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
2buss comparison: Fatso vs. Oxford Dynamics Dynamics Plugins
Old 18th October 2006
  #91
Lives for gear
 

Quote:
Originally Posted by norman_nomad View Post
Paul,

The info you provide is greatly valuable to me as an eager hobbyist in the field and I have no problem with you clarifying my half-witted analogies! heh

I understand exactly what you're saying as I've done plenty of reading.. and I was trying to speak to your exact point by making an abstract analogy that obviously didn't hit home!

My point about "continuity" was referring to Nyquist's thrust that perfect analog reconstruction can be performed with a sampling rate double that of the highest frequency to be represented.

More sampled steps on a continuous sine wave (aka 96khz sampling) are not necessary for your D/A to represent the original sine wav perfectly, just as more DPI, after a certain threshold, is not necessary to provide for a perfect representation of visual continuity.

So this was my half-witted analogy... ... to say that more isn't necessarily better; it can be just the same as less.

Hope this makes sense!
Yes - I understand - no worries :-)

The thing about optical pixel quantisation is that it cannot readily be fixed by dithering and filtering like digital audio can - because it is a physical reality and not a function that resides purely in the mathematical domain. There is no reasonable method to dither the pixels (apart from perhaps physically and randomly shaking the display in 2 dimensions) and no equivalent of an optical phase linear and steep spatial lowpass filter (a bit of milky translucent film has too slow a roll-off and loses too much detail) :-)

The other interesting thing is that the spatial freq content of the pixelated picture in something like an LCD panel is greater than the resolution - because the edge of the pixels are hard fixed boundaries (most often square), with a greater rate of change than the image signal itself. This can make the image seem like it's got greater detail than the original picture - but of course all that extra 'detail' is actually error that is not necessarily in the picture signal..

The only 'fix' is to have more and more pixels and stand far enough away so that regular pixel boundaries are way too small to ever see - and let the eye do the filtering....

In many ways this is much the equivalent of early digital systems that did not yet include dither - driving early ADCs and DACs that for technical reasons (or sometimes even 'deliberate' ones) did not have sufficient filtering. The idea that more and more quantisation steps and higher and higher sampling rates were required, was a logical trap one could easily fall into when encountering the situation for the first time..

And indeed if we wind back in time 20 years this was how many people in 'high office' were actually thinking!! Arguments about resolution, dither and time domain issues commenced for me right back in the mid 1980's mostly around the coffee machine when I was all too often accosted by 'senior techs' drafted in on high salaries, who were horrifed and perplexed that I was actually adding noise to everything- to improve signal accuracy!! :-( But for me, a seasoned analogue engineer - it wasn't much of a great leap of understanding or a genius revelation. Whenever us analogue guys came across something hoplessly and unfixably non linear, the hot plan was to 'rattle the hell out of it' with something we either couldn't hear or was at least a whole lot nicer than the error - after all, thats what we'd been doing with analogue tape bias for some 40years.

I actually remember trying to explain this for the n'th time to a perplexed senior who kept on annoyingly plaguing me with disbelief - by explaining that adding a noise signal to the DAC and removing it again afterwards could be thought of like randomly 'shaking the DAC around' so that the DACs quantisation steps never repeatedly appeared the same place twice wrt the signal - and therefore the conversion became continous without any steps - with a bit of added noise instead of distortion. I'm convinced that guy thought I was barmy and was more than relieved when I finally left to help form ODL. And do you know what? Within weeks of starting ODL, another high office notable turned up and darned well started the exact same objections - "how could adding noise to something ever improve anything"!? He plagued me continuously on every spare moment during every trip he made to our office. I'm sure the guy positively hated me when I doggedly 'failed to mend my ways' in the face of superior rank, learning and completely obvious logic.

No wonder I get sensitive about it sometimes!

Off topic - but interesting....?
Old 18th October 2006
  #92
Lives for gear
 
norman_nomad's Avatar
Quote:
Originally Posted by Paul Frindle View Post

The only 'fix' is to have more and more pixels and stand far enough away so that regular pixel boundaries are way too small to ever see - and let the eye do the filtering....

In many ways this is much the equivalent of early digital systems that did not yet include dither
Ha .. I got you to compare digital pictures with digital audio... I think my work here is done. *wipes hands*

I appreciate the insight and don't want to derail the thread...let's just agree I made an ill-suited analogy.

