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
Lets do it: The Ultimate Plugin Analysis Thread Dynamics Plugins
Old 24th April 2018
  #3181
Quote:
Originally Posted by meloco_go View Post
Maybe a more complicated test tone e.g. modulated sines with varying modulation periods can be used. Attack/release behavior should interfere with the phase of the modulation and after some analysis, these aspects could probably be calculated. But that's way beyond my level of math prowess...
That's a square wave (set to a very low frequency, say, 5 or 10Hz).

In my experience, sine bursts and white noise is also sufficient.

Sine bursts for example give good info on frequency dependency. White noise gives an idea of what a flat and full bandwidth provokes.

The missing part is that the resulting signal should be further rectified and smoothed a bit (AC to DC conversion) to generate nice smooth plots. No idea if any audio editor offers AC to DC conversion. Could be worth building a small plugin doing just that, to produce nice DC plots from AC signals.



Finally, you should also try very low frequency triangle waves, as they have the ability to directly "draw" the effective gain transfer curve.
Old 24th April 2018
  #3182
Gear Addict
 

Quote:
Originally Posted by FabienTDR View Post
That's a square wave (set to a very low frequency, say, 5 or 10Hz).
The problem IMO is that such waveforms stress the analog part of the compressor , which makes analysis difficult.

Quote:
Originally Posted by FabienTDR View Post
Finally, you should also try very low frequency triangle waves, as they have the ability to directly "draw" the effective gain transfer curve.
It is a nice test with plugins but again, hardware would just roll off these frequencies.
Old 24th April 2018
  #3183
Maybe they do, but then, that's what they do. And likely the reason why they sound different to the mathematically more likely correct digital pendants (is that an English word?).

They limit the bandwidth, cut the lows, add phase shift. No true dynamics processing business, but that's what they do.

The AC/DC problem is independent of analogue or digital, it's simply necessary to produce proper step response plots from an AC device (except if you can access the internal gain control signal in a way or another, i.e. directly use this DC signal).

Last edited by FabienTDR; 25th April 2018 at 12:36 AM..
Old 25th April 2018
  #3184
qwe
Lives for gear
 

Quote:
Originally Posted by FabienTDR View Post
pendants (is that an English word?).
Lets do it: The Ultimate Plugin Analysis Thread-pedants-wikipedia-source-public-domain-release.png

I think you meant to say pedants. ;-)

Which, clearly, I am one of.
Attached Thumbnails
Lets do it: The Ultimate Plugin Analysis Thread-pedants-wikipedia-source-public-domain-release.png  
Old 25th April 2018
  #3185
Sorry, I meant counter-part/mirror image. It's a common word in German and French, so I expected it to be known in English, lol.
Old 25th April 2018
  #3186
qwe
Lives for gear
 

Quote:
Originally Posted by FabienTDR View Post
Sorry, I meant counter-part/mirror image. It's a common word in German and French, so I expected it to be known in English, lol.
Ah yes, that meaning is in the dictionary, just not in my head's dictionary.

Quote:
Originally Posted by FabienTDR View Post
The missing part is that the resulting signal should be further rectified and smoothed a bit (AC to DC conversion) to generate nice smooth plots. No idea if any audio editor offers AC to DC conversion. Could be worth building a small plugin doing just that, to produce nice DC plots from AC signals.
Hmm, the "JS - Gain Reduction Scope" plugin script that comes with Reaper could be modified to enable audio files of the smoothed gain reduction sample values to be rendered.

Here's what I did to achieve this:


Add the output pins after line 18:

Code:
in_pin:left input 1
in_pin:right input 1
in_pin:left input 2
in_pin:right input 2
out_pin:left output
out_pin:right output
options:no_meter
In the @sample section, change these lines...

