What does Waves Vitamin *actually* do? It ain't no saturator...

[imrae]

Since the effect implements a band-splitting filter I would say it is very likely to be unproductive to focus on phase shifting or similar linear effects. This is because it can take a lot of skill/knowledge to be able to identify a phase plot that matches perfectly to the expected error resulting from that band-splitting filter and so such an effect can end up being another red herring.

Indeed. I think the similarity in shape between my allpass setup and the phase response of HP/LP series is near enough that for now we should assume there is no additional phase trickery going on.

Waves has never put out a misleading campaign before in my experience, there's no reason why they would do it with Vitamin

I didn't mean to accuse them of fraud, just to warn that the information they supply is meant to sell the plugin rather than help us understand it. That 25 min demo video is really misleading because of poor level-matching. The dynamic behaviour of this thing is still different from any compression I've ever used, so it's not a regular parallel comp setup (or it's one with very interesting settings).

So: it screws with the phase as a side-effect of band-splitting, then does *something* that favours dynamic material and involves the "punch" control (envelope speed?), then feeds that into gentle(?) saturation. We're getting somewhere!

Hard to determine if saturation comes from the dynamics process until we know more about that part.

[aciddose]

I think really the only option to be certain of anything is to reverse engineer the code. It is reasonably trivial to set hard-coded coefficients to different values or to bypass portions of the effect using the appropriate tools, although the amount of effort to do so is insanely high.

That's normally what you'd do with a circuit: step 1) locate the signal path, 2) isolate different sections of the circuit, 3) independently analyze each section, 4) combine the different elements of the circuit and identify their interactions.

In software that is actually also possible ... just that you need a little more than a scope probe, soldering iron and wire cutters.

What I'd personally do though is try to measure the effect on different content. Can you identify the "maximum" effect the plug-in can have on a signal? Understanding what type of signal would for example produce the most harmonic content could provide a lot more insight.

Assuming that it is indeed a series of different processors however, and there is no reason not to consider what the marketing/authors said there, simple methods like passing in sine test signals are unlikely to pay off much more.

[imrae]

Understanding what type of signal would for example produce the most harmonic content could provide a lot more insight.

That is exactly the insight I'm after! If the documentation was more helpful this experimentation would be less necessary...

If something goes on the chain, I want to know what artefacts it is likely to cause and what kind of material it is sensitive to. That makes it easier to check for potential problems and debug them when they happen. Can't do that with an effect that "does something, just see if you like it".

[MogwaiBoy]

Don't forget it does its multiband with minimum phase crossovers, so that in itself will do something a little funky with your signal.

And there's this too:
https://www.gearslutz.com/board/music-c ... ancer.html

[aciddose]

Unfortunately I don't have the effect nor the trust/motivation to acquire and test it myself even as a demo version.

Well... my advice is to avoid using the wide band you suggested because we have absolutely no idea what effect if any the bandwidth might have on the underlying parameters.

Try instead using some "default" settings, does the effect have presets? Whatever the plug-in initializes with in a new instance after install should be fine. Send in plain sines at the center frequency and edges of each band, noting the different results you get if any on the spectrum.

Try to normalize the result and use a proper normalized analyzer. By sending in a 0 dB sine you should get a plot where the window is normalized and the peak is 0 dB, not -12 dB or similar like your graphs are showing which would probably be the result of the tool you're using failing to bother to do any of the steps necessary to get proper results.

For authors of such plug-ins and tools, for the love of... please just implement the most basic features like normalization of your windows. The normalized window is window * 1/mean(window), it really couldn't be any more simple! When computing the fourier transform you must normalize the magnitude by: 2/bins! It couldn't get any more simple.

xcope0.png

(I should really make a plug-in tool some day rather than wasting my time complaining toward deaf ears... This plot is with a Kaiser window: it approximates Dolph-Chebyshev's main lobe width but has a very steep roll-off instead of a flat one.)

Ensure you're using a proper window like Hann (raised cosine) as other windows (Blackman, etc) have numerous issues, side-bands and wide main lobes with poor resolution. Blackman-Harris-b is a reasonable compromise although it will rarely be present, Dolph-Chebyshev or Kaiser as an approximation to D-C are better.

