 24 bit beat 32 float in the final sample rate converted, dithered 16 bit file. People forget that what might not be audible after a single process can become audible upon additional processing. It's important to listen with our ears and not our eyes and logical sounding unproven theories. In the case of masking, it's totally non-linear, depends on the specific frequency spectrum involved and varies from individual to individual.
excellent post, as always from you Bob, and thank you.
I agree, I have been hearing a better result since doing this, which of course I never would have thought of doing, if I hadn't learned about it on GS . Thanks.
Old 12th February 2017
  #307
Motown legend
 
Bob Olhsson's Avatar
 

I also started out somewhat skeptical about both the need for 24 bit dither and the advantages of higher sample rates.
Old 12th February 2017
  #308
Lives for gear
 

Quote:
Originally Posted by JP__ View Post
Wow, not even one hour...
Im quite impressed.

Why be a troll when you can chose to be constructive instead?

I am seriously interested in how you can make a statement like that.

One reason is that I and people I know and respect have done tests with DC filters that indicates it can be 100% transparent. If you can tell more I/we can learn something. Is that something you are against?

In order to claim that a DC filter always is possible to detect with the ears would take some serious work which I have good reason to believe you have not done.
Old 12th February 2017
  #309
Lives for gear
 
edva's Avatar
Quote:
Originally Posted by Audiop View Post
Why be a troll when you can chose to be constructive instead?

I am seriously interested in how you can make a statement like that.

One reason is that I and people I know and respect have done tests with DC filters that indicates it can be 100% transparent. If you can tell more I/we can learn something. Is that something you are against?

In order to claim that a DC filter always is possible to detect with the ears would take some serious work which I have good reason to believe you have not done.
On the subject if DC filters, often I can certainly hear a difference. It seems to depend a lot on the source material. YMMV.
Old 12th February 2017
  #310
Airwindows
 
chrisj's Avatar
Quote:
Originally Posted by edva View Post
On the subject if DC filters, often I can certainly hear a difference. It seems to depend a lot on the source material. YMMV.
Depends on how steep it is, and how it's done.

I can make a perfect DC filter for some audio (and probably SHOULD because it'd have use for outputting CVs for modular synths, with some converters). Done properly, you certainly wouldn't hear a difference. It's called a bias current: just add a set number to everything. Will pop on the edges of the region unless the audio already popped on entry/exit of the region. You'd set it up using some sort of meter, for instance SPAN.

I can make a very good sounding highpass if you don't need it to be a brickwall. Heck, I just released just such a filter, 'Capacitor', and it has just such a highpass. It's not nearly as steep as people imagine when they go for 'DC offset' filters, but you could use it to cut a lot of DC and it wouldn't sound different. You just couldn't cut ALL the DC, is the thing.

The underlying point of contention here is (as always) people bringing up DSP processes that involve calculation, degradation of the audio through repeated math operations and trade-offs where you sacrifice something to improve something else… and a counter-argument that 'you can't hear any negative effect from doing this, because it's not a large enough change to consistently identify'. Implicit in that is this idea that nothing matters unless you can hear it every time under moderately hostile conditions.

And there's a place for that, but the place for that is proving to reasonable confidence that a phenomena, which might be mere imagination or random chance, DOES exist in the face of argument that it does not.

The converse is absolutely not true or supported in any way. You can't prove something doesn't exist, or that people don't hear it, by documenting that they don't reach a level of confidence with it, in testing. In fact sometimes (as with me and quantization noise, or Eric Sarafin hearing the 11.025K tone and then being unable to UNhear it) the threshold we're interested in is really far away from that ABX, 20 out of 20, blind testing threshold.

We have to be a lot more interested in audio than THAT.
Old 13th February 2017
  #311
Lives for gear
 

Quote:
Originally Posted by edva View Post
On the subject if DC filters, often I can certainly hear a difference. It seems to depend a lot on the source material. YMMV.
Nobody in his/her right mind with basic training and/or experience with audio engineering and human hearing would doubt that a DC filter can* be made to affect a signal audibly.

