Q for Paul Frindle

innesireinar · 9th May 2006, 02:14 PM

If you are still here I would like to clarify some concepts about the right level we have to work with in a DAW.
In the thread that we have had in the past autumn on PSW you wrote:

"Imagine for instance a loudish instrument with rich HF percussive harmonic content that for one reason or another only just reached peak values at the outp[/font]ut of your ADC in record. You then EQ it a bit (perhaps rolling off the HF a bit) changing the relative phases of the freqs in the spectrum, noticing that the peak sample value level has dropped a dB or so, you increase the gain to max once more. The drop in peak sample value resulting from a slight re-arrangement of the phase of the freqs - may still have resulted in an almost flat out signal when reconstructed - before you added the gain - now it could overload even though no red light is on."

Because most of samples libraries available are all normalized at full level, in this case I use to insert the trim plugin in the first slot before any process set to -6dB. Does this trip work properly in order to avoid illegal peaks (obviolusly without hi gain in eq etc)?

Thank you

ranierisenni

bob katz · 9th May 2006, 02:35 PM

Dear ranierisenni: I'm sure Paul is still here, but I'll chime in. It's the ultimate mix level and how you process the mix that will determine the level. I assume you're going to be mixing samples, reverb, eq, etc. The complex sum of all those elements will have a completely different transient response and level than that measly little sample that might be a small basis for your mix.

A conservative, "very safe" maximum peak level for your 24-bit mix as measured on a simple sample-reading digital meter is -3 dBFS, because the simple digital meter does not read the illegal levels.

If you are mixing totally digitally, then the "intrinsic" level of the samples as they would be played over a D/A converter, for example, does not enter in. It is only an issue if you are feeding a D/A converter for external analog processing. This is what I believe to be true, and just to be safe, I'd like independent confirmation from an authority like Paul.

When you send your mix to the mastering studio----

If the mastering engineer is processing everything digitally, he doesn't have to worry about "intrinsic overs" until the last stage in his processing chain, or until he reaches a sample rate converter, since all digital processors except for sample rate converters work on the "internal" digital value of the samples and whatever comes out, even if it is an illegal signal is not important in the all-digital processing world. I would like Paul to confirm that, but everything I believe says this is true. So your mix could be up to 0 dBFS (if you have an accurate digital meter). At the last stage in his digital chain, the mastering engineer can employ a true brick-wall peak limiter such as the TC System 6000 limiter, which prevents intersample peaks in D/A converters and hopefully also in sample rate converters. Limiters still add distortion, of course, but that's another issue.

If the mastering engineer is processing via analog, and is concerned about distortion, then the first piece he has to worry about (and maybe the only piece) is the D/A converter he uses to feed out to his analog gear. And worst-case scenario in that case is he attenuates the digital level going into the D/A by 0.3 to 3 dB. It avoids the degradation of multiple re-quantizations if you ran your mix at that level to begin with, avoiding one more calculation at the mastering house, but it's not the most serious thing in the world to drop the level and redither it to 24 bits in the digital domain.

Hope that gives a bigger picture of the issue.

Paul Frindle · 9th May 2006, 03:48 PM

Quote:

Originally Posted by innesireinar

If you are still here I would like to clarify some concepts about the right level we have to work with in a DAW.
In the thread that we have had in the past autumn on PSW you wrote:

"Imagine for instance a loudish instrument with rich HF percussive harmonic content that for one reason or another only just reached peak values at the outp[/font]ut of your ADC in record. You then EQ it a bit (perhaps rolling off the HF a bit) changing the relative phases of the freqs in the spectrum, noticing that the peak sample value level has dropped a dB or so, you increase the gain to max once more. The drop in peak sample value resulting from a slight re-arrangement of the phase of the freqs - may still have resulted in an almost flat out signal when reconstructed - before you added the gain - now it could overload even though no red light is on."

Because most of samples libraries available are all normalized at full level, in this case I use to insert the trim plugin in the first slot before any process set to -6dB. Does this trip work properly in order to avoid illegal peaks (obviolusly without hi gain in eq etc)?

Thank you

ranierisenni

Well possibly - but only if you don't add gain, EQ or saturation down line. The real purpose of reducing the input is to avoid having to deal with overloads every time you do something :-(

There's a great thread on PSW forums where I go through lots of this stuff with some interesting examples of things that give errors and demonstrations you can do to show the problems.

http://recforums.prosoundweb.com/index.php/t/4918/2578/

innesireinar · 9th May 2006, 04:29 PM

Quote:

Originally Posted by bob katz

Dear ranierisenni: I'm sure Paul is still here, but I'll chime in. It's the ultimate mix level and how you process the mix that will determine the level. I assume you're going to be mixing samples, reverb, eq, etc. The complex sum of all those elements will have a completely different transient response and level than that measly little sample that might be a small basis for your mix.

A conservative, "very safe" maximum peak level for your 24-bit mix as measured on a simple sample-reading digital meter is -3 dBFS, because the simple digital meter does not read the illegal levels.

