### Author Topic: Home Brew Analog Computer System  (Read 89487 times)

0 Members and 2 Guests are viewing this topic.

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Home Brew Analog Computer System
« on: November 28, 2012, 09:05:12 pm »
Ever since stumbling across the awesome stuff on show at this website:

http://www.analogmuseum.org/english/

...... I've been working like a man possessed to design and build my own solid state hybrid digital/analog computer to modestly put virtually all of those extinct commercial offerings to shame (complexity, size, precision, etc).

I've just completed the preliminary design of the Sine/Cosine Function module. The completed computer will be equipped with 8 of these modules. On the Patch Panel it is simply two-port black box which spits out either (depending on the mode setting) the Sin or Cosine of the applied input voltage (Ein), multiplied by a scaling factor (K) of 100. It has (for an analog circuit) a very high degree of accuracy and stability).

Transfer function equations and limits are given in the text on the schematic *.pdf. I'm just sharing the little circuit because I'm rather chuffed with it. Log/Antilog, dividing, squaring, square rooting is all easy-peasy stuff. The sin/cosine function module however is one I've been scratching my head over the most. Once I figured out how to practically implement the necessary slope inversion (at the 90 degree and 270 degree points for sine) with a high degree precision, the cat was in the bag.

Unlike in the current simulation file, the final circuit will have a power hybrid discrete HV op-amp with a short circuit proof output stage and a +/- 10mA drive capability. The input impedance is 1M, shunted with 20pF or so.

The attached charts show the transfer characteristic in sine and cosine mode, with a full scale Ein ( 0 to 360 degrees).

« Last Edit: November 29, 2012, 09:27:16 pm by GK »

#### poptones

• Frequent Contributor
• Posts: 711
• Country:
##### Re: Home Brew Analog Computer System
« Reply #1 on: November 29, 2012, 01:19:14 am »
That's very cool. I'd love to see a detailed writeup about the theory and operation.

#### SeanB

• Super Contributor
• Posts: 14678
• Country:
##### Re: Home Brew Analog Computer System
« Reply #2 on: November 29, 2012, 03:03:00 am »
I used to have a box of mechanical sine/cosine resolvers and hundreds of other styles of synchros and resolvers, they were parts of anb autopilot. We were cannibalising them for spares to fix moving map displays and other synchro indicators

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #3 on: November 29, 2012, 09:12:09 pm »
Those must have been interesting devices.

Don’t have time for a explanatory write up right now, except that the circuitry based on the LT1097 op-amps in the left hand half of the schematic forms a “function generator” (in analog computer speek) that produces a dual (polarity) slope, asymptotic approximation (with break points at 90 degrees and 270 degrees) of either a sine or cosine function from the applied, uni-polar input voltage which represents phase angle. A bipolar, multi break-point curve-shaper then does the rest.

I’ve started refining the circuit, to begin the transition from LTspice to Protel. Firstly, I had to get rid of that huge series string of LT1634 voltage references to bias the break-points of the curve shaper. That is NOT very economical; and especially so when one intends to build 8 of the things. I’ve decided to use a separate pair of dedicated supply rails of +/- 50V, exclusively for biasing the break-point shapers of all sine/cosine modules instead. The intermediate break-points between +/- 50V are now provided by a 0.1% tollerance resistor ladder/divider.

Much cheaper and even more accurate as the +/-50V can be accurately trimmed for “perfect” symmetry. The regulator will be a hybrid op-amp/discrete deign based on the LT1097 precision op-amp for the error amplifier, deriving the +/-50V form the separately regulated +/-105V rails.

I’ve also added another pair of break-points to the curve shaper to improve the accuracy of the curve fit around the +/- peaks.

Attached is the revised preliminary schematic, along with a plot showing how well (for now) the break-point shaper approximates an ideal sine curve, by plotting it against a perfect/ideal sinewave source.

« Last Edit: November 29, 2012, 09:28:38 pm by GK »

#### alanb

• Frequent Contributor
• Posts: 308
• Country:
##### Re: Home Brew Analog Computer System
« Reply #4 on: November 29, 2012, 09:34:40 pm »
I don’t know if anyone else remembers but a long time ago ( late 1960’s I think) Practical Electronics published a series of articles on the construction of a simple analogue computer. It was very basic but was a good introduction to analogue computing. It kept me occupied for many months and was where most of my pocket money was spent during the period.

If anyone still has a copy of the articles they would be great to see again.

