The No.1 Website for Pro Audio
 All  This Thread  Reviews  Gear Database  Gear for sale     Latest  Trending
Do I need SUM? For you guys who mix ITB and OTB...
Old 4th October 2005
  #91
Lives for gear
 
djui5's Avatar
 

Quote:
Originally Posted by RKrizman
Thanks. It's a great read, and I also reviewed that Paul Frindle thread. The only thing I don't understand from Nika's paper is why a compressed signal would have even more danger of overs. It seems to me that if dynamic range of the samples is compressed, then the associated inter-peak overs would also be reduced proportionately.
I wondered the same thing myself. The TL paper speaks of this also. It dosen't make sense, but Nika knows more about the "behind the scenes" working of digital systems than I do...so I'm not one to argue it. Does anyone know? Nika...are you still around here?

Quote:
It's interesting. In the past I had all these debates with people about how I couldn't hear the advantage of using a clean analog summing box versus just mixing ITB. As time went on I realized that I usually recorded and mixed at very conservative levels, and therefore was not generating the problems that might be solved by having the individual tracks reconstructed by the A/D before bing summed.

-R


Yeah
Old 4th October 2005
  #92
Lives for gear
 
gwailoh's Avatar
 

Quote:
Originally Posted by JP11

don't really see why you would need meters everywhere, though...why not just pretend that -6 is 0, and stay safely on the conservative side...when adding a plugin, make sure you're not adding any gain...wouldn't that work fine?
I know this might not seem obvious at first -- it wasn't for me! -- but you want to meter after every plugin because you don't know for sure what the plugin has done to your levels until you meter. Where 'meter' means, meter with an oversampling meter which emulates the reconstruction process of your downstream DAC.

The key to the whole thing -- for me! -- was to understand that your DAW meters measure sample values, but sample values aren't the same thing as the reconstructed signal generated by the DAC. In other words, your DAW meters lie. This is the issue in a nutshell.
Old 4th October 2005
  #93
Lives for gear
 
gwailoh's Avatar
 

Quote:
Originally Posted by RKrizman
Why 6 db and relative to what?
-6db was suggested by Paul Frindle, in one of his super-generous posts on PSW, as a work-around for lack of oversampling meters. The 'proper' solution is to meter with an oversampling meter after every processing stage, meaning in practice every plugin. But if you haven't got an os metering plug, the workaround is to insert the trim plug first, set it to -6db, then always keep your DAW meters in the green, not yellow, definitely not red. If you do all of this then you're 'probably' safe. My suggestion after experimenting with the meters is that you might be a little safer at -8db. YMMV.
Old 4th October 2005
  #94
Lives for gear
 
gwailoh's Avatar
 

Quote:
Originally Posted by djui5
RK,

So where are these meters? Is anyone making them? Does the waves meter work like this?
The list so far. If anybody knows of more, much appreciated for the lead:

1. TL MasterMeter
2. Sony Oxford Limiter
3. TCE 6000 MD3 algorithm
4. TCE MD3 PowerCore plugin (TDM forthcoming)

That's all, folks!
Old 4th October 2005
  #95
Lives for gear
 
gwailoh's Avatar
 

Quote:
Originally Posted by RKrizman
The only thing I don't understand from Nika's paper is why a compressed signal would have even more danger of overs. It seems to me that if dynamic range of the samples is compressed, then the associated inter-peak overs would also be reduced proportionately.
-R
I assumed he meant compressed-then-turned-up, for instance with makeup gain, as many people tend to do. In that scenario you'd be in even greater danger of producing legal sample values which are too hot for the DAC downstream.
Old 4th October 2005
  #96
Lives for gear
Quote:
Originally Posted by JP11
don't really see why you would need meters everywhere, though...why not just pretend that -6 is 0, and stay safely on the conservative side...when adding a plugin, make sure you're not adding any gain...wouldn't that work fine?
with meters inserted it's easy to see that the 6dB 'headroom' is easily lost by adding or subtracting EQ. so we can't apply the same rules we're used to in analogland.

if you produce intersample overs in this fashion time and time again w/different plugs in a dense mix, it's starts to sound like crap. hence the 'digital summing' deal. i'm not sure i won't opt for an analog board in the end but whatever i do, i'll be dealing with things ITB differently than i did before; that's for sure, and if i do go for an analog console it'll be for different reasons.
Old 4th October 2005
  #97
Lives for gear
 
djui5's Avatar
 

Quote:
Originally Posted by gwailoh
I assumed he meant compressed-then-turned-up, for instance with makeup gain, as many people tend to do. In that scenario you'd be in even greater danger of producing legal sample values which are too hot for the DAC downstream.