Code:
 recpos[0]=rmsb.rms.get_l() - rmsa.rms.get_l(); 
 recpos[1]=rmsb.rms.get_r() - rmsa.rms.get_r();
to the following:

Code:
 spl0 = recpos[0]=rmsb.rms.get_l() - rmsa.rms.get_l(); 
 spl1 = recpos[1]=rmsb.rms.get_r() - rmsa.rms.get_r();
 spl0 = 2 ^(spl0 / 0.696468);
 spl1 = 2 ^(spl1 / 0.696468);
EDIT: Corrected scaling coefficient from 0.692 to 0.696468.

Feed uncompressed signal to channels 3/4, compressed to 1/2. Positive values = amount of gain reduction by the compressor under test.


Attached: Sample view of JS - Gain Reduction Scope, rendered floating point WAV file opened in iZotope RX6.
Attached Thumbnails
Lets do it: The Ultimate Plugin Analysis Thread-reaper-gain-reduction-scope.png   Lets do it: The Ultimate Plugin Analysis Thread-exported-rms-gr.jpg  

Last edited by qwe; 25th April 2018 at 06:23 PM.. Reason: Corrected scaling coefficient from 0.692 to 0.696468.
Old 25th April 2018
  #3187
Cool!
Old 25th April 2018
  #3188
Looks pretty smooth! Now one should be able to use all sorts of bursts, from sine over sum of sines to noise based signals.
Old 26th April 2018
  #3189
Quote:
Originally Posted by qwe View Post
Hmm, the "JS - Gain Reduction Scope" plugin script that comes with Reaper could be modified to enable audio files of the rectified and smoothed gain reduction sample values to be rendered.
Thank you for a very useful technique to analyze compressor’s envelope. I used rms = 3 ms for quicker response. This test is much more precise, then looking at the waveform of the rendered track.

After 3 hours of testing, I found NI vc76 to be the closest to UAD 1176. (Tr5 is second closest) I was shocked how ridiculous every slate fg166, waves cl76, psp fet, overloud fet - 1176 is not easy to code.

I tested dbx160 style. Waves, ni, and uad are actually really close to each other. A little bit different frequency dependency of compression, but the envelope is the same.

Tr5 ssl, distressor, glue has pretty much the same envelope curve. But the tone is different.

That was intense testing.

Once again it’s only compressor’s envelope test with precise metering.

Upd.: oh, and uad la-2a legacy is actually a peak-compressor. crazy, right?

Last edited by EvgenyStudio; 27th April 2018 at 06:49 AM..
Old 1st May 2018
  #3190
Quote:
Originally Posted by qwe View Post
Possibly the analogue hardware has overall "smoother" action?
Yes, you absolutely right.
I checked a lot of compressor demos from hardware compressor thread in shootout section. It seems like peak compressor exists only in digital domain.

This method (gain reduction slope) also can show you if there’s some harmonic and frequency coloration (saturation and eq). And even some big units like shadow hill mastering compressor doesn’t have any eq coloration and also have not to much saturation. Retro 176, distressor... i mean there’s a lot of samples. And every single one is clean and open.

I was really shocked when uad la-2a start acting like a peak compressor. Thoughts like “everything is a lie” came in.

I check hardware la-2a samples and “the closest to real thing” plugins are tla100a and vc2a.

Kotelnikov is a mastering compressor beast. With linking control set to 50% it sound alive on mix buss.

Once again, qwe, thank you for showing this method.
Old 2nd May 2018
  #3191
Gear Addict
Quote:
Originally Posted by EvgenyStudio View Post

After 3 hours of testing, I found NI vc76 to be the closest to UAD 1176. (Tr5 is second closest) I was shocked how ridiculous every slate fg166, waves cl76, psp fet, overloud fet - 1176 is not easy to code.
Did you also test the Softube FET? I ask because your results seem to favor Softube code (also for your LA2A tests) and they also coded the VC76.
Old 2nd May 2018
  #3192
Quote:
Originally Posted by nonstatic View Post
Did you also test the Softube FET?
Yes, it also really good (i mean its envelopes looks like reference envelope from hardware 1176). If you have it - use it. It behaves like 1176.

But for me it looks like they did "fet circuit" for vc76 a little better. So envelope follower looks closer to reference.

