How to test a compressor plugin ...
-
- KVRAF
- 3369 posts since 16 Jan, 2005 from Ottawa, Ontario
..in any case, what is it the DIFFERENCE that you hear in this example...and by the way, I'm getting some pretty sharp gating effects with some compressors!
-
- KVRAF
- 5913 posts since 17 Aug, 2004 from Berlin, Germany
Thats interesting. Voxengo Marquis is the only compressor in my test, where the signal is loud with this phase cancelation test.
VintageWarmer - some deeper frequencies
Crunchessor - some higher frequencies
Sonalksis SV315 - silent
WaveArts Trackplug - silent
Cubase SX Multiband - silent
Voxengo Marquis - broadband frequencies and loud
VintageWarmer - some deeper frequencies
Crunchessor - some higher frequencies
Sonalksis SV315 - silent
WaveArts Trackplug - silent
Cubase SX Multiband - silent
Voxengo Marquis - broadband frequencies and loud
| Links-
- KVRAF
- 6478 posts since 16 Dec, 2002
greendoor:
this phase cancellation trick is a really good learning tool for beginners. You will even tually learn to hear these things happening in the original signal. And trust me it's a much more valuable skill than this simple trick.
this phase invert comparison trick has been used for 50(or more) years by designers of analog compressors as well. Someone rather famous designer once said he uses it to hear exactly the way his circuits affect the release and attack times. He can them make more precise adjustments to the circuits.
One thing is for sure. You still can't make objective conclusions based on what you hear when using this. When you say kjaerhus audio is "most transparent", based on the fact it creates perfect cancellation, there will always be someone who will say: "but dude, polysquasher actually sounds more transparent".
Or something along those lines.
this phase cancellation trick is a really good learning tool for beginners. You will even tually learn to hear these things happening in the original signal. And trust me it's a much more valuable skill than this simple trick.
this phase invert comparison trick has been used for 50(or more) years by designers of analog compressors as well. Someone rather famous designer once said he uses it to hear exactly the way his circuits affect the release and attack times. He can them make more precise adjustments to the circuits.
One thing is for sure. You still can't make objective conclusions based on what you hear when using this. When you say kjaerhus audio is "most transparent", based on the fact it creates perfect cancellation, there will always be someone who will say: "but dude, polysquasher actually sounds more transparent".
Or something along those lines.
-
- KVRAF
- 2049 posts since 18 Sep, 2003 from Seattle USA
This could be a really fine thread if we can nudge it back on track a bit to what the title implied about testing a compressor plugin. The implication is some type of objective test. Many of my plugins behave when I do the "phase" test, many don't including compressors, EQs, and harmonic distortion warmers. I think it's a fine way to quickly get in there and fine out what the "delta" is for processed and unprocessed sound - in some cases.
Also I think it might be important to be able to judge how certain compressor controls work according to some standard tests. This RANE note points out common features among compressors that we see in our compressor family here at KVR:
http://www.rane.com/note141.html
(there was another RANE note I can't find that accounts for differences in attack/release between different compressors).
Regardless of the controls on the compressor interface I'm thinking we can all have a kind of objective test (besides the fabulous bmanic inspired subjective tests using drums) that simply tests attack, hold, release, THD. I think there was a pretty good one on the internet when UA and Bomfactory were at war - pulses and tones were uses.
Compiling some type of test results using tones and pulses would allow quick comparison of specs (haha - if you like specs) to possibly judge how fast or colored a compressor is - relatively speaking in objective terms. Also does 1ms attack time on one compressor match the 1ms attack time on a different one? Any ideas on objective type tests?
Of course hearing a compressor take down a snare hit while leaving a hi-hat alone is a pretty impressive test also, in subjective terms!
Also I think it might be important to be able to judge how certain compressor controls work according to some standard tests. This RANE note points out common features among compressors that we see in our compressor family here at KVR:
http://www.rane.com/note141.html
(there was another RANE note I can't find that accounts for differences in attack/release between different compressors).