If you are mixing totally digitally, then the "intrinsic" level of the samples as they would be played over a D/A converter, for example, does not enter in. It is only an issue if you are feeding a D/A converter for external analog processing. This is what I believe to be true, and just to be safe, I'd like independent confirmation from an authority like Paul.

When you send your mix to the mastering studio----

If the mastering engineer is processing everything digitally, he doesn't have to worry about "intrinsic overs" until the last stage in his processing chain, or until he reaches a sample rate converter, since all digital processors except for sample rate converters work on the "internal" digital value of the samples and whatever comes out, even if it is an illegal signal is not important in the all-digital processing world. I would like Paul to confirm that, but everything I believe says this is true. So your mix could be up to 0 dBFS (if you have an accurate digital meter). At the last stage in his digital chain, the mastering engineer can employ a true brick-wall peak limiter such as the TC System 6000 limiter, which prevents intersample peaks in D/A converters and hopefully also in sample rate converters. Limiters still add distortion, of course, but that's another issue.

If the mastering engineer is processing via analog, and is concerned about distortion, then the first piece he has to worry about (and maybe the only piece) is the D/A converter he uses to feed out to his analog gear. And worst-case scenario in that case is he attenuates the digital level going into the D/A by 0.3 to 3 dB. It avoids the degradation of multiple re-quantizations if you ran your mix at that level to begin with, avoiding one more calculation at the mastering house, but it's not the most serious thing in the world to drop the level and redither it to 24 bits in the digital domain.

Hope that gives a bigger picture of the issue.

Thank you Bob.
Therefore is it a reconstruction of the wave issue only?
How good are the PT meters?
These days I'm transfering all my dat tapes into PT for storing given that I've decided to sell my DAT recorder and during the transfer the meter of my DA45 never reached over while the PT meters of the stereo track always indicated over with the red light on. The connection is via spdif, therefore there is not a level calibration issue. What is strange is that in PT only the meters of the track shown over, while the master meters (where the track is routed to) don't. Obviously all fader to 0dB all pan full opened.
Why?

innesireinar · 9th May 2006, 04:40 PM

Quote:

Originally Posted by Paul Frindle

Well possibly - but only if you don't add gain, EQ or saturation down line. The real purpose of reducing the input is to avoid having to deal with overloads every time you do something :-(

There's a great thread on PSW forums where I go through lots of this stuff with some interesting examples of things that give errors and demonstrations you can do to show the problems.

http://recforums.prosoundweb.com/index.php/t/4918/2578/

Exactly. I know very well this thread, is one of my favorite. I had
some posts with Bob on it (I think on page 9/10).

Does the "GAIN" knob on the Sony eq affect the input stage of the eq
or the out?

Bob Olhsson · 9th May 2006, 04:57 PM

There are numerous reasons that it's better to error on the low side rather than the high side when making a 24 bit digital recording. Clipping plug-ins is a very common problem as is running out of "balls" when using digital gear at the very top of its analog output capability. I highly recommend taking the time to experiment with levels and listening to the results so that you know precisely where the sweet spot is in your signal path.

Paul Frindle · 9th May 2006, 05:46 PM

Quote:

Originally Posted by innesireinar

Thank you Bob.
Therefore is it a reconstruction of the wave issue only?
How good are the PT meters?
These days I'm transfering all my dat tapes into PT for storing given that I've decided to sell my DAT recorder and during the transfer the meter of my DA45 never reached over while the PT meters of the stereo track always indicated over with the red light on. The connection is via spdif, therefore there is not a level calibration issue. What is strange is that in PT only the meters of the track shown over, while the master meters (where the track is routed to) don't. Obviously all fader to 0dB all pan full opened.
Why?

I can answer this perhaps.

In the days of DAT machines it was commonplace to set over indicators to only read after a consecutive and repeated number of overload samples.

For instance, in the case of the Sony 3324/48 DASH machines, 5 consecutive sample overs were required to put on the red light!!

When we made the R3 we were forced to follow this utter madness (after more than a year of bitter argument) so that the console would not show overs before the tape machine :-(

innesireinar · 9th May 2006, 06:11 PM

Quote:

Originally Posted by Paul Frindle

I can answer this perhaps.

In the days of DAT machines it was commonplace to set over indicators to only read after a consecutive and repeated number of overload samples.

For instance, in the case of the Sony 3324/48 DASH machines, 5 consecutive sample overs were required to put on the red light!!

When we made the R3 we were forced to follow this utter madness (after more than a year of bitter argument) so that the console would not show overs before the tape machine :-(

Thank you Paul
But why in PT when a track (0dB full panned) shows over the master track doesn't?
What does this mean?

Paul Frindle · 9th May 2006, 06:15 PM

Quote:

Originally Posted by innesireinar

Exactly. I know very well this thread, is one of my favorite. I had
some posts with Bob on it (I think on page 9/10).

