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 26th July 2006
  #91
Gear Addict
 
bonne's Avatar
 

I'll look into the Cool Edit Pro wave editor when I get a chance to see what the reported oversampled drawing of the waveform is all about.

Quote:
Originally Posted by Paul Frindle
IME the greatest cause of recon errors are actually quiet parts of the music that have been boosted in the mix or by comps and limiters. E.g. Sparse portions of Shania Twain produce much larger and more frequent overs (if boosted) than Green Day at full pelt!! The corellation between the overs and what one would expect from the material itself is not at all intuitive at first sight.

Unfortunately a 1mS burst of something nasty is actually very audible indeed. The Oxford recon compensation overcomes this by dealing with the errors before they arrive at the output (i.e. using a fairly long look ahead). This is why it has quite a long latency..
What you're saying about the quieter parts on the Shania Twain record is quite weird, isn't it? Not what one would expect. Do you think this has to do with reckless filtering and stuff in the production of that particular record or could this happen on any program just by boosting the level of quiter parts, say, by automation or by comps and limiters?

In that case it would be difficult, if not impossible, after tracking from your ADC into the digital domain to stay ITB for all your mixing, processing and mastering from start to finish on a project, and get sound results with no illegal overs (-unless you had the Oxford Limiter, the TC 6000 brickwall or similar).

There doesn't seem to be another way for ITB people to "avoid making a file that could inexplicably start sounding rubbish on your client's (or other people's) systems!".
Old 26th July 2006
  #92
Lives for gear
 

Quote:
Originally Posted by bonne
I'll look into the Cool Edit Pro wave editor when I get a chance to see what the reported oversampled drawing of the waveform is all about.



What you're saying about the quieter parts on the Shania Twain record is quite weird, isn't it? Not what one would expect. Do you think this has to do with reckless filtering and stuff in the production of that particular record or could this happen on any program just by boosting the level of quiter parts, say, by automation or by comps and limiters?
Yes that is indeed so - even increasing a fader in the digital domain can produce an illegal reconstructed output without showing overs on the DAW meters.

I wrote some stuff about this on the website:

http://www.sonyoxford.co.uk/pub/plug...tail.htm#recon

Quote:
In that case it would be difficult, if not impossible, after tracking from your ADC into the digital domain to stay ITB for all your mixing, processing and mastering from start to finish on a project, and get sound results with no illegal overs (-unless you had the Oxford Limiter, the TC 6000 brickwall or similar).
No not at all - everything is ok providing that your levels are ok within the DAW (and plugs) and your final output is error free :-) One sure fire way to ensure the output is error free is to reduce it to compensate. For instance if you reduce it to -3dBr you will avoid all but the most stubborn errors (good enough for Shania Twain over EQ'd stuff), if you reduced it to -6dBr there would be virtually no possibility of an error even under the worst reallife conditions you can imagine :-)

However the main advantage of something like the Oxford limiter is that it can do this without reducing the average modulation (and loudness) of your programme. It can even remove the errors whilst significantly boosting the loudness of the output :-)
Old 26th July 2006
  #93
Gear Nut
 
innesireinar's Avatar
 

Hi Paul, nice to hear you.
I've just finished my last pj where every tracks have been processed by sonylimiter.
BTW I've noticed some occasional reds with recon meter even with autocomp enabled. In the manual you reffered to something wrong that could happen on LE version about this, but could these (autocomp-ed) reds cause illegals in the converters?
After finish my mixes I've listened to them into my worse sys (my car) and I've noticed some distortions, expecially when the the BD hits even at low level, therefore I think this could be caused by my carstereo converters, but I've notice less problems with commercial material, even with this very louder.
A good analog meters (something like Dorrough) can help?
Another Q. Is there anyone at sony who can port the plugins on the new mactell platform, now that you no longer are there?

Cheers

ranieri senni
Old 26th July 2006
  #94
Gear Addict
 
bonne's Avatar
 

Quote:
Originally Posted by Paul Frindle
- everything is ok providing that your levels are ok within the DAW (and plugs) and your final output is error free :-) One sure fire way to ensure the output is error free is to reduce it to compensate. For instance if you reduce it to -3dBr you will avoid all but the most stubborn errors (good enough for Shania Twain over EQ'd stuff), if you reduced it to -6dBr there would be virtually no possibility of an error even under the worst reallife conditions you can imagine :-)

However the main advantage of something like the Oxford limiter is that it can do this without reducing the average modulation (and loudness) of your programme. It can even remove the errors whilst significantly boosting the loudness of the output :-)
Thanks for clearing this up, Paul.

Checking with the RME Digicheck, as Bob K. suggested, or possibly with the Cool Edit Pro oversampled wave editor display should indicate if dropping the output level is neccesary or not. With good level practices in the mix and processing one could avoid dropping the output level ceiling and get a good strong and clean signal peaking at -0.3dBFS.
Old 26th July 2006
  #95
Lives for gear
 

Quote:
Originally Posted by innesireinar
Hi Paul, nice to hear you.
I've just finished my last pj where every tracks have been processed by sonylimiter.
BTW I've noticed some occasional reds with recon meter even with autocomp enabled. In the manual you reffered to something wrong that could happen on LE version about this, but could these (autocomp-ed) reds cause illegals in the converters?
I am a bit perplexed by this meter problem you are having - this is a bit like the previous one you reported where the buss meter read overs that channel meters didn't?
It is true that the LE meters do not read exactly the same as the TDM ones - they can show an over which does no correspond to max at the fixed point ouput *i.e. 16bits etc).. Providing you go by the meter on the Limiter plug-in all should be ok.

Quote:
After finish my mixes I've listened to them into my worse sys (my car) and I've noticed some distortions, expecially when the the BD hits even at low level, therefore I think this could be caused by my carstereo converters, but I've notice less problems with commercial material, even with this very louder.
To understand if this is the limiter setting or your system I would ned to know the actual settings you are using. Generally provided the enhance is not used too heavily I would not expect distortions of this kind - unless you are using a fast release time without the auto gain function selected? In this case gain excursions can happen fast enough to create LF distortion. This is sometime wanted and there is at least 1 bundlued setting that does this on purpose.

If you can send me the set up I'll look at it for you?

Quote:
A good analog meters (something like Dorrough) can help?
Another Q. Is there anyone at sony who can port the plugins on the new mactell platform, now that you no longer are there?
I'm not sure what the Durrough meters actually do these days - but yes they could help if they are fast enough and you can calibrate them as I suggested earlier :-)

As for porting to mactell and new PT releases; it wasn't me that did this stuff (there is another programming expert who handles this), so my departure should not stop this happening at all. My job was only to conceive and design the applications themselves :-)

Cheers

ranieri senni[/QUOTE]
Old 23rd September 2006
  #96
Gear Addict
 
bonne's Avatar
 

Hi again, Paul.
A couple of months have passed and I've been trying to wrap my head around the good info you have offered here and elsewhere. Hope you're still visiting here from time to time.

I have a few more questions, if I may:

One has to do with the ITB headroom issue when processing with fixed point plug ins in a floating point DAW.

