Gyraf SSL Brickwall Limiter Modifications?
- Thread starter Gearsix
- Start date
Help Support GroupDIY Audio Forum:
Aleguitarpro
Well-known member
EDITED: You can easily add a passive clipper to gssl made with zeners directly soldered at the output of your gssl.
Majestic12
Well-known member
Aleguitarpro said:
That's a clipping circuit, not a brickwall limiter.
Aleguitarpro
Well-known member
Majestic12 said:
Yes, you'll right.
It's a clipping circuit easy to implement in a GSSL.
I think that for "something similar" Gearsix meant something to use to prevent clipping the output.
RuudNL
Well-known member
I think you will be unable to construct a limiter with an attack time of 'zero milliseconds'.
In broadcast limiters, usually a combined approach is used: A limiter with a reasonable fast attack, and a clipper/filter that only cuts the overshoots of the limiter. In practice, these overshoots are so short, that you don't get serious distortion because of the clipping.
In broadcast limiters, usually a combined approach is used: A limiter with a reasonable fast attack, and a clipper/filter that only cuts the overshoots of the limiter. In practice, these overshoots are so short, that you don't get serious distortion because of the clipping.
dirtyhanfri
Well-known member
What about copying the signal in the DAW and then send it to the sidechain input of your gssl? Attack at it's fastest setting and apply negative delay (sorry, I don't find the english word) to the copy in daw to compensate the attack time.
I did it a few times, just as a demonstration, not sure how it really works in a mastering task.
Hope it helps..
I did it a few times, just as a demonstration, not sure how it really works in a mastering task.
Hope it helps..
Thank you to all!
I posted the question because I consider the GSSL a fantastic compressor and it would be great to use it to obtain in analogue domain something similar to what makes Waves L2 and other plugins in digital domain
I posted the question because I consider the GSSL a fantastic compressor and it would be great to use it to obtain in analogue domain something similar to what makes Waves L2 and other plugins in digital domain
pucho812
Well-known member
I don'[t think there is any analog dynamic unit DIY or otherwise that is a true brickwall limiter. Pendulum audio pl-2 comes cvery close and is advertised as such...
jensenmann
Well-known member
pucho812 said:
There is one. It´s the EMT266 transient limiter. Audio is delayed 300us through an allpass filter while the sidechain gets it´s signal straight from the input without delay. Time enough for the sidechain to do it´s work until any transients arrive at the VCA/transistor array.
Everytime I mention this it reminds me on how many great inventions in audiobusiness actually came out of EMT.
SSLtech
Well-known member
Think about it. There have been many; just different approaches. The FCC has fines for overmodulation, and there's money in getting maximum level.pucho812 said:
dbx have built several. Clipping circuit was incorporated to catch brief 'overs'.
Brickwall is not difficult in analog. Many ways depending on how hard (in level and time) your brick must be.
But the SSL is an *RMS* detector. Not peak. If it does the right thing on rounded waves, it must under-react to peaks of spike waves. -----wrong.....
This isn't even new. Go back a few years dBx was selling a brickwall RMS limiter. First the classic dBx RMS limiter. After that a *fast* peak limiter (somewhat like a Fairchild's fastest setting, but faster, and just a couple chips). Then because they didn't want customers to see ANY over-spikes, a straight clipper. All in one box.
When set up right the RMS stage did the vast majority of the work. Low-duty spikes that didn't RMS-up but had high peaks were whacked-down and released quickly; this would be semi-inaudible. And anything which got by that got sliced-off so your OVER light never lit. Since the RMS section kept the overall average under control with some emphasis on broad peaks, the later two brutal stages would normally hardly-ever happen, and then too quickly to hear.
THAT corp's dBx-heritage systems have become so pervasive that I am not sure what modern good limiter does peaks fast and well.
The clipper is just a guitar fuzz, flattened and scaled to line level.
And even my analog mind is accepting putting it in the PC where we can "look-ahead" and plan a squashing strategy before the peaks happen, instead of reacting in real-time with no information about what is about to come. (That's for recording. Live Sound is different, but rarely very adverse to occasional OVERs.)
This isn't even new. Go back a few years dBx was selling a brickwall RMS limiter. First the classic dBx RMS limiter. After that a *fast* peak limiter (somewhat like a Fairchild's fastest setting, but faster, and just a couple chips). Then because they didn't want customers to see ANY over-spikes, a straight clipper. All in one box.
When set up right the RMS stage did the vast majority of the work. Low-duty spikes that didn't RMS-up but had high peaks were whacked-down and released quickly; this would be semi-inaudible. And anything which got by that got sliced-off so your OVER light never lit. Since the RMS section kept the overall average under control with some emphasis on broad peaks, the later two brutal stages would normally hardly-ever happen, and then too quickly to hear.
THAT corp's dBx-heritage systems have become so pervasive that I am not sure what modern good limiter does peaks fast and well.
The clipper is just a guitar fuzz, flattened and scaled to line level.
And even my analog mind is accepting putting it in the PC where we can "look-ahead" and plan a squashing strategy before the peaks happen, instead of reacting in real-time with no information about what is about to come. (That's for recording. Live Sound is different, but rarely very adverse to occasional OVERs.)
jensenmann said:
There are more ... BBC also had zero overshot limiters.