I have a question: Wildcowboys suggested that there should be a test to measure the kinds of losses one might or might not suffer when recording signals at less than full scale. I also think this is a good idea. Can you propose such a test?

All the geek talk is fun, but I really just want the best information so I can use it make better recordings...
Old 18th October 2006
  #93
Lives for gear
 
minister's Avatar
Quote:
Originally Posted by WildCowboys View Post
guys, without giving my opinion here...I have one major suggestion:
...
If someone says a 16bit track peaking at -24dB does not sound worse (has no smaller resolution) that the same track peaking at 0dB then go and prove it.
you can try this yourself.

take a 24 bit file that peaks close to zero and dither to 16/44.1 now take that same 24 bit file and bring it down -18dbB. (or -12 or -6) dither to 16/44.1

listen to the one close to zero, then, turn up your monitor 18dB and listen to the other. it'll be hard to A/B (and our sonic memories can reset real fast).
Old 18th October 2006
  #94
Lives for gear
 
WunderBro Flo's Avatar
Quote:
Originally Posted by minister View Post
you can try this yourself.

take a 24 bit file that peaks close to zero and dither to 16/44.1 now take that same 24 bit file and bring it down -18dbB. (or -12 or -6) dither to 16/44.1

listen to the one close to zero, then, turn up your monitor 18dB and listen to the other. it'll be hard to A/B (and our sonic memories can reset real fast).
such a test would be too hard, it would have to be much ruffer to hear differences. I do not doubt that changing levels plus/minus 30dB would make an audible difference in quality in a 24bit file. that´s the great thing about 24bit, the quality is so good that we do not have to worry about this stuff anymore.
the purpose of a test would be to prove the "esoteric" theory right or wrong, that says bit resolution is no factor for fidelity (when used with dither).

personally, this seems unlogical to me. all you have to do is to take the theory to the extreme, for example 8bit. It is totally logical that the size of the quantization error in 8bit is much higher than in 24bit for example. the higher this error, the higher the distortion/noise of the resulting signal. the interesting thing is the definition of the result of quantisation error - it can be seen as distortion because it changes the wave of the original waveform at each samplestep. this is no "clipping" distortion that we usually mean when we say distortion. BUT, if taking the original waveform away from the "distorted" one what remains is noise. So it is also "just" a noisefloor. I remember from my days as a kid with an amiga computer and a sampler, that when the SR was set high and the bitdepth at 8bit, the resulting sound sounded very hifi, but had an incredibly loud amount of noise. And that makes sense. What really sounds "hifi" to us are high sampling rates - they cause us to hear all frequencies and that we can hear them well & close to the original. We hear noise as noise.....so basically low-resolution recordings at high SR are perceived as hifi sounding recordings mixed with lots of noise.

to me this is the exlpanation why dave´s theories, which are solid, logical basic rules of digital audio, are true, while at the same time paul´s thoeries about the "misconception of resolution vs. fidelity" are also partly true - because perceived fidelity still remains high even though technical fidelity does not.

does that make sense to anyone? it does to me...

Rock on!
Pat
Old 18th October 2006
  #95
Lives for gear
 

Quote:
Originally Posted by WildCowboys View Post
such a test would be too hard, it would have to be much ruffer to hear differences. I do not doubt that changing levels plus/minus 30dB would make an audible difference in quality in a 24bit file. that´s the great thing about 24bit, the quality is so good that we do not have to worry about this stuff anymore.
the purpose of a test would be to prove the "esoteric" theory right or wrong, that says bit resolution is no factor for fidelity (when used with dither).

personally, this seems unlogical to me. all you have to do is to take the theory to the extreme, for example 8bit. It is totally logical that the size of the quantization error in 8bit is much higher than in 24bit for example. the higher this error, the higher the distortion/noise of the resulting signal. the interesting thing is the definition of the result of quantisation error - it can be seen as distortion because it changes the wave of the original waveform at each samplestep. this is no "clipping" distortion that we usually mean when we say distortion. BUT, if taking the original waveform away from the "distorted" one what remains is noise. So it is also "just" a noisefloor. I remember from my days as a kid with an amiga computer and a sampler, that when the SR was set high and the bitdepth at 8bit, the resulting sound sounded very hifi, but had an incredibly loud amount of noise. And that makes sense. What really sounds "hifi" to us are high sampling rates - they cause us to hear all frequencies and that we can hear them well & close to the original. We hear noise as noise.....so basically low-resolution recordings at high SR are perceived as hifi sounding recordings mixed with lots of noise.

to me this is the exlpanation why dave´s theories, which are solid, logical basic rules of digital audio, are true, while at the same time paul´s thoeries about the "misconception of resolution vs. fidelity" are also partly true - because perceived fidelity still remains high even though technical fidelity does not.

does that make sense to anyone? it does to me...

