Electronics > Projects, Designs, and Technical Stuff
analog video effect
vk6zgo:
--- Quote from: Alex Eisenhut on November 20, 2019, 11:49:41 pm ---
--- Quote from: mikerj on November 20, 2019, 05:36:10 pm ---You could do this pretty easily with a bunch of comparators.
--- End quote ---
That's the idea I suppose, create a bunch of reference voltages for the bank of comparators with a bit of hysteresis, and then re-create the video signal. The usual analog video problems need to be solved, clamping to the black level to leave the syncs untouched and delaying the chroma so it still lines up with the processed luminance.
--- End quote ---
Hardly major problems-- it was done all the time, back in the day!
--- Quote ---
I guess, it's just idle speculation right now.
--- End quote ---
SiliconWizard:
--- Quote from: macboy on November 20, 2019, 06:03:54 pm ---
--- Quote from: SiliconWizard on November 20, 2019, 05:49:59 pm ---Either way, wouldn't you need some kind of memory to deal with 2D blocks? Averaging luminance over a certain period on a given scan line (thus over a certain width) is easy, but averaging it over a certain square area? Can you do it without memory?
--- End quote ---
You misunderstand, there is no averaging involved. It is just reducing levels of quantization to some small number of levels. It's something like having a 1-digit voltmeter that can only display "0" through "9" instead of "0.000" through "9.999".
--- End quote ---
Well, I was actually thinking about a bit fancier "posterization" effect (I happen to have worked on one a long time ago). Just quantizing the signal "pixel" by "pixel" doesn't look very good. By "posterization", I was understanding any effect that would quantize luminance (or colors, depends on the effect you're after) to a limited number of levels, which is exactly what you said above. But there are various ways of doing it, some that will look much "nicer", and involving averaging over a certain area around every pixel (or otherwise some other kind of filtering), thus my question. You can discretize luminance with a much better looking effect.
Now if we are talking about basic quantization pixel by pixel without anything fancier, obviously I agree, you don't need any memory or any kind of filtering.
And for this, yes, once you have a sync detection circuit, that can be done with a relatively simple analog front-end and a flash ADC (+DAC, the DAC can be just a few resistors though, not that accuracy is probably a concern here.)
Obviously discrete comparators can also be used; their number will quickly add up though unless you're after a very small number of levels (admittedly, 16 levels for instance, would "only" require 4 ICs using quad comparators). Flash ADCs are actually made of fast comparators (plus some logic). https://en.wikipedia.org/wiki/Flash_ADC
I'm trying to think of a purely analog way of doing this. Sure you can argue that using comparators would be purely analog. I wonder if we can find a way of getting a similar "posterization" effect without comparators, with a reduced part count, using some clever trick.
Alex Eisenhut:
"using some clever trick."
Attenuate video signal, pass through a stack of Josephson junctions, amplify?
jhpadjustable:
--- Quote from: SiliconWizard on November 21, 2019, 08:38:46 pm ---I'm trying to think of a purely analog way of doing this. Sure you can argue that using comparators would be purely analog. I wonder if we can find a way of getting a similar "posterization" effect without comparators, with a reduced part count, using some clever trick.
--- End quote ---
Is it cheating if I use a diff pair as a comparator? I offer this unoptimized brute-force design, just a set of voltage buffers and a flash ADC and DAC without the encoding/decoding nonsense. It's cost-reduced if not part-reduced. Three resistors and one CA3046-type monolithic transistor ensemble make one stage. n stages provide n+1 levels of brightness. The capacitor represents a cheap level shifter. Color and sync sold separately. :D
If you had a fast sawtooth generator, one fast comparator (or equivalent), one DFF, and one fast edge-triggered sample/hold, you could clock the DFF n times faster than the sawtooth to get n (n+1?) levels of posterization.
Navigation
[0] Message Index
[*] Previous page
Go to full version