What is arguable is if a DC filter always is possible to detect with the ears.


*A single DC blocking cap is a DC filter and you can place the knee at 2kHz if you want to.. Lots of heavy drugs needed to "unhear" that. :-)
Old 13th February 2017
  #312
Lives for gear
 
Mixerman's Avatar
 

Quote:
Originally Posted by chrisj View Post
Yeah… one of the proudest moments I can remember, is getting into a 'dither and sample rate' argument not so long ago in which JJ was present, and having him be more or less on the same page as me. Mind you, I'm happy to entertain wilder notions too, but that was a good day.

My own 'golden eared guitarist' experience was in fact with Mixerman. (for a while there he was enthusiastic about maybe branding some of my stuff Mixerman-brand and splitting the take: he's got his favorites, NOT always the recentest versions of the plugs which is a darn good wake-up call for a dev)

He was using some interface, and there was a noise that was driving him insane and the company was stonewalling him, so he turned to me and provided the 'horribly ruined' mixes as evidence.

Sounded absolutely fine to me, no matter what I did. Great mixes, too.

I picked some lulls in the music, cranked up the sound, still couldn't hear it, and used the old program 'Amadeus' by Martin Hairer to do a sonogram of the audio. And there it was, plain as day: an obvious artifact at 11.025K exactly, very subtle but if you saw it depicted as a spectrum analysis over time there was no mistaking the little artifact making a faint, perceptible line. It proved conclusively that something was happening, and I know from some of Mixerman's preferences in plugins that he has bionic hearing around 10K (I had a compressor he liked, turns out it had great ability to not generate issues around 10K: the original Pressure, now free VST)

I still couldn't hear the problem, even over stripped Sennheiser HD600 headphones driven directly by a Lavry Black. But then I just have silver or possibly bronze ears. Not golden

I care about doing stuff right for people who can hear a lot better than me. The variation is amazing in people, it really is. And to my mind it's also a conclusive repudiation of statistical double blind tests as the only path to truth. Because I think, just possibly, Mixerman would not have been able to pass an ABX with high confidence, over that tiny 11.025K artifact, and this is important to consider.


It was extended listening that tipped him off, something niggled at him, and then he couldn't unhear it and it got worse and worse until he turned to me for some kind of help, thinking it'd be just as obvious to me. I never could hear it, even fugitively. I doubt he'd have been able to pick it out 10 out of 10, and demonstrate its presence with statistical confidence.

But that had nothing to do with it. The artifact was there. I measured it with a computer and metering of an unusual sort, incontrovertibly. And it was exactly what he'd complained of, whether or not he could pick it out every time. Sometimes it's not a matter of opinion.
This happened around the time Apple released the new Quad Cores around 2007. They switched to PCIe cards, and so my computer was cutting edge at a time when there really weren't any interfaces using the new PCIe.

Anyway, just to tell you, I could have easily picked it out blind were there actually a way to listen to the mix without the artifact. It was literally painful to me. I couldn't understand how no one else could hear it.

It wasn't until I sent them your Sonogram that they realized something really was wrong.

Enjoy,

Mixerman
Old 14th February 2017
  #313
Lives for gear
 
JP__'s Avatar
 

Quote:
Originally Posted by Audiop View Post

One reason is that I and people I know and respect have done tests with DC filters that indicates it can be 100% transparent. If you can tell more I/we can learn something. Is that something you are against?
In my world there's no 100% transparent processing (we had this discussion nearly a hundred times and I am not willing to repeat it again and again...), its in fact completely unlogical to make such a claim. So, all processes should be used within a reason and not by second guessing. Thats all.
Old 14th February 2017
  #314
Lives for gear
 
JP__'s Avatar
 

Quote:
Originally Posted by Audiop View Post

What is arguable is if a DC filter always is possible to detect with the ears.
Who had said this? In german we say: du drehst mir die Worte im Mund herum...
Such argumentation looks completely trumpish to me (and its not the first time you have done so). So, please do ignore my comments in the future. Thanks.
Old 14th February 2017
  #315
Lives for gear
 

Quote:
Originally Posted by Audiop
One reason is that I and people I know and respect have done tests with DC filters that indicates it can be 100% transparent. If you can tell more I/we can learn something. Is that something you are against?
Quote:
Originally Posted by JP
Yes, I am.
That's an interesting mindset on a discussion board..