Rock on!
Pat
This is the absolute crux of the misunderstanding - it's not esoteric at all.

If you take a 24 bit file and re-dither it for 8 bits (by adding the right amount of dither) then throw away all but the top 8 bits, you will have a perfectly valid sound with similar noise levels to a compact cassette - but absolutely NO distortion or any other degradation of the signal :-)

That is the whole point :-)

It will look more grainy and steppy on your W/S waveform display - but that's because the display is showing you only discrete numbers - NOT the reconstructed signal. :-)

This is not some pet opinion that should require discussion (certainly not these days) - it has nothing to do with 'opinion and logic'. These are the rules of sampling and it's industry standard stuff we are all using everyday. As someone pointed out here in a previous post - if this were not the case we simply could not make digital audio and processing work at all...

The fact that we ar having this discussion still 20 years after the event shows just how damaging industry hype has been - and to what degree the public have been duped - IMO it's tragic :-(
Old 18th October 2006
  #96
Lives for gear
 

Quote:
Originally Posted by norman_nomad View Post
Ha .. I got you to compare digital pictures with digital audio... I think my work here is done. *wipes hands*

...
I thought you would appreciate this :-)

I have a big failing which is - if you give me something to think about I just can't help myself, I simply can't prevent myself from thinking about it. This is what keeps me up until 4am most nights :-(
Old 18th October 2006
  #97
Lives for gear
 
minister's Avatar
Quote:
Originally Posted by WildCowboys View Post
the purpose of a test would be to prove the "esoteric" theory right or wrong, that says bit resolution is no factor for fidelity (when used with dither).
i hear where you are coming from and i can see how you can conceive of digital operating this way. there was a time when i sort of thought of it this way. (one thing is certain, 24 bit sounds better for recording than 16 bit ! ) but once you start actually reading what was written by people who design this stuff, make it work and advance the industry, you realize that it is not an esoteric theory, but how digital works.


on to make music!
Old 18th October 2006
  #98
Lives for gear
 
norman_nomad's Avatar
Quote:
Originally Posted by minister View Post
you can try this yourself.

take a 24 bit file that peaks close to zero and dither to 16/44.1 now take that same 24 bit file and bring it down -18dbB. (or -12 or -6) dither to 16/44.1

listen to the one close to zero, then, turn up your monitor 18dB and listen to the other. it'll be hard to A/B (and our sonic memories can reset real fast).
This is good, but is there any other test I can perform with more percise control?

It would be hard to attenuate my monitors by exactly 18db....
Old 18th October 2006
  #99
Deleted bd1be4f
Guest
Seems to me the best test (since the debate also involves converters) would be to take a repeatable signal, say a track from a CD or other signal that you have control over the levels going into your converters. Record one version of it into your DAW with the record levels peaking just below 0dbfs. Then record another version of the same signal with the record level peaking at say -30dbfs.

Then inside the DAW normalize or gain change the lower recorded file so that it too peaks at the exact same level as the first one. If in fact recording at lower volume degrades the sound, the normalized file should sound noticably inferior to the louder one.
Old 19th October 2006
  #100
Lives for gear
 
minister's Avatar
Quote:
Originally Posted by zboy2854 View Post
Seems to me the best test (since the debate also involves converters) would be to take a repeatable signal, say a track from a CD or other signal that you have control over the levels going into your converters. Record one version of it into your DAW with the record levels peaking just below 0dbfs. Then record another version of the same signal with the record level peaking at say -30dbfs.
that file is already dithered. that is why i suggested a 24 bit file as a source. and creat 2 or more dithered 16 bit files.
Old 19th October 2006
  #101
Deleted bd1be4f
Guest
Quote:
Originally Posted by minister View Post
that file is already dithered. that is why i suggested a 24 bit file as a source. and creat 2 or more dithered 16 bit files.
Doesn't matter what the signal is, since you're comparing two copies of the same exact signal, one recorded through the converters hot and one recorded at low level. What resolution the source was is irrelevant since it's entering the DAW via analog conversion. Whether that signal happens to come from a CD or, say a guitar, is irrelevant.
Old 19th October 2006
  #102
Lives for gear
 