I believe they didn't improve code for FET plugin from the release. They fixed bugs, gain reduction meter, and some minor things, but "circuit" is the same as was released back in... 2008?
But it still an impressive plugin.

Also, keep in mind, softube plugins run with oversampling, and native instruments plugins don't. If you like pushing plugin compressor a little too hard, softube fet would be a much nicer tool for that.
Old 2nd May 2018
  #3193
Lives for gear
 
JoeyM's Avatar
Quote:
Originally Posted by EvgenyStudio View Post
Yes, it also really good (i mean its envelopes looks like reference envelope from hardware 1176). If you have it - use it. It behaves like 1176.

But for me it looks like they did "fet circuit" for vc76 a little better. So envelope follower looks closer to reference.

I believe they didn't improve code for FET plugin from the release. They fixed bugs, gain reduction meter, and some minor things, but "circuit" is the same as was released back in... 2008?
But it still an impressive plugin.

Also, keep in mind, softube plugins run with oversampling, and native instruments plugins don't. If you like pushing plugin compressor a little too hard, softube fet would be a much nicer tool for that.
Any of you running a recent Presonus Studio One and test the Fat channel? That's an 1176 and LA2A emulation onboard and would be very helpful to see how it stacks up. I'd volunteer but my heart meds make me out of your element

Probably piece of cake for those of you who are good at routing.
Old 2nd May 2018
  #3194
qwe
Lives for gear
 

Quote:
Originally Posted by FabienTDR View Post
Looks pretty smooth! Now one should be able to use all sorts of bursts, from sine over sum of sines to noise based signals.
I think what's needed is a nice list of such test signals to use. :-)
Old 3rd May 2018
  #3195
Quote:
Originally Posted by qwe View Post
I think what's needed is a nice list of such test signals to use. :-)
One approach that I often use to verify my stuff are sine-bursts in a bed of noise.

1. Create an empty file
2. Add a white noise floor around say, -40dB.
3. Now add high level sine bursts to it, of varying frequency and durations. All with generous space between them.

The idea is to track the release phase by watching how the noise floor behaves after different types of bursts stop.

Good sounding compressors typically vary the release phase in dependence of burst duration and frequency.


Generally, it's rather difficult to analyse a dynamics processor as a black box. Given sufficient complexity, it can likely be impossible. Imagine a compressor showing a moon phase dependency, it could take years to even observe these effects 1, 2 times before they disappear again for years.
Old 25th May 2018
  #3196
Gear Nut
 

Quote:
Originally Posted by FabienTDR View Post
Imagine a compressor showing a moon phase dependency,...
now THAT sounds like some hardware one could sell to audiophiles
Old 6th June 2018
  #3197
Here for the gear
 

Freeclip

Here's what 'Freeclip' clipper and waveshaper modes look like.

Hopefully someone finds this useful.
Attached Thumbnails
Lets do it: The Ultimate Plugin Analysis Thread-arctangent.png   Lets do it: The Ultimate Plugin Analysis Thread-algebraic.png   Lets do it: The Ultimate Plugin Analysis Thread-hyperbolic.png   Lets do it: The Ultimate Plugin Analysis Thread-cubic.png   Lets do it: The Ultimate Plugin Analysis Thread-quintic.png  

Lets do it: The Ultimate Plugin Analysis Thread-hard.png   Lets do it: The Ultimate Plugin Analysis Thread-freeclip.png  
Attached Images
Lets do it: The Ultimate Plugin Analysis Thread-image.png 
Old 7th June 2018
  #3198
Here for the gear
 

Those images above are clipping at -1db

Each mode had different thresholds before aliasing occurs. Its handy to know how far you can push things with this plugin and remain alias free.

