Home » R/E/P Archives » Reason In Audio » Dynamic Range of Plugins and clarity with level changes
| Dynamic Range of Plugins and clarity with level changes [message #63309] |
Tue, 03 May 2005 11:35  |
Timeline
Messages: 215
Registered: November 2004
Location: Arcadia CA |
Active Member |
|
|
24bits? 32bits? 64bits?
When I reduce level to the input of an 1176 analog limiter it stays clear IMO.
When I reduce level to an 1176 plugin it takes clarity away to me.
Is there a way to create plugs that don't do this by design with the current plugin methodology?
Would there be an advantage to having more bits all the way through the path?
Gary Brandt
Timeline
|
|
| |
| Re: Dynamic Range of Plugins and clarity with level changes [message #63507 is a reply to message #63431 ] |
Wed, 04 May 2005 07:37  |
Timeline
Messages: 215
Registered: November 2004
Location: Arcadia CA |
Active Member |
|
|
When you chain plugins there can easily be distortion for instance there are several plugs by MOTU with no gain input or output and some have both.
The point is, I wondered why I hear a severe definition loss when dropping levels from unity on most plugs and wondered if there was a better way to design plug chains to be more like analog.
I remember hearing a 'Sound Stream Digital' format demo years ago that inpressed me as it kept low level bits a priority and handled digital differently.
That would have sounded better where digital plugins and level changes are concerned but the format died.
As it is now it seems that once recorded to Digital, everything would sound its best if NO plugs were used at all and the level came out individual IO's at the exact level they went in.
This just doesn't sound like very good planning and it's no wonder many are going back OTB to do mixing.
Why not change it for the better.
Gary Brandt
Timeline
|
|
| |
| Re: Dynamic Range of Plugins and clarity with level changes [message #63560 is a reply to message #63546 ] |
Wed, 04 May 2005 10:20 |
 captain54
Messages: 17
Registered: July 2004
Location: Chicago |
Member |
|
|
| blairl wrote on Wed, 04 May 2005 09:41 |
| Timeline wrote on Wed, 04 May 2005 06:37 |
I remember hearing a 'Sound Stream Digital' format demo years ago that inpressed me as it kept low level bits a priority and handled digital differently.
|
.
As a side note on your question: Not all plug-ins are created equal. Some plug-ins do not apply dither after processing and some do. Some plug-ins process at 24 bits, some at 32 bits and some at 48 bits depending on the host application and hardware. Technically, for a clean and accurate signal, dither should be used any time quantization occurs after a digital signal process.
|
ok, I follow you completely...now here's where I'm confused...what is the exact definition of "quantization"?
1)if you are running a session with a combination of 24 bit files and say, 16 bit loops, are you running the risk of degrading the 16 bit files when a applying a 16 bit dither to the master bus upon mixdown??
2)if you are applying processing (Lin band EQ/Limiter, for example) across a 16 bit submix, then is it necessary to dither upon mixdown? do any digital calculations to a 16 bit file automatically make it a 24 bit file??
|
|
|
| Re: Dynamic Range of Plugins and clarity with level changes [message #63572 is a reply to message #63309 ] |
Wed, 04 May 2005 10:35 |
maxdimario
Messages: 3811
Registered: December 2004 |
Platinum Member |
|
|
I get the same results.
analog attenuators have the same resolution at all levels, digital systems don't.
you can get away from it, hide it, perfect it, but that's the way it is.
better plug ins will work better I guess.
making up for gain in the digital stage is a no-no though.
|
|
|
| Re: Dynamic Range of Plugins and clarity with level changes [message #63577 is a reply to message #63546 ] |
Wed, 04 May 2005 10:38 |
Timeline
Messages: 215
Registered: November 2004
Location: Arcadia CA |
Active Member |
|
|
| Quote: |
As a side note on your question: Not all plug-ins are created equal. Some plug-ins do not apply dither after processing and some do. Some plug-ins process at 24 bits, some at 32 bits and some at 48 bits depending on the host application and hardware. Technically, for a clean and accurate signal, dither should be used any time quantization occurs after a digital signal process.
|
That's very interesting. So it was a similar digital format as todays?
You mention some plugs use 48bit. This might be the reason I hear some plugs handle the level issues better than other.
Starting to make sense.
Thanks
Gary Brandt
Timeline
|
|
|
| Re: Dynamic Range of Plugins and clarity with level changes [message #63585 is a reply to message #63560 ] |
Wed, 04 May 2005 11:00 |
 |
chrisj
Messages: 959
Registered: April 2004
Location: Vermont |
Gold Member |
|
|
No and yes. If you have a mixdown with 16 bit files in it, you don't have '16 bit files', you have a MIXDOWN. Unless your (already dithered?) 16 bit file is being passed through to the final mix with no gain or any other change at all (and sure that is possible- but with no other sounds happening at the same time?) you must dither, or get quantization distortion. Which can be faint to incredibly faint but ALWAYS sounds awful.
