"KPPS" technical definition/question

[havok1919]

Hi All,

I'm totally new to lasers, but I'm a pretty technical sort otherwise-- I wandered into the laser side of things after working with CRT based vector display systems (ie, arcade games) for way too long...

The question I have is related to galvo "speed" ratings. What is actually the 'working' (for lack of a better word) definition for "KPPS" as associated with galvo deflection systems?

To be a bit more clear, is a "point" considered to be an endpoint of a vector (two "points" make a line)? Or, is a "point" more of a single step along a vector with multiple points comprising a line?

Simple illustration: with some arbitrary coordinate grid, if we have one line with endpoints at (0,0) and (100,0) is that two "points" or is that possibly 100 "points"? (It seems to me like the definition would have to be relative to total beam distance traveled rather than arbitrary 'endpoints'-- but if that were the case it would seem to have to be related to the output resolution of the device feeding the scanner as well, so I've over-thought it now and have myself confused.)

I didn't have any luck 'search'ing for an answer on the forum, so apologies in advance if there's a simple and obvious definition posted somewhere and I missed it.

I'm wanting to drive the ILDA input directly from a little microcontroller/DAC setup (just because I think it'd be neat ;-) and I just want to be sure that I'm understanding the slew rates involved.

Thanks in advance for any wisdom on the subject!

-Clay

[Things]

This usually depends on your software setup. Most softwares add "optimizations", which includes adding more points to a line/shape.

For example, if you had a square, and 1 point in each corner, your galvo's would be told to go to that point, but then they'd be told to go to the next one, and you'd end up with more of a circle than a square. Extra points are added at the corners and along lines to make sure they don't distort too much.

KPPS is just that, how many of those points the galvo can get to in a second. If you had an image with 20,000 points, it would scan the image once every second, which is wayy too slow for your persistence of vision to make an image from.

[drlava]

No, an ILDA kPPS rating has a direct relationship to the response speed of the scanners. Scanners are said to be 30kPPS if they can scan the ILDA test pattern (with no points added or subtracted) correctly at a DAC point rate of 30,000 points per second, typically at 8 degrees full scan angle. This scanner rating determines how complex an image you can scan without experiencing flicker or distortion.

See this excellent expansion on the ILDA test pattern and scanner kpps ratings here:
http://www.laserist.org/Laserist/tech_tips_5.html
and here:
http://www.pangolin.com/ILDAtest.html

[havok1919]

For example, if you had a square, and 1 point in each corner, your galvo's would be told to go to that point, but then they'd be told to go to the next one, and you'd end up with more of a circle than a square. Extra points are added at the corners and along lines to make sure they don't distort too much.

KPPS is just that, how many of those points the galvo can get to in a second. If you had an image with 20,000 points, it would scan the image once every second, which is wayy too slow for your persistence of vision to make an image from.

(First off, thanks for the reply!)

Is a "point" defined as a move of a certain percentage of the display angle or something? (If there isn't a distance that a "point" implies I'm having a hard time understanding how the galvo could be rated in "point" units? Are these "20KPPS" points close together or far apart or what? ;-)

If we had, say, a 'random' starfield-- just random dots scattered in the display area-- the order that the dots are visited would affect the distance the galvos had to move and the time taken to draw the frame. An optimal path (shortest distance) to all the dots could dramatically reduce that time and increase the frame rate even though the number of displayed "points" didn't change? (just the order that they were drawn changed-- presumably something like Dijkstra's Algorithm or another path finder being applied by the software to reduce the total 'distance/cost' of the positioning)

I guess I'm basically saying that conceptually you could have a frame with 1000 'points', that with an optimal path could maybe be displayed at 20fps with a 20KPPS galvo setup... But that same set of points with a less efficient path could potentially not be displayed at 20fps on the same scanners due to all the 'extra' time spent positioning the mirrors to the new location?

Or is it more of an empirical metric where it's given a standard frame and how fast it can complete it accurately defines the speed of the galvo?

Thanks for putting up with the technical-newbieness. ;-)

-Clay

[havok1919]

No, an ILDA kPPS rating has a direct relationship to the response speed of the scanners. [...]

Ahhh! Excellent. Had my reply written before seeing yours posted! I'm going to read up (and probably be back with another question or two). Thanks for the pointers!

-Clay

[havok1919]

I read up some more and have a follow-up question. :-)

Assume someone has an ILDA 30KPPS capable scanner tuned for 30K.

Let's say my hardware has a DAC update rate of 30KHz (new value output on the DAC 30,000 times/sec) and I have an 8 degree scan angle. For the sake of discussion, I want a 30Hz refresh rate. 30K/30Hz = 1000 points per frame.

Are those 1000 "points" *all* potentially worst case full deflection? (ie, could you have 1000 points alternating between opposite corners and expect the scanner to keep up?) To rephrase that another way; are the points *any* possible positions within the deflection angle?

(sorry if my terminology isn't quite right-- feel free to correct me. I'm used to 'deflection' of CRT electron beams from the center of the screen.)

