ILDA Format BS

[The_Doctor]

Does your car have traction control? Antilock brakes?

Haven't got a car, so I can't get too deep into that one. My point is that even now, there are parts of the system that don't have that kind of 'intelligence' and they work fine. Short of striving for a neural control system like in our bodies, which would be overkill, I think there's no point to going beyond a careful separation of high and low order functions.

A hard drive is basically a magnetic head controlled by a galvo. They manage to be fast and accurate to within hundredths of thousandths of a millimetre. They do this every time. They very seldom fail. They are controlled not by the computer sending them a voltage, but by the computer telling them which block to seek to.

They also cost fifty bucks a pop. Do you think they'd be that cheap if they had to do all the modelling in the computer? Everything in the mechanical system would have to be more accurate and the electronics would have to be an order of magnitude more complex to compensate for the fact that the drive couldn't compensate for its own inadequacies.

"The cost of a single transistor is a million bucks". Not exactly true, but if the million bucks had not been spent building transistors, no one transistor could have existed. Same with hard drives. Your logic will make perfect sense when laser scanners are made and sold on the scale of hard drives.

How much would you add to a scanner? A sensor to determine the angle and distance to the projection surface? A microcontroller 'brain' to allow it to use the info gained to modify the projection to get good scale accuracy? As this is so much harder than using our brains to do it and tweak a control to correct it, we might as well do it that way. Using digital pots so we can store values to make things faster and easier to set up helps, and is cheap.

[carmangary]

By your logic we would all still be using open loop scanners.

[James Lehman]

This is one of the most worn out disputes of modern electronic design!

Analog will ALWAYS be the ultimate solution because it is NOT bound to a time quantum.

As soon as you go from one sample to the next, you have thrown away any possibility of knowing what lies between them.

With a pure analog solution, the tiniest improvements in any part of the system might very well lead to better resolution.

James.

[drlava]

EDIT: I recommend S/PDIF. If you need more channels for speed and flexibility, use them. It's been invented to solve the problem you mentioned, so why not use it? It even has a SYNC feature built in!

S/PDIF is inadequate because it is a stereo protocol. Additional channels are gained through DTS or Dolby encoding only which cannot be applied to the application at hand due to the gross distortion that would result. The optical protocol that does exist that could handle it is ADAT lightpipe however it is a unidirectional protocol. Look around at the lengths of S/PDIF optical cables the common plastic cables max out at 50-100ft. You would need low loss glass fibers and you don't want to get into how much more that would cost over CAT cables.

Differential dignaling over CAT6 would be less expensive, easier to implement, naturally bidirectional, and 4 differential lanes would be available, one for a plain go/nogo hardware interlock.

Heroic is right, it is much more feasible to compensate for the nonlinearities and time constants of a system when you get direct feedback to the control system, not a predictive model. A DSP with high speed feedback is uniquely suited to add real-time compensation for many more poles, zeros, and nonlinearities than you would even want to think about modeling on a predictive basis or build into an analog control system. Try tuning an analog galvo control system with 10 intercorrelated pots!

[James Lehman]

Take a look at ADAT LightPipe.

James.

[The_Doctor]

By your logic we would all still be using open loop scanners.

Daft. I won't say any more, I already said plenty. I mentioned closed loops and what they do (simple physical compensation). I'm not arguing against those, am I? Misrepresenting my words with one-liners doesn't do anyone any favours.

[carmangary]

Daft. I won't say any more, I already said plenty. I mentioned closed loops and what they do (simple physical compensation). I'm not arguing against those, am I? Misrepresenting my words with one-liners doesn't do anyone any favours.

What I have been hearing from you is that what we have now is good enough and it would be a waste of money to make it more complicated. And that we can solve it with software.

You could have done the same with open loop scanners by modeling them. Then there would be no need to spend money on complex galvo amps that we have now.

If my one liner offended you I am sorry but that's basically what I was hearing from you.

[heroic]

Differential dignaling over CAT6 would be less expensive, easier to implement, naturally bidirectional, and 4 differential lanes would be available, one for a plain go/nogo hardware interlock.

And what's more, you can use commonly available ethernet PHY chips which are very cheap.

"The cost of a single transistor is a million bucks". Not exactly true, but if the million bucks had not been spent building transistors, no one transistor could have existed. Same with hard drives. Your logic will make perfect sense when laser scanners are made and sold on the scale of hard drives.

Take a look at the PicoP projectors. They use MEMS two-axis galvos- the mirror tilts on both X and Y axes. I would personally love to have one of these in my projector because it would mean I could make the projector much faster and much smaller. Oh, yeah- they're looking at building them into cellphones, which means volume volume volume. So it looks like my logic does, in fact, make perfect sense. 8-)

[The_Doctor]

S/PDIF is inadequate because it is a stereo protocol.

So use more stereo channels. Some tak the high road, others tak the low road, but someone will get there afore the idealists are done. There are multichannel digital busses too, like the ADAT that James mentioned. I think you're right about the digital signalling via cheap network cable though, but then you have the capacity for a simple fast data stream anyway.

A PID controller (which scan amps are) IS partly predictive. If you're going to take that whole idea and convert it into digital, then maybe it's worth looking to see what other fields have done this and hitch a ride, avoiding any more wheel inventing that required. Anyway, this isn't the same issue we were at earlier, where the focus was on what amounts to vector models whose language extends to the scan amps. THAT's my only real contention, that we should avoid doing that because it amounts to an exclusion, defeating the kind of interoperability we mostly seem to want from laser stuff.

EDIT: I'll withdraw from this at this point because I think that last point is what's at the core of anything else I can say on this.

[mixedgas]

James wrote:

With a pure analog solution, the tiniest improvements in any part of the system might very well lead to better resolution.

---------------

DSP has one advantage over pure analog. It can know the past and thus react to the future better, google "recursion". That's why Cambridge is claiming 3X small jump over pure PID. And btw, according to Dr Lava, the last bunch of fast Chinese amps he looked at were P-I-I, no D.

With water cooling, Lightspeed/LFI supposedly hit 80-90K. so with DSP, that would put 120-160K within range if the stators can take the heat.


Carmongary's idea works, but if you do it, your putting the Quadmod card into the galvo amp, and saying, hey, display this image as best you can. You can tell a QM2K to go vector, and compute what it thinks is the best image from a given ilda frame. So somebody here reading this thread want to show us a few pics of complex frames in both optimized vector and point by point mode on a QM2000? I only have a Q32 point dumper, so I cant do it. So by his reasoning, brains plus DSP on the galvo amp will give you art from art, system indepedent.

BTW, here is the equation of state for a galvo, from Laser Beam Scanning, Gerald F.Marshall , Editor Marcel Dekkar inc, the equation comes from jean Montagu, the cofounder of GSI. He edited the LIS scanner chapter of the book.

Enjoy:


Steve