Analogue Protection Filtering for ILDA XY/Galvos

[mixedgas]

Its something else.
~
A point is one position of the laser beam in a frame. Unlike a video frame, a vector frame can have 1 point or 1000, or 10,000, and has a variable repeat rate. Until now, good galvos would support 700-1200 points at their very best, and something just came on the market that hits the physical limits of what a galvo can do.
~

You can see as much of the standards as possible without joining ILDA here:
~
https://www.photonlexicon.com/forums...ILDA-STANDARDS

You should see three PDFs in that link... One has the glossary...
~
A point in its simplest form is one XY position output by the digital to analog converter in a graphics system. Old scanners that the industry was founded on would support 12,000 small signal, small angle, position updates in one second... Hence 12Kpps.. Then an amplifier was introduced that got that to 24K small angle. Then another company came up with the standardized galvo design that got to 30K, by eliminating a torsion wire in the galvo. Again, has little to do with bandwidth and everything with small step response.
~
The standard was composed by a committee of artists and engineers interested in getting rid of "bandwidth, and angle, and company dependent tuning" long ago, where everybody had their own means of tuning the analog PID loop on a galvo. That was bad for business as no one could interchange artwork. They did not want to have test gear in the field, and they all could set the rate that vectors or points were output to a galvo, so PPS is the natural outcome of something that was a brilliant compromise, but needed no specialized test equipment, other then the standard test pattern, a means of measuring the projection angle, and a human eye.
~
The ILDA Tuning scheme frustrates every outsider that ever comes upon it, but works brilliantly once they actually tune the PID loop on their own amps. As long as the software/hardware maker has kept to a few basic concepts. Brilliant engineers wrote a procedure and published the standard "frame".
~
This is something a friend keeps on line for me as a reference. Drawing 2073 is typical of what the Chinese cloned, it is missing notch filters and other components that are in a modern galvo scanner amp..... It shows what was used by one company before the standard pattern, and what came when everybody got on the same page...
The feedback signal is from two or four photodiodes that look at a mask on the shaft... However the old G120s (12K) used a capacitive feedback sensor..
~
http://www.skywise711.com/lasers/scanner/scanner.html
~
http://www.laserist.org/StandardsDoc...n95_rev002.pdf
~

Like I say, if you want to understand this industry from a engineering perspective, you need the hardware... The scanner dies or is damaged when its drive coil, which is a few dozen turns of #30 or so wire, expands to touch the moving rotor, from Ohmic heating... The other possibility is exceeding the de-magnetizing current... Hence the coil temperature calculator.
The other choice is a cheap poly-fuse resettable device, but those don't always protect in time...
~
The analog control loop around the scanner handles position feedback, overshoot, undershoot, and damping of mechanical resonances. There are digital loops, but few entertainment companies can afford these... Using a PID or PII loop that is hand tuned by the factory or user...
~
Your scope graph below would be fine for scan angles of about 5-10 degrees at that frequency...

~
Points disappear when you switch to analog waveforms, which are infinite collections of points, if your abstract is a sine wave.
~
Now the graduate students I build equipment for, would like their laser working, so I need to go for a while.

Steve

[creatorlars]

Thank you for all of this.

Your scope graph below would be fine for scan angles of about 5-10 degrees at that frequency...

Walk me through the rationale and variables at play, here? Is there any relation to KPPS, or should I get myself out of that headspace entirely? If I increase the time constant of the slew limiter, does this increase the possible scan angle?

Just to be clear: My design goal is to come up with something that's universally safe for relatively low frequency abstract generation -- designing it to pass the highest possible frequency waveforms is a secondary concern. It just needs to be good enough to have some fun generating circular/square abstracts with a modular synth, and safe for laser galvos in a universal way. Users of my interface will have access to a whole world of voltage sources in their modular systems, and there is an unlimited number of beautiful abstracts to be created with triangle and sine waves up to ~500Hz. This slew limiter circuit I posted above essentially performs triangle waveshaping of sources with rising or falling edges equal to or less than the period of the time constant, and then attenuation of the scan angle beyond the cutoff threshold. So if I lose some fine detail due to slewing, it's okay.

As far as zener clipping goes, I suppose it makes the most sense to put the clipping before the slew limiter, so that any clipping harshness that is produced ends up being slewed along with everything else.