norman_nomad's Avatar
Quote:
Originally Posted by zboy2854 View Post
Doesn't matter what the signal is, since you're comparing two copies of the same exact signal, one recorded through the converters hot and one recorded at low level. What resolution the source was is irrelevant since it's entering the DAW via analog conversion. Whether that signal happens to come from a CD or, say a guitar, is irrelevant.
Yes, but how do you control the level at which the signal enters your converter... if you turn down the CD player, you've lost D/A bit depth, if you use a pad, you're introducing another component to one of the chains rendering it an unequal test (generally a pad doesn't have a "sound" but I'm being nitpicky).

See what I mean?
Old 19th October 2006
  #103
Deleted bd1be4f
Guest
Quote:
Originally Posted by norman_nomad View Post
Yes, but how do you control the level at which the signal enters your converter... if you turn down the CD player, you've lost D/A bit depth, if you use a pad, you're introducing another component to one of the chains rendering it an unequal test (generally a pad doesn't have a "sound" but I'm being nitpicky).

See what I mean?
Run the output of the CD player through a line mixer and into the converters for both passes, turning down the line mixer for the low level pass. Assuming you have a clean high quality mixer it shouldn't unduly influence the signal, and any coloration it adds to the signal would be added to both the high and low level versions.
Old 19th October 2006
  #104
Gear Maniac
 

I've been watching this thread (painfully!) and I felt compelled to get sucked in. First, I agree with Paul, resolution (in bits) is a bad way to describe audio quality, and that with proper dithering, you shouldn't get any nasty digital distortion artifacts with signals recorded at lower levels.

However.... if you record at low levels you WILL have a noisier signal (with or without dither)! Record 30 dB below fullscale, and your track will be 30 dB noisier than one that was recorded at a hotter level. This is just a fact.

It would be nice if we could record nice and hot into Pro Tools, using up most of the bits (which I don't think is a terrible idea, as long as you don't clip), but still have some headroom left for plug-ins. (You can do this with a trim plug-in, but this takes a lot of thinking that we shouldn't have to do.) Digi should have put at least a couple of bits of headroom into the the bus architecture to prevent digital overs, instead now we have folks thinking that driving your ADCs at -30 dB is a good thing. Crazy!

-- Ken

Ken Bogdanowicz
Plug-In Maker


Quote:
Originally Posted by zboy2854 View Post
Then inside the DAW normalize or gain change the lower recorded file so that it too peaks at the exact same level as the first one. If in fact recording at lower volume degrades the sound, the normalized file should sound noticably inferior to the louder one.
Old 19th October 2006
  #105
Quote:
Originally Posted by norman_nomad View Post
Yes, but how do you control the level at which the signal enters your converter... if you turn down the CD player, you've lost D/A bit depth, if you use a pad, you're introducing another component to one of the chains rendering it an unequal test (generally a pad doesn't have a "sound" but I'm being nitpicky).

See what I mean?
Here we are touching on a digital "hot potato" IMHO

In our keenness to present the highest possible level to our converters should we use pre converter limiters or compressors? - or not?

My point - To many engineers, a non compressed, naturally dynamic signal like a piano or vocal - with level set to the converter so as not to cause METER overload during the loudest passages, can look SO DAMN LOW that it can cause the engineer to panic and believe they are doing a poor job of delivering the 'highest possible level' to the converters (as suggested by my friend Mr Derr above) I think what some engineers SEE on meters affects their "dynamic decisions" and they compress to feed hotter, closer to full scale signal to their converters and I believe this practice to be bogus.

To make a bad pun....Should people take the 'use up the most bits possible' advice - if they are only going to then run their signal into a 'brick wall'?

Its as if everything can be ruined by obsession with serving hot level to converters....

Have you any thought on it Paul? Dave? Others?

Best regards to all,
Old 19th October 2006
  #106
Deleted bd1be4f
Guest
Quote:
Originally Posted by KBToys View Post
However.... if you record at low levels you WILL have a noisier signal (with or without dither)! Record 30 dB below fullscale, and your track will be 30 dB noisier than one that was recorded at a hotter level. This is just a fact.