That would make sense.
Old 4th October 2005
  #98
Lives for gear
bottom line is an oversampling meter will clearly delineate what to do and what not to do, and where the problem may lie. it's like a diabetic that wants to guess his sugar levels without measuring them.

AFAIK the problem is incremental. the less it happens, the better the mixes are likely to sound. i'm sure after awhile one would be able to ballpark in order to stay out of trouble but until i get to that stage, i'm using the meter!
Old 4th October 2005
  #99
Lives for gear
 

Beyond all the technical "mumbo jumbo" summing didnt make that much of a difference for me and none after mastering.
Old 4th October 2005
  #100
Lives for gear
 

I could see if you're mastering how meters would be crucial...Voxengo's Elephant deals with intersample peaking...
Old 4th October 2005
  #101
Lives for gear
 
Stick's Avatar
 

Quote:
Originally Posted by theblue1
You can't worm out of it that easy.


heh
Well, after seeing where this thread has gone, looks like I can, in fact worm out of it!

Generally, I've been mixing to a lower level on my master fader for awhile now, but this thread is making me realize I need to be more careful about what's happening plug to plug on each individual channel. I don't generally record super hot anyway, but sometimes I push plugs hard thinking I'll get more effect out of it... for example, Phoenix works harder the farther you push it, but according to this thread, there's probably a diminishing return...

Regardless, great discussion guys... keep it up.
Old 4th October 2005
  #102
Lives for gear
 
gwailoh's Avatar
 

Quote:
Originally Posted by Stick
I don't generally record super hot anyway
One tip from Paul Frindle's posts on PSW is that it's ok, even good to record hot, so long as you stay below clipping. If it hasn't clipped, the signal from the ADC can always be assumed to be legal. The only issue is, what happens when you start processing that legal signal. The fact that your DAW meters lie makes it easy to produce sample values which will be translated into illegal signals downstream by the DAC. I hope this distinction is clear!

So, after digesting all of this, my tracking habits have remained the same. I still try to record pretty hot, below clipping. Then insert the trim plug on every track and knock it down -6db to -8db for processing. Then monitor every processing stage with an oversampling meter. Then monintor the master stereo bus with an oversampling meter. Then give it to the mastering engineer to make it loud. The result has been consistently more focused stereo imaging and punchier bottom end.
Old 4th October 2005
  #103
Lives for gear
 

Now wouldn´t recording hot but reducing the level 6 db pre fade be the same as recording 6db lower?

If you turn down the level 6db you´ll lose 1bit, it seems like it doesn´t matter where it´s lost, but record at alower level would prevent the use of extra plugins.
Old 4th October 2005
  #104
Lives for gear
 
gwailoh's Avatar
 

Yes I think so. But, you might have artistic reasons for wanting to drive the input. I have API pre's for instance and they sound fatter when you hit them hard.
Old 4th October 2005
  #105
Lives for gear
 
djui5's Avatar
 

What would be really nice, and help out some, is if


DIGIDESIGN PUT F"N NUMBERS ON THEIR METERS


Old 4th October 2005
  #106
Gear Maniac
 

Quote:
Originally Posted by Fredrik
Now wouldn´t recording hot but reducing the level 6 db pre fade be the same as recording 6db lower?