[creatorlars]

12,000 small signal, small angle, position updates in one second... Hence 12Kpps

I've heard the word "small signal / small angle" used multiple times. I assume that this would mean a fraction of the total scan angle possible. Is there a formal definition on what constitutes a small angle update? For example, say there is a scanner rated at 20KPPS and has a total scan angle of 30 degrees. Is there a "maximum difference" that the XY angle can change between one point and the other? A period of transfer for 20KPPS would be 50 uS. So if I know how this relates to small signal angle, I could calculate what the slew rate should be for a "safe number," even if it's not the most optimal one.

[james]

In LaserBoy there are settings for maximum lit vector and maximum blank vector (in points).

Points are relative to the whole space defined by signed 16 bit integers, (-32767 to +32767) in 3 axis 3D space.

So when you optimize a frame for display on the scanners, LB adds points along straight lines that are equidistant and no more than these settings. This limits the velocity of rotation of the scanners; how far the scanners can rotate per sample.

It's kinda of a weird thing if you think about it. The optimization technique is done in the graphical domain, but the end result is a wave file so it ends up being digital audio signal processing.

[creatorlars]

Thanks James, the practical example is really helping me understand this a bit better, because I can visualize what's happening. Can you help me understand how the numbers you're talking about end up applying to the output itself?

In LaserBoy there are settings for maximum lit vector and maximum blank vector (in points).

When it comes to the display, can these parameters then be translated to "maximum scan angle change per point"? How would you determine a good value to use?

So, say I want to sweep the maximum scan angle of a 20KPPS scanner (ILDA -10V to +10V) along a linear path... how how many "points" (50 microsecond increments = 20KPPS) should it take to travel from one end to the other of the maximum sweep? To put it another way, how many points would be output if I wanted to draw a line between -32767 and +32767 in LaserBoy?

(How am I doing here?)

[james]

You can set the values to whatever looks good.

You can tell if you are over driving your scanners. They make scary sounds and the image they draw becomes distorted.

It really comes down to trial and error.

Also you need to know that there is a sensitivity setting on your scanner amp; sort of like the volume control on your stereo. This controls how far your scanners will swing when driven with that -10 to +10 volts.

Plus, if you use a LaserBoy DAC (a modified sound card with a correction amp) you can set the gains on the X and Y channels on the correction amp.

In your case you could control the peak to peak amplitude of your X and Y signals. This has everything to do with ultimate velocity.

Another important factor that has not yet been discussed here is angle dwell. This is points added in the corners of images with sharp angles. These added points give the scanners a chance to get to the corner point, sit there and settle for a bit before ripping off in another direction for good looking corners.

Before I wrote LaserBoy, I messed around with a set of scanners, a single green laser and an ordinary sound card. (The laser and scanners belonged to Steve as a matter of fact). I used an app called GoldWAVE that has an expression evaluator. So I could write functions for X and Y using sin and cos and whatever (t for time in seconds, n for sample number) to "draw" mathematical abstract images for the scanners. Kinda' like Spirograph. I think this is what you want to do with analog function generators. I just had to make sure that the fundamental frequencies of the functions were not too high as to not over drive the scanners. Again; trial and error. A hand on the master volume of the sound card was a must!

That was way back in 2003. Seems like a lifetime ago!

[creatorlars]

You can tell if you are over driving your scanners. They make scary sounds and the image they draw becomes distorted. It really comes down to trial and error.

In your case you could control the peak to peak amplitude of your X and Y signals. This has everything to do with ultimate velocity.

That makes sense. Currently my interface contains input attenuators for X & Y, and offset controls. Maybe it's good if I bring out a master zoom/scale control and a control for the slew limiting time constant to the frontpanel so they can be adjusted manually and on the fly -- and include a disclaimer with instructions about this tuning process in the documentation. I'm just concerned that I'm going to ship a product that ends up destroying scanners, but maybe I'm somewhat ignorant of what the users of laser scanners' expectations are in terms of protection.

Another important factor that has not yet been discussed here is angle dwell. This is points added in the corners of images with sharp angles. These added points give the scanners a chance to get to the corner point, sit there and settle for a bit before ripping off in another direction for good looking corners.

Ahh, that makes sense. I guess you could achieve that kind of effect in an analogue interface if I put Zener clipping before the master zoom/XY gain control.

Here's a photo of my current prototype, being driven by some Roland and Malekko oscillators. The Malekko Richter Oscillator II has triangle outputs and phase modulation of the secondary output, so you can generate a sin/cos function. If any of you guys helping me out here are interested in any of this stuff, I'll hook you up however I can.

[james]

DZ on this forum designed and built a really nice analog abstract console for laser display.