Does the "GAIN" knob on the Sony eq affect the input stage of the eq
or the out?

It affects the input. Please note that there is lot's of internal headroom in the EQ - so for instance you can max boost a freq in one band and then max cut it in another at full level without causing internal overloads - even without turning the input gain down :-) (BTW, this may not be true of some other EQ plug-ins). Therefore the real purpose of the gain control is to avoid clipping the output when you boost stuff.

Some people have requested that the gain control should go positive as well - in order to make up for gain lost when you cut things. This would be something I would put on a new version if I were doing one.

Paul Frindle · 9th May 2006, 06:21 PM

Quote:

Originally Posted by innesireinar

Thank you Paul
But why in PT when a track (0dB full panned) shows over the master track doesn't?
What does this mean?

If you mean that the master on another target machine (not PT) shows no over even though the PT meters did (?) then this may be because the target machine requires successive samples to overload before the red light comes, on whilst the PT red light comes on for only a single over sample.

Other possibilities are that the calibration of what exact sample code gives and over may be different between devices.

Less likely possibilities include things like DC offset removal in the target device causing extreme LF phase shifts etc..

innesireinar · 9th May 2006, 06:30 PM

No, I'm referring to the master track of PT.
Follow me, by doing a session like the one I've done for the dat transfer with only a stereo track and a stereo master track with the stereo track directly routed to the master track without any plugs in the chain, I've noted that sometimes the overled of the stereo track while those on the stereo master don't. Is it possible that inside the same DAW the meters of the recording tracks are calibrated different than those of the master, and aux channels?

innesireinar · 9th May 2006, 06:36 PM

Quote:

Originally Posted by Paul Frindle

It affects the input. Please note that there is lot's of internal headroom in the EQ - so for instance you can max boost a freq in one band and then max cut it in another at full level without causing internal overloads - even without turning the input gain down :-) (BTW, this may not be true of some other EQ plug-ins). Therefore the real purpose of the gain control is to avoid clipping the output when you boost stuff.

Some people have requested that the gain control should go positive as well - in order to make up for gain lost when you cut things. This would be something I would put on a new version if I were doing one.

Mhh, things are becoming more complicated.
In the thread on PSW you are referring to illegal clips that can occur
by processing a signal that previously was recorded close to 0dB (and
therefore your advicese were to work with signal around -6dB), right?
But if a processing such as eqing with very hi boosting doesn't cause
illegal peaks (because of hi internal headroom in the sony eq) what
(and where) can cause illegal peaks?
You were refer to many illegal tracks, each with illegal peaks mixed
together that could cause a degradation of the programme, therefore
it's not a summing issue, then, where is the problem? In the sum? Or
in any track?
If is a DA convertion issue like Bob said, it could be adjusted in the master track before going to the DAC?

Paul Frindle · 9th May 2006, 08:12 PM

Quote:

Originally Posted by innesireinar

Mhh, things are becoming more complicated.
In the thread on PSW you are referring to illegal clips that can occur
by processing a signal that previously was recorded close to 0dB (and
therefore your advicese were to work with signal around -6dB), right?
But if a processing such as eqing with very hi boosting doesn't cause
illegal peaks (because of hi internal headroom in the sony eq) what
(and where) can cause illegal peaks?
You were refer to many illegal tracks, each with illegal peaks mixed
together that could cause a degradation of the programme, therefore
it's not a summing issue, then, where is the problem? In the sum? Or
in any track?
If is a DA convertion issue like Bob said, it could be adjusted in the master track before going to the DAC?

It has nothing to do with summing at all, neither does it have anything much to do with EQ particularly :-(

Oh dear - there are several issues here which are quite separate - but ALL involving level, some seen on meters and some not seen on meters - some that occur within the worksation and some that only occur in the play out system. It's difficult to know where to start - but I'll try to summarise briefly:

The different level situations are:

1. Reducing input levels to the mixer to avoid hard sample value overloads every time you try to do anything. These are visible on meters and you can see them, but they detract from the artistic process because you need to keep destroying balances every time you reduce gains to fix them. This is a convenience for operational reasons - it simply makes the mix 'go better'.

2. Reducing levels throughout the channel to avoid internal overloads on less capable plug-ins (that may not appear on meters). This is a technical issue with some plug-ins I have seen, but it does not happen with any Oxford plugs, as they all have internal headroom and cannot clip internally before maximum output is achieved.

3. Reducing levels at the OUTPUT of the mix in order to avoid reconstruction overs at the DACs (either yours, the consumers and/or both) that almost certainly will NOT be displayed on any meter within your workstation. The reason this occurs is that people are aiming for maximum modulation and the meters on your workstations only display sample values - NOT signal levels. This is why the limiter includes a reconstruction meter (which shows actual SIGNAL levels) and a dynamic method to 'fix' such overs without losing average level.