Quote:
In my world there's nothing 100% transparent processing (we had this discussion nearly a hundred times and I am not willing to repeat it again and again...),
Yes there are processing that is 100% transparent to the human ear. Your beliefs has little effect on that.

Quote:
its in fact completely unlogical to make such a claim.
The opposite actually. Completely logical since the ear/brain has limited ability to detect a change on a signal. Once the change fall below JND no detection can occure.

Of course people differ in their abilities but some things we can say for sure, and some things we can say highly unlikely.

If we test 100.000. people for the audible effect of a change of a resistor value in a preamp that changes gain 0.001dB and nobody can detect it it's fair to assume it's well below the threshold of human hearing. If somebody then shows up claiming to be able to hear it, it's interesting becasue if it's true we can learn something new. That person would need to either show up for a proper test to believed though or at least be willing to share the details about the test that led to the conclusion.

People believe and claim lots of things in this world.
Old 14th February 2017
  #316
Lives for gear
 

Quote:
Originally Posted by Audiop
What is arguable is if a DC filter always is possible to detect with the ears.
Quote:
Originally Posted by JP__ View Post
Who had said this?
In post 300 you wrote:
Quote:
But a DC filter is audible too.
And that triggered my interest.

Quote:
In german we say: du drehst mir die Worte im Mund herum...
Such argumentation looks completely trumpish to me (and its not the first time you have done so).
I don't speak german and I don't know what "trumpish" means so I don't know what this "not the first time" means either. Maybe you could use english and tell me what you are accusing me for in common established words?
Old 14th February 2017
  #317
Lives for gear
 

Quote:
Originally Posted by JP
Man... Winters in sweden must be long and lonely...
If you are interested in talking about the nordic climate or my personal life, please drop me a PM. I think this thread is perhaps not the right place.
Old 20th February 2017
  #318
Lives for gear
 
Lipps's Avatar
 

So after 11 pages I've got it! But someone should make a clear post on when to dither into a sticky. Save everyone else time.
Old 21st February 2017
  #319
Gear Maniac
 
Yuri Korzunov's Avatar
 

Quote:
Originally Posted by Lipps View Post
So after 11 pages I've got it! But someone should make a clear post on when to dither into a sticky. Save everyone else time.
1. Dithering is method of suppressing of non-linear distortions, that risen after decreasing bit depth (truncation) with/without rounding.

2. Before truncation and rounding low level noise added to signal.

3. Dither don’t work properly without rounding.

4. As result energy of non-linear-distortion «sticks» distributed into nearest frequency area. Total level noise is increased slightly, but «sticks/peaks» of non-linear distortions drown in surrounding noise.

5. Sometimes applied noise shaping (pushing noise energy to high frequencies about ultrasound) for decreasing of noise level in «more audible» lower range.
Old 21st February 2017
  #320
Airwindows
 
chrisj's Avatar
Quote:
Originally Posted by Yuri Korzunov View Post
1. Dithering is method of suppressing of non-linear distortions, that risen after decreasing bit depth (truncation) with/without rounding.

2. Before truncation and rounding low level noise added to signal.

3. Dither don’t work properly without rounding.

4. As result energy of non-linear-distortion «sticks» distributed into nearest frequency area. Total level noise is increased slightly, but «sticks/peaks» of non-linear distortions drown in surrounding noise.