Earlier you wrote ("sticky" thread, Reason in Audio, Prosoundweb):
With a complete floating point math system values above notional maximum are not clipped so it is admissible to overdrive signals - they can still be recovered by reducing level down line. This is also true of the Oxford EQ RTAS or LE - no clipping will occur within the plug-in.
However if you use a fixed point math processing expansion system the signals WILL get clipped as they go into and out of the expansion processor - even though the host mixer application is floating point. SO in this case you must avoid the clipping - as it cannot be recovered down line.


Could you safely ignore the -6dB step down precaution before each fixed point plug in provided you reduced level before leaving the floating point system? For instance Waves plugs in Nuendo.



A related question has to do with dithering in a floating point DAW. Could you by using the dither function in a fixed point plug in, for instance Waves L2, avoid the known difficulties with dither in float and achive proper dither this way?

Kind regards

Jørn Bonne
Old 2nd October 2006
  #97
Lives for gear
 

Yes, I'm still around and still vaguely sane :-)

Quote:
Originally Posted by bonne View Post
Hi again, Paul.
A couple of months have passed and I've been trying to wrap my head around the good info you have offered here and elsewhere. Hope you're still visiting here from time to time.

I have a few more questions, if I may:

One has to do with the ITB headroom issue when processing with fixed point plug ins in a floating point DAW.

Earlier you wrote ("sticky" thread, Reason in Audio, Prosoundweb):
With a complete floating point math system values above notional maximum are not clipped so it is admissible to overdrive signals - they can still be recovered by reducing level down line. This is also true of the Oxford EQ RTAS or LE - no clipping will occur within the plug-in.
However if you use a fixed point math processing expansion system the signals WILL get clipped as they go into and out of the expansion processor - even though the host mixer application is floating point. SO in this case you must avoid the clipping - as it cannot be recovered down line.


Could you safely ignore the -6dB step down precaution before each fixed point plug in provided you reduced level before leaving the floating point system? For instance Waves plugs in Nuendo.
I'm not completely sure what you mean here - it could be 2 things; clipping or headroom? But the fundamental point is that any sample values generated in the float system that are greater than +/-1 (flat out wrt fixed point) will get clipped off at the transfer to fixed point - and therefore will not make it accurately to the plug-in at all. Therefore they cannot be recovered anywhere in the system - not even within the plug-in itself.
Whereas such values if remaining completely in the float domain (i.e not passed to a fixed point processor or interface) would remain unclipped and could be recovered by reducing faders aferwards etc.. So really to avoid clipping using fixed point expansion cards like PowerCore you must strictly observe levels at the fixed point boundaries - it doesn't really matter where the excess level is lost as long as it IS lost before the float to fixed transfer.

O.K. now if we are talking about headroom (level where excess signals can go without being clipped) it can be seen that a system running completely float using no fixed point processing anywhere at all (including within plug-ins) has a natural built in headroom. I.e. you can pass levels that are way too big around the place without drastic loss - provided you reduce it all correctly before it finally goes out to the fixed point media (CD etc).

If you want to get a similar situation (of enhanced level freedom) with fixed point systems (such as TDM etc.) you can only do that be electing to reduce levels at the start of your processing (by losing level) then aiming for modulation levels that are averagely below max modulation - then bumping it up at the end to get a fully modulated final mix output..


Quote:
A related question has to do with dithering in a floating point DAW. Could you by using the dither function in a fixed point plug in, for instance Waves L2, avoid the known difficulties with dither in float and achive proper dither this way?

Kind regards

Jørn Bonne
Hmm.. Im not at all impressed with this much advertised 'dither difficulty' in float systems. I am not sure where this idea comes from and why people believe it. It really isn't true in practice..

To understand all this (quite simply with common sense) it is only necessary to think of float and fixed point math conventions as nothing more than numerical representations. Providing that the data is presented at the required precision with the required dither for the eventual target output data width, it is quite irrelevant which particular number system happens to be used for the processing itself. They are just numbers - and there is no magic :-)

So accordingly, if a float processing system must finally present data to a fixed point media (such as CD - or anything else in the real world) all that is needed is to add the dither at the correct level within the float domain as the final process, so that when it gets transfered to the final output data media the dither is correct for that output data width. Providing the float system can calculate to an equal or greater accuracy than the output fixed data width, there is nothing preventing the dither from working identically - regardless of whether it was generated in float or fixed math representation.

It really is as simple as that - I hope this makes sense :-)
Old 2nd October 2006
  #98
Lives for gear
hello paul,
as always thank you for your informative and generous posts. i did a search and have not been able to find concrete answers for these questions, so if you've already answered them i apologize:

1) would a 32 bit floating point mixer be better than a 48 bit mixer? i read, but can't remember the source, that 48 bits is theoretically better but 32 bit floating point is easier to program for.

2) if i use the PT TDM mixer and use plugs like Oxford or the new Waves SSL, would the -6db cushion be able to be done away with? i've noticed (or imagined) that SSL plugs actually sound better when being hit harder.

if the answer is yes, do you think converters such Digi 192 or Apogee 16X are able to reconstruct that information accurately before going out to analog gear?

i hope these questions make sense and thank you for your time in advance.
Old 2nd October 2006
  #99
Lives for gear
 

Quote:
Originally Posted by raal View Post
hello paul,
as always thank you for your informative and generous posts. i did a search and have not been able to find concrete answers for these questions, so if you've already answered them i apologize:

1) would a 32 bit floating point mixer be better than a 48 bit mixer? i read, but can't remember the source, that 48 bits is theoretically better but 32 bit floating point is easier to program for.
In the end there is nothing intrinsically 'better' about the math itself (float or otherwise) - as ever it's a question of the quality of the application
My opinion would be that 32 bits float may be pushing it and could make life difficult for the designer as it has a 24 bit mantissa. However please note that most commercial host processors are running very considerably higher precisions than this (64 bit float) - at which point it doesn't matter anymore. Programming for high precision float is easier than fixed point, as scaling and relative levels are of virtually no consequence..

Quote:
2) if i use the PT TDM mixer and use plugs like Oxford or the new Waves SSL, would the -6db cushion be able to be done away with? i've noticed (or imagined) that SSL plugs actually sound better when being hit harder.

if the answer is yes, do you think converters such Digi 192 or Apogee 16X are able to reconstruct that information accurately before going out to analog gear?

i hope these questions make sense and thank you for your time in advance.
From the point of view of sample value overloads, I would allow some headroom myself. It may be the case (as with the Oxfords) that the plug-in itself has loads of internal headroom, however the artistic freedom some headroom provides is still very useful inthe mixing process.

From the point of view of intersample overloads in the DAC playing your final product, the issue is to deal with this as a final process - last thing before the mix output.
Your best choice is to use a limiter that compensates dynamically for intersample overloads (such as the Oxford limiter and some others).
The next best thing is to at least get a meter plug that shows you the overloads and adjust levels manually to avoid them (this however will cost you some loudness).
The last option in the absence of either of the above is to err on the side of caution and aim to modulate at least 3dB below maximum, not least of all since this should avoid YOUR DACs clipping and give you a chance to hear it like it is :-)

This is particularly important if your mix is going to an additional mastering phase where more processing will inevitably occur. The mastering engineer should be better placed to decide the final levels and should have the tools and experience to judge this correctly (hopefully?).. I can't really say how the Apogees or the Digi converters deal with intersample overloads - but I wouldn't trust it personally..