I own 4 units, AM6/??. Sorry, I don't remember the exact numbers. Again the main path is delayed around 330usecs. I've pushed these HARD on room mics and peaks NEVER go past their set amount.
Michael
PRR said:
You are the first person I've heard say this; every other mention I've seen refers to it as peak detection, and that's what it looks like to my amateur eyes. Yes, it has slow settings that don't seem so 'peak', but it appears to be. It acts much more like ye ole' vari-mu peak detecting limiter than any DBX RMS product I've ever driven. Love to hear more thoughts.
Agreed almost everything brickwall in analog either used diode clipping or all pass delay as part of the scheme. General Electric had 'brickwall' with all pass delay in the late 1940's.
SSLtech
Well-known member
PRR, I'm reading it as a peak detector...
http://gyraf.dk/gy_pd/ssl/ssl_sch.gif
Looking at the schematic from the sidechain VCA (which has a "louder" volotage added to set the "AC threshold" by effectively varying the sidechain's audio "magnification") -and its accompanying I-to-V op-amp stage- the signal goes through a precision rectifier op-amp stage, and then a final op-amp stage which sets the gain (for ratio purposes) and a secondary DC bias to set the "DC threshold" and then compares the remaining (unipolar after the rectifier) maximum peak excursion against that. -Any peaks which overcome the DC bias (fixed slightly differently for different ratios) will pass through the diode to point 'C' and onwards to the time constant network. -So far there's been no time constants involved, and it's peaks only which have any effect.
-Or that's how I read it.
Mind you... The linked image above is just the GSSL variant... which raises the subject of the immense over-simplification of the original question in the first post: We've been told NOTHING of the requirements of this particular 'brick wall'.
In a multi-channel limiter, do you want the sum total of all 2/4/5/6 channels to be limited, or do you want all channels to be squeezed equally so that no single channel exceeds? -Is it to meet an FM stereo broadcast limitation, where both L+R and L-R must be considered, so that no part of the matrix is overmodulated? (And if neither channel may exceed, you 'd need to have TWO sidechains. -Better buy a Turbo board pronto).
Is it really peak level which counts (as is common when considering delivery medium constraints), or is it RMS (which is far more useful for power limitation, eg physical playback systems)?
Without some answers to clarify the question (if the original poster even fully understands what they want)... then... all in all we're all just another brick in the wall.
http://gyraf.dk/gy_pd/ssl/ssl_sch.gif
Looking at the schematic from the sidechain VCA (which has a "louder" volotage added to set the "AC threshold" by effectively varying the sidechain's audio "magnification") -and its accompanying I-to-V op-amp stage- the signal goes through a precision rectifier op-amp stage, and then a final op-amp stage which sets the gain (for ratio purposes) and a secondary DC bias to set the "DC threshold" and then compares the remaining (unipolar after the rectifier) maximum peak excursion against that. -Any peaks which overcome the DC bias (fixed slightly differently for different ratios) will pass through the diode to point 'C' and onwards to the time constant network. -So far there's been no time constants involved, and it's peaks only which have any effect.
-Or that's how I read it.
Mind you... The linked image above is just the GSSL variant... which raises the subject of the immense over-simplification of the original question in the first post: We've been told NOTHING of the requirements of this particular 'brick wall'.
In a multi-channel limiter, do you want the sum total of all 2/4/5/6 channels to be limited, or do you want all channels to be squeezed equally so that no single channel exceeds? -Is it to meet an FM stereo broadcast limitation, where both L+R and L-R must be considered, so that no part of the matrix is overmodulated? (And if neither channel may exceed, you 'd need to have TWO sidechains. -Better buy a Turbo board pronto).
Is it really peak level which counts (as is common when considering delivery medium constraints), or is it RMS (which is far more useful for power limitation, eg physical playback systems)?
Without some answers to clarify the question (if the original poster even fully understands what they want)... then... all in all we're all just another brick in the wall.
I was wrong, you are correct. Peak detect.
The fastest attack is 0.38 milliSecond. Is that not fast enough? Then reduce a 0.47u cap until it is. (You can't go much lower on the 820r resistor until you use something beefier than a '072 to drive it.)
Is it a problem that sub-mS transients can cause a long-lasting dip in the signal? Sure. But this gets into what we mean by "brick wall". Do we just want to NEVER get tickets for overmodulation? Or is the real point to cream the audio like painting a ceiling smooth and shiny?
Are music consumers so deaf we can't leave a few dB margin? If they are that deaf, how would they know we were OVER?
The fastest attack is 0.38 milliSecond. Is that not fast enough? Then reduce a 0.47u cap until it is. (You can't go much lower on the 820r resistor until you use something beefier than a '072 to drive it.)
Is it a problem that sub-mS transients can cause a long-lasting dip in the signal? Sure. But this gets into what we mean by "brick wall". Do we just want to NEVER get tickets for overmodulation? Or is the real point to cream the audio like painting a ceiling smooth and shiny?
Are music consumers so deaf we can't leave a few dB margin? If they are that deaf, how would they know we were OVER?
Similar threads
