Laser Projector (Red, Green, Purple) "Laser 3D Party Light" from Spencer's

[tribble]

Well, here's the confusing thing about "speed." There's the speed that the DAC iterates through the points defined in the frame, and there's the physical speed that the galvo achieves in response to whatever waveform the DAC generates due to, not only point timing, but point values.

The ILDA12K pattern has a smaller square with points only at the corners, a large outer square with points along the edges (spaced closer together near the corners than at the centers, I might add), and a 12-point circle that fits just OUTSIDE the inner square. And, it has a bunch of other stuff. I realize that each of these features is designed to produce different waveforms to test different parameters of the amp/scanner set so you can adjust them. Fine. Let's say I adjust things so that the ILDA test pattern is as good as I can make it. What if I then just feed the DAC high-point-density graphics that never get the scanners moving because they don't generate high frequency control signals? The images will be drawn more slowly than they could be and still be within the scanners' performance envelope.

I guess this is why there is still so much art as science in tuning. It all depends on what you're displaying and how it relates to the test patterns you used for tuning.

I would be curious to run a WAV file of the ILDA test pattern through a spectrum analyzer (like that built in to foobar2000, the media file player) to see how demanding it is at different "playback rates" and then compare that to a WAV file of other ILD files we show and compare them.

I think it would be SUPER COOL to have the laser playback software automatically adjust the point density or playback rate (Kpps) so that the highest-frequency control signal generated by the show is at the same peak as those used to tune the scanner system. THEN I think we would have the best possible display of any arbitrary frame without a lot of manual rejiggering. :-)

BTW...

I put 6800uF on my power supply rails and saw zero improvement. Either there is too much impedance between where I attached the caps and where the power is needed, or the visuals are not affected by power supply stiffness given the default tuning.

Also BTW...

I notice what I can only describe as "jitter" in the test patterns at scan speeds of 25 and higher. It's as if the DAC sometimes puts values into one bin, but sometimes puts them in to the next-lowest bin instead. The jitter is pretty much only between two positions for any given vector. It makes the "2" in the "12/30K" portion of the ILDA test frame dance from left to right, for example. Do you see that? It seems to come and go.

Later today I will post pictures of my "before adjustment" test patterns at various speeds for comparison to yours and anyone else who gets one of these units... I wonder how much inter-unit variation there is...

[james]

When it comes to laser signals one of the important factors is "rate of change". This can also be seen as the slope of the line that connects one sample to the next. It's actually the derivative of the function.

The weird thing about laser signals is that they often have dead stops in them. It is quite common for the scanner to move for several samples in a row and then stop and hang at a sample value for several samples. This kind of instantaneous motion to stop puts no real stress on the scanners, but a spectrum analyzer will see it as an infinite series of high frequency harmonics.

A whole lot of work and analysis went into the way LaserBoy optimizes a set of frames for animation in a wave.

There is also a utility to examine a wave for the largest step from one sample to another. It tells you the step and the sample number at which it occurs.

Similar information is available for every frame when this information is displayed with the / and ? keys in LaserBoy.

To be able to see some of the fine details in a LaserBoy wave, you can open it back into LaserBoy as unformatted. There is a setting in the [Tab] menu that determines the fixed number of samples that will be chunked into each frame. So you can see multiple ilda frames connected together in the resulting wave with the optimized transitions, that would normally not be visible.

There are added elements to the frame set that make smooth transitions from one frame to the next and in the process of repeating a frame to achieve a frames per second appearance in the wave. These are called the intro, bridge and coda.

In menu u you can set options f and g to be able to see the front, side (Y) and top (X) of the art stretched out over time (as Z). If you go back to main, you can press and hold the digit 4 and rotate the image to be able to see how X and Y come together to form the image.

James.

[tribble]

OK, here are some test images from my copy of the Spencer Laser 3D Party Light, out of the box.

I had some problems uploading the images, so I used the URL method. However, the pictures are huge. Sorry... I will fix it if someone can tell me how.

I changed the color of the ILDA12K test pattern to all red so that (1) the exposure would be more even and (2) so we would not be dealing with any DPSS lag. I wanted a direct-drive diode for the tests, and red won over violet. Same story for the lasermedia pattern and the quadrature pattern. I repeated the frames a bunch of times into a single .ild file so as not to be dominated by frame reload times. The projection is on to a blanket hanging from a pole, so it's not perfectly flat. A small amount of distortion is due to the blanket, but most of it is due to the scanners/tuning/speed settings. The "focus" knob is probably at about 80% of max for these shots.