As for pushing plug-ins harder - for linear processes (like EQ) of high quality, providing the level is within legal range the sound should not change in relation to input level. For lower quality plugs (that may have low level inaccuracies, noise and non-linearities) pushing the level may overcome some of these shortcomings - and they may (or may not) improve.

For non-linear processes (such as the Limiter, some settings of the Dynamics and all settings of the Inflator) there is a direct and deliberate relationship between harmonic content (distortion) and the internal plug-in level - this is why generous input gain control (and make-up) is provided. In this case, providing the signal gets into the plug at a legal level, the appropriate sound should be obtainable by changing plug-in input gains to get the right internal mod levels etc..

I hope this helps :-)
Old 2nd October 2006
  #100
Lives for gear
Quote:
Originally Posted by Paul Frindle View Post
For non-linear processes (such as the Limiter, some settings of the Dynamics and all settings of the Inflator) there is a direct and deliberate relationship between harmonic content (distortion) and the internal plug-in level - this is why generous input gain control (and make-up) is provided. In this case, providing the signal gets into the plug at a legal level, the appropriate sound should be obtainable by changing plug-in input gains to get the right internal mod levels etc..

I hope this helps :-)
it helps alot. thank you very much again paul.
Old 3rd October 2006
  #101
Gear Addict
 
bonne's Avatar
 

Quote:
Originally Posted by Paul Frindle View Post
Yes, I'm still around and still vaguely sane :-)



I'm not completely sure what you mean here - it could be 2 things; clipping or headroom? But the fundamental point is that any sample values generated in the float system that are greater than +/-1 (flat out wrt fixed point) will get clipped off at the transfer to fixed point - and therefore will not make it accurately to the plug-in at all. Therefore they cannot be recovered anywhere in the system - not even within the plug-in itself.
Good to see you back on the forum, Paul.

What I was getting at has to do with information Bob Katz supplied on page 1 of this thread.

Bob descibed a test where he chained two fixed point plugs, Waves C1 and L2, in a floating point DAW. He found that when increasing the level out of the C1 to well above O dBFS the following L2 read the floating point "signal" from the C1 (via the DAW?) without causing any distortion/clipping.

My concern is whether this would be the case with any chained fixed point plugs in a floating point DAW, e.g. Nuendo.

If not one would still have to "drop level" between plugs like you have recommended, even in a floating point DAW.



About the floating point dither difficulty, Paul.

My concern stems from a recent thread here, "Exporting 24 bit masters from 32 bit Nuendo Projects", where one of the posters supplied this link:

http://www.cadenzarecording.com/imag...tingdither.pdf

In this paper by Nika Aldrich the floating point dither difficulty is explained, with the conlusion that "floating can't be properly dithered".

Cheers

Jørn
Old 3rd October 2006
  #102
Lives for gear
 
Nordenstam's Avatar
 

Verified Member
I'm the one waving the 'dither and floating point is incompatible' flag while refering to the Nika Aldrich link in the other thread. My humble apologies to Jørn, BK and everyone else if it was wrong! As Paul presented the problem above, it didn't seem to be any problem at all.

Though, Nika Aldrich's explanation was at least as convincing. Partially because of the strong source credits:
Quote:
Originally Posted by Nikas floating dither paper
Special thanks are due to Dr. Stanley Lip****z, and John Vanderkooy of the Audio Research Group at the University of Waterloo, Glenn Zelniker of Z-Systems Audio Engineering, and Paul Frindle of Sony Corporation’s Oxford Division for their comments and editing of this paper.

.. this is getting interesting!
Old 4th October 2006
  #103
Lives for gear
 

Quote:
Originally Posted by bonne View Post
Good to see you back on the forum, Paul.

What I was getting at has to do with information Bob Katz supplied on page 1 of this thread.

Bob descibed a test where he chained two fixed point plugs, Waves C1 and L2, in a floating point DAW. He found that when increasing the level out of the C1 to well above O dBFS the following L2 read the floating point "signal" from the C1 (via the DAW?) without causing any distortion/clipping.

My concern is whether this would be the case with any chained fixed point plugs in a floating point DAW, e.g. Nuendo.

If not one would still have to "drop level" between plugs like you have recommended, even in a floating point DAW.
Since the DAW is a float format and signals are passed as float values, the plug's internal headroom scaling can allow larger than full level signals to be passed (up to some limit) into the float interfaces.
The key here is that the signals are not passed between plugs via a fixed point interface :-) Although I must say that it beats me why anyone would make a fixed point plug aimed to be used on a float DAW?
And you are right, that in this case the level drop to avoid clipping would not be needed :-)

Quote:

About the floating point dither difficulty, Paul.

My concern stems from a recent thread here, "Exporting 24 bit masters from 32 bit Nuendo Projects", where one of the posters supplied this link:

http://www.cadenzarecording.com/imag...tingdither.pdf

In this paper by Nika Aldrich the floating point dither difficulty is explained, with the conlusion that "floating can't be properly dithered".

Cheers

Jørn
Yes I am aware of this. Scores of emails passed between Nika and I before he decided to write his book etc.. He is not wrong in principle, there is a theoretical issue with dithering float signal when it remains in float, regarding the discontinuity caused when scaling boundaries are crossed. BTW this can be avoided anyway if the design scales in such a way as to only cross the boundaries in the headroom overload part of the range. Remember that a floating point representation is only a fixed point mantissa with a scaling exponent - it is possible to treat it like a fixed point system and avoid crossing the scaling boundaries. In this case a 32 bit float's 24 bit mantissa could yeald 144dB of dithered dynamic range without cossing a scaling boundary for all signal below flat out for the fixed point target. A 64 bit float with say for argument a 48 bit mantissa, could yield 288dB.


However in practice it is obvious that providing sufficient accuracy is achieved in the float domain - no dithering is required in the high precision float domain itself (because the error is vanishingly small - several 100's of dB down) - and by the same token - dithering signals that are aimed at fixed point conversions can be easily done in the float domain without problem. The problem only becomes significant IF the intrinsic accuracy of your float representation is less than that required for the intended fixed point output data width..
Old 4th October 2006
  #104
Lives for gear
 

Quote:
Originally Posted by Lupo View Post
I'm the one waving the 'dither and floating point is incompatible' flag while refering to the Nika Aldrich link in the other thread. My humble apologies to Jørn, BK and everyone else if it was wrong! As Paul presented the problem above, it didn't seem to be any problem at all.

Though, Nika Aldrich's explanation was at least as convincing. Partially because of the strong source credits:



.. this is getting interesting!
Please read above - there is no conflict between Nika's paper and the practical situation I am describing. There is nothing wrong with Nika's explanation.