If you're doing any processing that involves changing the shape or amplitude of the 16 bit waveform then it is no longer the same 16 bit waveform- it's something else, which can be described with higher word length to whatever degree of accuracy you want, and dithered down to 16 bit again.
theoretical meltdown, ENGAGE...
Digital exception- neatly doubling or halving the volume. This equates to a bit shift in linear encoded PCM fixed point digital audio- increasing volume (if you don't clip) gets EXACTLY the same information as before, halving volume loses you the least significant bit but I don't think dither would actually gain you anything.
Hmmmm... which would mean logically that you could keep doing it- like having a high-res sinewave and continually bit-shifting, and it would still sound right. I don't think that's correct, so doh! I was wrong about the halving.
...and doubling the volume that way would exactly correspond to the same information, but I'm wondering if that's really a better sound than having one additional bit with dither noisefloor. Again with the reductio absurdum- if you kept on doing it, without dither, you'd get exactly the same you started with, but if you were starting from very low levels it would sound real grainy. If you dithered, it would sound like noise. So far so good.
But PROPER dither, namely TPDF applied in exactly the right way, makes this noise indistinguishable from high res noise anyway. I've done the double blind test on it and I can tell raw noise from quantized raw noise but NOT if the noise is dithered before being quantized (this is also an argument against 'self dithering', you gotta dither even if your sound IS pure noise itself). So if your low res sound being amplified over and over with bit shifts is itself dithered you gain nothing from adding extra dither each bit shift, and if it isn't dithered you can't help it that way either...
Okay, okay, I'll stop! *whew* damn. Anyway- executive summary is, you have to dither if you are DOING anything. The one narrow little exception is, you can always neatly double the gain of a digital file, do nothing else, and get away with it, because of the properties of how bits work. If you then mix using that file, you still need to dither. If you have the bit-identical file playing through to something, you don't dither the pass-through. As soon as there is any other sound changing the 'mix' bits by overlaying another sound onto them, you dither.
I'm not EVEN going to get into how equalization can produce sample sequences that can't exist in a directly sampled recording done using properly designed converters  that's more in the frequency domain, anyway, and all that you're talking about is in the resolution domain...
airwindows AU plugins | airwindows mastering
|
|
|
| Re: Dynamic Range of Plugins and clarity with level changes [message #63602 is a reply to message #63309 ] |
Wed, 04 May 2005 12:12 |
Timeline
Messages: 215
Registered: November 2004
Location: Arcadia CA |
Active Member |
|
|
OK chris, I think I understand.
When halving the level? Would not the lowest level hi frequencies be effected and drop out of the signal path changing the sound dramatically? Raising gain seems like a better choice in that low level high frequencies are at least in an accessible range. Changing levels on faders in a DAW mix I would think still have that same issue compared to analog, as was mentioned in an earlier post.
I wonder if some sort of DAW pre-emphasis would be a good idea within the DAW like on the early Sony Dash machines? Maybe not as dramatic but somewhere above 6k or so. If the post fader OP's or group pre-faders or send OPs of the DAW reversed the emphasis I wonder what that would sound like.
It would obviously need more processing and CPU but I wonder what it would sound like on a limited level.
Maybe one could then increase the headroom of the plugs slightly too and avoid overload in a chain of plugs while still keeping the low level HF stuff clear, or would this help at all?
Gary Brandt
Timeline
|
|
| |
| Re: Dynamic Range of Plugins and clarity with level changes [message #63705 is a reply to message #63611 ] |
Wed, 04 May 2005 21:07 |
captain54
Messages: 17
Registered: July 2004
Location: Chicago |
Member |
|
|
| blairl wrote on Wed, 04 May 2005 12:34 |
| captain54 wrote on Wed, 04 May 2005 09:20 |
ok, I follow you completely...now here's where I'm confused...what is the exact definition of "quantization"?
1)if you are running a session with a combination of 24 bit files and say, 16 bit loops, are you running the risk of degrading the 16 bit files when a applying a 16 bit dither to the master bus upon mixdown??
2)if you are applying processing (Lin band EQ/Limiter, for example) across a 16 bit submix, then is it necessary to dither upon mixdown? do any digital calculations to a 16 bit file automatically make it a 24 bit file??
|
Here's a link to a paper that explains dither in great detail. I think most questions you have will be answered here:
http://www.cadenzarecording.com/dither.html
|
very informative....thanks....if I read it correctly, it answers my questions as follows...
1)16 bit loop mixed in a session with other 24 bit files technically becomes becomes a 48 bit session when any type of processing is applied....if the entire mix is bounced as a 24 bit file, a 24 bit dither is applied...
