Oh Bollox

[The_Doctor]

Ok heres something.

I realised that this test - the Mamba one was done with blanking. So I repeated with no blanking and guess what I get similar results to you. Same scan speed and angle just red laser as the vid camera seemed to overecxpose the green making it difficult to see. First vid shows red scan complete then shows a close up.
second vid - and heres the thing shows the scope trace set to XY and it displays a similar shape to that being projected. This shows that the scanners cant track accurately with this input.

http://www.stanwax.plus.com/laser/rednoblank.wmv

http://www.stanwax.plus.com/laser/scope.wmv

The thing I find most interesting though - and maybe someone has an explanation for this - is that when I lower the scanspeed to 1000pps so the intermediate points are shown more clearly, the return seems to have as many points as the scanned line. I would have expected several points as the object is drawn with a quick return to the starting point without any intermediate points. I am guessing this is needed for better return accuracy but dont know.I expected the scope trace of YT to show a sawtooth wave but it in fact is a triangle wave.

Rob

Very interesting. I guess there ought to be a specified scan speed and angle to do this with so scanners are compared more accurately.

I think the effect might partly be inertial, as the scanners are moving at different rates at the time the error is most prominent. Even so, the whole point of closed loop feedback is to compensate for that.

Is your scope looking at the feedback output of the scan driver? I guess it must be.

My rotating line WAV file is a triangle, with a little pause at each peak for dwell points. That's how I get sharp vertices in my polygons, by allowing both scanners a tiny moment to stabilise at each before moving on. The fact that this error still occurs is significant because that means it's not caused by residual momentum from the previous move.

In Mamba, I guess it's drawing one, returning on the other pass, but it still has to avoid stressing a scanner so it will still define a step size and count to keep the return trace within safe bounds. The actual waveshape might be an asymmetric sawtooth/triangle form, or might even contain a curve to accelerate it in each draw, either to compensate for inertia to maintain an even line weight on the drawn line, or to increase the speed of the return line. I kept my line as simple as possible, strictly linear, with dwell points at each end, so it's easy to replicate, though I see now that software might try clever things that actually mask what is really going on in favour of rendering some specific line.

All that aside, the fact that even CT scanners show this effect at all IS interesting. I'd suspected they might, but had no way to know till now. Thanks again, this is really cutting to the chase now.

[The_Doctor]

Sorry for the large slightly off-topic post but I wasn't sure whether to start a new one or not.

Cheers,

Matt.

Don't mind me, I do it all the time. I suspect most of this stuff comes under the general heading of 'Oh Bollox' anyway... Quite a discovery, when we're not expecting stuff like this from stuff that costs us so much.

[Lampy]

Yeah, bit of a bugger really. I don't know why there is a "Don't return to centre" function. I can't really see why you'd want your galvos to jump back to the centre after every frame anyway. The disable function worked in the previous version but the MIDI didn't work properly so I don't really want to down grade back. C'mon Old Grumpy, sort it out! Other than that I'm very happy with Mamba Black, especially the MIDI functions.

Back to The Doctors' thing, I tryed it with the DT-40s and they performed well. I'll video it at some point if you need to see but you can hardly see any problems. I haven't even re-tuned them yet. I've tried this on Mamba and Beam Sculptor. The Beam Sculptor gives you easy control over what is projected without interpolating things for you if you don't want it to.

I also tried the ILDA test pattern on a slow rotate and they keep consistant tuning at any rotational angle. I wouldn't have thought they'd be able to do that.

[The_Doctor]

I also tried the ILDA test pattern on a slow rotate and they keep consistant tuning at any rotational angle. I wouldn't have thought they'd be able to do that.

I would but what about the small text stuff? Any sign of wandering aberrations in that?

If you have a copy of Sound Forge or similar tool, and can load and loop that WAV file I linked to in that other thread I linked to from here, a sound card can feed the signal I used to test with to the scan drivers. It's symmetrical about zero so DC blocking isn't a problem for it. This might be easier than trying to make sure the right signal is being made some other way.

And sure, pics or detailed descriptions are always useful.

[stanwax]

I suspect most of this stuff comes under the general heading of 'Oh Bollox' anyway...

That made me laugh out loud!

as for the scope no doc this is on the output of the easylase board. Hence why I thought it was interesting - or more so.

Rob, have you checked the output settings in mamba? You can change the interpolation for blanking as well as adding more blanking points.

