Dithering is a Mathematical Process - NOT a psychoacoustic process.

[mansr]

My goodness this is a disappointing exchange between smart guys I respect.

It's disappointing from a smart guy I used to respect. I'll stop now. Sorry for the interruption.

 

[j_j]

My goodness this is a disappointing exchange between smart guys I respect.

What am I to say? He denies that ISP's are an issue, then equivocates around the whole issue, then tries to frame my position (incorrectly) while scolding me for malpractice, ...

 

[bennetng]

What am I to say? He denies that ISP's are an issue, then equivocates around the whole issue, then tries to frame my position (incorrectly) while scolding me for malpractice, ...

Quite the opposite, I meant prevention of ISP during mastering does not guarantee clip-free operation on the other end (playback/boardcasting etc), there are other processes producing other kinds of clipping, for example, lossy codecs, which can also be solved by using software volume controls, ReplayGain and the like. I don't assume the volume controls on end user's side are always safe and I posted test signals on forums for them to test their volume controls.

So it is actually you who framed my position to think that clipping (not restricted to ISP) is not an issue.

I visit this forum for technical exchange, not for a drama like this.

 

[j_j]

Quite the opposite, I meant prevention of ISP during mastering does not guarantee clip-free operation on the other end (playback/boardcasting etc),

Unless you have two accounts, that wasn't directed at you.

 

[bennetng]

But not if you're starting with e.g. 16 bit source and the output is 24 bit. ;-)

Then think about it, modern software DAWs are at least 32-bit float, when mastering engineers want to take care of the uncertainty of ISP and other forms of potential clipping, then they either need to reduce the overall level, or lower the ceiling of their limiter, and export to a 16-bit file. In this way, exporting with a lower level actually introduce more noise, and/or introduce more limiting when compared with the floating point source, which is the "source of the source". Reducing overall level is probably harmless, lower the ceiling of the limiter? Perhaps not a good idea, but who knows, the temptation of "louder is better" in the mind of some engineers.

 

[solderdude]

I think everyone agrees that preventing ISP's is important and realize that things like poor mastering exists and things like post processing (EQ, correction) can make things worse.

Once the fact (poor file) exists there are only so many options: find another recording that is mastered better or use DACs that can handle it or live with the consequences or attenuate digitally before it is sent to the DAC that does the conversion.
People with DAC's that drive power amps directly do not have this problem.

 

[xr100]

Then think about it, modern software DAWs are at least 32-bit float, when mastering engineers want to take care of the uncertainty of ISP and other forms of potential clipping, then they either need to reduce the overall level, or lower the ceiling of their limiter, and export to a 16-bit file. In this way, exporting with a lower level actually introduce more noise, and/or introduce more limiting when compared with the floating point source, which is the "source of the source".

Yes, true. I'd be quite happy if the distribution format was 64-bit float, but I don't think the rest of the world would be too keen. ;-)

Regarding ISP/limiters, a factor to keep in mind is whether or not it's ideal for "brickwall" peak limiters (or rather their sidechains) to only "see" sample values.

Funnily enough, the idea behind the most famous "brickwall" limiter (L1) came from Michael Gerzon. The idea was to take care of the odd peak here and there, so the (untouched) bulk would then be "normalised" (say, to -0.1dB) and then dithered and noise shaped to the target bit-depth. Not sure if the original version had the "Analog" mode (which is a "True Peak" mode but not per BS1770, albeit another "True Peak" mode per BS1770 has been added for the latest "25th Anniversary" overhaul of L1) but it would seem that even Gerzon wasn't thinking about ISP's back then. Nor the way brickwall limiters would be used and abused...!

Perhaps not a good idea, but who knows, the temptation of "louder is better" in the mind of some engineers.

Well, up to a point pop/rock/EDM/etc. requires dynamic range compression. Everyone is quite used to hearing this, even if they don't realise it.

Truly excessive amounts of dynamic range processing... more likely... requirements of clients...

 

[xr100]

People with DAC's that drive power amps directly do not have this problem.

Sorry, you've lost me there?

 

[pma]

With 32-bit decoding enabled:

Disabled:

Thank you. I could not find this option ("enabled") in CoolEdit Pro. I will try in Adobe Audition, which is installed on another PC, that I usually do not use.

 

[bennetng]

Sorry, you've lost me there?

Because doing so the digital attenuation being used is usually more than enough to avoid any form of digital clipping. Of course, that requires a high DNR DAC, not a noisy one like a low end Realtek.

 

[xr100]

Because doing so the digital attenuation being used is usually more than enough to avoid any form of digital clipping. Of course, that requires a high DNR DAC, not a noisy one like a low end Realtek.

Oh I see, I wondered where the "volume control" was.

Are Realteks (assuming a good "implementation") that noisy?

EDIT: Well this one certainly is: https://www.audiosciencereview.com/...8930-realtek-hd-audio-interface-review.10403/

Oh, and if all is well "digitally," ISP's can still be responsible for havoc in analogue stages... ;-)

 

[xr100]

Thank you. I could not find this option ("enabled") in CoolEdit Pro. I will try in Adobe Audition, which is installed on another PC, that I usually do not use.