4. Similarly, being aware that some plug-ins have the equivalent of reconstruction INSIDE their processes and therefore can make a perfectly legal sample value signals (i.e. no red lights) into and illegal sample values (like the DAC above) but actually within the workstation. These may be seen on meters, but confuse people badly and may end up forcing you to make a final mixes much less loud than they could have been. Some plug-ins can do this without any modification to the freq response or loudness - even when set to 'flat' or in 'bypass'. I demonstrated this effect using the noise PT noise generator. The GML EQ is an example of this which was noted in the PSW thread - and it is not unique :-(

Paul Frindle · 9th May 2006, 08:22 PM

Quote:

Originally Posted by innesireinar

No, I'm referring to the master track of PT.
Follow me, by doing a session like the one I've done for the dat transfer with only a stereo track and a stereo master track with the stereo track directly routed to the master track without any plugs in the chain, I've noted that sometimes the overled of the stereo track while those on the stereo master don't. Is it possible that inside the same DAW the meters of the recording tracks are calibrated different than those of the master, and aux channels?

Hmm... not too sure about this. The PTLE has a very slight issue with it's calibration of the meters wrt the final truncated output (this is because the mixer is running float but the output must be fixed). This means that an app like the Oxford limiter will red light very slightly before the PTLE meters - because the limiter is measuring the correctly truncated fixed point output.

Also some intermediate versions of PT TDM seemed to have meters that were measuring the top 16bits of a 24bit signal (for a while until they apparently fixed it). We had an issue with this when designing the inflator.

But I personally have not seen exactly what you are describing, but I would not rule out the last of these issues.

Are you running LE or TDM?

yeloocproducer · 10th May 2006, 03:03 AM

Paul, couple questions... do you only advise trim plugs at the top of the channels, or just lowering the "input" stage on your first plugin (such as an eq), or would either be sufficient?

I think a lot of folks just want an easy to follow rule of thumb... like, if the tracks were recorded near 0, put a trim in at -6 before other plugins so that you get some headroom for your compression, eq, and harmonic distortion plugs... and that should get rid of most of your "invisible" overs. Also keep the outputs of your plugs at -6 peak. Plus, keep an eye on your master faders anytime you have audio getting converted back to analog... whether stemming out and summing, or running your 2 mix out to an outboard compressor, etc. In this case, make sure that level is again peaking (on the digital meters) at no greater than -6db on any given stem or your master so that you don't create reconstruction overs. And when using "upsampling" plugins, be very very careful that you set thier outputs lower as well b/c they might be creating reconstruction overs as well when they bring things back down.

Am I oversimplying, or is this the overall "how to"? I've been following the threads here and at prosoundweb and this is what I've put together. Please correct me if I got any of that wrong.

Paul Frindle · 10th May 2006, 12:49 PM

Quote:

Originally Posted by yeloocproducer

Paul, couple questions... do you only advise trim plugs at the top of the channels, or just lowering the "input" stage on your first plugin (such as an eq), or would either be sufficient?

I think a lot of folks just want an easy to follow rule of thumb... like, if the tracks were recorded near 0, put a trim in at -6 before other plugins so that you get some headroom for your compression, eq, and harmonic distortion plugs... and that should get rid of most of your "invisible" overs. Also keep the outputs of your plugs at -6 peak. Plus, keep an eye on your master faders anytime you have audio getting converted back to analog... whether stemming out and summing, or running your 2 mix out to an outboard compressor, etc. In this case, make sure that level is again peaking (on the digital meters) at no greater than -6db on any given stem or your master so that you don't create reconstruction overs. And when using "upsampling" plugins, be very very careful that you set thier outputs lower as well b/c they might be creating reconstruction overs as well when they bring things back down.

Am I oversimplying, or is this the overall "how to"? I've been following the threads here and at prosoundweb and this is what I've put together. Please correct me if I got any of that wrong.

I think you've got the points well - but I would put it even more simply:

- I would lose the level (-6dB or so) right at the input of the mixer from the track.

This in order to allow me to work without worrying constantly about overs. I don't want to keep destroying my balances by adjusting levels all over the place.

- Maintain the output of plugs in the channel at around -6dB or so too.

This because I don't want plugs down line to start clipping when I instantiate them and have to start fiddling with my precious balances again. (watching out for plugs that increase levels even though they are not doing anything much - i.e. the intersample peaking stuff)

- Do the mix aiming for around -6dB or so for the final mix output.

This because I don't want to risk clipping of converters and stuff and hearing distortions whilst I am getting my balances and mixing.

- Do my buss compression and maximising and what not, by making up the gain with my prefered output processing as a final process.

This to give me more control over the final result (by having a clean mix reference) and avoid thrashing the output processing.

Also I would mix with my intended limiter/buss processing in place on the master buss - and unbypass it from time to time to see how the mix I'm doing survives the processing. You can get much better results if you 'work into' a comp/limiter by balancing stuff in it's favour.

