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
Q for Paul Frindle Dynamics Plugins
Old 24th October 2006
  #121
Lives for gear
 
abit's Avatar
 

Paul, thanx for doing this.
Great stuff, informative.
Pleasure to read.
thumbsup
Old 24th October 2006
  #122
Lives for gear
 
Nordenstam's Avatar
 

Verified Member
A few more Q's, if you have the time to answer. =)


Regarding hidden overs, peak limiters and coding in general:

Was pondering the subject of, quote, "intersample auto-correct function". Wondered if you could shed some light on how it's implemented? Oversampled sidechain?

The only way to do it with my available tools is to oversample all the processing in the peak limiter, which works well. Found that at 8x, the sample peak and "real peak" dB level where correct to a few hundreth dB's. (Voxengo Elephant limiter, R8Brain SRC and RME digicheck meter) This signal seems to keep peak level much the same when downsampled to 44.1 too. 4x oversampling usually does the job as well. Is there a generally rule to how much oversampling is needed to avoid hidden overs?

Oversampling also helps supress aliasing harmonics. Are there other less CPU intensive ways to supress aliasing? Do you feel it's worth a cycle of (good) up- and down-sampling to higher rates in the limiter to avoid aliasing distortion?

Last Q, if you happen to know: MP3'ing changes peak levels like SRC does, especially at lower rates. The decoder synthesis often clips, even though the input samples are all below 0dBFS. Is there perhaps some handy rules to apply level vise to be safe from the overs in the lossy coding many consumers applies? Like sub-200kBPS max peak at -1dBFS, sub-100kBPS at least -1.5dBFS, sub-50kBPS at least -2dBFS and so forth. <- These where random number picked for the example, but I got a feeling there may be some general rule like that, that may apply to lossy coding. Would be very useful! Any other way than lowering gain to avoid this sort of clipping? Mucho oversampling? Haven't had time/energy to give this enough testing yet, would be interesting to know if you happen to have any ideas on this subject and/or could tip on any sources to look up! Not that Mp3 is so terribly exciting, but it does help to know what happens in the average consumer listening environment.


Thanks,

Andreas N
Old 24th October 2006
  #123
M2E
Lives for gear
 
M2E's Avatar
 

Hi Paul,

Thanx for the incredible answers. You've been really opening my eyes to a lot.

You mentioned that when moving the effects slider on the Inflator that it controls the warmth of the plugin but, what if I totally turn it down and raise the input?
Will it just be considered at this point a fader? Also if the effect fader is down the the curve do anything at this point?

Also, maybe an idea for you as I haven't seen is a Pre Amp Plugin. Since 90% of people complain about the warmth of digital, how come no-one hasn't tried this? Is it that hard to do?
Also since the latency isn't a factor then, it sounds like this would be the next major plugin.
What would be the downside to that?

Thanx again Paul,

M2E
Old 28th October 2006
  #124
Lives for gear
 

Quote:
Originally Posted by M2E View Post
Hi Paul,

Thanx for the incredible answers. You've been really opening my eyes to a lot.

You mentioned that when moving the effects slider on the Inflator that it controls the warmth of the plugin but, what if I totally turn it down and raise the input?
Will it just be considered at this point a fader? Also if the effect fader is down the the curve do anything at this point?
Thanks for the encouragement - late response is because I have been away..

Raising the input with the effect down will operate like a fader and do nothng interesting.

Also with the effect fader down the curve won't do anything either cos the effect isn't being sent to the output.

Quote:
Also, maybe an idea for you as I haven't seen is a Pre Amp Plugin. Since 90% of people complain about the warmth of digital, how come no-one hasn't tried this? Is it that hard to do?
Also since the latency isn't a factor then, it sounds like this would be the next major plugin.
What would be the downside to that?
Yes you are right, what's perhaps missing in the digital domain is the character that analogue used to give to the art. Although digital can be transparent and ultra-reliable, what's now needed is to put some controllable artistic character back into it - with the great advantage that it will always be the same, will not age with use and all plugs of the same model would be the same etc. etc..

This is a complex subject - but just to say that much of what is widely considered to have character in the analogue domain actually does not and much of what does have 'character' (or at least distortion) in both the digital and analogue domains really shouldn't. For this reason I don't really subscribe to the idea of making apps that have character and distortion as an almost incidental and originally unwanted side effect.

I consider that artistic character is far too important to leave to chance or bury within another main application trying to do something else entirely... I would want to bring it out into the open in it's own right and make something stand alone where such characters could be generated as a separate issue - to be used wherever and whenever you liked - without fighting with some other main effect.. In other words in my book an amp is something that just makes signals bigger and a compressor should only compress - with full control of any added 'character' etc..

I have a great long list of stuff I would like to make and this kind of thing is on that list too. Why? Because I want it myself :-)

My first departure into this (murky) domain was the inflator and the next so far is the limiter which uses another kind of manipulation. The inflator really is pure distortion designed to sound loud. And the odd parts of the limiter is a different kind of 'distortion' that lets you hear transients - without passing the peak levels of those transients to the output..