"CoolEdit" was "cool" back in the day but I really would suggest using up-to-date software. I'm also surprised by how many people seem to be using Audacity? I have it installed but rarely use it. I could be wrong, but it rings "alarm bells" in my head...

 

[solderdude]

It is free software and handy when you don't want or need a daily workhorse yet have enough functionality.
The audio equivalent of the Gimp for image manipulation.

 

[pma]

Because doing so the digital attenuation being used is usually more than enough to avoid any form of digital clipping. Of course, that requires a high DNR DAC, not a noisy one like a low end Realtek.

With a good, low noise preamp, I get substantially better S/N than with digital attenuation, speaking about my Cambridge Audio DacMagic Plus. I believe that with something like RME ADI-2 DAC the S/N would be better and comparable to a good linear preamp (I get about -130dBV(A) noise voltage from my linear preamp). Still, IMO cleanliness of highly attenuated signal is better with analog than with digital attenuation, in case that very low noise analog circuits are used.

"CoolEdit" was "cool" back in the day but I really would suggest using up-to-date software. I'm also surprised by how many people seem to be using Audacity? I have it installed but rarely use it. I could be wrong, but it rings "alarm bells" in my head...

CoolEdit Pro 2.1 version is almost comparable with Adobe Audition. It is another league than CoolEdit 96 (which you are probably referring to) and cannot be compared to Audacity. As I said, I have Adobe Audition 3.0 on another computer, but I use that computer rarely, as it is an old XP machine.

 

[xr100]

It is free software and handy when you don't want or need a daily workhorse yet have enough functionality.
The audio equivalent of the Gimp for image manipulation.

GIMP is actually better than Photoshop at certain things, especially given that there are many free plug-ins and scripts available.

OTOH, it is pretty useless at other stuff, certainly even basic colour correction/contrast/brightness controls are hopeless, and, for instance, Photoshop eats it for dinner if doing detailed retouch work, etc.

Anyway, I digress. I expect Audacity is OK if your needs are trimming files or exporting to FLAC etc., but I wouldn't trust it for much more, e.g. using high pass filters or something I'd want to test before use!

 

[xr100]

CoolEdit Pro 2.1 version is almost comparable with Adobe Audition. It is another league than CoolEdit 96 (which you are probably referring to) and cannot be compared to Audacity. As I said, I have Adobe Audition 3.0 on another computer, but I use that computer rarely, as it is an old XP machine.

I honestly can't remember which version of CoolEdit I'm referring to. I certainly remember using CoolEdit as far back as the "96" version, but I also recall using much later versions and then Audition 1.0. I've used recent versions of Audition, though not installed on this machine.

I don't remember ever being too impressed by Audition. Keep in mind that as far as I'm concerned, digital processing should be as close to theoretical perfection as the CPU (aka "patience" for non-real-time processes) can reasonably handle. Digital nasties soon await with DSP, though, so caution...

 

[j_j]

... In this way, exporting with a lower level actually introduce more noise, and/or introduce more limiting when compared with the floating point source, which is the "source of the source". Reducing overall level is probably harmless, lower the ceiling of the limiter?...

Since it's not really that hard to calculate a rough oversample by 4 (does not have to be an audio quality oversample, only a rough estimate), using that to drive a limiter would do fine. Small inaccuracies in the estimate are fine, 1% doesn't cause any DAC I've measured (or a couple of my coworkers have measured) to go nutters, so a small inaccuracy isn't a big deal. 10%, not good. Some commercial CD's hit 3dB. Now THAT is a problem!

 

[xr100]

Some commercial CD's hit 3dB. Now THAT is a problem!

Here is a not-so-good example from the late 90's:

And a small section of its waveform:

Yow!

And an example from within the past 5 years:

Overall compression is excessive, but "True Peak" level is kept in check. And none of that horrendous clipping, either.

 

[KSTR]

"CoolEdit" was "cool" back in the day but I really would suggest using up-to-date software. I'm also surprised by how many people seem to be using Audacity?

Audition (I use 1.5 and 3.0) is very up to date even today. The core of digital audio processing theory and practice haven't changed that much over the years and the software was written by true audio processing experts. Also, user interface is excellent, technically oriented, not bloated (at least up to version 1.5) yet extremey powerful. Power tool for power users.
I've tried many audio editors over the years and nothing comes close (Audacity? meh). For multitracking and mixing, real studio use, of course there is much better software (I was a big fan of Samplitude, but its days are over, Reaper is my current fav), but as an editor (and basic analyzer) toolbox Audition is absolutly fantastic. Together with SoX, REW, DeltaWave, etc, you can have a zero $ yet very professional software suite.

 

[KSTR]

It better be free, how much can you charge for using one of 2 machine language instructions?

The time consuming part is getting the environment (VST etc) to work, not the (trivial) payload code. I for one am very thankful for this kind of simple stuff.