Are we close to getting plugin compressors that sound like real ones?

[himalaya]

In fact, most posters here have taken the view that there's no 'correct' or 'incorrect'. 'Live and let live,' someone said.

Agreed!

But! Agemo's sentiment is misplaced in context of conducting tests which is what even Shy has recommended:

I think it's essential that people compare things themselves and not care much about what anyone says (including me)

At the end of the day I do not care one bit about compression and compressors, I'd rather tweak the filter cutoff , but when taking part in a comparative test, there must be results, and no 'airy fairy' exclamations of "live and let live".

[Trakstar]

Before I got banned I was trying Shys test so I will post tomorrow

[Uncle E]

when taking part in a comparative test, there must be results, and no 'airy fairy' exclamations of "live and let live".

Yes, and the results must be comprehensive, otherwise they don't tell us anything.

[Trakstar]

proper comparison with software compressor of choice

[jens]

This thread is mighty cute.

[herodotus]

when taking part in a comparative test, there must be results, and no 'airy fairy' exclamations of "live and let live".

Yes, and the results must be comprehensive, otherwise they don't tell us anything.

But how can they be comprehensive when

Unfortunately precise definitions for the terms attack and release do not exist due to a lack of industry standards. Moreover, manufacturers make this worst by not explaining how they define the terms. Most don't, they just list a range of settings, leaving the user to guess if the time shown represents how long it takes to get to the end of the gain change, or to the middle, or the 3-dB point, or what -- caveat emptor.

(From here).

[himalaya]

when taking part in a comparative test, there must be results, and no 'airy fairy' exclamations of "live and let live".

Yes, and the results must be comprehensive, otherwise they don't tell us anything.

So let's see if they are.
The initial premise was, quoting Shy:

It's not about software compressors in general sounding "significantly different", it's mostly about them sounding significantly worse, because instead of providing smooth gain reduction, they give a sound that's broken and stuttered compared to a good analog compressor's.

my underline with larger font.

If my demo, where I try to match the output of the Slate Dragon compressor, does not exhibit that "broken and stuttered" sound (which is also demonstrated by Shy, so we have reference as to what he means), can it be regarded as taking the whole of the premise (and challange)? If so, can it be thought of as being comprehensive?

[Uncle E]

But how can they be comprehensive when

Unfortunately precise definitions for the terms attack and release do not exist due to a lack of industry standards. Moreover, manufacturers make this worst by not explaining how they define the terms.

That's why we do null testing.

[Uncle E]

If my demo, where I try to match the output of the Slate Dragon compressor, does not exhibit that "broken and stuttered" sound (which is also demonstrated by Shy, so we have reference as to what he means), can it be regarded as taking the whole of the premise (and challange)? If so, can it be thought of as being comprehensive?

Think about it in terms of sound design. If I span a single C3 sample across the keyboard, it's going to sound wrong on the majority of keys, correct? Yet, if you play a C3 on the original instrument and I play the sample at C3, they'll sound the same.

The case is similar here. Compressors react differently based on how they're set, you have emulated a certain characteristic of the sound but you can't say your results are comprehensive until you've emulated every aspect of it. Also, I get it that you've met Shy's conditions but those conditions weren't sufficient.

[herodotus]

But how can they be comprehensive when

Unfortunately precise definitions for the terms attack and release do not exist due to a lack of industry standards. Moreover, manufacturers make this worst by not explaining how they define the terms.

That's why we do null testing.

I am all for null tests. They are as objective as any test can be in a place like this.

But null tests don't help with matching compressor settings. And unless you are matching settings, how can you be sure that differences in sound between two compressors are due solely to differences in quality and not to mismatched settings?

With equalizers, you can at least theoretically know how to match settings; 3db of gain at 2kHz with a Q of 4.2 should match the same setting on another equalizer. Even then, you could have difficulties due to different kinds of filter design, but there is at least a universal reference. Compressors don't even have that.

[Uncle E]

But null tests don't help with matching compressor settings. And unless you are matching settings, how can you be sure that differences in sound between two compressors are due solely to differences in quality and not to mismatched settings?

Perhaps. As far as this thread goes, the kick is so much more prominent in Shy's Slate example than any of the software examples that I don't think it'd null. I also don't think the sharp transients from the software examples will null. Anyway, it's easy enough to find out on our own.

