Analog Correlator Circuit.

Correlator made me think of an analog multiplier. But the cheap part is harder.

George H.

Sum (subtract) the two voltages and compare with a window comparator?

Why the rectifiers? A window comparator takes care of all stuff after the subtractor. I don't remember the numbers but there are some cute ones out there. You basically set the trip points on separate pins and when it's within the voltages the output is in one state, outside, the other.

Why the rectifiers? They would make the position compare ambiguous.

A subtractor with gain followed by a window comparator should work.

--

John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser drivers and controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro   acquisition and simulation

If it's _absolute_error_, the percentage specification makes it "interesting" :-) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

Sure, you need to put the two signals into a differential front end and use the sum as the error. Near zero out, assumes near 0 error, this can then be used to fire your laser.

However, the problem I see here is, it seems you have to many components from the feed back? It appears, you may get some phase margin error and that may have to be corrected by changing how your feed back is getting back.

Also, you said it has sloppy linearity at the ends? I am familiar with that :) You can use a log circuit to reshape that. This assumes that when it swings out to the far edges, it is repeatable?

In our beam scanning, we insert a trapezoid pulse at the top tip of each pulse just prior to the return scan. This is to help prevent the beam from sitting on the out edges of the scan longer than in the middle.

Something to think about..

Jamie

Once upon a time, John Bergoun (SP?) of Solid State Music designed and developed various devices for music synthesis, and some had or included multiplier cells. If i remember correctly, Exar carried that electronic baton forward in regards to manufacture and sales. SSMnnn was the product series,where nnnn stands for product number. But they did not last long. Even the baby bird (goo gull) dropped out. The Gilbert Cell was the prime "mover and shaker" in those chips as i remember. There should be something that has that on a chip which would help reduce the cost. And then there are a few "programmable" analog chips..

If you use a modern micro with internal clock and sensible design practice then noise from the micro should not be an issue as far as your signals (which seem quite big) are concerned. The EM radiation from an internally clocked micro can be very low. I've used such parts successfully in very low noise systems without problems.

Michael Kellett

assuming the percentage is percentage of the other signal's current values and not percentage of full scale or percentage of envelope.

4 resistors, dual comparitor and an xor gate _ /| / A ------+-/\/\/--+--/\/\/-+ | / | _|_ | | - | | |\ | `-------------|+\ __ | | >-----\ \ .---------------------------|-/ )=1>--- out | | |/ .--/__/ | | | | | | | | |\ | | `----------------------|+\ | | _ | >--' | /| .------------------|-/ | / | |/ B -+-/\/\/--+--/\/\/-+ / _|_ -

--
?? 100% natural

Isn't there already an error signal that drives the servo? Can't you just detect its magnitude and lock out the laser if it's too high?

--
John
>
>
>
>
>
>

What acuracy? What cost? Power? What response time? Just how much latency can you stand?

I [think] can do it for around a $1-3 in parts/100 and obtain approx

100 ppm relative accuracy and 0.1% absolute accuracy. Circuit noise lost in background noise. absolute noise below uV. LatencySHOULD be less than 10mS, but could be up to 30mS. Circuit having the advantage of being easily changed during development. Power consumption less than 100mW. from memory 30mW?

Now the analog approach has costs: extra one time engineering $25k divided by 100 units plus additional components and test during manufacturing of $10ea representing a 'trade-off' budget of $260 ea, hmmmm. can do an LOT with that kind of money.

t

s i

d text -

We use the AD633 ~$10 in onesies. And also a more expensive offering from AD.

The summing, window comparator solution looks cheaper. Unless there's some 'gotcha' I'm not seeing.

George H.

[snip]

Great, Jasen! I'm not sure if the signs are correct, but the concept is spot-on! ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

Looks hazardous. For example, if A=0 and B=0, the output is undefined and probably oscillates. There are other nasty cases.

But Jim loves it, so it must be OK.

--

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

I questioned the signs, did I not? But it was a good initial grasp of the problem on Jasen's part... something you _never_ exhibit.

John Larkin, I don't know why you have such an intense desire to be a turd, but why don't you move yourself (the turd) outside with your "parentage", the dogs? And STFU! ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

Sorry for not replying earlier.

The issues are three.

1. The clients who will be installing said circuit and adjusting it, are lucky to own a decent multimeter. Very few even own a scope. They will be adding it to a existing scan set, and said scan sets' feedback signals very greatly from unit to unit. This is a laser entertainment (Light Show) product, and the beams need to hit bounce mirrors across a room. The safety issue is taken care of by masking of the possible beam positions.

1. The scanners have two photodiodes and a flag on the scanner shaft, the feedback is thus differential. The issue is some makers of the scanners bother to drive the leds that illuminate the position flag with a temperature compensated AGCed current source. Others went the cheap route and just used a resistor. It depends which US designed amplifier they copied. Linearity of some of these copied sensors is not very good.