If you turn down the level 6db you´ll lose 1bit, it seems like it doesn´t matter where it´s lost, but record at alower level would prevent the use of extra plugins.
Not exactly... in tracking you always want to hit each part of your chain optimally, which with *some* ADCs means staying 6 dB below clipping. But once the signal is encoded, *good* trimming software allows you to keep the code very pristine by adding all those extra bits to avoid processing-math errors. I mean, that's why software ads tout "double precision" (i.e. 48 bits) or other huge bit depths. Still, too many plugs automatically dither or truncate back to 24 bits for compatability, so there's a consideration.
Old 4th October 2005
  #107
Lives for gear
 

Quote:
Originally Posted by Samuel L. Lord
Not exactly... in tracking you always want to hit each part of your chain optimally, which with *some* ADCs means staying 6 dB below clipping. But once the signal is encoded, *good* trimming software allows you to keep the code very pristine by adding all those extra bits to avoid processing-math errors. I mean, that's why software ads tout "double precision" (i.e. 48 bits) or other huge bit depths. Still, too many plugs automatically dither or truncate back to 24 bits for compatability, so there's a consideration.
Ok I now about the double precision part. My point is since 6db = 1bit you will lose that bit regardles if you record 6db lower or take the signal down in the daw.
Old 4th October 2005
  #108
Lives for gear
 

Quote:
Originally Posted by gwailoh
Yes I think so. But, you might have artistic reasons for wanting to drive the input. I have API pre's for instance and they sound fatter when you hit them hard.
You could pad the signal and still retain the sound of the preamp.
Old 4th October 2005
  #109
Lives for gear
 
gsharp's Avatar
 

Quote:
Originally Posted by Fredrik
Ok I now about the double precision part. My point is since 6db = 1bit you will lose that bit regardles if you record 6db lower or take the signal down in the daw.
You'll lose the bit on the way in only if your source requires a full 24 bits (144dB dynamic range) to accurately and correctly represent it. If your source has, say, 75 dB of dynamic range you're not losing anything at all by recording WELL lower than clipping.

You don't lose any bits at all pulling the fader down 6dB in the DAW. That's the point of the 48 bit mixer. Think of those extra 24 bits (48-24) as a cushion around the data. You can move your 24bit audio up and down in level without losing anything. In ProTools for example you can pull your fader down to about -96 before you start lopping anything off the bottom of your data.
Old 4th October 2005
  #110
Lives for gear
 

Quote:
Originally Posted by gsharp
Think of those extra 24 bits (48-24) as a cushion around the data. You can move your 24bit audio up and down in level without losing anything.

Well everything that is done inside the daw will affect the data, adding 24 bit in the digital domain wont give anything that wasn´t there, but will give higher precision in the calculations done in the daw. I see it like this i you have a digital picture 320*200 pixels and upsample that to 640*400 the picture will look exactly the same, but if changes are being made at this stage they will be at double precision.
Old 4th October 2005
  #111
Lives for gear
 
gsharp's Avatar
 

I was just referring to the (erroneous) idea that pulling the fader down causes a loss of bits.
Old 4th October 2005
  #112
Lives for gear
 
RKrizman's Avatar
 

Quote:
Originally Posted by gwailoh
I assumed he meant compressed-then-turned-up, for instance with makeup gain, as many people tend to do. In that scenario you'd be in even greater danger of producing legal sample values which are too hot for the DAC downstream.
The graph with the white paper showed the signal as compressed without being volume compensated.

-R
Old 4th October 2005
  #113
Lives for gear
 
RKrizman's Avatar
 

Quote:
Originally Posted by Fredrik
If you turn down the level 6db you´ll lose 1bit,
My understanding is that, at least in Protools, that doesn't happen. The 24 bit wide signal just gets shifted lower but retains its resolution.

-R
Old 4th October 2005
  #114
Lives for gear
 

Quote:
Originally Posted by RKrizman
My understanding is that, at least in Protools, that doesn't happen. The 24 bit wide signal just gets shifted lower but retains its resolution.

-R
Perhaps this is true, this is all very confusing times for a AE.

Would this be special to protools HD then or would it also apply to 32bit floatingpoint?

Also are there some links that could explain this concept of retaining the bits while lowereing the volume?
Old 4th October 2005
  #115
Lives for gear
 
RKrizman's Avatar
 

Quote:
Originally Posted by Fredrik
Perhaps this is true, this is all very confusing times for a AE.

Would this be special to protools HD then or would it also apply to 32bit floatingpoint?

