Have my cheap galvos hit their limit, or mine?

[rooksoup]

Hi all, long time listener, first time poster.

I'm looking to better understand what I'm seeing in the behaviour of some cheap Chinese clone galvos + driver boards, and whether it means what I think it does...
Long story short, I've built a setup to scan Lissajous figures onto a CMOS camera at various voltages/frequencies using an arbitrary waveform generator, being somewhat careful not to fry the hell out of everything. The driver has three-pin +10/0V/-10V signal input for each axis, to which I've shoved my bipolar signal input onto the +ve pin and grounded the negative, because if it's stupid, and it works, etc. etc.. Normally I drive with +-2V, but I also made a very simple op-amp circuit to boost that to +-10V, with near-identical results except an obvious boost in optical deflection.
I've done some simple raster scans where I just send a stepped DC voltage to each axis to sanity-check for voltage deflection, but I expect higher speed is an entirely different beast.
Anyway, from the Lissajous figures I can determine whether the shape I got out was the shape I intended to draw (in my case, shoving a sine wave onto X and a cosine onto Y for a nice circle). What I get, above a certain frequency (~800Hz), is a sort of circle with squared corners. A squircle?
This is consistent between individual galvos and driver boards when swapped out. I know the six pots are set per driver/galvo pair, but I wanted to be sure.# I also see what I think are resonance effects at certain frequencies, so have been trying not to stick around there where possible. The deflection range drops off with frequency in a way which makes sense to me for a second-order system.

Each galvo itself has six pins, and the classic IR LED/two photodiodes-and-a-flag-on-the-back closed-loop feedback system. I infer the first two pins are the positive and negative signals for the motor coil, and the last four are some combo of LED power, current from each photodiode, cathodes etc.
Getting an oscilloscope trace on the positive motor signal while it's running, I notice something that might explain what I'm seeing: while the frequency of the driver output is dead-on what I put in, the waveform amplitude looks funky and a bit triangular, and past the 800Hz mark it starts clipping to resemble a square wave. Interestingly the X/Y sine/cosine signals also appear to not be quite in opposing phase. I've got nothing on that one, as they're more circular than I'd expect to see if that were true.
Questions which I hope can be answered by anyone with better electrical engineering knowledge than me (which is most people):

1. Am I understanding what is happening even slightly correctly?
2. Is the waveform clipping a deliberate feature of the driver board, is is some sort of fundamental rate limit (slew limit?), is it simply overshooting and correcting via the closed-loop, or is it just a, ehh, inertia will carry the mirror forward, good enough for laser shows, situation that would explain why I still get a continuous repeatable shape, even at much higher frequencies?
3. If this is an actual frequency/voltage output limit, would I be right to think it's not possible to tune your way out of this via the six driver pots? After reading about enough tuning disasters here, I don't want to touch the tuning unless it's the best option...

4. Any chance this is me reaping what I've sown for driving on only the +ve pin and somehow overloading the driver with less than it would usually need to do so? I gathered that I'd improve noise rejection by driving on the -ve pin too, but I will take any excuse not to make a terrible circuit.

5. Not entirely sure I understand the actual driver voltage-out amplitude relative to voltage-in. I naively assumed to start with it would be a simple up to +-10V driving signal in, movement out, modulated by whatever closed-loop feedback there is. I assume there's much more here?

Would appreciate any advice/heckling/etc. you can provide, and would like to give thanks for the many threads here that helped me get this far. My expertise lies in putting some pieces of glass in an order and then writing some Python about it.

[rooksoup]

Update on the "I've got nothing" phase issue: mea culpa, I just misunderstood what I was seeing, it is as expected mathematically, I think. I have no idea how to fix the text formatting. Sorry.

[mixedgas]

Galvos have a performance curve of angle vs frequency. For very small angles they can move very fast, but as you widen the jump, they start to act more or less as a variable frequency low pass filter. They also have shaft resonances which can wreak havoc on scanner amps that do not have notch filters in the feedback loop.

A simple test is to make two marks on the far wall for about 2.5 degrees optical angle, and sweep slowly with a sine wave from a signal generator. You'll see a rapid fall off in response at some point. Don't push the test by increasing the angle. Now do the same test at 5 degrees optical. You'll start to see a pattern.

Do the same thing with a triangle wave and you'll only see 85% of the performance of a sine wave.

Do it with a square wave, its depressing.

Don't push hard on driving the scanner amps wide angle.

Steve

[buffo]

Welcome to PhotonLexicon, rooksoup!