Here are the ILDA12K, LaserMedia, and quadrature patterns at a Spencer setting of "20:"





The rest of the LaserMedia patterns look very similar, but the edges get progressively rounder. I'm going to omit them unless someone wants to see them. The rest of the quadrature patterns look similar too, with increasing overshoot and increasingly large missing left side.

Here is ILDA12K at Spencer setting "30:"


Here is ILDA12K at Spencer setting "40:" You can see there was one twitch that drew the circle in a completely different place. This happens more and more often as the speed setting goes up, starting about speed 25.

[tribble]

OK, after several hours twiddling pots, I can say for sure that the CENTER pot in each group of three is the servo gain for that axis. The INNERMOST pots are the Low Frequency Damping. The OUTERMOST pots I have no idea. They don't appear to do anything. They are thirty turns, and after ramping from one extreme to the other and seeing perhaps only the tiniest fraction of a difference in High Frequency Damping behavior (or maybe just imagining it), I believe they are either disconnected or shunted out of the circuit by a fixed resistor somewhere. The tiny pots near the outer corners of the board are some sort of image scaling, as are the two pots right next to the inputs that are oriented differently from the rest (those may be input scaling?).

While adjusting the pots I noticed that if I got too close to the X galvo there would be noise on the signal. The X galvo seems to be very sensitive to noise and is a little jittery. Its cable goes directly over the red LD; I don't know if perhaps it is getting some modulation interference or if there is an iffy connection in one of the pins. Maybe that galvo is just off-spec; don't know.

I was able to eliminate most of the skew in my circles and my hypocycloids, so that made me happy.

I am too tired to take another round of photos, but when I'm happy that the adjustments are as good as they can get, I'll post another set of 20-speed test images.

[james]

I think what you have there is a classic example of a Chinese copy! They have no idea about how anything works! They just know how to make things that "look" the same as the items they are copying. Good luck!

[DogP]

I had some problems uploading the images, so I used the URL method.

I've been having the same problem uploading images for the last couple days.

If you send me the exact ild/prg files used, I'd be glad to run the same tests and post pics of mine... PM sent w/ my email address. About the incorrectly drawn points... does it always hit the same wrong points, or is it random? The DAC is an 8-bit parallel DAC... I wonder if the slew rate on (one or all of) the digital lines is slow (too much capacitance somewhere, bad solder joint, etc). An incorrect bit when the /WR strobe is asserted would cause these discrete jumps (if it's the same jump every time, then it's probably just a single faulty bit).

OK, after several hours twiddling pots, I can say for sure that the CENTER pot in each group of three is the servo gain for that axis. The INNERMOST pots are the Low Frequency Damping. The OUTERMOST pots I have no idea. They don't appear to do anything. They are thirty turns, and after ramping from one extreme to the other and seeing perhaps only the tiniest fraction of a difference in High Frequency Damping behavior (or maybe just imagining it), I believe they are either disconnected or shunted out of the circuit by a fixed resistor somewhere. The tiny pots near the outer corners of the board are some sort of image scaling, as are the two pots right next to the inputs that are oriented differently from the rest (those may be input scaling?).

I think the outermost pots are the Zero Offset of the driver... is that possible? It probably wouldn't be very noticable visually, but I believe it'd shift the DC offset of the image a little bit. To adjust it properly, you'd probably need to use a scope (or maybe a multimeter would work), to set the the output to 0V with an input of 0V. That'd be hard to set visually, since there's no reference for what is "centered". And yes, I believe the ones that are oriented differently are the input scaling pots (I had tweaked them a little bit on the other projector).

Oh yeah... I got a new camera for Christmas (Sony NEX-C3), and WOW, what a difference it makes.
Couple pics in the dark, playing with the manual settings:



And auto mode from across the room with the blacklights on:


DogP

[mixedgas]

Mr Tribble, PM me with a email. I commented your LM test pattern pic to give you a idea of what I think is going on, but I can't upload images today for some reason, no matter how I resized it.

Whenever the diagonal lines grow apart, the pattern is showing a difference in x and y speeds. This is what LM is designed to do.
Since they are majorly different in two quadrants, (And really screwed in one) I have a few questions.

