Electronic analog computers 3

[nicabod]

You might be surprised to learn that electronic analog computers are still being made. Afaik, the only company is Comdyna, <

http://www.comdyna.com>.

I'm not advertising them, but I do think Ray deserves credit. I do hope his knowledge of how to build good ones is being preserved for posterity.

I've never had the honor of setting up an electronic analog computer, but have some "feel" for what it's like to do.

Electronic analog computers give you a great "hands-on" ability to modify the ongoing problem (often a simulation (if not always?)) as it runs. They are great, so I read, for teaching control system theory.

It's interesting to consider how op amp performance has improved spectacularly since electronic analog machines went out of use.
Given careful attention to shielding, grounding, thermal offsets, component linearity, and such minor(?) sources of error, it should be possible to make some impressively-accurate e.a.c.'s.

One interesting item, only somewhat related, is that no digital simulation of an analog circuit is valid if the circuit's normal behavior is chaotic. Such circuits have to be analog, period.

Regards,
NB


Nicholas Bodley |*| Retired technician
Eastern Mass.

 

[SlideRuleEra]

Outstanding posts on analog computers, Mr. Bodley. Many thanks for sharing this information.

About 1969 the professor of a controls class took us students to a little used engineering lab. In a small room was an electronic analog computer connected to a very large x-y plotter. The computer had a removable panel that looked like the front of Comdyna machines in Mr. Bodley's link. Only these panels were about 2 feet square and covered with a grid of plug connection points about 1 inch on center in each direction. A box contained a "grab-bag" of electronic components (resistors, capacitors, inductors, and patch cables). The professor wrote the math equations to simulate a spring & dashpot system then proceeded to program the equivalent electical analog for overdamped, critically damped, and finally underdamped situations. I'll never forget watching the x-y plotter slowly "go crazy" trying to plot the underdamped case. Then he had us students give it a try - I never got it to work at all, but that was a day that sticks in the memory. We all knew that we were participating in something that might never be seen in use again. In my case that has been right.

 

[GregLocock]

The SR71 (Blackbird) had an analogue computer for its automatic stability system, and as late as the F16, the fly by wire was based on an analogue computer. The F16 is inherently unstable, if the analogue computer goes down, so does the plane.

I think your distinction between analogue and digital chaotic systems is rather arbitrary. Either can be chaotic. Neither can accurately simulate the other.



Cheers

Greg Locock

 

[nicabod]

To Greg Locock--

Thanks for the info. about aircraft that use analog systems; that was unexpected.

As to my comments about chaotic circuits, it's good to be reminded that digital circuits can be chaotic, although I can't think of any at the moment; I'm far from being an expert! Also need to do some homework...

What I had in mind was that the current belief that digital is better, no matter what, (I exaggerate) doesn't always hold true.



Nicholas Bodley |*| Retired technician
Eastern Mass.

 

[rmw]

SlideRuleEra,

My graduating quarter, I had a design course that required us to program an electronic analog computer to design a shock absorber/spring system for an imaginary car of given mass, so that it would dampen within one and one half cycles upon hitting a curb of given height at a given speed. I did get it to work, as my subsequent graduation depended upon it. It was interesting, and I learned much more about first and second law functions as part of first and second order equations than I did about the benefits of analog computers. I think I might have left my slide rule sitting on top of that computer, now that I think about it.

Many times since, as I have watched the temperature indicators for the back end of a gas turbine climb during acceleration, I have thought back to what the analog computer class taught me (about energy dissapation during acceleration).

rmw

 

[GregLocock]

Examples of chaotic digital circuits would be any valid calculation of fractals, or even solutions to 3 (or more) body gravitational problems.

http://www.cuug.ab.ca/~kmcclary/ORRERY/index.html

would be chaotic, but it is so slow that you probably won't ever find out. This is one case where the analogue computer (a mechanical orrery), is NOT an accurate model of the real world system, whereas the digital simulation can be shown to be accurate for thousands of years (not millions though).

Oh, I may be confusing chaotic with fractal. So far as I know all fractal systems are chaotic (that is, future iterations must be calculated, the solution for an arbitrary number of iterations ahead cannot be predicted correctly), but not all chaotic systems are fractal (in which the fractality is constant irrespective of scale, that is, the structure of the solution is self-similar at different scales).

If you are interested in exploring chaotic and fractal systems graphically then the most wonderful program is

http://spanky.triumf.ca/

http://www/fractint/fractint.html

Cheers

Greg Locock

 

[zeitghost]

For that wonderfully obscure British degree level qualification known as the Council of Engineering Institutions Part II, as late as 1979 it was a requirement in the Computer course that you should understand how to program an analogue computer.