As a general rule, a set of 30K scanners has an absolute maximum bandwidth of 2500 Hz for very small steps (less than 2 degrees optical, which is actually less than 1 degree of actual rotor deflection). And even at that rate, you'll see about 3 dB of distortion in the scanned image. This is a fundamental hardware limit, assuming perfect scanner tuning; if your tuning is off, then your bandwidth will be reduced. (When it comes to the maximum bandwidth for full-travel movement, the frequency limit is an order of magnitude lower.)

The artifacts you described are caused by the input signal exceeding the maximum acceleration that the scanners are capable of. (Or at least, exceeding the maximum acceleration that the current tuning can provide - just keep in mind that most scanners come from the factory already tuned to within a few percent of optimal, if not dead-on.) At 800 Hz, I suspect you'll need to be well below 10 degrees optical to avoid distortion, if not lower than that.

Consider the acceleration of the X and Y galvos when you feed them a standard sine/cosine signal (to draw a circle): as you increase the frequency, eventually the scanners will start to lag behind the input signal at the ends of their travel (when they must reverse direction). This is what causes the circle to distort into that rounded-square shape you described. Sustained operation like this is not recommended.

Most scanner amplifiers include a poly-fuse that will go open-circuit if you try to push the scanners too hard. However, you can cook a set of scanners by operating them right at the ragged edge of that over-current trip point for extended periods. Likewise, there are certain resonances that will kill a set of scanners. If you hear a screeching sound, that's bad. Reduce the amplitude of your input signal and/or reduce the frequency of the signal generator.

As Steve posted above, square waves are brutal. Be very cautious with any waveform that has a large jump; start with lower frequencies and amplitudes and slowly work your way up.

Regarding the relationship between the input signal to the scanner amplifier and the output signal sent to the scanners, recall that scanner amps are closed-loop feedback amps. That means the output signal is constantly being modified in an attempt to bring the position feedback signal into alignment with the input signal. Thus, the output waveform will often be very different from the input signal. And yes, there is a lot going on in that feedback circuit (That's why each channel has 6 adjustment pots.)

You mentioned that you were using inexpensive scanners, but keep in mind that all scanners have a maximum acceleration limit. The rotor and mirror are made to be as light as possible, but they still have inertia that must be overcome, and the faster you want them to change direction, the more force you'll need. Eventually you'll need more force than the coils can produce, even with excessive current flowing through them. The current state-of-the-art scanners only offer about a 3-fold increase in performance vs the standard 30K scanners.

Adam

[rooksoup]

Thanks buffo and mixedgas for your wisdom!

I'm relieved that my laser-assembler's intuition of "stop if the setup makes a terrible noise" is correct. Amazingly I've not managed to actually break any of my galvo sets yet after quite a bit of 2000Hz+ circle-drawing/accidental dropping/placing-on-ungrounded-metal, though one did eventually start squeaking if forced to deflect more than about 5 degrees from neutral, and I also had to superglue a mirror back on. Can't complain for the price.

I'm also glad that it shouldn't be necessary to mess with the tunings. Given that my sets of galvos have performed impressively consistently when mixed-and-matched with driver boards, I'm inclined to believe the boards are set about as well as they could be out-of-the-box. Or maybe I just got lucky!

I'm going to try to find the sweet spot for speed for my purposes without square-waving - thankfully I only need an extremely small range of optical deflection, so that probably previously limited the amount of damage I could inflict. I suspect the sweet spot might be "800Hz, final offer, and be thankful it only cost £40".

[mixedgas]

If you have an oscilloscope, even the cheapest of amps usually has a position sensor Testpoint, usually TP1 is ground and TP2 is POS. TP3 will sometimes be after the integrator in the PID loop if it exists.

TP 1 and X-TP2, Y-TP2 are usually brought out on a three position 2.54 mm header on dual amps. '

Usually one header is power supply VCC-ground-VSS and the other is the feedback for tuning. YMMV...

The other place to scope is the top end of the galvo current sensing resistor.

But be careful if you probe.

Make sure the galvos are well heatsinked. The usual cause of death is the drive coil expanding and hitting the rotor, when overdriven or oscillating. Which is why I monitor the PSU currents when setting up. Position sensors should burn about 20-30 mA each on the positive rail, BTW.

Very much like a 4 Ohm speaker as coils go.

Most scanner amps are based on the attached Cambridge Schematic. Various parts are removed/changed to reduce costs.

Usually a good quality audio amplifier chip is used out of specification as a power opamp to reduce costs. Even more so an NPN-PNP Darlington stage as output to really reduce costs.

Ia and Ib to the three opamps in the upper left corner of the drawing are the position sensor photodiode currents

AGC out is the position sensor LED drive.

You likely will not have the coil current calculator / overdrive shutdown. The RC4200 multiplier chip is no longer made and there is NOT a low-cost substitute. Closest available chip adds $12 USD per channel, so the protection circuit is often deleted first.

Steve