- If my mix is going to a professional mastering engineer, I would avoid too much output buss processing and provide a file with which the guy can actually work! Leave off odd forms of noise shaping dither (use conventional triangular PDF only), print the file aiming for -3dB or so, so that it doesn't immediately start overloading his DAC when he first listens to it (this is a courtesy measure)..

- If I your are mastering it yourself (i.e. for client approval and stuff), use a meter (or comp/limiter) that lets you see the actual reconstructed signal.

This to avoid making a file that could inexplicably start sounding rubbish on your client's (or other people's) systems!

As a courtesy measure, I would also provide this (home mastered) file for the professional mastering engineer, as a demonstration of the sound I was trying to get for his reference :-)

bob katz · 10th May 2006, 01:52 PM

Quote:

Originally Posted by innesireinar

Thank you Bob.
Therefore is it a reconstruction of the wave issue only?

I believe so!

Quote:

How good are the PT meters?

I haven't seen anything on the LE system that would qualify for more than an "idiot light". I suggest Inspector XL or in your case something better that is not yet available as a plugin for Pro Tools, that interpolates and calculates intersample peaks.

These days I'm transfering all my dat tapes into PT for storing given that I've decided to sell my DAT recorder and during the transfer the meter of my DA45 never reached over while the PT meters of the stereo track always indicated over with the red light on. The connection is via spdif, therefore there is not a level calibration issue. What is strange is that in PT only the meters of the track shown over, while the master meters (where the track is routed to) don't. Obviously all fader to 0dB all pan full opened.

[/quote]

My contention is that as long as you are:

a) making perfect clones (which your technique seems to imply is true... but always test your DAW for its ability to clone sources... never even trust a panpot until you know it is safe)

or

b) processing digitally and inspecting an over-counting meter for consecutive samples

c) NOT processing via analog

Then you can continue with great happiness :-)

Did Paul chime in to confirm or deny my statement?

innesireinar · 10th May 2006, 02:03 PM

Quote:

Originally Posted by Paul Frindle

Hmm... not too sure about this. The PTLE has a very slight issue with it's calibration of the meters wrt the final truncated output (this is because the mixer is running float but the output must be fixed). This means that an app like the Oxford limiter will red light very slightly before the PTLE meters - because the limiter is measuring the correctly truncated fixed point output.

Also some intermediate versions of PT TDM seemed to have meters that were measuring the top 16bits of a 24bit signal (for a while until they apparently fixed it). We had an issue with this when designing the inflator.

But I personally have not seen exactly what you are describing, but I would not rule out the last of these issues.

Are you running LE or TDM?

I swear, Paul
This is what happens.
I'm on LE
If anyone out there can confirm if this happens in HD sys also...

BTW
If I wish to use the Sony limiter for metering only - it could be the case when mixing the first and unmastered file - how does the limiter work? Does it introduce latency? Is it full transparent? It could be an idea to route every channels to an aux ch. before going to master ch. then open a send on this aux ch. and send the entire mix to another aux where it could be instatiated the limiter. Route the aux ch. directly to the master. In this case (to avoid latency) I could use a source before the limiter for mixing while I could watch to the limiter for metering.
Does it work?
I've seen that the limiter is heavy, expecially for me who am still on LE and expecially when enable the reconstruction meter and auto comp.
There is a thing that I don't understood. Given that the limiter is a digital tool, I tought that it works on samples values, right? THe reconstructed shapes of the waves happens during the DAC process(?) Therefore, how can the limiter's meter indicate the real recon waves, given that it is still in the digital domain? Does it try to assume what it could happen during the DAC process?

innesireinar · 10th May 2006, 02:10 PM

[quote=bob katz]I believe so!

I haven't seen anything on the LE system that would qualify for more than an "idiot light". I suggest Inspector XL or in your case something better that is not yet available as a plugin for Pro Tools, that interpolates and calculates intersample peaks.

But is it an LE issue only? Or even in HD systems this happens?

bob katz · 10th May 2006, 02:16 PM

I would say that Paul is being extremely accurate and conservative on this issue. If you follow his recommendations, there is little to worry about.

However, (and Paul knows this), you could stretch the envelope or be less worried if your system is TOTALLY floating point (which includes those testy issues of transfer between plugins and the main system) and only worry about over levels if you are feeding a fixed point interface or a converter.

I've tested my direct-x/VST SADiE and I was able to confirm by both measurements (distortion) direct observations of levels with music, that I could, for example, have two Waves plugins in a row, the first which I purposely pushed over so that its meters actually read OVER 0 dBFS and the second was the Waves L2. I could set up the gain structure two different ways and get identical results. It was clear proof that the threshold and input of the L2 were reading the floating point signal from the previous processor with no problems.

Attached is a bitmap screenshot of the demonstration. Please note this example is frozen in time after a maximum peak level has already occurred. The maximum peak the C1 saw was +3.5/+1.3 OVER 0 dBFS. And this was transferred accurately to the L2 as you can see on its meters showing maximum peak INPUT. This is also not a real example of amounts of limiting, but rather a demonstration that you could "overload" the C1 and set the threshold of the L2 accordingly and not have distortion passed between the two plugins.