We are not talking about dithering a float signal for it's own benefit - we are talking about adding dither to a float signal for the sole purpose of sending it into a fixed point (real world) target format. This is not the same thing at all :-)
Old 4th October 2006
  #105
Gear Addict
 
bonne's Avatar
 

Quote:
Originally Posted by Paul Frindle View Post
Since the DAW is a float format and signals are passed as float values, the plug's internal headroom scaling can allow larger than full level signals to be passed (up to some limit) into the float interfaces.
The key here is that the signals are not passed between plugs via a fixed point interface :-) Although I must say that it beats me why anyone would make a fixed point plug aimed to be used on a float DAW?
And you are right, that in this case the level drop to avoid clipping would not be needed :-)



Yes I am aware of this. Scores of emails passed between Nika and I before he decided to write his book etc.. He is not wrong in principle, there is a theoretical issue with dithering float signal when it remains in float, regarding the discontinuity caused when scaling boundaries are crossed. BTW this can be avoided anyway if the design scales in such a way as to only cross the boundaries in the headroom overload part of the range. Remember that a floating point representation is only a fixed point mantissa with a scaling exponent - it is possible to treat it like a fixed point system and avoid crossing the scaling boundaries. In this case a 32 bit float's 24 bit mantissa could yeald 144dB of dithered dynamic range without cossing a scaling boundary for all signal below flat out for the fixed point target. A 64 bit float with say for argument a 48 bit mantissa, could yield 288dB.


However in practice it is obvious that providing sufficient accuracy is achieved in the float domain - no dithering is required in the high precision float domain itself (because the error is vanishingly small - several 100's of dB down) - and by the same token - dithering signals that are aimed at fixed point conversions can be easily done in the float domain without problem. The problem only becomes significant IF the intrinsic accuracy of your float representation is less than that required for the intended fixed point output data width..
Thanks for the explanation.

I was puzzled by the "floating can't be properly dithered" claim and at the same time seeing Izotope Ozone's MBIT+ dither algo getting very high marks in dither shoot-outs. Ozone is 64-bit float.
Old 4th October 2006
  #106
Lives for gear
 

Quote:
Originally Posted by bonne View Post
Thanks for the explanation.

I was puzzled by the "floating can't be properly dithered" claim and at the same time seeing Izotope Ozone's MBIT+ dither algo getting very high marks in dither shoot-outs. Ozone is 64-bit float.

Yes - this argument is largely founded on the particular issue of whether float is more efficient than fixed if you have a limited data width which is too close to the required final accuracy. This is where the explanation in Nika's paper applies.

For instance, if you have only say 32 data bits (real internal data buss lines) and you were aiming at 24 bit output, for the most part a designer is most probably better off using them as fixed point rather than float - provided that the design can do the appropriate fixed scaling optimally. This is because by losing the 8 scaling bits assigned by float, you can re-use the data width as 'real' signal and do a more limited and optimal scaling yourself (i.e. not having 8 bits of permanent scaling you will never need, wasting your data width) etc..

You can think of it a bit like; having a 24bit mantissa and an 8 bit scaling - deciding you don't need 8 bits of scaling but you do need more mantissa accuracy, so deciding to reduce the scale bits to get more mantissa bits - then deciding that 'stuff it' I'll have all the bits as mantissa and do the darned scaling myself when I need values notionally greater than +/-1 :-) Of course this isn't possible with commercial processors that come with predefined represntations - but you get the idea?

However once you get to 48 or 64 bits it's all irrelevant. At the point where the mantissa has enough accuracy for ALL the needs of your algorithms, float is simpler to design with as you can throw numbers at it 'willy-nilly' and it sorts the scaling out itself, without damaging losses of accuracy (very simplistic explanation here).
Old 5th October 2006
  #107
Lives for gear
 
Nordenstam's Avatar
 

Verified Member
Hi!

You're truly a bountiful resource! Much appreciated.

So if I got this right, the only time there's any need to worry about dithering floating point is if the exponent scales the LSB of the mantissa above the LSB of the destination integer word length? Basicly the same problems as if scaling an integer word so the LSB goes above the truncation point.

So why is the paper on Nikas page still publicly available? Isn't it a moot point?


Andreas
Old 5th October 2006
  #108
Lives for gear
 

Quote:
Originally Posted by Lupo View Post
Hi!

You're truly a bountiful resource! Much appreciated.

So if I got this right, the only time there's any need to worry about dithering floating point is if the exponent scales the LSB of the mantissa above the LSB of the destination integer word length? Basicly the same problems as if scaling an integer word so the LSB goes above the truncation point.

So why is the paper on Nikas page still publicly available? Isn't it a moot point?


Andreas
Yes - you've got the gist of it.

One would want to dither the float if the mantissa was close enough to the final output wordlength to create significant quantisation distortion. The point is that it is the mantissa which describes the actual quantisation of any value - however large or small..

Nika's paper is a valid analysis of the situation which is aimed at explaining the issues. It's not (AFAIK) attempting to assert whether or not any particular application will actually suffer because of it. However I'm not altogether sure that it is straight forward enough? I prefer to explain it in numerical terms because I think it is more easily understood this way?

Generally though, I worry that people are being encouraged (primarily by pseudo scientific marketing blurb) to get hung up on the 'internals' of processes by exchanging 'buzz words' and forgetting that the quality of an application is (as ever) defined by it's whole.... This is a bit like looney fringe analogue people dismissing a product due to the existence of some 'un-blessed' component. It does not make sense at all and it's a very limiting trap to fall into if you start believing it...

Very intrusive and aggressive marketing 'spin' that seeks to capitalise on any feature (however irrelevant) which also attempts to promote products and 'keep the subjects alive' in the market place by provoking a deliberately manufactured and largely erroneous feeling of 'user involvement', does nothing but confuse the hell out of people. In my opinion this is totally reprehensible and amounts to direct lies!

One really effective method they use to tell lies without actually contravening the law is to use what I call 'implied comparison', where they encourage you yourself to create the lie rather than them. So something as innocuous as a statement like "64 bit floating math throughout, gives our product incredible 'resolution' and sonic purity" immediately provokes you to think that it would not do so without it. Therefore any product without it is by implication inferior - that's YOUR thought, not their text. So we end with a mostly pointless argument about 'which is best, float or fixed'. Sigh!

In the midst of all this at every level in modern life, it's now up to us to unravel the lies and deceptions. Incredible :-(
Old 19th October 2006
  #109
M2E
Lives for gear
 
M2E's Avatar
 

Hi Paul,

Thank you so much for taking the time out to answer these great questions. I have a few myself if you don't mind as a current Inflator TDM owner.
I also remember you on the Digisite posting and also uploading a SSL G series preset for the Sony Dyn. Plugin that I heard was great.

This is my system in short,
888/24 w/2 1622's
Pro Toos Mix ++++++ in a Mac g4 Dual 1.25
Magma 6slot
PT 6.4.1cs4 software

1. Anyways, I'm a current Pro Tools Mix Plus owner. I noticed alot in other threads as well as this one that you mentioned to turn down the input on plugins to give more so called headroom on tracks and master fader. As I'm a owner of a Mix + System and not a HD System, does this still apply to me as well? Will my system do something different to the sound that the HD wouldn't?

