Balaecing RGB?

[Diachi]

Um, not quite. You're never going to need double the green to mix with red! Just the opposite, in fact. (And then some!)

EDIT: Just notice that you posted a correction. Missed your post by 1 minute!

Lambogini8;

It should be 1X Green to 2X Blue to 4X Red.

I also need to point out that this is a gross approximation. It should be close enough to get your started though.

For more detailed discussion on the topic, search for the term "color balance" here on PL and you'll find numerous threads that will give you a lot more information (and more accurate numbers.)

Adam

I know, I corrected myself before you posted that was a small error .

[heroic]

I've seen it called both. True, Cyan is the term usually used to describe an ink of that color

Cyan's what physicists call it, so I'll stick with that

The colour made by mixing perceptually-equal parts of 473 blue and 532 green isn't quite a pure cyan, though, because 473 blue is already quite greenish. It's almost a sky blue, but more greenish than that.

Then again, my 635 red is quite orangey, which accounts for the fact that I don't need so much of it.

For my next trick, I'm going to add a 405 violet and a 670 red to expand the colour gamut, and do colour matrixing in an FPGA... should be interesting, no? :-) I need to build a new case for that though.

[buffo]

I've been thinking about adding some 405 to my projector too. (Currently running 660 for red, so I've got the low end covered, but with 473 for blue, I'd like to be able to add some 405 to deepen it a bit.)

Longer term, I have an argon projector project that is going to have 635 and 660 red mixed in after the PCAOM... Should be very nice, if I can ever get the stupid PSU on the argon to function reliably.

Adam

[tocket]

For my next trick, I'm going to add a 405 violet and a 670 red to expand the colour gamut, and do colour matrixing in an FPGA... should be interesting, no? :-) I need to build a new case for that though.

Sounds very interesting. Have you thought about what the color transformation matrix might look like?

I think you aren't really expanding the gamut though, unless you place the RGB coordinates outside of the spectral locus (so to speak), which can probably be done mathematically. If you do this you gain access to the full gamut of the laser combination, but at the loss of some accuracy. Linear algebra isn't my best subject though, so I might be mistaken.

It does sound like an interesting substitute for a real multi-channel color system though like the QM2000. I've given it some thought, but realized that microcontrollers lack the performance required and that I don't know anything about FPGA or DSP programming.

[heroic]

Sounds very interesting. Have you thought about what the color transformation matrix might look like?

I think you aren't really expanding the gamut though, unless you place the RGB coordinates outside of the spectral locus (so to speak), which can probably be done mathematically. If you do this you gain access to the full gamut of the laser combination, but at the loss of some accuracy. Linear algebra isn't my best subject though, so I might be mistaken.

Quite exactly so. My intent was to place the R, G and B axes at the extremes of the colour pyramid, and generate the matrix metrics from the perceptual properties of the specific frequencies available. I believe this is approximately what Pangolin's palette wizard does :-)

There are codes to do this that are used in printing- for example, ColorSync from Apple et al. I plan to combine this with the active power regulation scheme, which has rather run up against a dead end in using the microcontroller- turns out that even a 20 MHz AVR isn't fast enough. I was considering using a Propeller but it seems like FPGA can do it better since it is a straight combinatorics problem once the lookup tables have been computed.

As for the loss of accuracy, it seems like nonlinearities in the laser transfer function far outweigh the loss of a couple LSB of accuracy in colour rendering, at least for the cheap lasers I'm using. No doubt things are very different in the land of pro lasering

Anyway, I'm still getting to grips with what I can do to improve rendering of colours in the laser. I'm rather dissatisfied with the rather yellowish green of 532nm, and I'm rather tired of the rather greenish blue of 473... and the orangey red of 635 is on notice. I demand purity!

[mixedgas]

As for the loss of accuracy, it seems like nonlinearities in the laser transfer function far outweigh the loss of a couple LSB of accuracy in colour rendering, at least for the cheap lasers I'm using. No doubt things are very different in the land of pro lasering

end quote

Unless your Hugo or Tarm, you just buy a big honkin laser and do the best with what you have. Few people can afford spectral accuracy.

My buddy Karl's compact Hungarian system was roughly 20K$ with roughly 750 mW and 6210s in it. Nice,balenced RGB, a heck of a warrenty, but geeze!

Steve

[colouredmirrorball]

OK: a question for anyone of you with decent software:
I am thinking of buying a 100 mW 532 nm green laser. What powers of red, blue (445 or 473 nm, whatever is cheapest, but 445 is preferred) and violet (405 nm) would I need? The choise of red is free between 660, 650 or 635, I don't know which one is the cheapest. I am asking for the cheapest option here, 'cause, well, you know... money pit etc.

[allthatwhichis]

I am not sure how easy it will be to find a "lower" powered 445nm. The only one I can think of is the 500mW from Kvant.

For red, I am a 660nm man, although the 640s will have that beam spec but I have not seen it for myself, so... 660nm would take about 400 to 500mW to match with 100mW of 532. I am using 300mW with my 100mW green and 100mW of 473 and my white is a bit "cyanish". I think 100mW more, or if I had my green first instead of my red, the white would be more balanced. I would add 200 to 300mW of 405 to that for the "purple" in yo pallatte...

[Phritzler]

Hello!
Normally you couldn´t say a exact 1:2:4 for example, becouse the problem is the blanking and temperature of the DPSS lasers. One blue 100mW has blanked 50mW, the next 70mW and the other 120mW becouse of the temperaturadjustments what aren´t the best from the most DPSS suppliers/manufractorys from china, hungary,etc...
The same with the green. Only the red (if they are Didoes) haven´t this much differences, but mostly I say a TEC for the red is a must be.
Becouse if you scan a white square for example for one hour, at beginning it is white, after 1 hour it is more cyan becouse of the red diode. The red diodes will become more hot and with that the wavelength will go darker. Clear, with free eye you didn´t see the 3nm difference, but at mixedcolors (at best at white) you see the difference! So if you have a TEC it will be more stable...
For me I had and have all wavelengths here what for showuse is interesting (405nm UV blue not really, 445nm Nichia 500mW/1W Diodes,457nm DPSS lasers, 473nm DPSS lasers,...640nm Diodes,650/660nm Diodes) and I got the best results with this powers..
You must change between low power and higher power. At low power levels the balance is normally a bit other. Also you must calculate in the breakdowns if you use DPSS lasers, if you blank them with AOM, you need other powers.
With my experiments I get best white with this Parameters

This is for "small Power" (up to ~5W balanced white)
660nm Red :
0,8:1:3 GBR with 532/473/655nm
1:1:3 GBR with 532/457/655nm
1,2:1:3 GBR with 532/445/655nm
640nm Red:
0,8:1:1 GBR with 532/473/640
1:1:1 GBR with 532/457/640
1,2:1:1 GBR with 532/445/640

At higher power the levels are a bit other, for example my new RGB I build at the moment:
5W 457nm, 3W (after changing 5W) 532nm, 6-7W ~650(648)nm (cooled down selected 652nm diodes)
That should be a nice white setup normally,but I must test is (isn´t finished yet ). You must see that I use AOMs for blank the green and the blue and didn´t blank them direct!

The 640nm opnext diodes (150mW CW) have a bit better beamspecs as the mitsubishi (mitsubishi is the manufracturer of the open can diodes) diodes. Becouse the 640nm diodes have a bit a smaller emitter as the mitsubishi diodes typically. 640nm is also really much brighter than 660nm, but the price is still a bit hight if you need more than 1W or something


Greets,
Phil