But there are a good number of further classes of effect I would like to exploit :-)
Old 28th October 2006
  #125
Lives for gear
 

Quote:
Originally Posted by Lupo View Post
A few more Q's, if you have the time to answer. =)


Regarding hidden overs, peak limiters and coding in general:

Was pondering the subject of, quote, "intersample auto-correct function". Wondered if you could shed some light on how it's implemented? Oversampled sidechain?
In a way it is, but nothing is actually clocking faster as such. It calculates what the peak value for reconstructed signal will be by looking over a running number of passing sample values at base rate and presents this value at the end of each sample. This value is then used to make the correction - not the original samples themselves..

Quote:
The only way to do it with my available tools is to oversample all the processing in the peak limiter, which works well. Found that at 8x, the sample peak and "real peak" dB level where correct to a few hundreth dB's. (Voxengo Elephant limiter, R8Brain SRC and RME digicheck meter) This signal seems to keep peak level much the same when downsampled to 44.1 too. 4x oversampling usually does the job as well. Is there a generally rule to how much oversampling is needed to avoid hidden overs?
There is a complex trade off between accuracy in the pass band, latency and the measured intersample peaking. Sample rate isn't the only factor...

Quote:
Oversampling also helps supress aliasing harmonics. Are there other less CPU intensive ways to supress aliasing? Do you feel it's worth a cycle of (good) up- and down-sampling to higher rates in the limiter to avoid aliasing distortion?
IMVHO it is unnecessary to up/down sample a comp/limiter. It really should be possible to make these work without aliassing directly.. One should also take into account that with today's obessions with peak levels and loudness, running level concious processes at higher rates than the target output rate may cost you some modulation and peak level... The results may not be as you expect!

Talking about non-linear freq aliassing; if the limiting is done incorrectly (i.e. clipping or infinitely short attacks) upsampling will help as it will reduce aliassing. But the only sample rate that will lose it completely is an infinite one! So relying on this alone is not the best approach - the cost gets silly.

Quote:
Last Q, if you happen to know: MP3'ing changes peak levels like SRC does, especially at lower rates. The decoder synthesis often clips, even though the input samples are all below 0dBFS. Is there perhaps some handy rules to apply level vise to be safe from the overs in the lossy coding many consumers applies? Like sub-200kBPS max peak at -1dBFS, sub-100kBPS at least -1.5dBFS, sub-50kBPS at least -2dBFS and so forth. <- These where random number picked for the example, but I got a feeling there may be some general rule like that, that may apply to lossy coding. Would be very useful! Any other way than lowering gain to avoid this sort of clipping? Mucho oversampling? Haven't had time/energy to give this enough testing yet, would be interesting to know if you happen to have any ideas on this subject and/or could tip on any sources to look up! Not that Mp3 is so terribly exciting, but it does help to know what happens in the average consumer listening environment.
There are several mechanisms that could give rise to higher output peak values from different coding mechanisms like MP3, WMA etc.. I think it is far too complex for one to have a cast iron rule of thumb. The only real way is to code and decode it simultaneously and measure the output with a reconstruction meter in real time. However I agree that AFAIK there's no product that allows you to do this in one go - yet?
Old 28th October 2006
  #126
M2E
Lives for gear
 
M2E's Avatar
 

By the way Paul,

I was looking at the Oxford Dyn Comp and saw it had a limiter on it. How different is the limiter on the compressor than the byitself Limiter? Does the compressor have the same sound and feel as the alone Limiter plugin?

Also on the Oxford EQ, what did George Massenberg do with you guys? Did he help with the extra settings or did you guys emulate his 8900EQ? I think that's the version.

Thanx

M2E
Old 28th October 2006
  #127
Gear Maniac
 
Studio Dweller's Avatar
 

Thank you Mr. Frindle

I have been following this thread with great interest. Thank you so much, Paul for the insight and taking the time post all of this. Just great information that needs to get out.

-Larry
...a very happy owner of the Oxford EQ/GML, Inflator and Limiter.
Old 28th October 2006
  #128
Lives for gear
 

Quote:
Originally Posted by M2E View Post
By the way Paul,

I was looking at the Oxford Dyn Comp and saw it had a limiter on it. How different is the limiter on the compressor than the byitself Limiter?
There are some similarities, but the archtecture and gain scalng are different as it is part of a comp/limiter combo app. Also the dynamics does not have any of the enhance, recon meter and auto comp processing. Neither does it have the noise shaping dither processing (although that does not affect the character).

However you can still get great loudness and density by using the dyn comp/limiter with the warmth function - which does actually allow peaks to be heard without causing excessive overshoots if you use it carefully.. The warmth should really have been called an overdrive function..

Quote:
Does the compressor have the same sound and feel as the alone Limiter plugin?
No not really. The comp is a general purpose 'clean' programme and instrument compressor with much more control of many more parameters. It is quite different.

Quote:
Also on the Oxford EQ, what did George Massenberg do with you guys? Did he help with the extra settings or did you guys emulate his 8900EQ? I think that's the version.
I simply emulated the 8900EQ, with all it's control ranges, laws and responses. The only difference between the emulation and the EQ is that the analogue unit has a slight tendancy to roll-off a bit towards 20KHz (around 0.15dB). We emulated that too, but George decided that this was a bad side effect of the analogue circuit and asked for it to be removed.. By the time George arrived in Oxford we had done the emulation and he spent a good few hours A/B testing it with the original under all sort of conditions with programme he had brought with him and declared that he could hear no difference between the emulation and the EQ box using the R3 system. This sparked another interesting conversation (as you might imagine), but this is another subject...