2. If I turn down the input on a plugin or fader will it take away bits or degrade the original sound on my system?

3. Funny thing is that all this time I thought that my Mix Plus System was using a 48bit mixer plug as well. I was just told that is wasn't and that I have to be extra careful with using the input and output on plugins as it somehow messes with the image. Huh??? Really???

4. When I mix I usually don't push the master hard at all. I tend to use the Analog Channel then the Inflator then the Ozone (as it's 64bit internal). I would love to switch the Ozone and the Inflator around but, I use the Stereo Imaging in the Ozone a little and didn't know if, I put the Inflator after that will it thin out the imaging back to original width?

5. As I still use 24bit/44.1, I saw no need to update to HD. But you as a programmer for those incredible plugins that you've done, and knowing these programs like you do, am'I missing out? Will my plugins sound better going HD?
I produce and write R&B/Hip Hop and the lowend is very very important. If I move up to PTHD will I notice a huge difference in the Inflator and how it reacts to lowend?

6. Not sure if you've answered this question already which, I'm sure you've been asked before but, I email Colin @ MCDSP about his rtas/tdm plugins and the sound difference between the two and he said there was no sound difference in either. You said that as well I think in this post or another one on here. But, my question is, if I use an rtas plugin before a tdm, would that make a difference more than if, I chose to do that with tdm to tdm? Should I always try to use a tdm to tdm more than an rtas to tdm?
Hope I didn't confuse you on this one.

7. Are all the plugins you made for Sony, 48bit Double precision? And does that have a huge factor on the sound quality of the plugin? When I demoed the CraneSong I emailed him asking if that plugin was 48bit DP and he said no. It didn't need to be. I was told by someone at Digi that it's better to use DP plugins but, when I asked him for a list of the DP plugins for Digi, he didn't have a list and wouldn't tell me which digi plugins were DP. This was about 2 years ago. Now I noticed that all the new plugins that digi is bringing out are DP.
Does it truely matter? Especially for my system which is I'm told only 40bit.

8. Last and final question, do you have a site that people can download your presets of special units for the Sony plugins like the SSL G series buss compressor one you had for the Sony Dynamics plugin? And do you have a personal site of information about plugin design?

Again thanx ahead of time for your time. I'm going to go through all of your postings here as I've never heard it broken down by anyone like you. I totally get it and love how you really explain things. WOW!!!! Your knowledge is huge and inspirational as you've self taught yourself. Again WOW!!!!

Thanx, and please forgive me for any misspelled words or phrases.

M2E
Old 20th October 2006
  #110
Lives for gear
 

Quote:
Originally Posted by M2E View Post
Hi Paul,

Thank you so much for taking the time out to answer these great questions. I have a few myself if you don't mind as a current Inflator TDM owner.
I also remember you on the Digisite posting and also uploading a SSL G series preset for the Sony Dyn. Plugin that I heard was great.
Thanks for the kind comment. No problem with helping if I can :-)
Quote:

This is my system in short,
888/24 w/2 1622's
Pro Toos Mix ++++++ in a Mac g4 Dual 1.25
Magma 6slot
PT 6.4.1cs4 software

1. Anyways, I'm a current Pro Tools Mix Plus owner. I noticed alot in other threads as well as this one that you mentioned to turn down the input on plugins to give more so called headroom on tracks and master fader. As I'm a owner of a Mix + System and not a HD System, does this still apply to me as well? Will my system do something different to the sound that the HD wouldn't?

2. If I turn down the input on a plugin or fader will it take away bits or degrade the original sound on my system?
Yes I do advocate making yourself some headroom and not pushing things to max all the time. I have been having a discussion about this only today and I've set out my views here:

https://www.gearslutz.com/board/showt...t=72794&page=5

The last couple of pages of this thread is worth a read :-)

Quote:

3. Funny thing is that all this time I thought that my Mix Plus System was using a 48bit mixer plug as well. I was just told that is wasn't and that I have to be extra careful with using the input and output on plugins as it somehow messes with the image. Huh??? Really???
My understanding is that the interface between the plugs is 24 bits, exactly the same as the HD - so I don't believe there's a problem.

But older systems apparently drop down to 24 bits to pass signals between parts of the mixer. This might at the limit affect things in the mixer itself, but it will be a very small effect IMO.

Quote:
4. When I mix I usually don't push the master hard at all. I tend to use the Analog Channel then the Inflator then the Ozone (as it's 64bit internal). I would love to switch the Ozone and the Inflator around but, I use the Stereo Imaging in the Ozone a little and didn't know if, I put the Inflator after that will it thin out the imaging back to original width?
The inflator does affect the image dynamically to some extent because it deals with L/R and centre signals variously depending on setting and what is happening in the programme. The least imaging changes happen with the curve control in centre position. Not knowing much about ozone and how you are setting things I can't say what's best, but generally putting the inflator last is the best idea if you can do it IMO..

Quote:
5. As I still use 24bit/44.1, I saw no need to update to HD. But you as a programmer for those incredible plugins that you've done, and knowing these programs like you do, am'I missing out? Will my plugins sound better going HD?
I produce and write R&B/Hip Hop and the lowend is very very important. If I move up to PTHD will I notice a huge difference in the Inflator and how it reacts to lowend?
Not at all. The Oxford plugs perform identically on both systems :-) Obviously I can't vouch for others, but I would expect them to do as as well.

Quote:
6. Not sure if you've answered this question already which, I'm sure you've been asked before but, I email Colin @ MCDSP about his rtas/tdm plugins and the sound difference between the two and he said there was no sound difference in either. You said that as well I think in this post or another one on here. But, my question is, if I use an rtas plugin before a tdm, would that make a difference more than if, I chose to do that with tdm to tdm? Should I always try to use a tdm to tdm more than an rtas to tdm?
Hope I didn't confuse you on this one.
Firstly, obviously it's possible for plugs to be different between RTAS and TDM, the code is of course different and it's a different instantiation. All I can say is that we went to lots of trouble to ensure that the Oxford plugs are identical. It should make no difference which you use and in what order..

Quote:
7. Are all the plugins you made for Sony, 48bit Double precision? And does that have a huge factor on the sound quality of the plugin? When I demoed the CraneSong I emailed him asking if that plugin was 48bit DP and he said no. It didn't need to be. I was told by someone at Digi that it's better to use DP plugins but, when I asked him for a list of the DP plugins for Digi, he didn't have a list and wouldn't tell me which digi plugins were DP. This was about 2 years ago. Now I noticed that all the new plugins that digi is bringing out are DP.
Does it truely matter? Especially for my system which is I'm told only 40bit.
All the Oxford TDM plugs operate DP 48 bits, because we need it to get the required performance and control ranges etc.. From my experience every time I thought we might get away with single precision on various parts of processes, in the end we had to resort to DP to get the required performance.
I know this depends on the quality one is aiming for - but IME I can't think of a single thing you can do without DP and end up with the performance I was after - not even a volume control! I am looking for 24 bit completely transparent performance..