5. Sometimes applied noise shaping (pushing noise energy to high frequencies about ultrasound) for decreasing of noise level in «more audible» lower range.
I prefer a more functional/practical description (which you may find says similar things)

1 When you reduce word length, the audio gets forced to 'the nearest voltage' that can be recorded in the new word length. Like if it's one bit, your output voltage can only be ON or OFF, if it's two bits you get to add HALF and so on. This is coming from a continuously varying voltage (or as near as you can get it)

2 If you add noise just right, this output will kinda sputter back and forth at the speed of the sampling rate. The result is, it averages out to be similar to what the underlying voltage ought to be. Any given moment is still going to be ON or OFF, but if it's dithered, it will average out to be exactly halfway. Or, if you started with 3/4 voltage, then ON will be picked 3/4 of the time and OFF 1/4 of the time.

3 If you didn't dither, the signal tries to follow the desired curve but it fails. It gets stuck on 'the nearest voltage' and it's really easy to hear the presence of the resulting artifacts, if you're listening for the right thing. There's no one specific frequency that's wrong, the whole sound gets garbaged up and can't be put right. Dithering completely stops this happening because the signal's never, ever stuck on 'nearest voltage' anymore. It's always bouncing around among the nearest possible values, so it's always causing the impression of an output between them.

4 Noise shaping is if you keep track of how wrong you were, and try to fix it faster than it would be fixed by just waiting. Ordinary dither (TPDF) will return a perfect result if you filter away all the noise, but we like hearing higher frequencies. Noise shaping can force things to sound quieter noise-wise, but it can only do it by adding energy (typically way up top). Adding extra noise isn't good, but if it's right up near the limits of digital to reproduce (called the Nyquist frequency), often the digital playback will try to filter out that noise (it will never try to filter out midrange or ordinary treble noise unless you specifically ask it to). So, noise shaping is really dependent on the playback system's filters and treble roll-off.



It's not really about redistributing 'sticks' of digital artifacts (that's more commonly seen in aliasing, which is a whole separate issue! ), it's about avoiding any places where the signal wants to go one way, and the output file says 'no, we're gonna stay on this value for a while, even though it's wrong'. 'sticks' aren't a cause, but a result of this: they're what you get when spectrum analyzing the audio. And again, they're much more of a thing when you're talking about aliasing. I'm pretty sure if you just looked at spectrum analyzer 'sticks', it would look like truncation was the winner because adding noise just adds loads more 'sticks'. But it makes it possible to keep track of where the underlying voltage was supposed to be… statistically.

Covering anything up with noise is totally not what's happening. We're arranging so that you can still hear where the voltage was supposed to go, not concealing anything. It just so happens that putting in noise, lets us do this keeping-track, and putting in not enough noise or the wrong sort (or too much!) fails to let us keep track of where the voltage was. TPDF is exactly two noise sources of exactly one bit in size, so any TPDF variation gives us this statistical result where we can hear 'what was supposed to happen'.
Old 22nd February 2017
  #321
Gear Maniac
 
Yuri Korzunov's Avatar
 

Quote:
Originally Posted by chrisj View Post
I prefer a more functional/practical description (which you may find says similar things)
I try describe rules, that independent on dither implementation.

Main purposes of dither is:

1. Suppressing rounding distortions levels (it looks as "sticks" in level about -80 ... -90 dB at spectrum) to level of surrounding noise.

2. Minimal surrounding noise increasing after this process.

Control dithered 16-bit signal via spectrum analyzer and check it on ear too.

Adjust the using dithering algorithm or found other one for achieving better result by this aims.

If used noise shaping, look for this aims at spectrum analyzer and ear too.



Musical signal is too complex for learning of principles of applying of dither.
For tutorial purposes recommended to apply dither to pure sines at different non-multiple frequencies (may be spectrum displaying issues).
Old 22nd February 2017
  #322
Lives for gear
 
Hyder boy's Avatar
 

Can anyone recommend a dither plugin for use with hardware inserts on pro tools 12.
Old 22nd February 2017
  #323
Lives for gear
 
Lipps's Avatar
 

When guys. When!!