The gain levels vary dramatically depending on the rate 48/96 and the clip/waveshape modes;

Algebraic @48 gain +1.8db
Algebraic @96 gain +8.0db
Arctangent @48 gain +1db
Arctangent @96 gain +4db
Cubic @48 gain +3.4db
Cubic @96 gain +3.4db
Hyperbolic @48 gain +6db
Hyperbolic @96 gain +11db
Quintic @48 gain +1.8db
Quintic @96 gain +1.9db

I've excluded the default hard clip for obvious reasons.

Old 7th June 2018
  #3199
Lives for gear
 
dotl's Avatar
 

what about softube cl1b mk2? has anyone analysed it yet?
Old 8th June 2018
  #3200
Oh! Plugindoctor now supports external I/O!

Looks like we can finally launch a new "ultimate hardware analysis" thread!
Old 8th June 2018
  #3201
Gear Guru
 
UnderTow's Avatar
Quote:
Originally Posted by FabienTDR View Post
Oh! Plugindoctor now supports external I/O!

Looks like we can finally launch a new "ultimate hardware analysis" thread!
I'm in the process of measuring all aspects of my Silver Bullet. Quite interesting.

Alistair
Old 8th June 2018
  #3202
Lives for gear
 
zvukofor's Avatar
Oh my, Plugindoctor revealed all stuff i've heard in compressors! Now i can see how they react! And wonder, why there're no universal comp plugin that has any A/R curves besides log and sigmoid, basically a pot - concave/convex, and overshoot parameter, transient squash parameter...
Old 8th June 2018
  #3203
Gear Guru
 
UnderTow's Avatar
Quote:
Originally Posted by zvukofor View Post
Oh my, Plugindoctor revealed all stuff i've heard in compressors! Now i can see how they react! And wonder, why there're no universal comp plugin that has any A/R curves besides log and sigmoid, basically a pot - concave/convex, and overshoot parameter, transient squash parameter...
DMG Compassion...

Alistair
Old 14th June 2018
  #3204
I did a little analysis of my 1073 clone by Warmer (it's not Warm Audio, it's Russian custom builder). As you can see, it has a flat frequency response. When used as "normal mic preamp", there's little coloration in the spectrum after 60 hertz. I have REW file with deep analysis of harmonics, I attached it (with Babyface Pro and Apollo Twin measurements).

Looks interesting! Because of every single neve-emulation by slate, ua, arturia, etc has weird frequency response and harmonic saturation. Only true 1073 preamp plugin is Black Rooster 73

Plugin doctor is must have, right now!

notes:
1 when I test frequency response I forget to disengage HPF in RME line in EQ.
2 there's 50 and 100 hz hum happening from power supply.
Attached Thumbnails
Lets do it: The Ultimate Plugin Analysis Thread-1073-dist-mode-50-hz.png   Lets do it: The Ultimate Plugin Analysis Thread-1073-dist-mode.png   Lets do it: The Ultimate Plugin Analysis Thread-1073-freq-responce.png   Lets do it: The Ultimate Plugin Analysis Thread-1073-normal-condition-50-hz.png   Lets do it: The Ultimate Plugin Analysis Thread-1073-normal-condition-449-hz.png  

Attached Files
File Type: mdat RME vs UA vs 1073.mdat (7.46 MB, 46 views)
Old 14th June 2018
  #3205
Agreed on the Black Rooster VPRE-73. The first time I heard it i knew I had something special.
Old 21st June 2018
  #3206
Lives for gear
 

Quote:
Originally Posted by FabienTDR View Post
Andy answered already.

Allow me to add that the ultrasonic filter has been built in the progress of a GS thread where I questioned the trend to super high res (samplerates beyond 96kHz) in audio production. I showed that independently of anti-aliasing considerations, the THD of chains of nonlinear processors is negatively affected by the system's bandwidth (as soon it goes beyond the audible bandwidth).

More bandwidth = more IMD distortion.
The narrower the input signal's bandwidth, the more harmonic (i.e. "nicer") can nonlinear processors act.

And it's in this context that we quickly developed a small proof of concept allowing anybody to directly compare the effects of filtered and unfiltered 192kHz processing.

