Dust Analyzer (still "beta") [was: quickie signal analyzer proto]

[mystran]

Hi Mystran,

Any tips on how you get such a steady frequency display like in your analyzer?
In my own project the display is very jumpy even with the use of window overlapping and taking the maximum value of several windows.

thanks,
rola

I do absolutely NOTHING special: you simply get last 8192 samples at the time of redraw (which is controlled by WM_TIMER so not terribly exact, hence "peak hold" is currently unreliable "best effort" only.. I might add some "maximum interval" guarantee at some point), and log of squared complex magnitude for each bin. Since 8192 samples (at 44.1kHz) is roughly 200ms, and 50fps display would be roughly 20ms, there is about 90% overlap in the best case (assuming short enough audio buffers, as it updates at-most-once per process call).


Anyway.. Intel is driving me nuts. Why does Intel do such a crappy job with their integrated graphics? I'm currently seriously considering GPU render path for NV/AMD, with software fallback just for the crap that is Intel's "HD" graphics..

edit: basically any attempt to move more work to the GPU comes down to "yeah I could do that except for Intel"

[rola]

I do absolutely NOTHING special: you simply get last 8192 samples at the time of redraw (which is controlled by WM_TIMER so not terribly exact, hence "peak hold" is currently unreliable "best effort" only.. I might add some "maximum interval" guarantee at some point), and log of squared complex magnitude for each bin. Since 8192 samples (at 44.1kHz) is roughly 200ms, and 50fps display would be roughly 20ms, there is about 90% overlap in the best case (assuming short enough audio buffers, as it updates at-most-once per process call).

That's it. I was only squaring the real part. Now it works it really well.
Thanks!

[V'ger]

Great stoff, thank you very much

[antto]

yeah, i have trouble with this stupid Intel "HD" videocard too, grr
here's something interesting about their drivers:

[mystran]

yeah, i have trouble with this stupid Intel "HD" videocard too, grr
here's something interesting about their drivers:

Oh but isn't that GMA rather than "HD" graphics? GMA is even worse.

[antto]

ah, yes, it's Graphics Media Accelerator
i confused it with my on-board realtek "HD" audio card (oh boy)

[mystran]

Update 0.5.0:

• reworked tons of stuff for better performance

• now runs in a separate thread to hopefully reduce impact on other stuff

• probably plenty of new bugs (there's a ton of new code, which is why I'm updating this before other plugins)

• sadly no new features

--

oh and the GPU render path didn't make it.. instead I ended up hacking a special purposed CPU solution..

.. the threading stuff is known to be ... experimental..

edit: and quickfix 0.5.1 to work around a potential race condition (happened occasionally ilbridge64.. maybe elsewhere too)

[brok landers]

hi mystran,
first off, thanks for all your freeware, which imo is incredible quality, i always wanted to tell you that. really high quality stuff, honestly. and i am usually vey picky...

however, have to ask:
could you maybe add an alternative bargraph view with 128 bands (1/12 octave) into your analyser, overlaid with the fft curve? like thisone, (thisone will never go 64bit and also has "only" 64bands/1/6 octave):



also, could you maybe add an option to specify the amplitude range? when mastering a pop song, mostly -40db to -6db range is way enough, especially when wheighting is 6db (pink noise flat response), whereas if i have to analyse potential aliasing in a synth, -140db to 0db has to be used...
that would be it for me, as i'm struggling with _all_ vst analysers up to date, exept the mentioned one, which is the seven phases analyser, which sadly seems to be not supported anymore in any way...

just ignore my requests if you aren't interrested, no harm meant, no harm taken... but i had to ask...

again, thanks for all your great stuff!

edit:
dustanalyser sadly doesn't show up in studio one pro 2.5.2 and energy xt, both 1x and 2x...
i'm on win xp pro, sp3...

[eidenk]

Haha, that new GetThreadId function requirement has broken it for me. No big deal as the previous version runs fine and gives me no grief whatsoever.

[mystran]

Haha, that new GetThreadId function requirement has broken it for me. No big deal as the previous version runs fine and gives me no grief whatsoever.

Huh? Do you get some error and if so, what exactly?

[mystran]

hi mystran,
first off, thanks for all your freeware, which imo is incredible quality, i always wanted to tell you that. really high quality stuff, honestly. and i am usually vey picky...

Thanks.

however, have to ask:
could you maybe add an alternative bargraph view with 128 bands (1/12 octave) into your analyser, overlaid with the fft curve? like thisone, (thisone will never go 64bit and also has "only" 64bands/1/6 octave):

What should said bargraph show? Should it run 128 parallel band-passes or some derived version of the FFT information?

also, could you maybe add an option to specify the amplitude range? when mastering a pop song, mostly -40db to -6db range is way enough, especially when wheighting is 6db (pink noise flat response), whereas if i have to analyse potential aliasing in a synth, -140db to 0db has to be used...

Yeah. I guess something like that is on the list of things to do. With regards to weighting, you get flat response from pink noise with 3dB/Oct actually, though empirically a number of commercial tracks (especially dance music) seem to roughly fit 3dB/Oct at lower spectrum (below 1kHz or so) while the fit tends to be closer to 6dB/Oct at the higher frequencies.

edit:
dustanalyser sadly doesn't show up in studio one pro 2.5.2 and energy xt, both 1x and 2x...
i'm on win xp pro, sp3...

Will investigate.

[mystran]

Haha, that new GetThreadId function requirement has broken it for me. No big deal as the previous version runs fine and gives me no grief whatsoever.

Huh? Do you get some error and if so, what exactly?

Hey, thanks for reporting this!

It turns out I've indeed used GetThreadId (sort of half-accidentally) and it turns out that infact that's a Vista+ API (so breaks w2k and xp too, which I generally consider a bit higher priority).

Sorry about that, shouldn't be hard to fix.

edit: seems MSDN is quickly getting next to useless, as Microsoft tries as their best to prevent you from easily finding what actually works; they don't even bother listing supported systems beyond XP anymore.. every ancient Win3.1 API specifies "minimum XP" which isn't terribly helpful.. so I guess it's time to forget MSDN ...

edit2: turns out this is sillier than I though.. the only place I used the problematic function is a method that sets some debugging information (the thread name) and since it's a member of my thread wrapper class.. I'd already had the required "thread id" as a member variable anyway (or I could even use -1 for "caller thread").. so.. well.. yeah.. will update the download later today

[mystran]

Update 0.5.2:
- fix the "GetThreadId" thing on pre-Vista
- add "dark background" color mode to spectrum view

I check that it loads in eXT.. but for whatever reasons GUI performance is horrid when moving mouse over the plugin window.. I have no idea what would cause this, will have to investigate..

[mystran]

Ok, so the GUI lag issue has to do with mouse cursor, and how Windows automatically resets it back to an arrow.. which apparently causes something crazy to happen in eXT. Found a way to fix it (stop the automatic resets) but need to update my toolkit to restore correct cursors automatically.

[mystran]

0.5.3:
- fix the GUI lag issue with eXT


brok, can you try if this version works correctly on your XP system?