A similar process occured with the emulation of his dynamics - and both were installed in NY for final verification by George, Mick Gazauski and myself, before making a final S/W release...
Old 28th October 2006
  #129
Gear Head
 
Luna Sound's Avatar
 

Hi Paul. I would also like to thank you for your willingness to pass on so much useful information.
Please could you also tell me what's the difference if any between the processing done by the Inflator plugin and that done by the enhance section of the Limiter plugin.
Thanks
Old 28th October 2006
  #130
Lives for gear
 

Quote:
Originally Posted by Luna Sound View Post
Hi Paul. I would also like to thank you for your willingness to pass on so much useful information.
Please could you also tell me what's the difference if any between the processing done by the Inflator plugin and that done by the enhance section of the Limiter plugin.
Thanks
They are very different and sound totally different too. The inflator produces a harmonic profile continuously (i.e. distortion on steady state signal), whereas the enhance is a dynamic process which does not produce distortion on steady signal (even continuous tone). You can try this yourself using the oscillator in PT and just listening to them both :-) Perhaps the best way to describe the enhance is that it is a dynamically complex form of expansion and compression?

None of the plug-ins I have made are subsets or direct cut and pastes of parts of each other. All are different - because otherwise there would be no real point.
Old 28th October 2006
  #131
Quote:
Originally Posted by Paul Frindle View Post
I simply emulated the 8900EQ, with all it's control ranges, laws and responses. The only difference between the emulation and the EQ is that the analogue unit has a slight tendancy to roll-off a bit towards 20KHz (around 0.15dB). We emulated that too, but George decided that this was a bad side effect of the analogue circuit and asked for it to be removed.. By the time George arrived in Oxford we had done the emulation and he spent a good few hours A/B testing it with the original under all sort of conditions with programme he had brought with him and declared that he could hear no difference between the emulation and the EQ box using the R3 system. This sparked another interesting conversation (as you might imagine), but this is another subject...

A similar process occured with the emulation of his dynamics - and both were installed in NY for final verification by George, Mick Gazauski and myself, before making a final S/W release...
Hey Paul,

I've always loved the 8200 emulation. Will the 8900 dynamics emulation see the light of day?
Old 28th October 2006
  #132
Lives for gear
 

Quote:
Originally Posted by ambientdig View Post
Hey Paul,

I've always loved the 8200 emulation. Will the 8900 dynamics emulation see the light of day?
`

I cannot speak for Sony plug-ins as I no longer work for them, but I would say it's unlikely. George has never given permission to release it as a Pro Tools plug-in and I personally don't have any evidence that he will change his mind.

BTW the 8900 emulation is very different to the Oxford dynamics. The architecture and processing is very different and all the controls have very different dependancies and specific interactions. It does not sound the same and one cannot set the Oxford Dyn to make anything like the sound of the 8900, because of fundamental differences in the processing algorithms. It was very considerably more difficult to make than the 8200 EQ!!
Old 29th October 2006
  #133
Gear Maniac
 
knightsy's Avatar
 

Hi Paul,

I saw a trick a few months ago for EQing distorted electric guitars (in the Slipperman Sistorted Guitars thread on the MARSH). Basically it involves working bands of EQ against each other and using the resulting phase anomolies as a sonic tool.

Specifically, one part of it involved working a low pass filter (anywhere from 9k down to even 3k) against a large narrow band bell boost (as much as 15dB at 21k for example). In analog, I can make this technique work even on cheap EQs, but I haven't managed it with a plugin EQ yet... I've tried the Digi EQ III, the Waves Eqs, even the SSL E channel.... no luck. Not even close.

Is it the phase shift that drives this sound? Can it be modelled digitally?
Old 29th October 2006
  #134
Lives for gear
 

Quote:
Originally Posted by knightsy View Post
Hi Paul,

I saw a trick a few months ago for EQing distorted electric guitars (in the Slipperman Sistorted Guitars thread on the MARSH). Basically it involves working bands of EQ against each other and using the resulting phase anomolies as a sonic tool.

Specifically, one part of it involved working a low pass filter (anywhere from 9k down to even 3k) against a large narrow band bell boost (as much as 15dB at 21k for example). In analog, I can make this technique work even on cheap EQs, but I haven't managed it with a plugin EQ yet... I've tried the Digi EQ III, the Waves Eqs, even the SSL E channel.... no luck. Not even close.

Is it the phase shift that drives this sound? Can it be modelled digitally?
I'm not sure i understand this? DO you mean that you put on a whole load of boost then roll it off again afterwards - or is it the other way around? Are any of the EQ sections saturating etc.? If they are not saturating I can't think why a digital EQ would not achieve the same results? IMO it's unlikely to be phase shifting.

What sound are you after, could you post a few seconds of before and after?