You could use a AGC plug-in with slow settings on the output from "Vitamin" before the analysis tool to ensure the input to the tool is normalized. Also test the AGC plug-in to ensure it doesn't introduce any harmonics or other content on its own. AGC with zero distortion is a trivial thing to accomplish, worst case if you can't find a plug-in to do it for you you'll just have to trim the gain each time using a simple gain plug-in instead.

You can trim the gain to get a perfect 0 dB peak output on the analyzer so the results you're seeing make sense. What you want to be measuring is the difference in levels, not the absolute levels which don't matter at all.

Each measurement, record the gain compensation (+15 dB for example) and the spectrum graph. I'd name the screenshot (does your spectrum tool take screenshots for you?) with this information.

Example: Vitamin spectrum, default preset, sine input, 1kHz, 15 dB.png

Try:

• Plain pure sine input at various gains (-100 dB to +40 dB in 20 dB steps) and frequencies (the center and near the edge of each band.)
• AM modulated sines by sines, modulated at 1/10th Hz, 1 Hz, 2 Hz, 5 Hz, 10 Hz, 20 Hz, 50 Hz, 100 Hz.
• FM modulated sines by sines (same frequencies for modulation)
• The same of each with DC offsets added, note any differences and if none don't bother to record with offsets

So you need to iterate through a few steps with a few different test plug-ins to get a base-line response. Other tests include amplitude sweeps, frequency sweeps, varied waveforms (sine, cubic parabola(1/N^3), triangle(1/N^2), square(1/N), ramp(1/N), noise (brown, red, pink, white, gray/slate, blue) and impulse response.

As I said I doubt this will get you much further, but that's a good place to start. There really should be a VST plug-in to automate this entire series of tests and incorporate all these tools but I'm not aware of one. (...and too lazy to ever bother completing one.)

I've been thinking of creating my own spectrum + transient graph plug-in as most plug-ins seem aimed at real-time output, while that's really for the most part useless to do any proper analysis.

[simon.a.billington]

Since the effect implements a band-splitting filter I would say it is very likely to be unproductive to focus on phase shifting or similar linear effects. This is because it can take a lot of skill/knowledge to be able to identify a phase plot that matches perfectly to the expected error resulting from that band-splitting filter and so such an effect can end up being another red herring.

Indeed. I think the similarity in shape between my allpass setup and the phase response of HP/LP series is near enough that for now we should assume there is no additional phase trickery going on.

Waves has never put out a misleading campaign before in my experience, there's no reason why they would do it with Vitamin

I didn't mean to accuse them of fraud, just to warn that the information they supply is meant to sell the plugin rather than help us understand it. That 25 min demo video is really misleading because of poor level-matching. The dynamic behaviour of this thing is still different from any compression I've ever used, so it's not a regular parallel comp setup (or it's one with very interesting settings).

So: it screws with the phase as a side-effect of band-splitting, then does *something* that favours dynamic material and involves the "punch" control (envelope speed?), then feeds that into gentle(?) saturation. We're getting somewhere!

Hard to determine if saturation comes from the dynamics process until we know more about that part.

I think after a certain degree you just have to decide to roll with it. Does it sound good?? Is it doing it’s job?? Ultimately that’s really all that matters.

Of course that’s also a highly subjective thing that only we, ourselves can answer.

[imrae]

I think after a certain degree you just have to decide to roll with it. Does it sound good?? Is it doing it’s job?? Ultimately that’s really all that matters.

Of course that’s also a highly subjective thing that only we, ourselves can answer.

For effects that are not highly program-dependent, this is a fair point. I don't need to know what balance of pitch-shifting and modulation is used in a one-knob chorus, say.

But in a subtle mix effect it is really important to know if/when it can get loud and if/when it can get weird. If you put an upwards expander on a track you know to check what it's doing on the loudest and peakiest bits. I don't know where to check on Vitamin. I don't want to listen back to the whole track every time I move a knob, so a *little more* understanding is needed.

I really did think some simple qualitative testing would be enough, which is why the initial analysis was pretty cursory and missed the subtle THD. aciddose has outlined a thorough program of quantitative testing above, which does seem to be needed to get to the bottom of this. It's a pity Waves didn't provide some more enlightening examples in their own material, as they must have a pretty good idea of what the behaviour range is; instead they provided non-level-matched examples where it is pretty comparable to EQ.