From what I can see...

I agree with Dog P,
Somebody tweeked the amp design(or left out parts) to give really tight damping, but it looses some velocity loop when the galvos are moving fast.
You have great corners, which is where the error would be really objectionable in a low cost projector by a non geek.
I get the feeling they have some wrong integrator cap values, or went to a nearest stock value. Ie their way off stock value to keep parts count down.
I also think they might have sliced out one of the feedback paths in the amp.


As for your comment about TWO pots changing image size, OK, you should have a pair of pots for image size. The other pair is the position sensor LED drive current or sensor opamp gain. If they copied things right, there should be two photodiodes in the base of each galvo. A small square flag should block each photodiode's view of the led equally when the shaft is centered. Messing with the photocurrent changes the sensor linearity and image size. Usually things are scaled so there is one point where the sensor produces good linearity and a nearly constant volts per degree. When you say the one axis has so much more gain then the other, well, its the sensor calibration.

IF they went cheap and went to just a resistor biasing the led, that would explain a lot. On the galvo they copied, led current and the opamp gain are both adjustable.
The prototype uses a temperature compensated opamp current source for the leds.

I get the feeling one of your posit sensor flags is not quite on center or tilted and they used the DC offset to correct it, which is not good. Nothing you can do about that. That is the upper right where the velocity is so different that the lines are way apart. The amp also loading down the PSU in the upper right, which explains the sag that does not occur in the bottom horizontal edge line, but shows up in the top one. Its trying to correct, but struggling.

The problem is not all your errors are symmetrical in all four quadrants. This makes diagnosis really. really difficult.
One of them is symmetrical in two quads and might be fixable.

One error seems to come from a undersized power supply. Can you add stiffening caps, a couple of big electrolytics?

You might have reached the "best you can do stage" other then fixing the linearity error and nasty size offset with the led current.
If you can shoot me pics of the base of your galvo, I'll bet it matches the standard pinout. If so, we can find out if the led current is changing.

Does this use a real dac chip, R2R ladder, or just filtered PWM to drive the galvos? How much of yours is the same as DOG-P's?

Do you have access to a oscilloscope?

If you want, I can send you a 6800 Amp schematic, but I can't upload it either. I'm curious how much they could chop out and still get that level of performance.


Let me know how I can send you the commented pic. I usually have no problem posting pics here, so some update has changed something.

I need pics, and yes, a open/miswired shield would cause what your seeing when you move your hand.

Steve

[tribble]

If you send me the exact ild/prg files used, I'd be glad to run the same tests and post pics of mine...

I uploaded all of the text pics, a text file describing which images are of what, and the .ild files for test images into this directory:

http://www.likesgadgets.com/489143/

I'm missing the .prg, which I already re-wrote, but I can re-create it and post that too. I bet your first guess about what it looked like is correct anyway. :-)

Oh yeah... I got a new camera for Christmas

Sweet camera, and nice pics!

[tribble]

One error seems to come from a undersized power supply. Can you add stiffening caps, a couple of big electrolytics?

I put 6800uF across the PS leads where they came in to the board; from there the power cables go directly to the amp board. I saw no difference in any of the test patterns. I could probably put a pair of 1000uF on each of the rails at the amp board power jack, but that would take a little more doing and 1000 is not all that much. I don't have a pair of high-uF high-voltage caps around. (OK, I have a pair of 70,000uF at 40V, but those pair are twice as big as the projector in the first place!)

Does this use a real dac chip, R2R ladder, or just filtered PWM to drive the galvos? How much of yours is the same as DOG-P's?

From what I can tell, it's the same version as DogP's. I will try to post pics of the driver board, galvo backs, and other bits as soon as I can.

Do you have access to a oscilloscope?

I do, but my probes are fried. I need new probes. I can probably jury rig some probes but they would be 1x and I'll have to check the max input voltage on the scope at that level. It's an old tektronix unit with a CRT and whatnot. I can probably make it work, but it will take some time to either source new probes or make some.

Thanks for the information and assistance!

[mixedgas]

Ok, next up, see if there are filter or decoupling caps at the opamps. Small ones, .1 uf or so. Usually ceramic.
If those are there, its the sensor structure.


I got your PM, I'm on the way out of the house for a while. I will get back to you with a email.

Steve