But generally rolling off the HF in distorted guitars is a good thing to do as it softens the harsh high order harmonics that the distortion creates. This is in fact much of what guitar speakers do as they rarely reproduce high levels of HF.
Old 29th October 2006
  #135
Lives for gear
Quote:
Originally Posted by Paul Frindle View Post
`

I cannot speak for Sony plug-ins as I no longer work for them, but I would say it's unlikely. George has never given permission to release it as a Pro Tools plug-in and I personally don't have any evidence that he will change his mind.
that's bad news. i was really looking forward to that one.
Old 4th November 2006
  #136
Lives for gear
 
Nordenstam's Avatar
 

Verified Member
Hi!

Thanks for your reply. I understand there's more to clever DSP coding than brute force oversampling. Though that's way beyond me and my current soft and hardware platforms. Oversampling works fairly well in lieu of better options!

Thomas Lund of TC electronics recently published some AES presentation papers that shed light on the topic of intersample peaks and raised peak level through psychoacoustic coding. http://www.tcelectronic.com/Default.asp?Id=9249 The "stop counting samples" .pdf presentation have a table of mp3 overload figures at page 33 and another interesting table for oversampled PPM numbers on page 54. Thought it might be worth a mention in this thread.

Quote:
Originally Posted by Paul Frindle View Post
The only real way is to code and decode it simultaneously and measure the output with a reconstruction meter in real time. However I agree that AFAIK there's no product that allows you to do this in one go - yet?
A plug in for a loop through generic psychoacoustic codecs would be useful!


Regards,

Andreas Nordenstam
Old 6th November 2006
  #137
M2E
Lives for gear
 
M2E's Avatar
 

Hey Paul,

If you have a sec, check this out. He mentioned that there is no way to mix accurate using 32bit 24bit pcm. You have to use 64bit. What do you think about this and his graphs.
http://www.youtube.com/watch?v=-9EeW9WhNWA

Thanx

M2E
Old 7th November 2006
  #138
Motown legend
 
Bob Olhsson's Avatar
 

Verified Member
This is all true however I really don't see how one can assume competing applications have all been crippled to only perform 32 bit float calculations when all modern CPUs come equipped to perform 80 bit extended precision calculation.

I'm a total novice at the math but something sounds a bit fishy here.
Old 7th November 2006
  #139
Lives for gear
 

Quote:
Originally Posted by M2E View Post
Hey Paul,

If you have a sec, check this out. He mentioned that there is no way to mix accurate using 32bit 24bit pcm. You have to use 64bit. What do you think about this and his graphs.
http://www.youtube.com/watch?v=-9EeW9WhNWA

Thanx

M2E
Yes this is correct. If you want to maintain 24 bit accuracy and add 24bit signals together at full level you need slightly more than a 24 bit mantissa.

The reason (very basically) is that the 32bit float will represent a 24bit number all the time - at any scaling, regardless of real world value. So if you scale for 24bits = +/-1 (full level) and the signal gets bigger than that, the forced re-scaling will represent the 24bit mantissa at equivalent to +/-2 (because an exponent has incremented).

This means that back in the real world (your 24bit flat out world) the accuracy has fallen to 23bits effectively - when the signal is scaled back down again, because you lost the original 24bit LSB during the scaling.

Of course you can avoid this by making sure in the design that signals are always scaled to reach less than +/-1, but then you are left with the equivalent of a 24bit fixed point system - without any overload margin above +/-1, or a larger standing noise if you reduce signals so as not to overload when added together.

So essentially float is (wrt the real world) a representation that automatically sacrifices smaller increments as the number gets bigger - allowing an arbitary overload margin ulimately at the expense of real world accuracy. Like anything else - you don't get something for nothing :-)

However this is still very much better than simply allowing the signal to clip!! Remember a vanishingly small error at 23 or 22bits down (at around -130dB) is a whole lot less audible than a hard clip 'splat' at full level!! I doubt very much that anyone would ever hear this issue under any normal conditions....

In fact for a 32bit data system you are better off using fixed point for a mixer, as you can control the scaling yourself with the 8 bits of precision you have over and above the required output bit width..
Old 7th November 2006
  #140
M2E
Lives for gear
 
M2E's Avatar
 

Quote:
Originally Posted by Paul Frindle View Post
Yes this is correct. If you want to maintain 24 bit accuracy and add 24bit signals together at full level you need slightly more than a 24 bit mantissa.

The reason (very basically) is that the 32bit float will represent a 24bit number all the time - at any scaling, regardless of real world value. So if you scale for 24bits = +/-1 (full level) and the signal gets bigger than that, the forced re-scaling will represent the 24bit mantissa at equivalent to +/-2 (because an exponent has incremented).

This means that back in the real world (your 24bit flat out world) the accuracy has fallen to 23bits effectively - when the signal is scaled back down again, because you lost the original 24bit LSB during the scaling.

Of course you can avoid this by making sure in the design that signals are always scaled to reach less than +/-1, but then you are left with the equivalent of a 24bit fixed point system - without any overload margin above +/-1, or a larger standing noise if you reduce signals so as not to overload when added together.

So essentially float is (wrt the real world) a representation that automatically sacrifices smaller increments as the number gets bigger - allowing an arbitary overload margin ulimately at the expense of real world accuracy. Like anything else - you don't get something for nothing :-)

However this is still very much better than simply allowing the signal to clip!! Remember a vanishingly small error at 23 or 22bits down (at around -130dB) is a whole lot less audible than a hard clip 'splat' at full level!! I doubt very much that anyone would ever hear this issue under any normal conditions....