I've never used this since then, but it seemed remarkably complex at the time, what with scaling things in time and amplitude to ensure that the amplifiers didn't saturate.

As I recall, I avoided anything in the examination that touched on analogue computing....

rgds
Zeit.

 

[Skogsgurra]

Guess what!?

It looks like I have to put my hands on a Solartron TY1451 analogue computer real soon now. Yes, vacuum tubes and +/- 100 V working range. Chopper stabilised.

The reason is that I ridiculed a guy that had got strange errors in his embedded system when he tried to work outside the -32768/+32767 range. I told him that it would never happen in an analogue system - and now I have to prove it!

A friend (you need no enemies with such friends) happened to have the TY1451 in his institution and I was allowed to use it if I repaired it first. So, I guess it is time to blow the dust off the tube checker and be prepared to put out some minor fires during the following weeks. I will keep you posted - and probably yell for some help.

 

[GregLocock]

http://dcoward.best.vwh.net/analog/sol.htm

For a picture. Good luck (with 160 amplifiers to check and calibrate you'll need it).



Cheers

Greg Locock

 

[nicabod]

Skogsgurra,

Good luck with the project! You will keep one hand in your pocket, I trust.

Again, I'm reminded of how much op amp specs have improved since the days of non-stabilized tube op amps.

Indeed, it does seem that a good part of the skill needed to program electronic analog computers (and programmable mechanical ones, rare beasts...) was/is arranging the topology of the interconnections, and scaling everything to fit. Somewhere on the Web is a detailed exposition of a non-trivial problem, and the rethinking needed to set it up successfully.

Nicholas Bodley |*| Retired technician
Eastern Mass.

 

[GregLocock]

Yes, there is a link to it on the parent of that web page. It includes a circuit diagram of a breadboarded analog computer using op amps.

I am tempted to build it!

Cheers

Greg Locock

 

[Skogsgurra]

Greg, Thanks for the pictures. I certainly hope that there is a typo in the spec sheet. Onehundred and sixty amplifiers in that cabinet seems to be on the high side. I hope there are no more than ten or twenty - which already is plenty of work.

Nic, How shall I be able to "put my hands" on it if I shall keep one hand in my pocket? You see? Problems abound.

It will be interesting to find out in "modern terms" what those opamps can do. Terms like offset voltage, bias current, noise, bandwidth and so on. Looking forward to it - it is all about finding the necessary time to do it.

 

[SlideRuleEra]

skogsgurra - Good luck on your project.
Mr. Bodley was suggesting that you keep "one hand in your pocket" - as I remember, that was a safety measure. When working with the high voltages used by vacuum tube circuits it minimized the chance that you would have current flowing "in" one hand and "out" the other hand - directly "across your heart". Believe that it was quite effective - maybe because just doing this kept you continuously aware of the risk.

 

[Skogsgurra]

So I will have to put my hand (singular) on it. Soldering will be tricky with only one hand. I think that two hands are one too few when it comes to soldering (tin in one hand wire in one hand and solder iron in one ...?)

 

[SlideRuleEra]

Maybe that's the REAL reason the printed circuits were invented.

 

[ietech]

Just to state the obvious. I assume the unit would be de-energized if you were soldering and, of course, all energy storage devices, caps. etc discharged prior to work. If this is the case you can put your hands, head and feet in there if you want. I don't know of anyone who would make mechanical or electrical repairs on an energized system unnecessarily.

Component or circuit testing is another business, in this case keep one hand in your pocket.

Good Luck with your project,

ietech

 

[zeitghost]

Doesn't an amplifier run out of headroom in a similar way to the signed int that your embedded friend was using?

I suppose the difference would be that it stays solidly stuck to the rail, rather than generating curious and incorrect values.

Guess that must be why they invented the long int....

Would be nice if some of these programming languages actually took account of errors like that though.

rgds
Zeit.

 

[Skogsgurra]

With all these good advice, the project is bound to be a great success. ;-)

 

[IRstuff]

Bear in mind that "hand in pocket" only works for household level voltages.

I got zapped by a 25KV supply and I was wearing rubber soled shoes on a linoleum floor. Touched the door of the power supply and next thing I knew, I was flat on my back on the floor.

It's a good rule, but as with most rules of thumb, there are caveats.

TTFN

 

[ietech]

Another rule of thumb "test for voltage twice or you may only get to touch once". This even goes for doors in a troubled system.