Also are there some links that could explain this concept of retaining the bits while lowereing the volume?
It might be in Stan Coty's white paper at the DUC (sorry, no link).

-R
Old 4th October 2005
  #116
Lives for gear
Quote:
Originally Posted by Fredrik
Would this be special to protools HD then or would it also apply to 32bit floatingpoint?
it is my understanding there is no 'bit loss' with either 48 double precision (PT) or 32 bit floating point, unless the faders are pulled down by an extreme amount. if you pull the faders down by 40, 50, 60 dBs, no problem.

someone please correct me if i'm wrong.
Old 5th October 2005
  #117
Lives for gear
 
djui5's Avatar
 

Quote:
Originally Posted by Fredrik
Perhaps this is true, this is all very confusing times for a AE.

Would this be special to protools HD then or would it also apply to 32bit floatingpoint?

Also are there some links that could explain this concept of retaining the bits while lowereing the volume?


Like stated earlier, this is the reason for the higher bitdepth mixing busses. To retain your 24 bit's of dynamic range, while allowing you to move the signal up or down in volume.


Imagine an ice cube in a glass of water. Whether that ice cube moves up or down in the glass, it's still the same size and displaces the same amount of water. 24bit's is the ice cube, the rest is the water.
Old 5th October 2005
  #118
One with big hooves
 
Jay Kahrs's Avatar
Quote:
Originally Posted by djui5
What would be really nice, and help out some, is if


DIGIDESIGN PUT F"N NUMBERS ON THEIR METERS
Why on Earth would they want to do that?

Who needs references? Once it's in PT it's perfect, 'ya just can't screw it up...'cause its' like Pro Toolz 'ya dig?

Ok, someone gag me with a putty knife.

FWIW, I plunged today and picked up Sound Forge 8 (with CD Architect 5.2) and the metering is really great. Totally scalable and there's both PPM & VU style with peak & valley holds!!! If a $300 piece of software can have that then there's no reason the $3000 piece shouldn't.

More seriously, 'bout that white paper....(which I haven't read lol)

Quote:
What is basically says is that due to sampling of A/D converters, your signal can be up to 6db louder at peaks than shown on your meter. Your D/A converters compensate for this so you get proper metering on the output. The problem is when mixing ITB, plug-ins can be clipped without seeing it on meters. Have you ever had a plug-in clip even though you keep the signal within proper range? I have..and it's really really f'n annoying. I always wondered why it did that, and always figured the signal was too fast for the meters, but now I know. It's due to sampling of the A/D converters.
I'm totally with you until the very last sentence or two.

Sure, an analog meter will miss all kinds of peaks because it's ballistics aren't fast enough to catch 'em...but a good digital PPM should show all the peaks no matter how fast they are right? Or are we talking about the leading edge of something with a sharp transient like a snare drum...which a PPM meter should still show. And besides, once the signal has passed through the A/D and is in the digital domain then things like proper, realistic metering should be easy.

Right? (???)

You can only push things so far in either analog or digital before you run out of steam and it starts to sound like poo. The real difference is that when analog distorts it compresses a little and can sound "rock & roll" before it turns to total mush. When digital hits the wall...it's literally a wall and it's pretty damn unforgiving. Maybe what we need are real dummy lights, stuff that lights up red when the signal is 3dB away from the point of clipping. I changed the meter colors (if supported) on any software I own to reflect that and my all my gear, analog and digital is much happier.
Old 5th October 2005
  #119
hey I tried to download the white paper ...
link is DEAD
Old 5th October 2005
  #120
Lives for gear
 
gwailoh's Avatar
 

Quote:
Originally Posted by Jay Kahrs

Right? (???)
No! Sorry! The whole issue is the difference between sample values and signal as reconstructed by the DAC. They're not the same. The DAW meters measure the former, which is why they lie. To approximate the latter you must have an oversampling meter which emulates the DAC's reconstruction process. That means TL MasterMeter, Oxford Limiter, or TCE.

Sorry that there's no shortcut for reading and understanding that white paper, and the PSW thread with several contributions by Paul Frindle. Links are above in this thread. I know everybody's impatient but the read is well worth your effort.
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