In fact for a 32bit data system you are better off using fixed point for a mixer, as you can control the scaling yourself with the 8 bits of precision you have over and above the required output bit width..
Yeah but if Pro Tools is a 48bit or in my case 40bit internal system, wouldn't that make up for all of the overs and anything else that a 24bit system does? Pro Tools is said to have a 56bit full internal something over all system or something in that realm. They say that Pro Tools stays at 48bit until the final output. Maybe I got this info wrong or something. Just trying to get an understanding about mixing with or without plugins and making my system giving me what I need to get the job done and better. But also not be fooled by the hype and the digital lingo that tricks everyone into buying something that's really not there yet.
So if 64bit is the tell all and it's the only way to get a true 24bit mix, what's the downfall. There's got to be a downfall, there always is. Hahaha

I'm quite sure if 64bit is the tell all. Pro Tools now has a reason to resale new Hardware and an all new system just off those bases alone, and there we go at square one again with having to buy everything all over again because of missed info or held back info just to slowly get people to by 16bit, 20bit, 24bit, 32bit, 48bit and now 64bit.
Where does it really end?

Thanx Paul,

M2E
Old 7th November 2006
  #141
Lives for gear
 

Quote:
Originally Posted by M2E View Post
Yeah but if Pro Tools is a 48bit or in my case 40bit system internal system, wouldn't that make up for all of the overs and anything else that a 24bit system does? Pro Tools is said to have a 56bit over all system or something in that realm. They say that Pro Tools stays at 48bit until the final output. Maybe I got this info wrong or something. Just trying to get an understanding about mixing with or without plugins and making my system is giving me what I need to get the job done and better.
So if 64bit is the tell all and it's the only way to get a true 24bit mix, what's the downfall. There's got to be a downfall, there always is. Hahaha

Thanx Paul,

M2E
The current PT TDM mixer uses a fixed point processor with a 56bit accumulator, which means it can add up as many 24bit signals as you like without ever encountering an error at 24bits :-) Once you have enough accuracy - you have enough, beyond this more doesn't get you anything at all. It works perfectly well - don't waste your money!!

Don't listen to the hype - 64bit float is NOT the only way to get an accurate 24bit mix - of course it isn't. All the guy is saying is that if you MUST use a floating system (because your processors are so designed) you are better off using 64bit than 32bit. So no surprises there hey? Not exactly news to anyone in the know is it?

Except (yep - here comes the hype), he is making far too much of the vanishingly small error associated with 32bit float.. As I said, it is down at around -130dB ref full scale and I'd be astounded if anyone could hear that on real DACs, amplifiers and speakers in a physical room.
Old 8th November 2006
  #142
M2E
Lives for gear
 
M2E's Avatar
 

Quote:
Originally Posted by Paul Frindle View Post
The current PT TDM mixer uses a fixed point processor with a 56bit accumulator, which means it can add up as many 24bit signals as you like without ever encountering an error at 24bits :-) Once you have enough accuracy - you have enough, beyond this more doesn't get you anything at all. It works perfectly well - don't waste your money!!

Don't listen to the hype - 64bit float is NOT the only way to get an accurate 24bit mix - of course it isn't. All the guy is saying is that if you MUST use a floating system (because your processors are so designed) you are better off using 64bit than 32bit. So no surprises there hey? Not exactly news to anyone in the know is it?

Except (yep - here comes the hype), he is making far too much of the vanishingly small error associated with 32bit float.. As I said, it is down at around -130dB ref full scale and I'd be astounded if anyone could hear that on real DACs, amplifiers and speakers in a physical room.

Thanx Paul for your quick answer and for you breaking that down. Urr the greatest!!!

By the way, do you know if my older Pro Tools Mix Plus System has the same 56bit accumulator as the new systems? I think I remember me hearing that about my system.

Thanx Paul

M2Eheh
Old 8th November 2006
  #143
Lives for gear
 

Quote:
Originally Posted by M2E View Post
Thanx Paul for your quick answer and for you breaking that down. Urr the greatest!!!

By the way, do you know if my older Pro Tools Mix Plus System has the same 56bit accumulator as the new systems? I think I remember me hearing that about my system.

Thanx Paul

M2Eheh

It will have the same 56bit accumulator as it uses the same class of processor.

My understanding is that the only issue with early systems was a forced reduction to 24bits that occured when the the mixer got big enough to need to spill over to another processing chip (the native interface of the processors is 24bits). AFAIK this is possibly the case for your system, but I can't remember exactly at which system revision this issue went away and what were the numbers of channels needed for use of more than one chip.

Sorry :-(
Old 8th November 2006
  #144
M2E
Lives for gear
 
M2E's Avatar
 

Quote:
Originally Posted by Paul Frindle View Post
It will have the same 56bit accumulator as it uses the same class of processor.

My understanding is that the only issue with early systems was a forced reduction to 24bits that occured when the the mixer got big enough to need to spill over to another processing chip (the native interface of the processors is 24bits). AFAIK this is possibly the case for your system, but I can't remember exactly at which system revision this issue went away and what were the numbers of channels needed for use of more than one chip.