Quote:
8. Last and final question, do you have a site that people can download your presets of special units for the Sony plugins like the SSL G series buss compressor one you had for the Sony Dynamics plugin? And do you have a personal site of information about plugin design?
I no longer work for Sony, but some of my set ups have been put on their site in the downloads section. I have done various pre sets from home as part of experiments and discussions on gearslutz and I have put them up from time to time for people to fiddle with them. I will search through and put up whatever I can find if that's helpful :-)

Quote:
Again thanx ahead of time for your time. I'm going to go through all of your postings here as I've never heard it broken down by anyone like you. I totally get it and love how you really explain things. WOW!!!! Your knowledge is huge and inspirational as you've self taught yourself. Again WOW!!!!

Thanx, and please forgive me for any misspelled words or phrases.
Thanks for the kind comments. I do run on perhaps too much on forums and I find myself saying the same things again and again as new threads fire up on past subjects etc.. But I have a passion to cut through the hype and try to install the truth. I too get lots from posting as it forces me to think about and verbalise my thoughts. In doing so I discover things as well, because I have to ask myself if something I take for granted is really true! A statement on-line can never be fully retracted!

I'm really glad that all this may have helped people :-)
Old 20th October 2006
  #111
Lives for gear
 
Empty Planet's Avatar
 

Well, I always read your posts with interest, and I'm fairly sure there's a silent army of lurkers that do so as well. The value of your contributions is inestimable. Thanks!

Cheers.


Old 20th October 2006
  #112
M2E
Lives for gear
 
M2E's Avatar
 

Quote:
Originally Posted by Paul Frindle View Post
Thanks for the kind comment. No problem with helping if I can :-)


Yes I do advocate making yourself some headroom and not pushing things to max all the time. I have been having a discussion about this only today and I've set out my views here:

https://www.gearslutz.com/board/showt...t=72794&page=5

The last couple of pages of this thread is worth a read :-)
Yes indeed. Thanx soooo much.

A few other points to talk to you about.

1. When using the inflator TDM/RTAS on the master fader. Does it make a difference if I use it in Multi-mono stereo or Multi Channel stereo? Does using it in Multi-mono make the stereo image a little bigger? Seems like it to me for some reason.

2. When using a plugin on a channel. In your experience, would it be better to turn down the volume on the plugin output and leave the channels fader at unity or turn down the channels fader so that you still have your headroom on your master fader? Or does it really make a difference?

3. You mentioned before that you were not familiar with the Izotope Ozone. Well, it's a kind of "do all plugin" where it has a Paragraphic EQ, Mastering Reverb, Loudness Maximizer, Harmonic enhancer, Multi channel compressor, and a Stereo Imager. I mainly use the Imager to widen out my songs and the Harmonic.
Well I tried to use the Inflator after this plugin and the image seemed to be intak but, sounded like there was a little distortion when turning on the Inflator. So that's one of the reason why I'm asking my second question. Would it be better if I turn down the master fader or turn down the output on the 64bit Ozone plugin.

4. Here's an outer/inner Pro Tools Question. Should it make a huge difference if I send the master fader output to a Alesis Masterlink like Bruce Sweiden does or do a Bounce To Disk in Pro Tools. On both mixes using only the Inflator on the master fader?

5. If I'm hearing distortion but, the input of the Inflator is not hitting the red, what would be better to to do? Turn down the Effect? Or input?

6. Did you have anythig to do with the Oxford Reverb plugin? If so, did you guys model it after anything?

Again, Thanx for all your help. Just thinking about questions as I'm working.

M2E
Old 20th October 2006
  #113
M2E
Lives for gear
 
M2E's Avatar
 

Sorry Paul, A couple more questions....

1. What is the ramifications for recording through TDM plugins?
example: I create a Audio track and an Aux track. My mic is going through the Aux channel and bussed to the Audio Track. I put on the Aux track a Sony Oxford Dynamics plugin. So I'm recording my vocal through the Oxford Dyn.
What are the real downsides to this other than a small amount of latency?

2. Also, is there a way that we can get rid of Latency in plugins? Can it be built with in the plugin or the program?

3. Would you ever think about doin what Steven Massey did when he left TL Labs and just start your own company and create new plugins?

Thanx

M2E again...
Old 21st October 2006
  #114
Lives for gear
 

Quote:
Yes indeed. Thanx soooo much.

A few other points to talk to you about.
Ok, I'll try to answer them as concisely as possible and not ramble on - as there are quite a few :-)

Quote:
1. When using the inflator TDM/RTAS on the master fader. Does it make a difference if I use it in Multi-mono stereo or Multi Channel stereo? Does using it in Multi-mono make the stereo image a little bigger? Seems like it to me for some reason.
Well it shouldn't theoretically make a difference. Although the process does cause our relational L/R perception to be modified, L and R channels are in fact processed entirely separately and independently.

Quote:
2. When using a plugin on a channel. In your experience, would it be better to turn down the volume on the plugin output and leave the channels fader at unity or turn down the channels fader so that you still have your headroom on your master fader? Or does it really make a difference?
The important thing is to avoid clipping the output of the plug-in into the TDM buss - this has no headroom of it's own. So if there's a risk of that, it's better to turn down the plug-in output.

Quote:
3. You mentioned before that you were not familiar with the Izotope Ozone. Well, it's a kind of "do all plugin" where it has a Paragraphic EQ, Mastering Reverb, Loudness Maximizer, Harmonic enhancer, Multi channel compressor, and a Stereo Imager. I mainly use the Imager to widen out my songs and the Harmonic.
Well I tried to use the Inflator after this plugin and the image seemed to be intak but, sounded like there was a little distortion when turning on the Inflator. So that's one of the reason why I'm asking my second question. Would it be better if I turn down the master fader or turn down the output on the 64bit Ozone plugin.
Provided that the Ozone is not clipping it should not matter where the gain change is made.
One thing that's worth remembering is that the inflator is a kind of distortion generator - it may not work well on some kinds of programme. Also you need to avoid using 2 distortion generation processes in series. It's possible that the Ozone maximiser (or other process) is generatng it's own distortion.

If you are trying to maximise the volume of highly processed programme from plugs that generate character distortion, the Oxford limiter is a better bet as this does not generate the same kind of long term distortion. It works in an entirely different way.

Quote:
4. Here's an outer/inner Pro Tools Question. Should it make a huge difference if I send the master fader output to a Alesis Masterlink like Bruce Sweiden does or do a Bounce To Disk in Pro Tools. On both mixes using only the Inflator on the master fader?
As far as I know the masterlink is a kind of integrated CD compilation and mastering box that lets you add effects, edit, compile and splice etc. If any of the internal effects are used then obviously this changes the sound. But if you are using it as nothing more than a digital recorder (bit bucket) then there should be no difference - numbers are numbers. But you should check that no internal effects are permanent or cannot be bypassed.

