Analog type analyzer?

[Vitaly]

Hi Vojtech,

I'm rare guest, here, so my apologies if it was already asked.

I wonder if you have any plans for an analog type analyzer? You could make it with Melda's twist - selectable number of bands, filter slope. Basically, it is a crossover with a meter at each band's output. And if we, users, are able to set a crossover's split freq-cy and a level's of each band, we get a great tool. One could set such an analyzer to be anything - 1 oct, 1/3 oct, Bark scale and so on.

What do you think?

Vitaly.

[MeldaProduction]

Hi Vitaly,

back in business?
Ok, well, analyzers... right now there are so many interesting ideas, that this doesn't seem to be so interesting to be honest. Anyway the main reason people are using FFT instead is the efficiency - with crossovers it is almost impossible to handle it with current CPUs, unless you are ok with say 10 bands. More than that it just takes too much CPU and the accuracy is bad, because the crossovers are not steep enough (they can be, at the cost of even more CPU...)...

[Nokenoku]

Anyway the main reason people are using FFT instead is the efficiency - with crossovers it is almost impossible to handle it with current CPUs, unless you are ok with say 10 bands. More than that it just takes too much CPU and the accuracy is bad, because the crossovers are not steep enough (they can be, at the cost of even more CPU...)...

Huh?

Fruity Vocoder for example has 128 bands, and the CPU consumption is absolutely no problem nowadays (even with the 3rd order filters).

I really wish there'd be an analyzer plugin, which uses a similar technique like the Fruity Vocoder. Maybe with up to 512 bands, a good display and similar behaviour.

[MeldaProduction]

I'm afraid it's not that simple. Vocoder is quite something different, though similar tricks could be used in an analyzer. But note, these are tricks, it's not true, that now 512 bands is ok, that is not true at all. And quite frankly the income for an analyzer is minimal. Check the "super-resolution" mode on our analyzers. That gives you basically the same thing, but with minimal CPU requirements.

[Nokenoku]

I'm afraid it's not that simple. Vocoder is quite something different, though similar tricks could be used in an analyzer. But note, these are tricks, it's not true, that now 512 bands is ok, that is not true at all. And quite frankly the income for an analyzer is minimal.

???
Sorry, I don't understand you.

The Fruity Vocoder has a display of its bands. And with the right settings it shows what's going on much better than all the analyzers I've tested so far.
The detection of that plugin works really well.

Haven't checked out your analyzer yet though.
I don't want to install a whole bundle to test it (portable would be great).

Just checked Youtube to see, if there's a video on it. There isn't.
(But there's something else, you probably want to get removed. Just search for MAnalyzer on Youtube and you will see.)

[MeldaProduction]

Ok, that's up to you, I really don't understand and you should check our analyzers anyway. Just check the MDynamicEq for example, it has the analyzer too. And if it's not good enough, you can tell me what exactly is wrong with analyzers, in other words what's the advantage of the vocoder.

[Nokenoku]

And if it's not good enough, you can tell me what exactly is wrong with analyzers, in other words what's the advantage of the vocoder.

For FFT-Analyzers I've testes so far:

1. Too slow:
Not even speaking of the delay from the FFT window size, but most analyzers only display a "dia show".
SPAN for example is terrible, its update-rate of the display feels like 10fps or something like that.

Now I have just installed yours and I think it's not much better. By using "overlap" the image gets smoother, but I guess it's just some sort of interpolation between the existing pictures, so it makes a fluid picture, but with more latency than it had to be.
(By the way .. are there manuals for your plugins?)

The best FFT analyzer in that regard I've seen so far seems to be that one from rs-met, but it has other disadvantages.
I guess it processes multiple overlapping windows, which would also explain its CPU hunger with higher window-sizes.

In my opinion every analyzer should do it like that, preferably with an option to set up the rate.
Also I think analyzers could use different window sizes depending on the frequency area.

2. Bad display:
In most analyzers that thin line is really not good to read.
2 big plus points for MAnalyzer here. It has coloring below the line, and there is smoothing for the line.
(Even better though it would be to have something like what happens in Fruity Vocoder, when you increase the bandwidth. This options seems to preserve the high energy parts of the spectrum better than your current implementation of line smoothing.)

By the way, I have seen a screenshot of MAnalyzer with bars?
Where has this option gone?

[MeldaProduction]

Ok so:

1) Yes, that's exactly what is super-resolution for. Btw. there's a physical limit - 20Hz for example has a period of 50ms, so even with the maximum overlap (btw. it uses overlap, no interpolation or whatever are you talking about) you cannot get quicker response. That's just physics.
And when it comes to drawing, this seems to be problem with your HW. On my computer the update rate is faster than I can see, so I always use averaging to slow it down (btw. averaging is enabled by default, another thing you should check before making these statements).
What computer are you using?

2) So basically you want bars instead of a line. Well, some people prefer it, but in my opinion it's just a matter of what you are used to. And with smoothing you get much better resolution. Bar analyzers are essentially unusable for some surgical tweaking. And "keeping energy in higher frequencies"??? What is that? You are probably talking about the attack/release in the vocoder... That is done to make the vocoder sound good, not to create a realistic analyzer.

All in all, sorry, but this has nothing to do with this topic. Both complaints are about drawing, when first one seems like some kind of problem on your computer, second is a matter of taste. We have been talking about filtering instead of FFT. This is simulated with in fact higher precision and less CPU consumption by super-resolution mode.

[Nokenoku]

And when it comes to drawing, this seems to be problem with your HW. On my computer the update rate is faster than I can see, so I always use averaging to slow it down (btw. averaging is enabled by default, another thing you should check before making these statements).
What computer are you using?

Ok, that might be something worth checking out. I just noticed, that my whole display is slowing down, when I open the GUI of MAnalyzer.
I'm using a Q9550 and a GTX460. My host is FL Studio. OS is Windows 7 64bit.
Maybe it's because the graphics card powers down in 2D mode?

I still don't know, what "super resolution" is supposed to do and what exactly "overlap" means. Are there really no manuals for your plugins?

And "keeping energy in higher frequencies"??? What is that? You are probably talking about the attack/release in the vocoder... That is done to make the vocoder sound good, not to create a realistic analyzer.

I wasn't talking about high frequencies, but rather peaks with a high energy.
What I mean is probably just a different way of smoothing. Nothing about attack/release times. If you turn up the bandwidth in the Fruity Vocoder you get a smoothed up curve, but it behaves in a different way than the smoothing in MAnalyzer.
In some cases it's more helpful.

[MeldaProduction]

The analyzer sometimes takes down other drawing. Basically with all plugins it§s like a fight for drawing resources. You take more and get quicker updating, you take less leaving it for the others and your FPS goes down...

Docs are not available for freebies. Though there are "?" buttons everywhere, and you can get context help by pressing F1 with mouse over a control.

About the second point - that's definitely a different smoothing. But what is correct is a question. Try also enabling "analytical smoothing", maybe you'll like that better. It more realistically responds to separate spectral peaks. That's however rarely useful with musical signals.