Sorry :-(
Wow thanx. Yeah, I think it 59 tracks then it switches to the other processor. I remember a Digi guy telling about the switch.
Don't know what AFAIK means though.

Again thanx a million. Your a huge help in understanding this bigtime.

M2E
Old 12th January 2007
  #145
Gear Addict
 

Newbie question ... trim plugs

Hi Paul,

I'm a late joiner to this thread... so I hope you get this post. I've been reading a lot of your posts in regard to meter levels, illegal signals, and the common problems with mixing ITB. I have to admit that a lot of this stuff is still a little bit over my head (eg. the difference between sample value and actual signals). However, at least I can understand the HOW TO part, in terms of using trim plugs, etc.

I was wondering if you would be kind enough to answer one or two questions for me, although I have no doubt that these questions might be redudant to you. I tried to get as much info in your various threads, but here goes anyway..

First a simple question... What is an example of a trim plug?

I work on an older Pro Tools system (Mix Plus 5.1), on OS 9.1. I don't know if that's relevant at all... but I don't think I have a "trim plug"... or maybe I do, but don't know it. Mostly, I use Waves Renaissance plugs, Waveshell 3.0 (for the time being, that's all I have to work with). You had mentioned (I think) that in the absence of a "trim plug", one can use another plug that has an input control. Well, I think that the Waves Renaissance plugs (the EQ for example) only have an output level control... not an input level control. Does that make the Waves plugs obsolete in this matter?

Bottom line... can you please let me know exactly what plug I can use (given my older Pro Tools system) to trim my levels? Thanks.

I also wanted to make sure I understood something correctly... in regard to recording levels. If I get this right, it is best to record into Pro Tools at no higher than -3dB (in order to use the available bits and get the best signal to noise ratio)... and THEN, only after that, to reduce the level with a trim plug.... trying to keep all levels, at all stages of the mix, at no higher than approx -6dB.

Is this correct?

And finally, a general type of question regarding the benefits of doing things this way. My overall complaint about the sound I have been getting with my ITB mixes, is that the mixes sound very small and thin. Up until this point, I knew nothing about what you have been discussing (illegal signals, meters, etc). Up until now, I have been, for the most part, using the red clip light as my guide... although honestly, I tend to aim for -3dB, most of the time. I have been using this ballpark figure for individual channels, sub-mixes, as well as for the stereo buss (which I then boost with a Renaissance compressor and an L2). This is for "self-mastering", of course...

On a sidenote, funnily enough, some of my mixes have a very low level on the stereo buss (i.e. almost halfway down), EXCEPT for the snare hits. Let me explain... As I mentioned earlier, I have been using a Waves L2 to boost the overall level of my mixes (self-mastering). Of course, we all know that the L2 eats up the snare, amongst other things. So what I have been doing is buidling my mixes so that the overall level of the mix is relatively low, and then cranking the snare really loud, so that by the time everything reaches the L2, everything somehow evens out. What that means is that if I were to bypass all the plugs on my stereo buss, the mix would sound ridiculously low, except for ridiculously loud snare hits. I hope that makes sense...

Sorry for all the extra info... but I wanted to give you as clear a picture as possible, of what I have been doing (maybe I can learn a little bit from you)

So, my general question is... when you say that using trim plugs will improve the sound quality of ITB mixes... does this also translate to "bigger" sounding mixes? Or is this strictly an issue of clarity? Or all of the above? Once again, my main concern is that everything seems to sound very small and thin.

I was also considering the possibility that the older version of Pro Tools (mix plus 5.1) has an inferior mixer to the HD mixer. I'm still a little green about that topic as well (i.e. does my Pro Tools version have a 24 bit mixer? and does that have less headroom than the HD mixer, which if I understand it correctly, is a 48 bit mixer, which then, at the very last stage, truncates back down to 24 bit?)

For the record (and if this is any help, when answering my questions)... my system consists of the following:

- PT Mix Plus 5.1 (with expansion chassis and a total of 7 DSP cards)
- Apogee Rosetta 800 AD/DA
- Brent Averill 1084 mic pres (which I use to record everything)
- Waves Renaissance plugs (which I use mostly)
- Waves L2 (on stereo buss)

Also... for the record, I usually use the "bounce to disk" feature in Pro Tools. I set the dither on the L2 to 16 bits. And finally, I import that file into Toast, and burn a CD.

Thanks for bearing with me. I would appreciate any input you might have for me.
Old 1st March 2007
  #146
Lives for gear
 

Oh gosh - you make so many points here it could take a day to answer them :-)

However let me try briefly.

There are 2 reasons to record, process and master at less than flat out dBFS:

1. To avoid hidden overs not shown on metering due to the reconstruction of the signal. This occurs mostly in D/As (variably depending on how they are designed) and to some extent within some plug-ins and digital processes. Error from this range from limiting all the way to loud 'splats' as values may fold over completely.

A safe margin for normal programme to avoid most of this is -3dBFS. Although some artificially maximised programme can create more than this and some test material can create 6dB of over - so real safety is only gained at around -6dBFS.