Ken Bogdanowicz
Plug-In Maker
Ken,

There is no doubt that any time you are dealing with analog signals (including those feeding an ADC), a quieter signal will have a lower signal to noise ratio than a hotter one, due to the simple fact that you're dealing with analog signals, which all have inherent noise. However, this has nothing to do with the actual issue of whether the digital process itself introduces degraded sonics when capturing a signal at lower levels. Is it your contention that it does?

Further, while I don't think anyone here is advocating recording into your ADC's peaking at -30dbfs, the fact remains that as Jules pointed out, recording a very dynamic and uncompressed signal may mean that certain parts of the signal and performance may in fact reside at -30dbfs for example. Again, to say that this portion of the recorded digital signal is therefore sonically inferior to the louder portions is to say that digital as a recording medium is essentially fatally flawed, and I would categorically reject such an assertion.
Old 19th October 2006
  #107
Lives for gear
 
minister's Avatar
who is suggesting recording with an average level of -30?

i record with levels close to 0 VU which is optimal for the analog compnents. if i record close to 0dBFS, then i am WAAAAAY over 0VU. ever since i switched to recording with peaks around -18 up to -14 (or -12) -- this not in stone -- i not only have more hedroom in case the perfomace gets BIG, my analog is not distorting and the digital sounds better.

once i keep my peaks going into plug-ins under 06, my digital sounds better.

Pro Tools mixer has WAY more headroom than an anlog mixer.

don't get me wrong, i love analog, and am upgrading my analog capabilities with nice eq's and compressors. we are talking digital levels and this notion of 'resolution'.
Old 19th October 2006
  #108
Gear Maniac
 

Quote:
Originally Posted by zboy2854 View Post
Ken,

However, this has nothing to do with the actual issue of whether the digital process itself introduces degraded sonics when capturing a signal. Is it your contention that it does?
...
Again, to say that this portion of the recorded digital signal is therefore sonically inferior to the louder portions is to say that digital as a recording medium is essentially fatally flawed, and I would categorically reject such an assertion.
It was just starting to sound like there was no downside all to recording at a low level. There is, it's noisier, just like in the analog world. That's not to say that we should obsess about every last bit, or squash the hell out of your signal before you hit your converters. Converters are quite good now, and we don't need to do that. However, for the BEST quality, every component in your signal chain should be driven at it's optimum level. For analog stuff, like mic pres, that depends on the piece of kit your using and is sometimes a trade-off between distortion and nosie. For MOST converters, that means making sure you never ever clip, but being sensible about getting good signal levels. Louder will be quieter.

-- Ken

Plug-In Maker
Old 19th October 2006
  #109
Gear Maniac
 

Quote:
Originally Posted by zboy2854 View Post
Ken,

However, this has nothing to do with the actual issue of whether the digital process itself introduces degraded sonics when capturing a signal at lower levels. Is it your contention that it does?
Just to answer this more directly, yes the digital process degrades sonics when capturing a signal at low level, it adds noise. And so do a resistor. Everything in your signal path introduces some noise or distortion. Dithering eliminates the distortion problem in A/D conversion, but doesn't eliminate noise.

-- Ken
Old 19th October 2006
  #110
Deleted bd1be4f
Guest
Quote:
Originally Posted by KBToys View Post
Just to answer this more directly, yes the digital process degrades sonics when capturing a signal at low level, it adds noise. And so do a resistor. Everything in your signal path introduces some noise or distortion. Dithering eliminates the distortion problem in A/D conversion, but doesn't eliminate noise.

-- Ken
Agreed, but as you pointed out, it should be made clear that the degradation made by the conversion process is simply the addition of noise, no different than saying any other piece of gear inserted into a signal chain degrades the signal when passing a low level signal through it. This is an important distinction from anyone asserting that there is something unique about the digital process which harms the signal, when in fact there is not.

So long story short, running a low level signal into an ADC is no more detrimental to the signal than running a low level signal through any other piece of analog gear.
Old 19th October 2006
  #111
Lives for gear
 

Quote:
However.... if you record at low levels you WILL have a noisier signal (with or without dither)! Record 30 dB below fullscale, and your track will be 30 dB noisier than one that was recorded at a hotter level. This is just a fact.
Sure, that's a fact, but the question is, does it matter? In most cases I don't think it does...if you record a signal at 110 dB SPL at -30 dBFS and the ambient noise in your room is, say, 40 dBSPL, and the noise floor of your converters is at -120 dBFS...in other words, the noise in your room is about 20 dB higher than the noise floor of your converters...do we really perceive it as noisier than it would be if we recorded at -20 dBFS or -10 dBFS or -1 dBFS? In any case our dynamic range stays the same...the ambient noise goes up as our recording level does, and sure, our converters' noise floor drops in relation to everything else, but do we hear it?

