Spectrogram (not spectroscope). is it possible with SM?

[datsound]

hello, would it be feasable to make a spectrograph in SM with bitmaps? i would want to write 2000-3000 pixels per second with the data coming from audio lines.

i want to make a bit- precise spectrograph with about 800x800 pixels for every 2-3 seconds of sound, it doesnt matter if i have to replay the sample 20-30 times to get all the data for the graph, but it would be good if i can write a bit for every 20 samples.

it is something i did in reaktor with audio tables but i would prefer to do it in SM.

[trogluddite]

replay the sample 20-30 times to get all the data

I take it from that, that you are not concerned with it being a real time scrolling type display - more of an 'off-line' analysis of a static sample?
The reason I ask, is that an 'off line' display will be much easier to create - scrolling an 800*800 area of screen, and plotting a big chunk of pixels at screen refresh rates, would be very very CPU intensive in SM!
I'd be happy to help you with the GUI side of your project (I have made scopes and a goniometer in the past) - but my knowledge of using FFT etc. for frequency analysis is non-existent!

[MichaelBenjamin]

heres a start, i found this some while ago, hands up who made it.
i havent done anything with it yet, but i remember it worked.

[datsound]

replay the sample 20-30 times to get all the data

Thankyou that is really nice of you to suggest, for the moment i am making a synth, however the temptation to make a HIFI sonogram with near bit precision is so high i couldnt help to ask.

The sonogram that micheal benjamin has posted seems very detailed, more even than ALL the free online ones which are lo-fi.

here is a pic of the one i did in reaktor, it is a bird song.

The advantage is you can zoom in to portions of audio 10ms long and 1khz in span and it would zoom with amazing bit precise detail so you could see audio waves the size of a pin dropping in any audio file. that is kindof what i thought could be fun in SM, save it could take less memory if the bit depth was just 6 bits rather than 32 bits(on reaktor 32 bit spectrogram means each point has 2^32 audio levels... OTT) !!!


the artifacts visible on my picture is because it is going through filters and i put the volume to high so there is major blurring of peaks across time and frequency.

[MichaelBenjamin]

very nice picture

[datsound]

Thanks, you shouldnt say that i will have to upload another audio pron picture.

i have spent hours drooling over high res scans of sines, percussion and various sounds but this isnt a very pretty sonogram as it is just a fuzzy recording of a bird, but for sonogram detailing i think it is as high res you can get in the world, and you can see all the individual sine peaks at each frequency, and i can even do quad resolution graphs higher than this.

sorry it is so green! yellowy was better. i am feeling really stupid for uploading all these green pictures, but it is not something that is known widely in the audio world, sonograms which show individual sine waves.

[datsound]

sorry here is a better one at double graph precision, 1khz x 50ms.

[MichaelBenjamin]

nice, can you disable the white bars?
makes me wonder if the sm spectogram can do similar resolution without choking too much..

[datsound]

yes i can disable the white bars, its theoretically possible to make a 3d map of a spectrogram, i wanted to make a spaceship version where you can fly around... i did a basic 3d version that turns the peaks into a z dimension.

a really amazing thing is you can take 2 sonograms, say a human voice and a bird call, and crossfade the graphs of the sines to morph the human voice and bird call together, i cant remember what the technical name is for that dsp trick but i heard one online.

it's for really buff programmers, with FFT it is less precise so easier to trace a graph of a sound but with the pics above it is really difficult to do because they even have each sine with phase, and a series of sines can also be more of a triangle wave and so on. would be very hi fi but challenging to do.

[dutchound]

http://www.wimp.com/audioediting/

spectroscope futuristic editor... i did think of this but it takes quite some wallop to impliment!

The best is if he makes a morph function on 2 spectroscope shapes.