[simon.a.billington]

I think after a certain degree you just have to decide to roll with it. Does it sound good?? Is it doing it’s job?? Ultimately that’s really all that matters.

Of course that’s also a highly subjective thing that only we, ourselves can answer.

For effects that are not highly program-dependent, this is a fair point. I don't need to know what balance of pitch-shifting and modulation is used in a one-knob chorus, say.

But in a subtle mix effect it is really important to know if/when it can get loud and if/when it can get weird. If you put an upwards expander on a track you know to check what it's doing on the loudest and peakiest bits. I don't know where to check on Vitamin. I don't want to listen back to the whole track every time I move a knob, so a *little more* understanding is needed.

I really did think some simple qualitative testing would be enough, which is why the initial analysis was pretty cursory and missed the subtle THD. aciddose has outlined a thorough program of quantitative testing above, which does seem to be needed to get to the bottom of this. It's a pity Waves didn't provide some more enlightening examples in their own material, as they must have a pretty good idea of what the behaviour range is; instead they provided non-level-matched examples where it is pretty comparable to EQ.

Yeah I can understand that. Normally I am the one to dissect things and how they work. However, a while ago I’ve come to realise that sometimes there are just no easy answers or any answer at all. I figured that it would be a much better outcome for myself to stop head butting brick walls, stop obsessing and make better use of my time.

I’m talking about me, by the way, and I can obsess quite a lot, I don’t always catch myself doing it either.

With Vitamin, personally, I decided to let it be. I don’t have to know how to put an engine together to drive a car. So I just use Vitamin. I either like what it is adding to the sound or I don’t. I’ll play with the controls a bit more and if I still don’t like it I’ll pass on it. At the heart of it, I know it’s achieving kind of doing an eq, compression, saturation thing... however it’s going about it. That’s enough for me.

I also use the Aphex Exciter without knowing much of what it’s doing. I understand eqs have resistors innthem chained in a particular way to do their thing, I’m sure it’s more complicated than that. Either way, i just use those too. Just another choice in tool to get the job done.

This is not to be dismissive of you’re need to understand, I do respect that. I’m also like minded, remember. You have to think to yourself though, what will you do if your investigations come up short. My suggestion would be to just use it and evaluate whether you like what it’s doing or not.

[ruffindajungle]

Glad they don't let those pragmatic math engineers near the music creation. Effects could well be described as colors and saturation width of the brush. And it sounds more compelling than any maths, in this field.

[stromkraft]

it sounds more compelling than any maths, in this field.

I have no idea what that's supposed to mean. Digital audio is all maths whether you think so, or not.

[ruffindajungle]

it sounds more compelling than any maths, in this field.

I have no idea what that's supposed to mean. Digital audio is all maths whether you think so, or not.

"it sounds more compelling than any maths" Sry, by "it" i meant thinking like a painter rather than mathematician. It's terrible to read threads like this for example :

https://www.steinberg.net/forums/viewto ... 3&t=155663

Close your calculator and eyes too while your at it and open your ears

[imrae]

What are you trying to say, that knowing physics makes my hearing worse or something?

[Ou_Tis]

it sounds more compelling than any maths, in this field.

I have no idea what that's supposed to mean. Digital audio is all maths whether you think so, or not.

"it sounds more compelling than any maths" Sry, by "it" i meant thinking like a painter rather than mathematician. It's terrible to read threads like this for example :

https://www.steinberg.net/forums/viewto ... 3&t=155663

Close your calculator and eyes too while your at it and open your ears

Guessing you don't actually know any mathematicians very well (or understand what advanced math and physics is like at all... or really basic axiomatic geometry and analysis for that matter).

Pattern recognition is a huge part of music. Mathematics is the essence of pattern.

[reggie1979]

Da maths are important! (especially when doing subs, 1 dollar a month for a year is a sin because paying 120 bucks is SO much smarter! )

Anyways, I have it, I rarely use it, I'd sell it but it's not under WUP.

[stromkraft]

"it sounds more compelling than any maths" Sry, by "it" i meant thinking like a painter rather than mathematician.

If you're trying to say that music should be created (& enjoyed) with little concern over the underlying technology, then I'm with you. But all audio technology, including effects like Vitamin, are inescapably math-based. Therefore all audible results and other aspects of such digital tools can be analyzed and explained within the digital math-based domain. That includes physical modeling.