[Agreed]

That leaves the digital advocates totally screwed, because analog advocates just have the goalposts strapped to their backs and can run around and make THIS the thing that they care about, wait no it's actually THAT, hold up the problem here is it's this OTHER thing, ad infinitum. There's also the inherent advantage that analog hardware has in the form of individual behavior due to parts tolerances, wear on the unit or lack thereof, etc., which invalidates the premise of perfect emulation of anything but one specific example of a unit.

You're stacking the deck if what's wrong is an intangible something that can be swapped around whenever the last one got made tangible and fixed. At some point, in order to have a meaningful comparison, it has to be results-based rather than some sort of impossibly broad identity criterion. Can the tools do the job as well, regardless of the various whack-a-mole "but it's not A! ..B!..C, D, etc." point by point objections.

I guess I'm saying until there's a way for analog diehards to say that software is good enough based on more rigid criteria which don't make the victory conditions impossible or freely movable, discussions like these are going to float around without accomplishing much.



Edit: To be fair, I think digital starts in a posture of needing to defend/improve by default, with the exception of things that aren't really possible in analog (linear phase, true look-ahead, etc.). I'm not at all saying that there aren't legitimate criticisms that can and should be leveled at software that doesn't live up to stated claims, especially. But I do think that for a productive ongoing discourse to happen, we need to step back a bit from begging the question against digital just by how the debate itself gets framed, and reframed, and reframed... I say results-based is the most effective criterion, personally, because while it's got a bit of "black box" to it, at least it sets up a genuine goal that doesn't move around. If someone says "digital emulations of the Fairchild 670 don't capture its real behavior," I'd say back that two different Fairchild 670s don't capture each-others' behavior! (And wouldn't null perfectly, either ) But if we set out a specific goal, then we can see how hardware and software compare. Hope that makes more sense.

[Uncle E]

That leaves the digital advocates totally screwed, because analog advocates just have the goalposts strapped to their backs and can run around and make THIS the thing that they care about, wait no it's actually THAT, hold up the problem here is it's this OTHER thing, ad infinitum. There's also the inherent advantage that analog hardware has in the form of individual behavior due to parts tolerances, wear on the unit or lack thereof, etc., which invalidates the premise of perfect emulation of anything but one specific example of a unit.

All true! I admit it!

You're stacking the deck if what's wrong is an intangible something that can be swapped around whenever the last one got made tangible and fixed. At some point, in order to have a meaningful comparison, it has to be results-based rather than some sort of impossibly broad identity criterion. Can the tools do the job as well, regardless of the various whack-a-mole "but it's not A! ..B!..C, D, etc." point by point objections.

Yes, I completely agree! Rather than discussing the results, which each person has their own subjective opinion about, we could simply post our examples and then let the end listeners decide for themselves without discussion.

[Agreed]

That leaves the digital advocates totally screwed, because analog advocates just have the goalposts strapped to their backs and can run around and make THIS the thing that they care about, wait no it's actually THAT, hold up the problem here is it's this OTHER thing, ad infinitum. There's also the inherent advantage that analog hardware has in the form of individual behavior due to parts tolerances, wear on the unit or lack thereof, etc., which invalidates the premise of perfect emulation of anything but one specific example of a unit.

All true! I admit it!

You're stacking the deck if what's wrong is an intangible something that can be swapped around whenever the last one got made tangible and fixed. At some point, in order to have a meaningful comparison, it has to be results-based rather than some sort of impossibly broad identity criterion. Can the tools do the job as well, regardless of the various whack-a-mole "but it's not A! ..B!..C, D, etc." point by point objections.

Yes, I completely agree! Rather than discussing the results, which each person has their own subjective opinion about, we could simply post our examples and then let the end listeners decide for themselves without discussion.

But then we don't get to have our say on the internet, and that's boring Now I made myself sad!

[kvaca]

Rather than discussing the results, which each person has their own subjective opinion about, we could simply post our examples and then let the end listeners decide for themselves without discussion.

well, than finally post your hardware examples you have promised already and let us enjoy it without any discussion at all / but I reserve the rights to make similar examples for you in my crappy software compressor /
and anybody who try to continue in discussion will be compressed to death