Don't try this at home! Make sure you know what you are doing and have a firm grasp of the concept and can prove it to yourself through testing ON YOUR SYSTEM as I have done. And regardless, this is not the practice that I go through routinely, I did it as an exercise and test of functionality. Seeing RED is something that should concern everyone.

BK

bob katz · 10th May 2006, 02:16 PM

[quote=innesireinar]

Quote:

Originally Posted by bob katz

I believe so!

I haven't seen anything on the LE system that would qualify for more than an "idiot light". I suggest Inspector XL or in your case something better that is not yet available as a plugin for Pro Tools, that interpolates and calculates intersample peaks.

But is it an LE issue only? Or even in HD systems this happens?

I'm not an HD expert, so I could only speak from my experience.

Paul Frindle · 10th May 2006, 03:41 PM

Quote:

Originally Posted by bob katz

I would say that Paul is being extremely accurate and conservative on this issue. If you follow his recommendations, there is little to worry about.

However, (and Paul knows this), you could stretch the envelope or be less worried if your system is TOTALLY floating point (which includes those testy issues of transfer between plugins and the main system) and only worry about over levels if you are feeding a fixed point interface or a converter.

I've tested my direct-x/VST SADiE and I was able to confirm by both measurements (distortion) direct observations of levels with music, that I could, for example, have two Waves plugins in a row, the first which I purposely pushed over so that its meters actually read OVER 0 dBFS and the second was the Waves L2. I could set up the gain structure two different ways and get identical results. It was clear proof that the threshold and input of the L2 were reading the floating point signal from the previous processor with no problems.

BK

Yes this is indeed so - the float math representation gives you headroom and this is also the case in PT LE because all the applications run on the host processor which is inherantly floating point. Interestingly RTAS plugs on TDM systems also run float - however the input and output of the RTAS plugs have to presented and rendered in 24bit fixed point for the rest of the application - and of course the intended output wordlength.

And that last point illustrates the rub: Even in floating math with arbitary headroom scaling, there needs to be a knowledge of the actual signal levels. SO for instance any plug that needs to know levels (i.e. dynamics, limiters, or any non-linear process) MUST have an internal reference of operating levels that matches the notional signal level - and the signal format at the end of the day is fixed point.
An example of this is the Sony Inflator, since this is highly a non-linear process that is signal dependant, in the RTAS and LE processing one has to actually enforce fixed point limitations and scaling within the processing itself - in other words we use extra processing to achieve this..

So to put this another way - all that is needed is an operating level that is something other than total and complete flat out - which is under your control. All float is giving you is an arbitary and uncontrolled headroom that sometimes gets you out of trouble (providing you don't bank on it) - simply because there is no sense in the current flat out operating level conditions everyone has been encouraged to use :-(

yeloocproducer · 10th May 2006, 04:44 PM

So Paul, this is pretty eye-opening... basically you're saying that a TDM (or any fixed point) system is slightly easier to "clip" with reconstruction errors than LE (or any float system) if you're not careful. Wow. One last clarification and I think I'm set.

Quote:

Originally Posted by Paul Frindle

I think you've got the points well - but I would put it even more simply:

- I would lose the level (-6dB or so) right at the input of the mixer from the track.

Sorry to keep coming back to this, but when you say "lose the level right at the input", does that mean in Protools: lower the 1st plug input, lower the track fader/automation, or throw on a trim plugin at -6? I'm pretty sure you mean just lowering the fader...

Paul Frindle · 10th May 2006, 05:29 PM

Quote:

Originally Posted by yeloocproducer

So Paul, this is pretty eye-opening... basically you're saying that a TDM (or any fixed point) system is slightly easier to "clip" with reconstruction errors than LE (or any float system) if you're not careful. Wow. One last clarification and I think I'm set....

Easier to clip with samples - nothing to do with reconstruction errors - they are another matter.

However as I have said - just because with float representation you can pass signals around at levels arbitarily above max, does NOT mean that everything can make good use of such signals. This is a very important thing to remember!

By far the best thing is to set your personal operating level at something sensible below max (i.e. -6dB or below) and be in charge of what's happening YOURSELF - rather than leave it to the whims (and unreported foibles) of processes driven beyond their design limits :-(

Quote:

Sorry to keep coming back to this, but when you say "lose the level right at the input", does that mean in Protools: lower the 1st plug input, lower the track fader/automation, or throw on a trim plugin at -6? I'm pretty sure you mean just lowering the fader...

I would throw in a trim at -6dB. That's what i do. There is an issue in that the PT trim plug does not re-dither the output as it should, but what happens at -140dB or so is obviously totally insignificant in comparison with what we are talking about.

yeloocproducer · 10th May 2006, 05:41 PM

Gotcha. Reconstruction errors are only on the D to A process. Thanks Paul for clearing things up, including the dither on the trim, I was wondering about that too.