The audible results were ambiguous, though. While complex chains really improved their sound impressively, simple examples mostly suffered under the side-effects of the filtering (which exist, too. No free lunch).

IMHO the main lesson one should memorize is simple: When messing with nonlinear devices, make sure the signal doesn't have a too wide bandwidth. Everybody likely already experienced the point that a distortion device can sound magnificent with the right input signal, say, a guitar input, vocals, bass. But once you apply it to signals of greater complexity (and bandwidth), it starts to sound horrible because the nonharmonic partials become too strong and numerous.

The ultrasonic filter essentially fights the same effect, but at a larger scale if you want (normal SRs vs "hi res" SRs). But it is in no way necessary in classic, "standard rate" setups. Below or at 96kHz, benefits can be hard to find. Don't use it blindly.
Fabien,

I'm still investigating your input here.

Your said:

Quote:
More bandwidth = more IMD distortion.
The narrower the input signal's bandwidth, the more harmonic (i.e. "nicer") can nonlinear processors act.
If that is true, then it would make sense to apply an butterwoth 2nd order filter on every channel at lets say 30hz and 16khz. Does it?

Also, does this also apply to analog gear?

I finding more and more reasons to track at 44,1 instead of 48k that I currently do.

Thanks
A.
Old 21st June 2018
  #3207
Quote:
Originally Posted by planck View Post
If that is true, then it would make sense to apply an butterwoth 2nd order filter on every channel at lets say 30hz and 16khz. Does it?
No, this wouldn't help I'm afraid.

First of all, such filters would grab deep into the audible area, audibly filter frequency is specified at the -3dB point, not where "it stops". They also wash out the phase.

Second, 2nd order simply aren't strong enough. 44.1kHz to 88.2kHz means one additional octave.

You need very steep linear phase filters.

Quote:
Originally Posted by planck View Post
Also, does this also apply to analog gear?
Generally yes and no. Yes because analogue is very nonlinear, no because analogue is "naturally" band-limited.

But as mentioned above several times, I do not think that any fixed filter setup would really help at the end of the day.

IMHO the important lesson behind this whole IMD discussion (there's also a dedicated thread here: Some thoughts on "high resolution" audio processing), is to be aware of these effects and including them into your decisions. Be it grouping as a bus vs not doing so, prioritize individual tracks over master bus processing, avoiding to saturate wideband material to excessively or serially. Keeping an eye on the ultrasonic area, things like that. Occasionally, a good filter can help. But not as a default bottleneck.

Please check the linked thread, this is primarily a problem of very hi resolution audio formats, 88.2kHz and beyond.
Old 18th July 2018
  #3208
Lives for gear
 
stinkyfingers's Avatar
 

Quote:
Originally Posted by EvgenyStudio View Post
Only true 1073 preamp plugin is Black Rooster 73
the saturation for this is a chebyshev waveshaper...(only generates up to 6 harmonics)
is it really "true" in regards to harmonic distortion ?
Old 18th July 2018
  #3209
Quote:
Originally Posted by stinkyfingers View Post
the saturation for this is a chebyshev waveshaper...(only generates up to 6 harmonics)
is it really "true" in regards to harmonic distortion ?
You can test it for yourself. I attached some pics and even REW files for reference in my previous post.
For me it’s the closest plugin in regards to harmonic distortion.
Old 18th July 2018
  #3210
Lives for gear
 
stinkyfingers's Avatar
 

Quote:
Originally Posted by EvgenyStudio View Post
You can test it for yourself. I attached some pics and even REW files for reference in my previous post.
For me it’s the closest plugin in regards to harmonic distortion.
well, i can't because i don't have a 1073 (whatever that is)...
are there audio files in the REW file ?
i opened it with REW but don't know how to get any audio/.wav (if possible?)
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
n8tron / Music Computers
258
timbreman / Electronic Music Instruments and Electronic Music Production
22
Realziment / So much gear, so little time
6
djpolaco / High end
2
jchadstopherhuez / So much gear, so little time
11

Forum Jump
Forum Jump