If you have a reconstruction meter you can monitor this effect yourself (for the mix output) and compensate manually, or if you have a suitably equipped limiting app (I.e. Oxford Limiter) you can correct these errors automatically. This does not help much with stuff within the channels of the mix itself - so prudence us still advisable.

2. To create headroom, using lower levels within your mix allows you to avoid clipping signals every time you do anything - it frees you up to concentrate on sound rather than red lights and radically eases the mixing process. Some plugs may actually sound better because internal overs may be avoided.

To do this you need to reduce levels to something sensible first thing in the playback channel - process at lower levels - end up with a mix at less than flat out - then make up the level at the very end of the mix. It sounds like you are doing this already :-) But please note this is NOT to avoid math overs in the PT summing buss - as these are catered for already in the PT mixer :-)

Ok - you talk of making mixes that are suitably modified by the output limiter? Yes, this is common practice and in fact mixing with the limiter in place is a really good idea as you instinctively adjust the mix for the best final sound. But these days we have to watch it as the industry is obsessed with loudness at the expense of absolutely everything else - we produce 2 dimensional programme that has no dynamic range. Therefore it isn't possible for you to create real dynamics in this current environment (if you want to stay in business) - instead you are limited to trying to create the impression of dynamics from the extra artefacts and distortions the limiter generates.

Is was with this in mind that I designed the Oxford Limiter - basically to create the impression of dynamic range when in fact there was none - and do it in a way that sounded as natural as possible. You can use this effect either to produce stuff that is loud as ever but sounds less artificial :-) - or you can use it to produce stuff that is as bad as before but is even louder :-(

I hope this is helpful :-)


Quote:
Originally Posted by Rockman View Post
Hi Paul,

I'm a late joiner to this thread... so I hope you get this post. I've been reading a lot of your posts in regard to meter levels, illegal signals, and the common problems with mixing ITB. I have to admit that a lot of this stuff is still a little bit over my head (eg. the difference between sample value and actual signals). However, at least I can understand the HOW TO part, in terms of using trim plugs, etc.

I was wondering if you would be kind enough to answer one or two questions for me, although I have no doubt that these questions might be redudant to you. I tried to get as much info in your various threads, but here goes anyway..

First a simple question... What is an example of a trim plug?

I work on an older Pro Tools system (Mix Plus 5.1), on OS 9.1. I don't know if that's relevant at all... but I don't think I have a "trim plug"... or maybe I do, but don't know it. Mostly, I use Waves Renaissance plugs, Waveshell 3.0 (for the time being, that's all I have to work with). You had mentioned (I think) that in the absence of a "trim plug", one can use another plug that has an input control. Well, I think that the Waves Renaissance plugs (the EQ for example) only have an output level control... not an input level control. Does that make the Waves plugs obsolete in this matter?

Bottom line... can you please let me know exactly what plug I can use (given my older Pro Tools system) to trim my levels? Thanks.

I also wanted to make sure I understood something correctly... in regard to recording levels. If I get this right, it is best to record into Pro Tools at no higher than -3dB (in order to use the available bits and get the best signal to noise ratio)... and THEN, only after that, to reduce the level with a trim plug.... trying to keep all levels, at all stages of the mix, at no higher than approx -6dB.

Is this correct?

And finally, a general type of question regarding the benefits of doing things this way. My overall complaint about the sound I have been getting with my ITB mixes, is that the mixes sound very small and thin. Up until this point, I knew nothing about what you have been discussing (illegal signals, meters, etc). Up until now, I have been, for the most part, using the red clip light as my guide... although honestly, I tend to aim for -3dB, most of the time. I have been using this ballpark figure for individual channels, sub-mixes, as well as for the stereo buss (which I then boost with a Renaissance compressor and an L2). This is for "self-mastering", of course...

On a sidenote, funnily enough, some of my mixes have a very low level on the stereo buss (i.e. almost halfway down), EXCEPT for the snare hits. Let me explain... As I mentioned earlier, I have been using a Waves L2 to boost the overall level of my mixes (self-mastering). Of course, we all know that the L2 eats up the snare, amongst other things. So what I have been doing is buidling my mixes so that the overall level of the mix is relatively low, and then cranking the snare really loud, so that by the time everything reaches the L2, everything somehow evens out. What that means is that if I were to bypass all the plugs on my stereo buss, the mix would sound ridiculously low, except for ridiculously loud snare hits. I hope that makes sense...

Sorry for all the extra info... but I wanted to give you as clear a picture as possible, of what I have been doing (maybe I can learn a little bit from you)

So, my general question is... when you say that using trim plugs will improve the sound quality of ITB mixes... does this also translate to "bigger" sounding mixes? Or is this strictly an issue of clarity? Or all of the above? Once again, my main concern is that everything seems to sound very small and thin.

I was also considering the possibility that the older version of Pro Tools (mix plus 5.1) has an inferior mixer to the HD mixer. I'm still a little green about that topic as well (i.e. does my Pro Tools version have a 24 bit mixer? and does that have less headroom than the HD mixer, which if I understand it correctly, is a 48 bit mixer, which then, at the very last stage, truncates back down to 24 bit?)