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #5 on: November 29, 2012, 11:26:48 pm »
The entire series is here:

EDIT:

The pages from the articles/issues are presented in random order on the site and you have to download each page individually, but they seem to be all there; I've just downloaded pdf files from the list on the left.

« Last Edit: November 29, 2012, 11:50:40 pm by GK »

#### alanb

• Frequent Contributor
• Posts: 308
• Country:
##### Re: Home Brew Analog Computer System
« Reply #6 on: November 29, 2012, 11:46:14 pm »
Thank you for that. I can see that the series actualy started in Jan 1968.

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #7 on: December 10, 2012, 11:43:04 pm »
Ever since stumbling across the awesome stuff on show at this website:

http://www.analogmuseum.org/english/

..............Surprisingly little comment/interest(?) in this kind of thing.........

The sine/cosine chassis of my analog computer has now grown to contain 9 sin/cos function generators. I've also decided to incorporate a precision 100V triangle waveform generator into the sin/cos chassis for calibration/test purposes. Here is my breadboarded prototype with schematic. This circuit generates a fixed frequency, 500Hz, 100V peak-peak triangle wave with a DC offset of 50V. With this signal as the input, my sin/cos function generator in the cos mode will produce a continuous sinewave output. This will aid part of the calibration procedure (by adjusting for minimum THD of the 500Hz sinewave with the aid of a distortion analyser).

The triangle wave generator works by integrating a precision amplitude (+/-5V peak) 500 Hz squarewave derived/divided down from a 2MHz crystal oscillator.
The triangle wave from the integrator is then amplifier to 100V peak-peak by a discrete high voltage op-amp. The 50V DC offset is regulated by a servo amplifier sensing the signal output, which steers the integrator reference. This produces a triangle wave at the output with a precise 50V DC offset and a duty cycle of exactly 50%. There is a trimpot to calibrate the DC offset in the voltage reference leg of the servo amplifier. The peak-to-peak amplitude of the triangle wave is calibrated with a trimpot varying the integrator input current.

The output current limits at +/-20mA, is short circuit proof and the HV op-amp is capacitive load stable. The integrator-steering servo loop is a “conditionally stable” system with a phase margin of 60 degrees.

edit - in the xtal oscillator part of the schematic there should be a 1M resistor terminating the MPF102 jfet gate to ground.

« Last Edit: December 11, 2012, 07:12:25 am by GK »

#### BravoV

• Super Contributor
• Posts: 5002
• Country:
##### Re: Home Brew Analog Computer System
« Reply #8 on: December 11, 2012, 01:52:30 am »

..............Surprisingly little comment/interest(?) in this kind of thing.........

Made self notes, watched, bookmarked, voted 5 stars for this thread since it was created, its just I'm too noob to involve in this wonderful topic. The closest thing to this topic that I'm learning and planning to work with is just the built single chip 4th quadrant multiplier.

Please, don't stop, and thank you.
« Last Edit: December 11, 2012, 01:55:12 am by BravoV »

#### saturation

• Super Contributor
• Posts: 4783
• Country:
• Doveryai, no proveryai
##### Re: Home Brew Analog Computer System
« Reply #9 on: December 11, 2012, 02:50:59 am »
Making an analog computer is quite a tour de force learning and implementing large amounts of analog technique that folks buy these days in pre-made ICs  . If you enjoy this level of work, I'll be a good way to get a job in analog IC design.
Best Wishes,

Saturation

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #10 on: December 22, 2012, 02:57:39 pm »
...... analog technique that folks buy these days in pre-made ICs  .

And digital ones just for fun

A bunch of prototype analog modules are currently on the back-burner until parts arrive, so I've made a start on the digital section of this "analog-hybrid" computer. Here is what I've called the "Iteration Counter". It's a 4-bit programmable-n counter whose primary purpose will be, receiving its clock input signal form the computers master timing/controller unit, counting problem solving iterations and, patched into an electronic analogue multiplexer network, switching different sets of "initial conditions" and/or constants for each problem solving iteration.
Just for example, suppose you "write" a program to compute and plot the trajectory of a projectile or missile. The problem can be solved/plotted/displayed successively with anything from 2 to 16 different (multiplexed) constants (just for example) for gravity. The completed computer will have four of these counter modules, which can be either daisy-chained for patched independently to perform other functions.