Is there a specific reason you shouldn't use the input on the first plugin at -6 if it's a double precision, high quality eq plug like sony or URS?

Paul Frindle · 10th May 2006, 05:46 PM

Quote:

Originally Posted by innesireinar

I swear, Paul
This is what happens.
I'm on LE
If anyone out there can confirm if this happens in HD sys also...

BTW
If I wish to use the Sony limiter for metering only - it could be the case when mixing the first and unmastered file - how does the limiter work? Does it introduce latency? Is it full transparent? It could be an idea to route every channels to an aux ch. before going to master ch. then open a send on this aux ch. and send the entire mix to another aux where it could be instatiated the limiter. Route the aux ch. directly to the master. In this case (to avoid latency) I could use a source before the limiter for mixing while I could watch to the limiter for metering.
Does it work?

This a bit problematic and the plug-in is only transparent when;

- The input level (set by gain) is less than 0db on the input meter,
- The safe mode is off and the enhance slider at minimum,
- The soft knee is at minimum,
- The intersample auto compensation

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Paul Simon	Elsteve9	So much gear, so little time!	2	4th October 2004 06:43 PM
Paul Franklin	slantbar	Q&A with Mike Shipley	2	11th June 2004 10:05 PM
Stephen Paul R.I.P	AudioGaff	The moan zone	1	24th December 2003 06:51 AM

9th May 2006, 02:14 PM	#1
innesireinar Gear Head Join Date: Sep 2004 Posts: 71	Q for Paul Frindle If you are still here I would like to clarify some concepts about the right level we have to work with in a DAW. In the thread that we have had in the past autumn on PSW you wrote: "Imagine for instance a loudish instrument with rich HF percussive harmonic content that for one reason or another only just reached peak values at the outp[/font]ut of your ADC in record. You then EQ it a bit (perhaps rolling off the HF a bit) changing the relative phases of the freqs in the spectrum, noticing that the peak sample value level has dropped a dB or so, you increase the gain to max once more. The drop in peak sample value resulting from a slight re-arrangement of the phase of the freqs - may still have resulted in an almost flat out signal when reconstructed - before you added the gain - now it could overload even though no red light is on." Because most of samples libraries available are all normalized at full level, in this case I use to insert the trim plugin in the first slot before any process set to -6dB. Does this trip work properly in order to avoid illegal peaks (obviolusly without hi gain in eq etc)? Thank you ranierisenni

9th May 2006, 02:35 PM	#2
bob katz Mastering Join Date: Mar 2006 Posts: 1,826	Dear ranierisenni: I'm sure Paul is still here, but I'll chime in. It's the ultimate mix level and how you process the mix that will determine the level. I assume you're going to be mixing samples, reverb, eq, etc. The complex sum of all those elements will have a completely different transient response and level than that measly little sample that might be a small basis for your mix. A conservative, "very safe" maximum peak level for your 24-bit mix as measured on a simple sample-reading digital meter is -3 dBFS, because the simple digital meter does not read the illegal levels. If you are mixing totally digitally, then the "intrinsic" level of the samples as they would be played over a D/A converter, for example, does not enter in. It is only an issue if you are feeding a D/A converter for external analog processing. This is what I believe to be true, and just to be safe, I'd like independent confirmation from an authority like Paul. When you send your mix to the mastering studio---- If the mastering engineer is processing everything digitally, he doesn't have to worry about "intrinsic overs" until the last stage in his processing chain, or until he reaches a sample rate converter, since all digital processors except for sample rate converters work on the "internal" digital value of the samples and whatever comes out, even if it is an illegal signal is not important in the all-digital processing world. I would like Paul to confirm that, but everything I believe says this is true. So your mix could be up to 0 dBFS (if you have an accurate digital meter). At the last stage in his digital chain, the mastering engineer can employ a true brick-wall peak limiter such as the TC System 6000 limiter, which prevents intersample peaks in D/A converters and hopefully also in sample rate converters. Limiters still add distortion, of course, but that's another issue. If the mastering engineer is processing via analog, and is concerned about distortion, then the first piece he has to worry about (and maybe the only piece) is the D/A converter he uses to feed out to his analog gear. And worst-case scenario in that case is he attenuates the digital level going into the D/A by 0.3 to 3 dB. It avoids the degradation of multiple re-quantizations if you ran your mix at that level to begin with, avoiding one more calculation at the mastering house, but it's not the most serious thing in the world to drop the level and redither it to 24 bits in the digital domain. Hope that gives a bigger picture of the issue. __________________ Bob Katz DIGITAL DOMAIN http://www.digido.com "There are two kinds of fools. One says-this is old and therefore good. The other says-this is new and therefore better." No trees were killed in the sending of this message. However a large number of electrons were terribly inconvenienced. Last edited by bob katz; 9th May 2006 at 02:37 PM.