-Duardo
Old 19th October 2006
  #112
Lives for gear
 
norman_nomad's Avatar
Quote:
Originally Posted by Duardo View Post
Sure, that's a fact, but the question is, does it matter? In most cases I don't think it does...if you record a signal at 110 dB SPL at -30 dBFS and the ambient noise in your room is, say, 40 dBSPL, and the noise floor of your converters is at -120 dBFS...in other words, the noise in your room is about 20 dB higher than the noise floor of your converters...do we really perceive it as noisier than it would be if we recorded at -20 dBFS or -10 dBFS or -1 dBFS? In any case our dynamic range stays the same...the ambient noise goes up as our recording level does, and sure, our converters' noise floor drops in relation to everything else, but do we hear it?

-Duardo
Which brings us full circle back to my first post of contention:

Quote:
Originally Posted by norman_nomad
I've been under the impression that I could even let my peaks fall lower if I wanted to, because even at lower levels, the noise floor and quantization distortion of a well implemented 24 converter is unlikely to creep into the audible range...
Old 19th October 2006
  #113
Lives for gear
 

Quote:
Originally Posted by norman_nomad View Post
Which brings us full circle back to my first post of contention:
Ok to start with, the very first and most important lesson - forget 'bits' and above all forget 'resolution'! All the bits in the data stream are always there at all levels - they do not get deleted when level is reduced. And quite simply there is no such thing as 'resolution' in an audio signal :-)


So to answer that original point:

For an ADC converter (you have 100 - 110dB SNR to play with)::

The penalty for reducing levels is only an increase in noise.

Leaving some useful headroom (around 6dB or more) by recording to peak at 6dB or less, is a good idea as this will avoid overs and in many cases reduce converter distortion...

This is particularly important with todays 'pro-sumer' converters that very often have compromised analogue front end circuits..

Many are not professional devices, in many of todays ADC systems, distortion may increase dramatically towards full input level.

And watch out - in some cases hitting the system with high levels will cause distortion that may not reduce even if the ADC's own input gain control is reduced. In this case your only option is to send less to it in the first place :-(



For the DAW (you have around 140dB SNR to play with) ::

The penalty for reducing levels into your digital channel is only noise.

Making some headroom by further reducing the level at the start of your channel (6dB - 18dB) is a good idea, as this avoids overs (even those you don't see on meters).

Also crucially it allows you to boost and cut signals (with EQ and such) without destroying your concentration and delicate balances each time you have to tweek levels to avoid red lights etc..



For the DAC - and/or final master output (for 24bit master you have 140dB SNR to play with - for 16 bit CD you have 90dB)::

The penalty for reducing the levels out of your mix is only noise.

Reducing your mix output to peak somewhere below -3dB is a good idea, because this will avoid most intersample overloads (which you don't see on your meters!), it will therefore stress converters less causing less converter error - ensuring that the result you and your listeners hear is much much more consistant. I.e. the sound will depend less on how your and your listeners DACs respond to intersample errors.

Also, if your mix is going to a mastering engineer, it will allow him much more freedom as he won't have to deal with clipped signals that cannot be repaired afterwards..

Printing to CD at levels below -3dB is an exactly similar situation - cheap comsumer devices are likely to perform very very much better indeed!!!! :-)

Simply put - It will sound much better - give it a try, you'll be amazed! :-)

What about that - is this concise enough? :-)
Old 19th October 2006
  #114
Lives for gear
 
Ruudman's Avatar
 

Well put, Mr. Frindle.




ruudman
Old 19th October 2006
  #115
Deleted bd1be4f
Guest
Yep, thanks for your wisdom and clarity on the subject Paul, I know that I for one have been helped immensely in my work since heeding your suggestions.
Old 19th October 2006
  #116
Lives for gear
 

Quote:
Originally Posted by KBToys View Post
Just to answer this more directly, yes the digital process degrades sonics when capturing a signal at low level, it adds noise. And so do a resistor. Everything in your signal path introduces some noise or distortion.
-- Ken

Hmm.... Not sure how best to answer this but:

Using good design techniques and taking proper advantage of double precision math, digital processing should NOT significantly increase noise (or unwanted distortion), - except in so far as it increases the low level gain of the input signal (i.e. boosting EQ, increasing levels, compression etc.).