2)if you are applying processing across the bus of a 16 bit submix, you are then in theory converting it to 48 bit by the very fact that you are processing it...if you bounce this file down to 16 bits, yes, you must apply a 16 bit dither....
|
|
|
| Re: Dynamic Range of Plugins and clarity with level changes [message #63727 is a reply to message #63572 ] |
Thu, 05 May 2005 00:05 |
 |
Ronny
Messages: 2739
Registered: April 2004
Location: Brunswick, Georgia |
Platinum Member |
|
|
| maxdimario wrote on Wed, 04 May 2005 11:35 |
I get the same results.
analog attenuators have the same resolution at all levels, digital systems don't.
you can get away from it, hide it, perfect it, but that's the way it is.
better plug ins will work better I guess.
making up for gain in the digital stage is a no-no though.
|
Setting analog mic pre's, mixers etc. to output at their optimal operating range and making up the gain in the digital realm is the way to go. Digital only knows gain as ones and zeros and from my tests and my experience there is no degradation when you change gain in the digital realm when gain varies by 20dB or less. Don't be fooled by thinking that you have to squeeze every bit out of a 24 bit system, you don't need to. There is much misinformation about losing sonics when you drop resolution and there isn't any basis for it. I'll explain. Some of the bit squeezers think that when you record a signal at -12dB into 24 bit, that you lose 2 bits of resolution and that this corresponds to a loss of signal quality. Not true, here's why. When you record into a 24 bit system at -12dB, yes you are utilizing only 22 bits, BUT it only takes 22 bits to represent that signal the same as when it's going full scale at -0dB and 24 bits are utilized. For example on a live orchestra recording where you have the conductor at soundcheck perform the loudest section so that you can set input levels, the soundcheck doesn't always give precise input levels to go by, often the excitement of the performance in front of the audience and longer build to the louder sections in the actual concert, will result in hotter levels coming from the instruments. It's by far better to give yourself more headroom by lowering the digital input gain a bit and than once the signal is digitized raise gain after the recording, than it is to risk overs that may cause irreparable distortion. Think about it, if you are recording peak at -0dB the RMS may be -15dB. Raw music is dynamical, gain is constantly rising and falling, when the lower sections go down to -24dB, you do not hear a lose in signal integrity, although the -24dB lower section is only utilizng 20 bits. Understand how it works? Less gain, less bits needed to capture and reproduce that signal at the same audible resolution as full scale peak. I guarantee you that no one can hear a difference if you record the same source peak at -0dB and also peak at -12dB and than raise gain digitally on the -12dB up to -0dB. I've done several blind tests with other engineers, producers and accomplished musicians and believe me on gear that is operating correctly and playing back through the same DAC, no one can pick the -0dB original over the recorded at -12dB peak and than gain raised to -0dB example. Don't take my word for it, run your own tests. Do it in the blind. If someone hears a difference and can pick between the examples every time, than you need to recheck your test parameters.
------Ronny Morris - Digitak Mastering------
---------http://digitakmastering.com---------
----------Powered By Experience-------------
-------------Driven To Perfection---------------
|
|
| |
| Re: Dynamic Range of Plugins and clarity with level changes [message #63907 is a reply to message #63843 ] |
Thu, 05 May 2005 18:38 |
|
Ronny
Messages: 2739
Registered: April 2004
Location: Brunswick, Georgia |
Platinum Member |
|
|
| Loco wrote on Thu, 05 May 2005 13:37 |
| Timeline wrote on Wed, 04 May 2005 08:37 |
When you chain plugins there can easily be distortion for instance there are several plugs by MOTU with no gain input or output and some have both.
|
Huh? What are you trying to say? On MOTU land there's no such thing as clipping on plugins or busses even if their meters are above zero, because they receive and send 32 bits all the way through the master fader where everything must go 24 or 16 bit.
|
That's the beauty of 32 float, you can exceed -0dBFs, you can measure and analyze above -0dBFs and you won't internally clip at -0dBFs, although there is no reason to process above -0dBFS, if you do have a few overs on multi-tracks, no problem as long as you attenutate the master section output, so that it isn't clipping.
------Ronny Morris - Digitak Mastering------
---------http://digitakmastering.com---------
----------Powered By Experience-------------
-------------Driven To Perfection---------------
|
|
| |
| Re: Dynamic Range of Plugins and clarity with level changes [message #63915 is a reply to message #63909 ] |
Thu, 05 May 2005 20:34 |
Timeline
Messages: 215
Registered: November 2004
Location: Arcadia CA |
Active Member |
|
|
Thanks Bob. That's what I was trying to say actually.
Was starting to feel kind of alone out here.
My descriptions could likely use some work.
|
|
|
Goto Forum:
Current Time: Thu Sep 2 13:19:44 CDT 2010
Total time taken to generate the page: 0.03126 seconds |
Members
Search
FAQ
Register
Login
Home