Before hardware inserts - got it
(What about digital hardware inserts?)
On monitor path - got it

Last in line on busses/submixes - ??? Or just the main 2 bus

What am I missing?
Old 22nd February 2017
  #324
Gear Guru
 
Karloff70's Avatar
 

Quote:
Originally Posted by Lipps View Post
When guys. When!!

Before hardware inserts - got it
(What about digital hardware inserts?)
On monitor path - got it

Last in line on busses/submixes - ??? Or just the main 2 bus

What am I missing?
Anywhere before hitting a hard output. That includes hardware inserts and monitoring of course. And I also do it when bouncing down plugins now, as I feel pretty certain Logic doesn't do it for me.

Basically, anywhere float becomes a 24bit file again. Plus if converting down to 16bit, but that is probably the only place everyone already did anyway.
Old 22nd February 2017
  #325
Gear Maniac
 

11 pages later.......

Old 23rd February 2017
  #326
Lives for gear
 
stinkyfingers's Avatar
 

Troll Alert!

Hi,
I noticed that when I use the gain plug in in Logic that the signal gets truncated right then and there on the channel strip. Should I place a dither plug in before or after the gain plug in ?
Thanks
Old 23rd February 2017
  #327
Airwindows
 
chrisj's Avatar
Quote:
Originally Posted by stinkyfingers View Post
Hi,
I noticed that when I use the gain plug in in Logic that the signal gets truncated right then and there on the channel strip. Should I place a dither plug in before or after the gain plug in ?
Thanks
It does? I haven't noticed that. Are you using Bitter, or some other sort of monitoring plugin that tells you this?

If it IS being truncated (to say 24 bit or 16 bit fixed point) then you'd put a dither before it.

If it's being truncated to 32 bit floating point (which it is, regardless) then no.
Old 23rd February 2017
  #328
Lives for gear
 
stinkyfingers's Avatar
 

Quote:
Originally Posted by chrisj View Post
It does? I haven't noticed that. Are you using Bitter, or some other sort of monitoring plugin that tells you this?

If it IS being truncated (to say 24 bit or 16 bit fixed point) then you'd put a dither before it.

If it's being truncated to 32 bit floating point (which it is, regardless) then no.
ah, i'm being a bit "cheeky" (as they say on peppa pig).

a while back i was testing random things in Logic v9 and noticed that you could generate DC by inserting the gain utility plug-in (visible on Span). more gain = more DC. iirc it was like 8 gain plug-ins stacked maxed out and one more with x amount of gain and you'd get DC at 0 DBFS. a little more testing and yadda yadda yadda it was determined that it was truncating.
it actually turned out to be pretty handy because i used it to generate the DC for my automation tests.
(talk about turning lemons into lemonade goof)

another of the many fascinating things i discovered about Logic (v9) is that when it does truncate, there are even harmonics added with the odd...it has tube truncation, so it's like a tube DAW ! and that could be the main reason why a lot of people hear such a difference with Logic. i know i did.
Old 23rd February 2017
  #329
Airwindows
 
chrisj's Avatar
Quote:
Originally Posted by stinkyfingers View Post
a while back i was testing random things in Logic v9 and noticed that you could generate DC by inserting the gain utility plug-in (visible on Span). more gain = more DC. iirc it was like 8 gain plug-ins stacked maxed out and one more with x amount of gain and you'd get DC at 0 DBFS. a little more testing and yadda yadda yadda it was determined that it was truncating.
it actually turned out to be pretty handy because i used it to generate the DC for my automation tests.
(talk about turning lemons into lemonade goof)


I'm going to make you a DC offset plugin.

Because that is just silly
Old 23rd February 2017
  #330
Lives for gear
 
stinkyfingers's Avatar
 

Quote:
Originally Posted by chrisj View Post


I'm going to make you a DC offset plugin.

Because that is just silly
thanks...that should help my bi-polar disorder.
Logic made me this way, you know...
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