LaserMAME

[swamidog]

on the laser projector side, things are pretty easy.

i'd get one of these RGB diode bundles:

https://sites.google.com/site/dtrlpf...gb-diode-budle

and then drive them with three of these:

http://illuminationsupply.com/laser-...l#.U0pXwa2xLw8

all your hardware has to do is output an analog voltage from 0-5v at low current and the driver will handle the rest.

i use the exact same components and 40k galvos (DT40W) in a small projector i recently built. the beam divergence and color modulation is really amazing.

Thanks! I thought I would probably be biting off more than I could chew. I did think about analog modulation, but as this is my first attempt, and didn't want to spend thousands on hardware or have to build much hardware myself, TTL seemed fine for a first try. Although I agree, proper analog modulation and a 40/60k scanner would be nice

I'll certainly try to contact Zoof, his asteroids clone looks ace! I didn't think that the persistence of vision would be too much of a problem, but then with a 20k scanner I'm not looking to draw that much complexity, just a few sprites (vectors) in a few colours should be enough to play with to get something cool working

Thanks again for the quick reply, this seems like a great community

[donavanboadboy]

I've already ordered one of these: http://www.ebay.co.uk/itm/2613951980...84.m1497.l2649 kits, and one of these: http://www.ebay.co.uk/itm/1210310519...84.m1497.l2649 and a couple of high speed DACs with differential outputs to interface the galvo scanner driver boards with the FPGA.

I really should have asked you guys on this forum first, but hey ho, itchy left mouse button, I couldn't resist as they seemed so cheap and looked relatively well built, and I'm not too interested in the optics as such, an RGB laser scanner seemed like something cool to try to control with an FPGA project, which is my primary motivation at the moment.


If these cheap Chineese jobs don't do the trick, I'll certainly look at alternative laser diode/galvo assemblies.

[donavanboadboy]

I've read mixedgas's explanation on this page: http://www.photonlexicon.com/forums/...p/t-17918.html and I think I've totally misunderstood the performance of Galvo scanning speeds. I had (naively) assumed that "kpps" meant "kilo points per second", where a "point" was *anywhere* on the scanning plane, LOL, how wrong was I! From what mixedgas says, it seems a 20 to 35k scanner would only be able to do about 2000 complete sweeps per second, which is 100x slower than I was assuming!

So, no wonder persistence of vision is a problem when trying to render a vector based game using a single X-Y galvo scanner:

a) Either the graphics (number of edges) would have to be kept very simple, or the a display algorithm would have to be very selective as to which edges to draw (not draw all edges of all sprites every frame), and even after it selects which lines to draw, it's an NP-complete (traveling salesman) problem to sort the optimal draw order of the lines it selects to be drawn, and of cause that has to happen at least 10 to 15 times per second or more.

b) Even when the display works out which lines to draw and in which order, to get a consistent brightness for all lines, all lines need to be scanned at a similar angular velocity, so non-linearities of the galvo acceleration curves takes into account etc.., nightmare, so equalising brightness probably have to be approximated by setting a limit on the maximum and minimum scan line length, i.e. split long lines up into shorter ones, maybe even try have a maximum turn angle between each line segment (feeds back into the traveling salesman problem making it even more constrained and complex to solve).

So, I think I really have bitten off quite a complex project here, I think I'll leave the speculation for now and wait until I have the hardware in front of me, than I can see and test for it's fundamental limits without faffing with equations

[buffo]

Actually, the limit is a lot lower than 2000 "complete sweeps". If you're talking large step movement (from one side of the scan field to the other), the limit is closer to 350.

The small signal bandwidth limit (small signal being defined as ~ 3 degrees of movement with 3 dB distortion) for a 30K scan set is 2500 Hz. But remember, that's for movements of 3 degrees or less, and that's with considerable distortion.

Adam

[donavanboadboy]

Ok, so I should be able to issue movement commands to the 20k galvos at about 2000Hz, providing I keep the max deflection angle to less than about 2 or 3 degrees per move.

I guess most scanning patterns are pre-computed and just run from a script, I wonder if there is any research or information available on real time "vectorisation" (for want of a better word) of say a set of splines into an ILDA command sequence? I guess I was right an that it's very hard to do, because you have an NP-complete "traveling salesman" problem to solve every frame, i.e. you would have to work out an optimal path to scan the vectors you generate from the source (a dynamic set of splines or something) every frame.

[buffo]

Ok, so I should be able to issue movement commands to the 20k galvos at about 2000Hz, providing I keep the max deflection angle to less than about 2 or 3 degrees per move.

No. For 30K galvos you would be OK at 2KHz. If you plan to use 20K galvos, the hard limit is 1.66 KHz, and to be prudent you should probably stay below 1.2 KHz.

I guess most scanning patterns are pre-computed and just run from a script

Depends on the software and the application.

I guess I was right an that it's very hard to do, because you have an NP-complete "traveling salesman" problem to solve every frame.

No, actually many of the more advanced laser show software packages do exactly this. Yes, it's an NP-complete problem, so chances are that you'll never find the optimal solution in real time (not enough processing power), but you can certainly get close, and that's better than no optimization at all.

Adam

[donavanboadboy]

Thanks buffo! You have been very helpful, I guess if I get this working I'll open source it or something, is there a gap in the market for a "random set of splines to ILDA" algorithm or has it been done to death and stuff already in the public domain?

[buffo]

There isn't much in the public domain. Most of the work in this area has been commercialized. One of the most affordable ones is something that Andrew Kibbler (DrLava here on PL) released a while back. It's a program called ILDA-SOS that will do frame-by-frame optimization. Not sure how much it costs by itself, but I know it's less than $100, and could be a lot less. (I think it might even be included for free if you buy his LSX show software).

The only free option I can think of is Laserboy, and I'm not sure how well it optimizes frames. (It's free, but the interface is keyboard-only, and is a bit difficult to learn.)

Adam

[donavanboadboy]

James (LaserBoy) has PMd me the public domain source code so even though I hope to work primarily with an FPGA, if I can contribute to the LaserBoy public code base I will do.

Great community, thanks everyone, Dan.