Thanks for your time playing Q&A with me...

-Clay

[JohnYayas]

No. The ILDA pattern is far from a worst case scenario. It's just an industry standard tuning pattern. If your 1000 points are too random or far apart, the scanners will not be able to scan them in a pleasing manner.

In fact, you may potentially be able to get better performance out of your scanners by NOT tuning them to the ILDA pattern. The ILDA pattern can dumb down a pair of scanners to allow point pulling to work. But, if your scanners can handle it and you don't want to use point pulling, you can set the gain and feedback higher so that they actually track truer to what they are told to do. It really all depends on whether you are shooting for interoperability or you want true tracking. By not tuning to the ILDA spec, you may have a better chance of scanning those worst case 1000 points that you asked about.

[mixedgas]

...If you hook a laser display system to a vector monitor (I do this for show programming) you see a bunch of bright dots with thin, dim, connecting lines, as we rely on the galvo's slow speed to "connect the dots".


...Because of the inertia of the scanner mirrors, vector laser is a bit different from stroke writing on a CRT. We tend to use more of the "travel time" between points to create a image, instead of stacking the points 1 bit apart like on a crt. . We also have a different "writing speed" between small jumps and large jumps. On a crt, you just see a thin trail between long jumps, unless you have one of the stroke writing systems that charges a capacitor to create a linear ramp waveform between points. On a laser system, the inertia of the mirror causes a gentle, even, line on short to medium jumps. Long jumps need various techniques to make the brightness even.


...To use the mirrors inertia to our advantage, the goal is to maximally, linearly accelerate the mirror for 1/2 the travel distance and then maximally, linearly, break for the other half. In reality, there is ringing, overshoot and undershoot, so the amplifier has a PID loop tuned to a given mirror/shaft inertia and there is a position feedback sensor on the shaft.

Steve



Other techniques such as adding anchor, corner, and blanking points are used to smooth out and sharpen up a image.

Anchor point, points added during long, visible jumps to guide the scanner to the next major change in the artwork.

Corner point. To get a sharp corner, often 2-3 points are added at the junction of the angle.

Blanking point: Guide point added on long jumps when the laser is off to ensure a even brightness in a image, by insuring the speed between all jumps is fairly constant.

In laser, if you have a huge 3D image or a image that is blanked at first and "revealed", ie the "write out" effect, we would tend to scan the whole point cluster even when its blanked, to keep the image brightness even.

Newer software has a lot of optimization tricks far beyond what I mention, but its a start.

Steve

[drlava]

Are those 1000 "points" *all* potentially worst case full deflection? (ie, could you have 1000 points alternating between opposite corners and expect the scanner to keep up?) To rephrase that another way; are the points *any* possible positions within the deflection angle?

No, 'ILDA 30k' scanners have a response bandwidth of DC to somewhere between 2kHz and 3kHz. So you have to stay within those limitations, often by adding guiding points and anchor points and pulled points (as Steve says). Open the ILDA test pattern in laserboy and trace around it to see what I mean.

In fact, you may potentially be able to get better performance out of your scanners by NOT tuning them to the ILDA pattern.

Sometimes you can overtune your 30k scanners beyond the ILDA test pattern at 30k, but that just means you can now run the test pattern at, say, 35k and it will look right, so you have tuned your scanners to 35k. You haven't really 'not tuned them to the pattern', you've just tuned them to a faster speed of the pattern.

[havok1919]

Thanks for the replies guys! I think I'm much more up to speed now-- really helpful posts!

On vector monitors we have sort of the opposite problem-- we actually need to keep a uniform write speed as the beam deflects along the entire vector with very little (if any) excess dwell time at endpoints (lest the beam burn the phosphor if left in one spot too long). With the lasers and mechanical scanning systems the slew rate of the mirrors essentially *are* your integration capacitor if we were to compare it to some of the older analog-ish vector gaming systems. ;-) (Ie, pretty much all the Cinematronics and later Atari vector games you see in laser MAME.)

I get the idea for acceleration/deceleration and increased dwell time and whatnot by interpolating with multiple points along lines. Makes sense!

One more question then... If the position systems in the drivers for the mirrors are closed loop, analog controllers, would there be any reason (other than data rate from the host) not to feed them with a continuous input? (ie, having the DAC system sweep the X/Y voltages between point "A" and "B" at a constant rate less than the maximum slew rate of the mirror/amp controller?)

(I'm basically wondering if I can take my constant write-speed CRT vector display controller hardware and essentially just turn the 'draw/move' speed way down and use it with the laser. I suspect that it would always result in the scanner running a bit slower than it's potentially capable of since there wouldn't ever be any "move from A to B as fast as you can" type jumps, but I assume that the PID loop is running *much* faster than the point rate and a continuous signal wouldn't cause it issues?)

Granted, I could just *try* this, but not knowing the conventional wisdom I'm trying to be a little cautious before I abuse anything *too* much. :-)

Thanks again,
-Clay