Regardless of the controls on the compressor interface I'm thinking we can all have a kind of objective test (besides the fabulous bmanic inspired subjective tests using drums) that simply tests attack, hold, release, THD. I think there was a pretty good one on the internet when UA and Bomfactory were at war - pulses and tones were uses.
Compiling some type of test results using tones and pulses would allow quick comparison of specs (haha - if you like specs) to possibly judge how fast or colored a compressor is - relatively speaking in objective terms. Also does 1ms attack time on one compressor match the 1ms attack time on a different one? Any ideas on objective type tests?
Of course hearing a compressor take down a snare hit while leaving a hi-hat alone is a pretty impressive test also, in subjective terms!
-
- KVRAF
- 12615 posts since 7 Dec, 2004
i'd like to note, one extremely foolish part of that document, which seems to be common with developers of plugins, is it lacks any mention of a feedback compressor. the implementation of a feedback compressor, and the way in which the controls interact is completely different than a feedforward compressor. also, most of the best sounding analog compressors, the ones which seem to be obsessed over, are feedback, not feedforward.
if you'd like to try a very basic form of feedback compressor, i wrote one (to my knowlage the only one in existance as vst) available at http://xhip.cjb.net/xhipeffects/
if you'd like to try a very basic form of feedback compressor, i wrote one (to my knowlage the only one in existance as vst) available at http://xhip.cjb.net/xhipeffects/
Last edited by aciddose on Sat Oct 29, 2005 9:42 pm, edited 1 time in total.
-
- KVRist
- 406 posts since 31 Jan, 2004 from Sweden
Cool test. 
UAD Fairchild - Silent.
My JP22 plug - Silent.
UAD Fairchild - Silent.
My JP22 plug - Silent.
Last edited by BobYordan on Sat Oct 29, 2005 9:49 pm, edited 1 time in total.
Cheers Bob
Back for plug development in Sonic Birth.
Back for plug development in Sonic Birth.
-
- KVRian
- Topic Starter
- 769 posts since 2 Apr, 2005
Well I expected this to be contraversial ... from a few comments, I can tell that some people understand how this phase inversion trick works, and a few simply don't understand it all.
I appreciate that a compressor can sound good, and fail this test. That means that you have to be very carefull if you are doing parallel compression with these plugins.
Voxengo is very good for adding Dry mix controls - so even though these compressors introduce phase anomolies, you can do parallel compression within the plugin.
Kjaerhus GCO-1 doesn't have a dry mix, but it seems that you can work around this by duplicating the track, and because GCO-1 doesn't screw with the phase you can mix the two tracks toghether with no phasing.
I've been testing EQ's with this trick too. Also, convolution based processors like TapeBus are really interesting to see what they are doing ...
The comments about using your ears, and "if it sounds good it is good" are valid. I think it's a bit pretenious to call this a beginners trick - I think it's very useful and identifies the strengths and weakness that we should know about.
I appreciate that a compressor can sound good, and fail this test. That means that you have to be very carefull if you are doing parallel compression with these plugins.
Voxengo is very good for adding Dry mix controls - so even though these compressors introduce phase anomolies, you can do parallel compression within the plugin.
Kjaerhus GCO-1 doesn't have a dry mix, but it seems that you can work around this by duplicating the track, and because GCO-1 doesn't screw with the phase you can mix the two tracks toghether with no phasing.
I've been testing EQ's with this trick too. Also, convolution based processors like TapeBus are really interesting to see what they are doing ...
The comments about using your ears, and "if it sounds good it is good" are valid. I think it's a bit pretenious to call this a beginners trick - I think it's very useful and identifies the strengths and weakness that we should know about.
-
- KVRian
- Topic Starter
- 769 posts since 2 Apr, 2005
No - you can't have this set right. For a start, this trick only works if your host has perfect plugin delay compensation. Maybe that's why some plugins fail ... hmmm.Really great topic!!! But i have one note.
