DDMF Plugindoctor

[DeathByGuitar]

I meant app, but yeah.. Same idea.

[Sam-U]

Like I said, Christian Budde is sadly not around this place anymore. But he'd definitely have something to say about that. My personal spin: it's 1kHz because how (easy?) it can be built in hardware form.

Because a) it's easy to remember and b) you can nicely spot it on a linear scaled grid Take whatever you want. I prefer 980Hz on 44.1kHz SR since this will result in a nice integer number of samples per period.

I'll buy it at the weekend. I've been searching a long time for such a tool with a nice interface and multiple measurement methods. And since it comes from DDMF I'm sure there will be more options added in future updates.

[plexuss]

I suggested a feature I hope he implements: one-click save all screens. I want to run my Nebula 4 library through it so I have reference plots to refer to but it would be a lot of work to do manually. so a one-click to export all screen shots would be really useful.

[Robert Randolph]

EDIT: see this post - http://www.kvraudio.com/forum/viewtopic ... 9#p7012489

So if I run Plugindoctor on MJUC Mk3 default settings I get this for linear analysis:



However, if I run a sine sweep and take the delta I get this:



Here's confirmation that my sweep+delta is probably correct (mostly) for filters with non-linear phase response:



A second test of my methodology:



(test_signal + -processed_signal) + -test_signal

EDIT: see this post - http://www.kvraudio.com/forum/viewtopic ... 9#p7012489

[docdued]

I get this with McDSP ML8000 - doesn't seem right. Any explanations?

I've just checked the demo version of the McDSP ML8000, and while it reports a latency of zero samples to the host, it actually creates 55 samples of latency. Hence the "wrong" phase display (which is actually correct, since, with an uncorrected latency, this is what you get).

[Compyfox]

So if I run Plugindoctor on MJUC Mk3 default settings I get this for linear analysis:

Two simple solutions:

1) raise the calibration to 0dB, because the test signal (what you call "delta") fires at maximum signal strength in PluginDoctor, so does the sine wave for the THD+N

2) activate istage for a "flat response" (also see the tool tip of the compressor)

3) realize that a compressor is not an EQ and that you won't always get a flat response




Also.. please explain for the other viewers (incl myself), how you create a frequency plot like with PluginDoctor in Voxengo SPAN (Free), but with a sine sweep? It's not conclusive what you actually did

[Robert Randolph]

So if I run Plugindoctor on MJUC Mk3 default settings I get this for linear analysis:

Two simple solutions:

1) raise the calibration to 0dB, because the test signal (what you call "delta") fires at maximum signal strength in PluginDoctor, so does the sine wave for the THD+N

2) activate istage for a "flat response" (also see the tool tip of the compressor)

3) realize that a compressor is not an EQ and that you won't always get a flat response

These aren't really solutions, because it appears that Plugindoctor is showing the wrong response.

I'm not interested in a flat response from MJUC, I'm interested in the _actual_ response.

Also.. please explain for the other viewers (incl myself), how you create a frequency plot like with PluginDoctor in Voxengo SPAN (Free), but with a sine sweep? It's not conclusive what you actually did

I did explain it partially at the bottom of my post, so I will elaborate:

The routing is as such: (test_signal + -processed_signal) + -test_signal. With - being a polarity inversion. The test_signal is a 1minute 10>22khz sine sweep.

You can set up the routing with sends if you want, but I just do it with Reaper I/O pins.

I then setup Span to the largest FFT size, with the "Max" type. This records the maximum bin values for each bin and displays it as such.

I've used this many times for frequency analysis, and I've found that it's spot on with AP testing on identical devices (except some inaccuracies near the low end, obviously). The result does need to be normalized though, but in this case only the relative values are of concern.

Here is a digital loopback ran through REW with MJUC at the same settings, this is much more accurate:



Red is a -18dbFS sweep. Green is a -3dbFS sweep.

This is about on par with the Span analysis, except for the obvious spectral resolution difference at lower frequencies that disguises some of the low-end bump.

I'm fairly certain that Plugindoctor's linear analysis is incorrect in this case regardless.