Matt
Yeah I have dicked about with the settings but to no avail. As I said this draws just fine with blanking its the odd behaviour of the returning 'blanked' scan thats very intriguing. I wasnt aware of the rtc bug though I do operate with it unchecked as the scanners seem much happeir with it off.

Please explain this "Don't return to center" option and what it is for.

Gary Im with you I dont really know what its for, as the scanners seem to work harder when its ticked and are much happier when its disabled...? If you are drawing a circle it seems natural for the scan just to track round in a circle - no reason on this earth for the scan even when blanked to deviate off this circular course until a new different frame is needed.

Rob

[The_Doctor]

as for the scope no doc this is on the output of the easylase board. Hence why I thought it was interesting - or more so.

Oh, I guess it's designed to compensate for inertia then, to get the blanked trace as quickly and safely as possible to the next draw start. Any chance of trying that WAV and a sound card? I guess that won't show any aberration at all. Shouldn't do anyway, beyond maybe a fraction of a line width given a beam divergence of better than 1 milliradian.


I looked up the Raytrack 35plus's that Zoof mentioned, seems they are DT-40's with a different scan drive. What Bill (Benner) said about stiffening the shaft might help, as this might be due to mirrors flapping around. WideMoves have especially spindly shafts... I'm not sure what other mods he's sending to Jian (I read the post but I haven't been following that discussion at length), but it looks as if DT-40's might be a good bet, if CT scanners can't be had cheaply enough. I need at least 30 K at 30° ideally but accuracy is far more important than speed to me.

[steve-o]

Would you believe, I have no idea what you are talking about?

__________________

Ditto

[The_Doctor]

You will, you will.. See Zoof's post and Rob's second video as well as mine, and you'll have the start points of it. All I had to go on to start with was what I saw.

Just bear in mind that Rob's second video appears to show something different, a software compensation for the effect, or more likely a way to get the return line drawn fast because in Mamba it's usually blanked.

[DZ]

Doc,

The issue that you are refering to comes from one scanner running faster than the other. There is a test pattern that specifically tests for this. Its the Laser Media test pattern. I can probably dig up the ilda file if you or anyone else needs it. You can usually eliminate the problem with tuning. Really though, unless its WAY mistuned, you wont notice a problem when running a show and certainly not if just running beam shows. It can be more pronounced if your making large steps with no points in between. Here is a quote from Greg Makov of LSDI, "My final test involved the Laser Media test pattern. This pattern was originally designed for relatively slow scanner speeds of 12K pps, but is still widely in use. One of the interesting aspects of the Laser Media pattern is that it has two diagonal lines that are retraced in each direction. This is a useful feature in evaluating the relative speed of the two scanners. If one scanner is running faster than the other, the retrace lines will not be superimposed, but will follow two opposing curves.

I took the Laser Media pattern with a scan angle of around 40 degrees and started a slow Z-axis rotation. This creates a condition where the diagonal retrace line approaches the scanner axis (vertical or horizontal) and one scanner makes a small step, while the other makes a large step. Essentially, this shows us the relation of small step response relative to the large step response."

DZ

[Zoof]

If one scanner is running faster than the other, the retrace lines will not be superimposed, but will follow two opposing curves.

Thank you !

Got up this morning and had to try this before going to work and it works perfect! The laserMedia and hysteresis file (which are included in the easylase test prog) lets you finely adjust the proportional gain (servo) on one axis to match the other. By turning the servo pot the hysteresis(*) simply disappeared. Then changes to low and high frequency damping are needed again but after going back and forth a few times between ilda test pattern and hysteresis got the sweat spot. Great improvement!
The rotating lines now almost perfectly match. Using different colors going back and forth revealed some separation but is invisible using a single color.

cheers!

(*) hysteresis is actually not the right term to use (correct me if I'm wrong). The separation appears when the galvos have different characteristics, mechanically and electronic tuning. If both galvo have the same hysteresis the separation should not be visible . Like DZ said the separation is introduced when one galvo is running faster than the other. This introduces a different lag; one galvo is lagging behind on the other (in the time domain) and causes a phase shift in the frequency domain.
Bill mentioned in another post some time ago that alot of scanners he sees are not tuned right: the prop. gain and damping are fighting each other causing the amp and galvo to run hot. So what I first had was an ILDA test image that looked quite good but on one axis the gain and damping were fighting each other not revealed in the ilda test pattern. The separation on the diagonal in the lasermedia and hysteresis images allowed me to match the gain and damping on both axis. I matched x to y since I had spent a lot more time tuning the Y axis following the exact instructions on laserfx, it must be the best tuned one