With this trick you're not hearing the compressed signal, but the uncompressed, where the compression works
you can't hear the compression, cuz it has lower levels.
Create to identical audio loops. Phase invert one. They should sum to total silence. The goal of this test is to hear only the difference between the two signals. If we want a compressor that doesn't harm the signal unless it is actually compressing, then we should hear silence except for the peaks. With a threshold of zero, there should be zero compression, and we should hear total silence.
I'm not saying that compressors that fail this test are useless. It's just that they are not perfectly transparent in a real digital audio sense. Yes, they might sound better - that's a seperate issue.
-
- KVRAF
- 2049 posts since 18 Sep, 2003 from Seattle USA
Right! That's why I was thinking some standard pulses & tones to judge how slow or fast the "actual" attack & release times are would make for some type of judgement criteria. Of course with super low threshold and very small ratios a really fast compressor could be made to work slow but that's a usage thing. How fast does it go from 0-60 (mph/kph) is what might be interesting!aciddose wrote:...the implementation of a feedback compressor, and the way in which the controls interact is completely different than a feedforward compressor. also, most of the best sounding analog compressors, the ones which seem to be obsessed over, are feedback, not feedforward.
There are a few feedback type compressors here at KVR: Voxengo Voxformer and Marquis, uhmmm I forget any others - maybe PSP MixPressor.
-
- KVRAF
- 12615 posts since 7 Dec, 2004
input -> amplifier -> output
--------- envelope <- scaling
assuming you have a fixed width font, that should be obvious.
but, basically since you're using a preset level (generally 0db) for your "threshold", (although it is a behaviour completely unrelated to its sidechain conterpart), and the control is based upon the output of the amplifier stage, obviously you wont be able to control a seperate signal accurately using this method.
if you used two signals with an average rms of the same value, you could get a sidechaining effect by controlling two amplifiers from one envelope. although, it is nothing like the typical feedforward sidechaining effect at all.
i suppose sidechaining is possible, but the effect would be quite different. generally feedback compressors are what you might find in a guitar insert or etc, while feedforward compressors are much more complex. since the goal is simplicity with the feedback design, and complex behaviour is the goal with the feedforward design, typically you wouldnt see the added feature of sidechaining and the extra stages required to perform it in a feedback unit.
--------- envelope <- scaling
assuming you have a fixed width font, that should be obvious.
but, basically since you're using a preset level (generally 0db) for your "threshold", (although it is a behaviour completely unrelated to its sidechain conterpart), and the control is based upon the output of the amplifier stage, obviously you wont be able to control a seperate signal accurately using this method.
if you used two signals with an average rms of the same value, you could get a sidechaining effect by controlling two amplifiers from one envelope. although, it is nothing like the typical feedforward sidechaining effect at all.
i suppose sidechaining is possible, but the effect would be quite different. generally feedback compressors are what you might find in a guitar insert or etc, while feedforward compressors are much more complex. since the goal is simplicity with the feedback design, and complex behaviour is the goal with the feedforward design, typically you wouldnt see the added feature of sidechaining and the extra stages required to perform it in a feedback unit.
-
- KVRAF
- 2049 posts since 18 Sep, 2003 from Seattle USA
OK thanks I see, that makes sense! It would seem to be pretty unpredictable then and the sound might not make sense...aciddose wrote: i suppose sidechaining is possible, but the effect would be quite different.
-
- KVRAF
- 11384 posts since 3 Feb, 2003 from Finland, Espoo
Which would indicate that UAD are not modelling the true behaviour of a fairchild correctly.. not even closely so then. Very disappointing indeed.BobYordan wrote:
UAD Fairchild - Silent.
- bManic
"Wisdom is wisdom, regardless of the idiot who said it." -an idiot
"They don't ban hate speech; they ban speech they hate." -an oracle
"They don't ban hate speech; they ban speech they hate." -an oracle