EDIT: see this post - http://www.kvraudio.com/forum/viewtopic ... 9#p7012489

[docdued]

I don't think Plugindoctor is showing an incorrect result. It seems that said plugin has a highly nonlinear response to transients so it makes a noticeable difference whether you are measuring with a delta peak or a sine wave. They are only equivalent if the plugin's response is strictly linear.

[Compyfox]

These aren't really solutions, because it appears that Plugindoctor is showing the wrong response.

I'm not interested in a flat response from MJUC, I'm interested in the _actual_ response.

I tried it with my usual methods of testing, and PluginDoctor shows the correct plot readout. The trick is to change the reference level of MJUC. Then you get similar readouts you have with your REW plot.

I did explain it partially at the bottom of my post, so I will elaborate:

The routing is as such: (test_signal + -processed_signal) + -test_signal. With - being a polarity inversion. The test_signal is a 1minute 10>22khz sine sweep.

You can set up the routing with sends if you want, but I just do it with Reaper I/O pins.

That does sound overly complicated if you ask me. If it works for you, great. But there are (way) simpler ways.

This is about on par with the Span analysis, except for the obvious spectral resolution difference at lower frequencies that disguises some of the low-end bump.

I'm fairly certain that Plugindoctor's linear analysis is incorrect in this case regardless.

Again, can't confirm drastic offsets happening here.

Proof:
Test Signal (WAV, *cough* "Delta Peak") -> Lightbridge Signalizer (resolution: 16384 samples)

MJUC_Test_01.PNG

MJUC_Test_02.PNG

versus PluginDoctor (normal resolution = 16384 samples) and MJUC mk3 mode (default settings, 0dB as reference)

MJUC_Test_03.png

It's all down to the scaling of the plot it seems (notice the first plot having a measured "bass boost" of 10dB, while the other plots all have a boost of about 3dB!). And the right settings on the plugin considering the input signal strength for measurements.

[cturner]

Anyone else find that Klanghelm's DC8C2 crashes PluginDoctor on switch to Expert mode, or I think setting threshold in any mode?

[cturner]

I tried it with my usual methods of testing, and PluginDoctor shows the correct plot readout.

I did a quick setup, viewing an impulse going through MJUC Mk3 into BC FreqAnalyst, and I get results much closer to Compyfox 's than the posted sine sweeps displayed in SPAN.

[Robert Randolph]

I don't think Plugindoctor is showing an incorrect result. It seems that said plugin has a highly nonlinear response to transients so it makes a noticeable difference whether you are measuring with a delta peak or a sine wave. They are only equivalent if the plugin's response is strictly linear.

Aha. That is the case.

Thanks.

[Compyfox]

Another test with MJUC mk3. This time, at 0dB ref and Istage activated


Once more: WAV -> Lightbridge Signalizer

MJUC_Test_04.PNG

And direct comparison with PluginDoctor

MJUC_Test_05-istage.png

Reason (directly from the tooltip)
"Istage mode: enables the interstage transformer, which is located between the GR-stage and final amp stage. Alters the tone and the characteristics of the compression. Helps to keep the lowend more intact. Also lowers the noise of the unit"


Yes, the device is highly nonlinear - it's a compressor after all. So measuring "what's going on" isn't necessarily easy and eyes can fool you. If all analyzers would use the same UI size/scale/zoom mechanics, then comparing the results would be super simple. But were it so easy.

[Robert Randolph]

Another test with MJUC mk3. This time, at 0dB ref and Istage activated


Once more: WAV -> Lightbridge Signalizer

MJUC_Test_04.PNG

And direct comparison with PluginDoctor

MJUC_Test_05-istage.png

Reason (directly from the tooltip)
"Istage mode: enables the interstage transformer, which is located between the GR-stage and final amp stage. Alters the tone and the characteristics of the compression. Helps to keep the lowend more intact. Also lowers the noise of the unit"


Yes, the device is highly nonlinear - it's a compressor after all. So measuring "what's going on" isn't necessarily easy and eyes can fool you. If all analyzers would use the same UI size/scale/zoom mechanics, then comparing the results would be super simple. But were it so easy.

It seems more so that my error was largely due to only checking with a sine sweep instead of also checking with an impulse response.

I should really sleep before I delve in to these things. Clear mental error on my part, and I updated my posts with a link to the correction.

BTW, thank you for sharing your analysis as well.

[Hermetech Mastering]

This is fab, thanks Christian!