Domestic made US galvos are world class, and no problems with linearity, on small, fast scanners. Big ones for laser marker systems usually have large magnitude sensor errors.

1. The third issue is the scanners are clones of a US made product for the most part. The usual cost cutting was done when the copies are made in the far east. I have no control over which set is in use by a given customer, and their "test" outputs for the feedback are all over the place in quality. So offset and magnitude vary greatly. Yet the window I need to hit to gate the beam is tight, because it will be used to ensure they hit the bounce mirrors.

I like using a MPU, however having a customer adjust constants in it is not a good idea, the need to add a LCD, EEPROM etc.

The origional idea of having the absolute rectifiers was so I could do vector velocity and use SQR X^2 + Y^2 a multipler as a velocity check. If the velocity is near zero and the sign is correct, magnitude may not be so important. If the beam is still, its still. This does not check the galvo for gross errors in position, though. This leaves the question of what to do when things are near zero. The system spends much time near zero.

I really like Jason's idea... I might just try that this weekend.

I'd prefer one or two trim pots. I know, trim pots are evil. , However, adjustment needs to be multimeter based, due to the variance in the customer's hardware choices. Few operators build their systems the same. That is a occupational hazard with laser stuff, it is the same fundamentally, but always has little differences in final implementation.

Before you say why .1%?, -5-0-5 out of a 16 bit DAC represents 30 degrees of travel if the customer uses the full range of the galvo/ servo. While this is not reccomended, some do. Hitting a 4x4 inch bounce mirror at 60 feet means you need to be pretty good on your error. Often customers do double bounces, hit one mirror, then fly across the room (or stadium) and hit another larger mirror. This means you need to be pretty good on your repeatability.

I also have no control over some of the software used. We do add breaking points, guide points, and corner points on vector images. That is very similar to adding trapezoid, so thank your for that suggestion. However we scan at a constant, but adjustable point rate, and the industry test pattern is for vector images, not stopping on a dime. So I just want to enable firing when the scanner is on the bounce mirror.

Steve

OWN OPINIONS ONLY, THE VIEWS EXPRESSED HERE DO NOT REPRESENT MY EMPLOYER.

Hmm, It would be interesting if you were getting an analog voltage reference from the galvo device or is it giving you parallel encoded line for ABS position?

Either way, the better ones should have an array system inside to report an abs position, much like using the grey code for decoding.

As for analog output, that too, can also be supplied by the galvo but it does not tell us if this reference is being derived from grey code array detection or, if it is using photo shaded disc and hope the lower and upper regions are linear! This can also lead to weak optics over time thus giving you non-linear response.

I also see a problem of response time. Being that galvos are mechanical devices, their response time is just about proportionally reduced as you increase the expected response time. For example, if you are pumping random audio into the reference input...

I don't know how fast they are these days but our units are not that fast... We merely use them for target hunting to auto align a critical position of some components in cable building in specific machines. The mirror's axes feed back has an array encoder that gives us a ABS position in analog output. A 16 bit level pos, which is more than enough.

In your case, it sounds you maybe working with other types of feed backs. Like a LVDT type maybe or analog photo disc ? Both can generate linearity errors..

Log amps can fix the output of these crappie galvo feed backs. Of course, this would mean an alignment process would have to be performed at set up, most likely..

If you don't want to go with a processor system, then a couple of buttons and cal pots must be presented so when you maximize the input signal, you can offset the feed back to match the input for the +/- sweeps.

Just and idea....

The external software should surely be able to let them tune the laser targeting against known fixed targets and then correct for any intrinsic non-linearity of crude hardware in the control software?

If you mean sqrt( x^2 + y^2 ) here you might find that a crude approximation like ABS(MAX(x,y)) + ABS(MIN(x,y)/2 is easier to implement in hardware (if you are picky the coefficients can be tweaked)

Is "zero" in this case nominal dead centre on one of the target mirrors or just straight ahead relative to the box containing the galvo?

--
Regards,
Martin Brown

I don't follow your application exactly, but a cheap scheme that provides some degree of information assurance is to send multiple tones with specific twist. This was used in electronic fuses (as in fuse for ordnance) back in the day. When your shell has an electronic trigger, you really really want to make sure you are the one that sets it off.

I don't see any analog correlator being simple. Also back in the day, you could use CCDs to perform the correlation, but that is essentially digital since you need a clock.

Correlation of two signals is an integral of a product; if your signals are known-amplitude square waves, this is equivalent to an XOR gate and a bit of resistor/capacitor filtering... A CMOS XOR gate like CD4030 or CD4077 and maybe an op amp and comparator, and you're in business. As for the 'matches within 0.1%', though, that's going to depend on a bit of cleverness. Anything in the analog realm will need some settling time after power-up, too.

PLL lock detectors do this kind of thing, you might want to crack a book and look 'em up.