All of the computers patchable digital logic will use open-collector inverted logic (0=1) with internal resistive pull-ups on all inputs. All I/O is designed to be blow-up proof, such that any I/O can be accidentally patched to a + or - 100V analog output without damage. The Iteration counter has a clock input, a reset input and 20 current-limited open-collector outputs with indicator LED's. There are sixteen decimal outputs (D0-D15) and four BCD outputs.

Here is a vid of it running; as shown, n-programming is achieved simply by patching the reset input to the appropriate decimal output. The 4 LED's on the far right are indicators for the BCD outputs.

Schematic is also attached, as well as a more readable form the Triangle Waveform generator schematic, with a couple of errors corrected.

« Last Edit: December 22, 2012, 03:27:17 pm by GK »

• Frequent Contributor
• Posts: 561
• Country:
##### Re: Home Brew Analog Computer System
« Reply #11 on: December 22, 2012, 09:23:31 pm »
@GK: A simple and probably stupid question from a software guy - what drove the design choice of 50V rails?, I can understand split supplies ( balanced ) just not the high voltage.

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #12 on: December 22, 2012, 10:31:00 pm »
Actually the analog modules run regulated supply rails of +/- 110V, with an operating output swing of +/-100V. In analog computer speak, that makes this a "100V computer" - 100V = 1 computing "unit". Back in the day most solid state analog computers were 10V computers, with the analog (op-amp) stages running on +/-15V rails. However the Rolls Royce machines from Systron Donner were 100V computers.
The reason for this is that it is much easier to maintain a high degree of accuracy when the voltage is ramped up. Consider an integrator stage (the heart of an analog computer) based on a jfet-input op-amp. It might have an input offset voltage of 10mV. 10mV is 10 times less error in a 100V computer than it is in a 10V computer. Also, "function generator" modules such as my sine/cosine module that rely on diode break-point / piecewise waveform shapers are much easier to make accurate in a 100V design than in a 10V design, as variations in diode voltage drops and transistor Vbe's are much less significant.

A 100V computer is also 10 times cooler than a 10V computer.

« Last Edit: December 24, 2012, 12:30:41 am by GK »

#### Jay_Diddy_B

• Super Contributor
• Posts: 1618
• Country:
##### Re: Home Brew Analog Computer System
« Reply #13 on: December 23, 2012, 01:35:54 am »
Hi GK,

You might like this high voltage op-amp

Jay_diddy_B

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #14 on: December 24, 2012, 12:20:49 am »
Hi GK,

You might like this high voltage op-amp

Jay_diddy_B

Unfortunately it only does +/-70V !

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

Finally got the design of this solid state sine/cosine resolver completed. Might get the board dropped into the etch tank tomorrow if all this damn Christmas crap doesn't get into the way.

« Last Edit: December 24, 2012, 12:30:00 am by GK »

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #15 on: January 06, 2013, 01:48:49 am »
Some progress photos. I recently burnt a big hole in my wallet buying bulk quantities of most of the semiconductors (thousands) needed for this computer project. I've got the triangle generator and a sin/cos module 90% loaded. I still need some more miscellaneous bits and pieces to finish them, but the next bulk order isn't going to happen until my funds recover. However there is still plenty of design work to do. I've actually got most of the electrical design worked out now on paper; its just a matter of drafting all of the schematics in Protel and laying out the PCBs.

Attached are pics of the sine/cosine module and the triangle waveform generator, as well as a PCB layout of the power supply board (which I have just completed).

The triangle waveform generator along with nine of the sine/cosine module PCB's, the power supply module and power transformers (1x 18v-0-18V sec. plus 1x 110V-0-110V sec.) will be mounted into a custom 3U 19" rack case to complete the "Sine/Cosine Chassis". The PSU board is self-contained with the exception of the series pass transistors which will mount onto the rear panel of the case on a large heatsink.

This completed chassis will only be a very small part of the complete computer which will completely occupy a ~2 meter tall, floor-standing 19" equipment rack (which I already have).

I'm starting on Protel files for the "LOG/EXP Chassis" now, which will contain five 100V logarithmic amplifiers and five 100V exponential (the inverse of a logarithm) amplifiers. The transfer function of these modules are based on matched transistor Vbe's (MATXX-series) with thermistor temperature compensation, rather than the piecewise approximation method as used in the sine/cosine module. In a LOG/EXP application this method has a much greater dynamic range.