Quote:
5. If I'm hearing distortion but, the input of the Inflator is not hitting the red, what would be better to to do? Turn down the Effect? Or input?
Now this is a bit more complex - good question. Because it is non-linear, the process is naturally sensitive to modulation level, so reducing the input level is not the same as reducing the effect. It depends what you are doing with it:

If you are using it to get some warmth say on a single track or sound, turning the effect down will reduce the contribution of the inflator and is likely to be the best approach.

If you are using it to maximise programme density whilst avoidng overs, you are better off changing input level and leaving the effect in place - as this will not reduce the overload performance or the relative contribution of the processing.


Quote:
6. Did you have anythig to do with the Oxford Reverb plugin? If so, did you guys model it after anything?
Yes I made it from scratch myself. It was built from the ground up following 'sounds in my head' that I wanted to make. No Oxford plug-in is modelled on any existing device, as with so much to be done and so many ideas to realise, modelling other devices would be a sad lmitation. I rarely even look at other people's apps when I'm trying to realise something because I do not want to be limited or persuaded by existing ideas. I do however listen to the 'art' people are producing and try to support and enhance the direction they are moving :-)

Quote:
1. What is the ramifications for recording through TDM plugins?
example: I create a Audio track and an Aux track. My mic is going through the Aux channel and bussed to the Audio Track. I put on the Aux track a Sony Oxford Dynamics plugin. So I'm recording my vocal through the Oxford Dyn.
What are the real downsides to this other than a small amount of latency?
There isn't really a technical penalty in recording through a plug-in. But there is the practical risk that the effect can't be undone or modified if you need to later in the mix. With 24bit recording the signal to nose ratio is so high that you shouldn't need to record the compressed signal you send to the foldback for the artist.

The extra latency from the dynamics plug-in is much less than the ADCs and DACs already in the signal chain between the mic and foldback, so it shouldn't cause any further trouble.

Quote:
2. Also, is there a way that we can get rid of Latency in plugins? Can it be built with in the plugin or the program?
On something like the PT platform a couple of samples are required to get the signal to the plug-in and back again, so there's nothing to be done about that it's part of the platform. Extra latency beyond this can be part of the plug-in operation and processing and it can vary greatly between types of plug-in. For instance the Dyn has 20 samples of look ahead delay, but the EQ has no excess latency at all. For some plugs with excess delay, a low latency mode can be offered with slightly limited operational range. But there are some processes that simply cannot be reduced as they require the signal look ahead in time to function..

Quote:
3. Would you ever think about doin what Steven Massey did when he left TL Labs and just start your own company and create new plugins?
Well I have a great many very exciting things I would still like to make - I have hardly touched the surface with the plugs I have been involved with so far. Some of these functions I've wanted since the 1970's when such technology was all but inconceivable. It would be great to realise some of these, but I haven't yet decided how best to do so.. For one reason or another I have been pretty much idle for around 6 months now. But, yes I have considered what you suggest :-)
Old 21st October 2006
  #115
M2E
Lives for gear
 
M2E's Avatar
 

Quote:
Originally Posted by Paul Frindle View Post

Well it shouldn't theoretically make a difference. Although the process does cause our relational L/R perception to be modified, L and R channels are in fact processed entirely separately and independently.
When putting the Inflator or the Dynamics, what do you recommend?


Quote:
Originally Posted by Paul Frindle View Post

If you are trying to maximise the volume of highly processed programme from plugs that generate character distortion, the Oxford limiter is a better bet as this does not generate the same kind of long term distortion. It works in an entirely different way.
I'm very very interested in this. I will try the demo for this. Thanx

Quote:
Originally Posted by Paul Frindle View Post

As far as I know the masterlink is a kind of integrated CD compilation and mastering box that lets you add effects, edit, compile and splice etc. If any of the internal effects are used then obviously this changes the sound. But if you are using it as nothing more than a digital recorder (bit bucket) then there should be no difference - numbers are numbers. But you should check that no internal effects are permanent or cannot be bypassed.
Yeah, I don't use the internal efx at all. I was just using it for an external high end CD Recorder.

Quote:
Originally Posted by Paul Frindle View Post

Now this is a bit more complex - good question. Because it is non-linear, the process is naturally sensitive to modulation level, so reducing the input level is not the same as reducing the effect. It depends what you are doing with it:

If you are using it to get some warmth say on a single track or sound, turning the effect down will reduce the contribution of the inflator and is likely to be the best approach.

If you are using it to maximise programme density whilst avoidng overs, you are better off changing input level and leaving the effect in place - as this will not reduce the overload performance or the relative contribution of the processing.
Aahhhhh, ok. So when I turn down the effects slider it adds warmth. When you say warmth, what do you mean? Also how does it go about this?


Quote:
Originally Posted by Paul Frindle View Post

Yes I made it from scratch myself. It was built from the ground up following 'sounds in my head' that I wanted to make. No Oxford plug-in is modelled on any existing device, as with so much to be done and so many ideas to realise, modelling other devices would be a sad lmitation. I rarely even look at other people's apps when I'm trying to realise something because I do not want to be limited or persuaded by existing ideas. I do however listen to the 'art' people are producing and try to support and enhance the direction they are moving :-)
WOW!!!! THAT'S AMAZING!!! I got to hear it. Thanx

Quote:
Originally Posted by Paul Frindle View Post

There isn't really a technical penalty in recording through a plug-in. But there is the practical risk that the effect can't be undone or modified if you need to later in the mix. With 24bit recording the signal to nose ratio is so high that you shouldn't need to record the compressed signal you send to the foldback for the artist.

The extra latency from the dynamics plug-in is much less than the ADCs and DACs already in the signal chain between the mic and foldback, so it shouldn't cause any further trouble.
This is great to know. So really, theres no advantage to recording in outboard compressors to Pro Tools unless of course you really like that compressor's sound than in ITB compressors if you happen to like that Pro Tool's compressor more.
Also you're saying that the latency is more in an ADC/DAC than latency in a plugin compressor like the Sony Dyn? WOW....Thats good to know.


Quote:
Originally Posted by Paul Frindle View Post

On something like the PT platform a couple of samples are required to get the signal to the plug-in and back again, so there's nothing to be done about that it's part of the platform. Extra latency beyond this can be part of the plug-in operation and processing and it can vary greatly between types of plug-in. For instance the Dyn has 20 samples of look ahead delay, but the EQ has no excess latency at all. For some plugs with excess delay, a low latency mode can be offered with slightly limited operational range. But there are some processes that simply cannot be reduced as they require the signal look ahead in time to function..
So let me ask you. Can this look ahead function latency be lowered some how. Can plugin companies give the user a way to control the latency? Also does this add so much better performance and/or quailty that the plugin/s have to have this to be the plugin/s they are? Thanx



Quote:
Originally Posted by Paul Frindle View Post

Well I have a great many very exciting things I would still like to make - I have hardly touched the surface with the plugs I have been involved with so far. Some of these functions I've wanted since the 1970's when such technology was all but inconceivable. It would be great to realise some of these, but I haven't yet decided how best to do so.. For one reason or another I have been pretty much idle for around 6 months now. But, yes I have considered what you suggest :-)
Hey Paul. On the real. How long does it normally take to create a finished plugin? Also, in terms of money, how much would it take to do a Inflator or Limiter or EQ or Compressor plugin? Take ya pick or answer one at a time. Just a round about number.
I've always wanted to know this.