He knows all about the analog side of this stuff.

I do everything in LB in the digital domain and not in real time. So it's a totally different thing.

In LB it's design the art, optimize and render the wave, play the wave.

So I can add extra points and do all kinds of stuff with no worries about time.

An analog console works in real time.

You can't really add extra points. There are no points! And you can't fiddle around because the function generators are going to just keep on doing their thing. (You can't make them wait).

I'm not sure if that makes any sense, but that's how it is.

[mixedgas]

Abstract consoles existed LONG before KPPS... It dawned on me last night how to explain this...

Forget KPPS!

!
!
!
Ok, When driven by graphics software, a Galvo operates in a ballistic mode. It is forced to its maximum performance by a carefully designed waveform designed to push the galvo to its limits. That performance is measured in industry using a standard waveform, the repetition rate of which is measured in KPPS.. In audio terms, the galvo-amplifier combination operates as its own reconstructive filter for the digital waveforms, which is why a 30 KPPS waveform really only has actual deflection components to 2.4 KHz or so, plus harmonics at the edges.. The test pattern pushes the Galvo to its performance limits by design. All graphics software does this, but classical abstract machines DO NOT. They have NO CPU processor to do it!!
~
~
KPPS divided by number of points in the image is the FUNDAMENTAL rate that the waveform repeats, in Graphics...
~
~
Forget all that, and read this:
~
When driven by a abstract generator outputting sine or triangle waves at up to 85% of its first resonant frequency, a Galvo runs in LINEAR mode and just tries to "keep up" with the waveform. There are no KPPS in that mode... There is no need to add corner points, anchor points, stretching points, or to linearize the velocity... Its just following the waveform like an oscilloscope trace would.
~
Keep the Galvo system filtered below 85% of its first mechanical resonance and your fine...
~
First resonance measured last night on a poorly tuned "30K" 6800 scanner with a 6850 amplifier with no performance enhancing notch filters installed on the amplifier, and standard mirror set was ~1600 Hz... 85% of that was 1350 Hz, or .707 times that is 1130 Hz... So have a switch selectable low pass at 500 Hz and 1000 Hz and you'll hit your target... Make sure it has low phase shift in SPICE simulation. Filter both axis exactly the same...
~

Shaft resonances that the scanner amplifier can't damp are what will overheat the Galvo fast, or cause its amplifier to oscillate...., and the edge of square wave from a music synth will contain all those frequencies... Driven hard at 11 KHz for example, the amp entered "protective shutdown" in a fraction of a second... Most modern scanner amps have notch filters in the feedback loop at the first and second mechanical resonances... So they can do the "digital" graphics..
~
~
Note: Measuring the mechanical resonance requires detuning the scanner amp feedback loop, as documented in the factory manual. It is not something an average user can do..
~
I can show you in published literature where the scanner has ballistic and linear modes, it will take a few days to dig up a paper by Brosens... One of the early galvo designers.... One of the goals of a good scanner amp is to not exhibit a noticeable change between either mode...
~
No need for a variable filter, that will just allow users to pass resonances. Sad thing about scanner resonances...At low amplitudes they add a cool looking "effect" to the abstract and users tend to dwell there, because the scanner uses less energy at first...
~
~
Steve

[creatorlars]

Abstract consoles existed LONG before KPPS... It dawned on me last night how to explain this...
Keep the Galvo system filtered below 85% of its first mechanical resonance and your fine...
~
First resonance measured last night on a poorly tuned "30K" 6800 scanner with a 6850 amplifier with no performance enhancing notch filters installed on the amplifier, and standard mirror set was ~1600 Hz... 85% of that was 1350 Hz, or .707 times that is 1130 Hz... So have a switch selectable low pass at 500 Hz and 1000 Hz and you'll hit your target... Make sure it has low phase shift in SPICE simulation. Filter both axis exactly the same...

Steve, you are the man! Thank you so much for putting your brainpower into this. I'm working on some filter topologies and I'll post the response curves for feedback.

So, to be clear, should the signal be entirely attenuated at 500Hz & 1000Hz or is this simply the maximum frequency for full swing at maximum scan angle? I'm guessing we're talking about the former, in which case I'm guessing in addition to lowest phase shift, I should design for steepest possible cutoff (maybe an 8 or 10 pole LPF) to minimize attenuation of the overall image as it reaches those mechanical resonance points. Sound about right?

Also, I'm assuming 500Hz & 1000Hz settings would be the assumed safe numbers for 20KPPS & 30KPPS, respectively, correct?