« Last Edit: January 06, 2013, 01:51:48 am by GK »

#### ftransform

• Frequent Contributor
• Posts: 729
• Country:
##### Re: Home Brew Analog Computer System
« Reply #16 on: January 06, 2013, 02:34:32 am »
can a ti-89 out perform your computer?

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #17 on: January 06, 2013, 02:51:21 am »
Can a TI-89 do any of this kind of stuff?

http://www.analogmuseum.org/english/examples/

..... or be patched into a prototype design to simulate/emulate simple through to highly complex control systems?

#### ftransform

• Frequent Contributor
• Posts: 729
• Country:
##### Re: Home Brew Analog Computer System
« Reply #18 on: January 06, 2013, 10:20:05 am »
Can a TI-89 do any of this kind of stuff?

http://www.analogmuseum.org/english/examples/

..... or be patched into a prototype design to simulate/emulate simple through to highly complex control systems?

Well I guess what I'm asking is how powerful of a modern computer do you need in order to equal the performance of your device?
Could it do something that a desktop PC can't?

I like your project I am just curious.

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #19 on: January 06, 2013, 10:58:35 am »
I think a modern desktop PC would have a many orders of magnitude greater computational power. Dunno how best to make a comparison to a digital computer; they are very much different things.

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #20 on: January 10, 2013, 08:53:35 pm »
Attached is the complete schematic for my logarithmic amplifier module. Due to the utility of log amps in a analog computer, I've decided to build 10 of these instead of only five. A custom 2U 19" chassis will house 10 of these along with their power supply. Another 2U chassis will house 10 exponential amplifier modules.

The transfer function is simply:

Eout = K * ln (Ein).

K = a fixed scaling factor of 20. The operating input voltage range is 0.01V to 100V, producing an output voltage in the range of +/- 92.1V.
The output amplifier clips at approximately +/-105V, so the dynamic range is a smidge over 4 decades.

The actual log converter based on the SSM2212 is a "fast" design I came up with on my own. Over the 4 decades it is very accurate (overall significantly better than 1% when trimmed).

A conversion to a logarithm of any base is simply a matter of external scaling by the programmer.

« Last Edit: January 10, 2013, 09:23:02 pm by GK »

#### SeanB

• Super Contributor
• Posts: 14678
• Country:
##### Re: Home Brew Analog Computer System
« Reply #21 on: January 10, 2013, 09:13:36 pm »
Looks nice.

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #22 on: January 10, 2013, 09:39:00 pm »
Looks nice.

Thanks!

BTW, if you saved the schematic, please delete it. There was a silly error and I had to take it down. I misplaced the decimal point calculating the resistor values for the offset trim/null range and it was inadequate by a factor of 10.

Attached is the fixed schematic (R8 increased from 10R to 100R and R10 added to restore balance).

EDIT: circuit revised - see reply#288

« Last Edit: February 15, 2015, 04:49:52 pm by GK »

#### SeanB

• Super Contributor
• Posts: 14678
• Country:
##### Re: Home Brew Analog Computer System
« Reply #23 on: January 10, 2013, 09:56:37 pm »
I do see the difference. You could probably replace D10-14 with 1N4148's as well, mount close to the heatsink inside a shrink tube to keep stray light from them.

#### GK

• Super Contributor
• Posts: 2409
• Country:
##### Re: Home Brew Analog Computer System
« Reply #24 on: January 10, 2013, 10:27:14 pm »
The D10-D14 string used for biasing the output transistors doesn't have to be thermally coupled to the heat sinks, as there is a large total of 132 ohms of emitter degeneration (R30-R33), which makes thermal runaway impossible. Suppose the output transistor heat up delta 50 deg C due to heavy loading. That will cause the emitter-emitter potential difference between Q15 and Q16 to increase by approximately 200mV, returning a quiescent current increase of only 1.5mA.  On +/-110V rails that will increase the quiescent power dissipation of each power output transistor by only ~165mW. This pretty much represents a worst case scenario; a 50 degree C delta heating in both power output transistors could only happen with amplifier being heavily ac driven into current limiting into an Eout short.

I used 1N4007's instead of 1N4148's for their lower individual voltage drop. Their total voltage drop along with the 132 ohms emitter degeneration result in a reasonably well defined (taking into account component variations) output transistor quiescent current of ~5mA.

I am though open to suggestions on how to best thermally couple the SOIC-8 SSM2212 dual monolithic transistor to the leaded thermistor for optimal temperature compensation of the log converter.
« Last Edit: January 10, 2013, 10:39:36 pm by GK »

Smf