The advantage that digital processing should have over physical components (resistors and other stuff bound by the laws of physics), is that digital processing can employ almost arbitarly high mathematical precision. In other words, it does not exist in the real world and is therefore not bound by real world limitations in that way :-) This is it's single biggest advantage, which goes a long way to allowing you to make applications that simply could not function well enough if made from analogue components...

For instance, I can categorically state that no Sony Oxford plug-in I have designed (that is all of them apart from one) adds any further noise of it's own to the signal (measurable at the 24 bit level) - apart from that which naturally occurs when the plug increases the gain (and noise) of the input signal.

The only source of excess noise from using the plug-ins on say the TDM platform is the dither required to interface to the 24 bit TDM buss. This sets the interface noise at around -143dB..
Old 19th October 2006
  #117
Gear Maniac
 

Quote:
Originally Posted by Paul Frindle View Post
Hmm.... Not sure how best to answer this but:

Using good design techniques and taking proper advantage of double precision math, digital processing should NOT significantly increase noise (or unwanted distortion), - except in so far as it increases the low level gain of the input signal (i.e. boosting EQ, increasing levels, compression etc.).

The only source of excess noise from using the plug-ins on say the TDM platform is the dither required to interface to the 24 bit TDM buss. This sets the interface noise at around -143dB..

Paul - I was talking about A/D conversion, not plug-in processing. BTW, your level recommendations sound about right to me - thanks for clearing that up.

For DSP, yes, we can do our math to whatever precision we'd like. However, I'm not sure it's safe to assume that EVERY plug-in out there uses double precision and dithering. There seems to be a pretty high market value placed on instance count (and vintage graphics) these days!

-- Ken

SoundToys
Old 19th October 2006
  #118
Deleted bd1be4f
Guest
Quote:
Originally Posted by KBToys View Post
There seems to be a pretty high market value placed on instance count (and vintage graphics) these days!

-- Ken

SoundToys
Well, I for one hope that all plugin makers these days are putting the highest priority on the processes and algorithms rather than graphics or instances, since today's CPU and DSP capabilities render most power considerations moot. Obviously the Oxford and Soundtoys plugs are among the best, so I assume you guys already do this. Speaking of which, when are the rest of the Soundtoys plugs going to be XP ready Ken? Daddy needs his Echoboy and Soundblender!
Old 19th October 2006
  #119
Lives for gear
 

Quote:
Originally Posted by KBToys View Post
Paul - I was talking about A/D conversion, not plug-in processing. BTW, your level recommendations sound about right to me - thanks for clearing that up.

For DSP, yes, we can do our math to whatever precision we'd like. However, I'm not sure it's safe to assume that EVERY plug-in out there uses double precision and dithering. There seems to be a pretty high market value placed on instance count (and vintage graphics) these days!

-- Ken

SoundToys
Yes - agreed. Like anything else, it's a question of design quality versus processing costs and the target market the apps are aimed at. BTW - I have made my thoughts known about the 'legacy rush' earlier..

At Oxford we took the attitude that these were indeed professional apps aimed at people employed in the profession of making commercial sound. So the same design considerations were used as for all our previous stuff - i.e. the best possible available quality and stability etc. But it has to be said that initially we did receive some surprise from customers about relatively heavy processing costs.

Crucially - what we did NOT do is assume that, because the kit people were now using was a fraction of the cost of systems we had built before, they would suddenly accept anything less in quality.. Regardless of how the market dynamics and economics change - fundamentally people are aiming at the best possible artistic results. The fact that this can be done at a fraction of the cost from only a decade ago is a fantastic bonus which we should all embrace and celebrate :-)
Old 19th October 2006
  #120
Lives for gear
 
norman_nomad's Avatar


Thanks to Paul Frindle and Dave Derr for contributing a great deal of high level content to this thread.

I think I'll stick to observing 0vu when tracking, and -6dbfs peaks ITB when mixing with plug-ins.

Knowing that, while recording at less than full scale will add noise to my signal, it's overall level will likely be masked by the noise floor of my recording environment therefore rendering it negligible.
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
XHipHop / Music Computers
20
Ghost Logic / Gearslutz Secondhand Gear Classifieds
0
benelli / Gearslutz Secondhand Gear Classifieds
0
Dutchmuzik / Gearslutz Secondhand Gear Classifieds
0

Forum Jump
Forum Jump