Again as always, Thanx a milion and a half!!!

M2E
Old 21st October 2006
  #116
Lives for gear
 

Quote:
Originally Posted by M2E View Post
When putting the Inflator or the Dynamics, what do you recommend?
The dynamics works best as a stereo instantiation as the side chain works on both L and R signals simultaneously. Whereas with the inflator it does not matter.

Quote:
Aahhhhh, ok. So when I turn down the effects slider it adds warmth. When you say warmth, what do you mean? Also how does it go about this?
In a word distortion, but hopefully nice distortion. When the effect slider is at zero it does nothing. As the silder is increased the effect is added. When it is at full the effect is fully added - and a side effect of this full setting is that it will not allow peaks beyond full sample value :-)

Quote:
This is great to know. So really, theres no advantage to recording in outboard compressors to Pro Tools unless of course you really like that compressor's sound than in ITB compressors if you happen to like that Pro Tool's compressor more.
Also you're saying that the latency is more in an ADC/DAC than latency in a plugin compressor like the Sony Dyn? WOW....Thats good to know.
Thats right - the only sensible technical reason for using an outboard processor is to get an effect you like that is unavailable in the box. But obviously if the unit is analogue the extra ADC and DAC loop will add some degradation and extra latency etc.. although it may still be worth it :-)

Quote:
So let me ask you. Can this look ahead function latency be lowered some how. Can plugin companies give the user a way to control the latency? Also does this add so much better performance and/or quailty that the plugin/s have to have this to be the plugin/s they are? Thanx
Controling the latency in real time would be a problem I think because this would change the timing of the programme - and with apps like PT that aim to correct all the other channels for the one that's late, it needs to have the latency declared as part of the plug-in instantiation process. So a control like you suggest would not work well..

If it could work with the platform, it is possible to make it obviously. For some apps like dynamics reducing latency would only limit performance of the very fastest settings (i.e. you could not catch peaks before they happened etc.).

For some applications it simply may not be possible - for instance something that falls into that category are phase linear EQ's, or anything that has one in it's internal processing like some multiband comps etc.. Basically anything that must use a time quantity of samples to develop the signal..

Quote:

Hey Paul. On the real. How long does it normally take to create a finished plugin? Also, in terms of money, how much would it take to do a Inflator or Limiter or EQ or Compressor plugin? Take ya pick or answer one at a time. Just a round about number.
I've always wanted to know this.

It takes a surpising amount of time to make a plug-in to a very high standard and the initial development stage varies depending on how much you know about how to make it in advance. The time it takes to do the actual coding of the application to run on something like PT varies depending on how complex it is and how many versions are needed to be supported. And TDM versions take up to 5 times longer to code than host RTAS versions. The testing and verification of the application varies similarly on complexity and number of variants.

For something like the reverb, I spent upwards of a man/year designing it (because I had never done a reverb before) and at least another man/year was spent between the guy coding it and myself trying to verify that the code actually did what my design did. So all in all we spent roughly 2 years on it from me starting to design it until we had something that worked sensibly. Another few man/months was spent on doing the all important included set-ups and the documentation.

For something like the limiter (which is more typical as I had much more idea about initially), I spent around 6 months experimenting with the novel ideas I was trying to apply and the processing it was going to contain. The code was started actually during this process (as the overall processing algorithms were largely understood about 70% through the design process) and the whole thing took the two of us around a year to make.. It then it went through a revision to add the intersample auto-correct function - and the noise shaping dither which the beta testers had asked for. Another later revision was done to put back functions that I had previously persuaded myself to remove - after we had feedback after sale from people having trouble with peak level control etc.. So all in all it took the two of us around 15 months before it was actually 'finished'.. I was moving this along at all speed since I knew it was most proabaly the last plug-in I would make in that environment...

Of course we could have made something good and highly useable in a fraction of the time had we have restricted ourselves to what we already new and had made before (like a subset cut out of the dynamics) - but as ever, I wanted something much much better than that - which broke new ground :-) Anyway it would seem really unfair to present something as 'new' that didn't break new ground - and expect people to pay for it - when they most probably already owned the original app the cut out subset is based on... Because of crushing taxation and revenue in the inflated economy of the U.K. employment here is amazingly expensive. So quite naturally a large part of the strain of my job was resisting ever-increasing pressure from all areas to err on the side of 'financial expediency'.
Old 21st October 2006
  #117
Gear Addict
 
RhOdEz's Avatar
 

I just need to say THANK YOU PAUL
Some of your toys made me say "YESSS" more then few times ,so your dedication for making musical tools in this harsh digital world is not being unnoticed ,i'm sure there's a lot of people feeling the same .
I hope you'll share some of your future plans /possible new developments with us

thumbsup
Old 21st October 2006
  #118
Lives for gear
 
Ruudman's Avatar
 

I couldn't use Sony Reverb TDM with the dithered mixer plugin.
That is, I could use it, but I would get a rippled sound once and a while (in the processed signal).
Especially with the Phase Mod set to high.
I don't get this scenario using the non-dithered mixer.

Why, I don't know....



ruudman
Old 22nd October 2006
  #119
Motown legend
 
Bob Olhsson's Avatar
 

Verified Member
Quote:
Originally Posted by bonne View Post
...Bob descibed a test where he chained two fixed point plugs, Waves C1 and L2, in a floating point DAW. He found that when increasing the level out of the C1 to well above O dBFS the following L2 read the floating point "signal" from the C1 (via the DAW?) without causing any distortion/clipping.

My concern is whether this would be the case with any chained fixed point plugs in a floating point DAW, e.g. Nuendo.
The L2 would have had to have been floating point too. That said, one can't assume how a plug-in is going to handle signals above full scale. Some plug-ins work fine and others break up.

Unless you are willing to tackle a science project testing everything you might possibly use, it makes a lot more sense to simply keep the levels down.
Old 22nd October 2006
  #120
Lives for gear
 

Quote:
Originally Posted by Bob Olhsson View Post
The L2 would have had to have been floating point too. That said, one can't assume how a plug-in is going to handle signals above full scale. Some plug-ins work fine and others break up.

Unless you are willing to tackle a science project testing everything you might possibly use, it makes a lot more sense to simply keep the levels down.

Yes indeed - you are completely correct.

Any process which has relationship to actual programme levels must align itself to the actual intended real world values of the signal, regardless of whether the DAW and math happens to be fixed or floating point. These levels are of course those that would leave the DAW as fixed point (real world) signal.

So for instance the dynamics, inflator and limiter must behave entirely similarly regardless of whether TDM or RTAS, as they relate to programme and not any particular math representation - and we must ensure that they do within the design..

Therefore the fact that some processes and plug-ins apparently can output float signals beyond full level, cannot be relied upon :-(

It is better by far to create your own headroom by lowering levels - because like that you are in charge and whatever happens is legitimate and can be relied upon :-)
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