9th May 2006, 04:57 PM	#6
Bob Olhsson Motown legend Join Date: Jun 2002 Location: Songwriter Gulch, Nashville TN Posts: 5,139	There are numerous reasons that it's better to error on the low side rather than the high side when making a 24 bit digital recording. Clipping plug-ins is a very common problem as is running out of "balls" when using digital gear at the very top of its analog output capability. I highly recommend taking the time to experiment with levels and listening to the results so that you know precisely where the sweet spot is in your signal path. __________________ Georgetown Masters 615 254-3233 Bob's room 615 385-8051 Music Industry 2.0

10th May 2006, 02:16 PM	#20
bob katz Mastering Join Date: Mar 2006 Posts: 1,826	I would say that Paul is being extremely accurate and conservative on this issue. If you follow his recommendations, there is little to worry about. However, (and Paul knows this), you could stretch the envelope or be less worried if your system is TOTALLY floating point (which includes those testy issues of transfer between plugins and the main system) and only worry about over levels if you are feeding a fixed point interface or a converter. I've tested my direct-x/VST SADiE and I was able to confirm by both measurements (distortion) direct observations of levels with music, that I could, for example, have two Waves plugins in a row, the first which I purposely pushed over so that its meters actually read OVER 0 dBFS and the second was the Waves L2. I could set up the gain structure two different ways and get identical results. It was clear proof that the threshold and input of the L2 were reading the floating point signal from the previous processor with no problems. Attached is a bitmap screenshot of the demonstration. Please note this example is frozen in time after a maximum peak level has already occurred. The maximum peak the C1 saw was +3.5/+1.3 OVER 0 dBFS. And this was transferred accurately to the L2 as you can see on its meters showing maximum peak INPUT. This is also not a real example of amounts of limiting, but rather a demonstration that you could "overload" the C1 and set the threshold of the L2 accordingly and not have distortion passed between the two plugins. Don't try this at home! Make sure you know what you are doing and have a firm grasp of the concept and can prove it to yourself through testing ON YOUR SYSTEM as I have done. And regardless, this is not the practice that I go through routinely, I did it as an exercise and test of functionality. Seeing RED is something that should concern everyone. BK Attached Thumbnails __________________ Bob Katz DIGITAL DOMAIN http://www.digido.com "There are two kinds of fools. One says-this is old and therefore good. The other says-this is new and therefore better." No trees were killed in the sending of this message. However a large number of electrons were terribly inconvenienced. Last edited by bob katz; 10th May 2006 at 02:21 PM.

9th May 2006, 06:30 PM	#11
innesireinar Gear Head Join Date: Sep 2004 Posts: 71	No, I'm referring to the master track of PT. Follow me, by doing a session like the one I've done for the dat transfer with only a stereo track and a stereo master track with the stereo track directly routed to the master track without any plugs in the chain, I've noted that sometimes the overled of the stereo track while those on the stereo master don't. Is it possible that inside the same DAW the meters of the recording tracks are calibrated different than those of the master, and aux channels?

10th May 2006, 03:03 AM	#15
yeloocproducer Lives for gear Join Date: Dec 2004 Location: los angeles Posts: 934	Paul, couple questions... do you only advise trim plugs at the top of the channels, or just lowering the "input" stage on your first plugin (such as an eq), or would either be sufficient? I think a lot of folks just want an easy to follow rule of thumb... like, if the tracks were recorded near 0, put a trim in at -6 before other plugins so that you get some headroom for your compression, eq, and harmonic distortion plugs... and that should get rid of most of your "invisible" overs. Also keep the outputs of your plugs at -6 peak. Plus, keep an eye on your master faders anytime you have audio getting converted back to analog... whether stemming out and summing, or running your 2 mix out to an outboard compressor, etc. In this case, make sure that level is again peaking (on the digital meters) at no greater than -6db on any given stem or your master so that you don't create reconstruction overs. And when using "upsampling" plugins, be very very careful that you set thier outputs lower as well b/c they might be creating reconstruction overs as well when they bring things back down. Am I oversimplying, or is this the overall "how to"? I've been following the threads here and at prosoundweb and this is what I've put together. Please correct me if I got any of that wrong.

10th May 2006, 02:10 PM	#19
innesireinar Gear Head Join Date: Sep 2004 Posts: 71	[quote=bob katz]I believe so! I haven't seen anything on the LE system that would qualify for more than an "idiot light". I suggest Inspector XL or in your case something better that is not yet available as a plugin for Pro Tools, that interpolates and calculates intersample peaks. But is it an LE issue only? Or even in HD systems this happens?

10th May 2006, 05:41 PM	#25
yeloocproducer Lives for gear Join Date: Dec 2004 Location: los angeles Posts: 934	Gotcha. Reconstruction errors are only on the D to A process. Thanks Paul for clearing things up, including the dither on the trim, I was wondering about that too. Is there a specific reason you shouldn't use the input on the first plugin at -6 if it's a double precision, high quality eq plug like sony or URS? Last edited by yeloocproducer; 10th May 2006 at 05:43 PM.