For the record (and if this is any help, when answering my questions)... my system consists of the following:

- PT Mix Plus 5.1 (with expansion chassis and a total of 7 DSP cards)
- Apogee Rosetta 800 AD/DA
- Brent Averill 1084 mic pres (which I use to record everything)
- Waves Renaissance plugs (which I use mostly)
- Waves L2 (on stereo buss)

Also... for the record, I usually use the "bounce to disk" feature in Pro Tools. I set the dither on the L2 to 16 bits. And finally, I import that file into Toast, and burn a CD.

Thanks for bearing with me. I would appreciate any input you might have for me.
Old 1st March 2007
  #147
16942
Guest
Quote:
Originally Posted by Paul Frindle View Post
Oh gosh - you make so many points here it could take a day to answer them :-)

However let me try briefly.

There are 2 reasons to record, process and master at less than flat out dBFS:

1. To avoid hidden overs not shown on metering due to the reconstruction of the signal. This occurs mostly in D/As (variably depending on how they are designed) and to some extent within some plug-ins and digital processes. Error from this range from limiting all the way to loud 'splats' as values may fold over completely.

A safe margin for normal programme to avoid most of this is -3dBFS. Although some artificially maximised programme can create more than this and some test material can create 6dB of over - so real safety is only gained at around -6dBFS.

If you have a reconstruction meter you can monitor this effect yourself (for the mix output) and compensate manually, or if you have a suitably equipped limiting app (I.e. Oxford Limiter) you can correct these errors automatically. This does not help much with stuff within the channels of the mix itself - so prudence us still advisable.

2. To create headroom, using lower levels within your mix allows you to avoid clipping signals every time you do anything - it frees you up to concentrate on sound rather than red lights and radically eases the mixing process. Some plugs may actually sound better because internal overs may be avoided.

To do this you need to reduce levels to something sensible first thing in the playback channel - process at lower levels - end up with a mix at less than flat out - then make up the level at the very end of the mix. It sounds like you are doing this already :-) But please note this is NOT to avoid math overs in the PT summing buss - as these are catered for already in the PT mixer :-)

Ok - you talk of making mixes that are suitably modified by the output limiter? Yes, this is common practice and in fact mixing with the limiter in place is a really good idea as you instinctively adjust the mix for the best final sound. But these days we have to watch it as the industry is obsessed with loudness at the expense of absolutely everything else - we produce 2 dimensional programme that has no dynamic range. Therefore it isn't possible for you to create real dynamics in this current environment (if you want to stay in business) - instead you are limited to trying to create the impression of dynamics from the extra artefacts and distortions the limiter generates.

Is was with this in mind that I designed the Oxford Limiter - basically to create the impression of dynamic range when in fact there was none - and do it in a way that sounded as natural as possible. You can use this effect either to produce stuff that is loud as ever but sounds less artificial :-) - or you can use it to produce stuff that is as bad as before but is even louder :-(

I hope this is helpful :-)

Paul would you say this advice holds true for a 32 bit float app like Logic as well as fixed point apps like PT?
Old 1st March 2007
  #148
Lives for gear
Quote:
Originally Posted by Paul Frindle View Post
I hope this is helpful :-)
all your posts are helpful. i for one am very thankful and mix differently after having studied your posts for awhile now. and the differences are noticeable.

imo it would be great if you were featured in 'Expert Questions & Answers', or a simple compilation of your posts to date would be a great help to many an ITB explorer.
Old 1st March 2007
  #149
Lives for gear
 

Quote:
Originally Posted by Ashermusic View Post
Paul would you say this advice holds true for a 32 bit float app like Logic as well as fixed point apps like PT?

Yes I would.

The intersample peaking reconstruction problem is the same at the output of the mix, as it must be represented in a fixed point output format anyway (i.e. CD or DVD).
Whilst with float it's possible to accomodate internally numbers bigger than flat out, any process that has need to refer to actual real values might be at risk of overload (or unspecified behaviour). Why take the risk?

From the point of the headroom issue, things might be different in that an entirely float system from start to finish might handle overs properly - however the meters will be calibrated to a fixed point reference (and will come on willy nilly, whether the signal is clipped or not). Some systems using expansion DSP pass and process signals in fixed point (PowerCore being one example) and may not have any of the float headroom and may mess up with overs.
Again, with 140dB or so dynamic real range at your disposal, why bother to risk it?

The most important thing to remember is that recording and processing at lower levels DOES NOT waste 'bits'. It doesn't work like that - all your 'bits' are there all the time at all levels :-)
Old 2nd March 2007
  #150
Gear Nut
 
innesireinar's Avatar
 

Hi Paul,
Happy to see you here.
After tons of posts about analog summing vs. ITB summing around various fora, which is your finally word about this war?
I think you, who have designed the best analog consoles and the most famous plugs for PT, could be the highest authority for saying something about that.
I remember the thread on PSW but now after two years is there something changed about that?

Cheeres

ranieri senni
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
Similar Threads
Thread
Thread Starter / Forum
Replies
Elsteve9 / So much gear, so little time
5
djui5 / Gear free zone - shoot the breeze
3
slantbar / Q+A with Mike Shipley
2
AudioGaff / The Moan Zone
1

Forum Jump
Forum Jump