Advertisement

Blog

Bring Back the Analog Computer

I was doing some reminiscing about my college days and comparing to what is being taught today. One of the things that I clearly see is that schools focus more on the digital side of electronic design than the analog. Also, there is more simulation work than hands-on circuit building. One of the interesting things I remember using was the analog computer.

This was a neat item to work on during the lab sessions of classes. For those that may be too young as to what it is, the analog computer was a nice tool to fully understand the basic circuit concepts to add, subtract, multiply, divide, differentiate, and integrate. The lab work basically takes an equation and the student would need to put the appropriate modules together to solve the equation. Though it sounds simple and mundane by today's standard of using a simulator or a computer, the challenge was figuring out the voltage levels one could use without saturating the output of the various parts.

The adding and subtracting parts are very easy to use, since they are in basic op amp circuits all around us. The challenges were the other devices. Using a differentiator could easily cause saturation if the step was too small.

The nice thing about today's circuitry is that it is faster and uses less power than before. The best part about analog computers is that the signal does not need to be digitized before processing. Only delay times kept things from being figured out immediately.

Thinking back through my design years I can recall two instances where the analog computer thought process was used extensively. The first was an image processing system and the other was used for mimicking a Wheatstone bridge.

The image processing project was looking for a rate of change outside a vehicle as the system was to detect an incoming projectile. The concept was to project a fan of light and then look for reflections off the incoming projectile, which was expected to see a set number of reflections based on the projectile geometry.

My rough sketch of the scenario is below:

The trick was that any processing time had to be quick since time to react was crucial. The program took two paths to solve the problem, an analog and a digital. I got to play with the analog design.

The thought process was to take a linear CCD and use it as an analog shift register. There would be successive scenes to compare with and then generate a signal if there was any consistent detection.

The circuitry would then compare the signals of the current imaging time and successive image times to see if there was a high reflection.

The analog circuit was not the real problem as much as obtaining the analog shift registers. This project was done back in the early 90s as companies were phasing out the analog shift register. I ended up using a linear CCD to feed a signal into the white balance input. This design did not go very far due to the issue of component acquisition for future use. This type of CCD configuration was being phased out.

The second analog circuit development was essentially making a DC strain gage act like an AC excited strain gage. The following circuit block is what was used:

Essentially, A1 acts as the subtracting amplifier, since it is a differential amplifier with divider resistors to drop down the high AC voltage. A2 is used as a phase shifter so that the output is always in phase with the input AC reference. The DC input goes into buffer A3 and A4 is used as a reference to handle the non-zero volt zero reference from the DC input. The multiplier circuit then takes the AC reference as the X1 signal. The X2 signal is grounded. The buffered DC input is fed into the Y1 signal and the A4 reference is fed into the Y2 signal. The multiplier then performs the simple task of (X1-X2)*(Y1-Y2). There is some additional signal processing, but this circuit worked out very well.

So, the big question to you readers, what type of analog circuits have you worked on that digital processing just would not be the best solution?

11 comments on “Bring Back the Analog Computer

  1. Bill_Jaffa
    January 27, 2014

    The analog peak detector is an application-specific analog computer, fast, cheap, effective, and real time; see “The peak detector: a classic analog circuit still in wide use” at:

    http://www.planetanalog.com/author.asp?section_id=396&doc_id=562072

     
     
  2. RedDerek
    January 27, 2014

    @BillS – Yes, the peak detector is something a digital circuit cannot really replace for high speed. Your article inspired me to get this one done. 🙂

  3. Victor Lorenzo
    January 27, 2014

    I think that most engineers, just like partialy hapened to me, are moving out from analog world to digital world. Analog computer functions were covered as part of analog electronics when I was a graduate student, but in my work as an analog&digital designer I've only used addition/substraction for level shifting and multiplication for scaling up/down.

    The peak detector can be implemented digitaly, but of course at a higher cost than its analog equivalent and requiring higher convertion rates.

    Oh, one correction, I did use logarithmic and exponential amplifiers, but only for hobby applications in one analog guitar effects processor for compression, AGC and distorsion.

    Once again, thanks for your posts, this has also been very interesting.

  4. amrutah
    January 28, 2014

    Long back, I had the opportunity to work on the OOK receiver which meets the AISG 2.0 standards, with bandpass filter, the envelope detector. I think implementation using digital is a very hard task with the inputs being of the order of 27mVp-p.

     

  5. Victor Lorenzo
    January 28, 2014

    I agree, it is sometimes complex to work with such low level signals, not because of signal levels (as we can add de-noising, level shifting and gain as needed) but for dynamic range. In some cases we end up using programmable gain amplifiers too.

    For instance, ECG signals:

    After signal conditioning you may desire to have 500mV from peak of P-wave to peak of S-wave and an ENOB of 10 bits for signal processing, but due to electrodes-to-skin interface and breath you'll have the ECG signal constantly moving (offset) by a time varying voltage which adds to ECG's baseline. That displacement forces you to increse dynamic range (e.g. 4Vp-p) and ADC resolution (12 bits or more) or include more signal conditioning stages to remove the varying offset (which could partly be an analog simil to a moving average filter for removing very low frequency components).

  6. Vishal Prajapati
    January 28, 2014

    I being one of the youngest engineer in the community, I have worked a bit on analog circuits in the lab sessions in the collage days. Even after that also, I have used those circuits in the analog front end for signal sensing. I have used adder, subtractor, integrator, differentiator, active filters and level shifters. I have extensively worked on microcontrollers. And from my experience I can tell it could be very expensive to implement such fuctions in the digital circuit in terms of computing power and cost involved. It is lot easier to implement such basic circuits in Analog circuits with minimum cost and least time delay.

  7. David Maciel Silva
    January 31, 2014

    Indeed, this is a circuit that relies solely on a treatment of the signal for proper processing.

    Digital processing, but has specific characteristics when treating a signal conditioning circuit necessarily the most part is analog.

  8. Bellhop
    January 31, 2014

    One of the most fun things that I did in my university days in the '70s was to use a Heathkit (vacuum tube) analog computer to regulate the speed of a 5 hp three phase motor. The analog computer was doing the PID feedback and regulating the motor field current via an amplifier. A tachometer reported speed so that a little two watt pot could adjust the speed of the big motor. Of course, speed was held constant over the full range of motor loads. Nothing illustrates PID like an analog computer!

  9. Victor Lorenzo
    January 31, 2014

    @Bellhop, I've seen many steam engines of many sizes and something that called my attention from the very begining was the centrifugal governor. It is probably the simples proportional (P) regulator ever. It is mechanically 'analog' ;).

  10. brook.d
    August 6, 2014

    It's true that back in the '90, students were able to better understand how an analog circuit should work, mainly because most computers were analog. Today, in the digital era people prefer a more advanced technology, they can find everything they want on http://www.perfdata.com/, the best source of hardware and accessories.

  11. JamesBryant
    August 14, 2014

    An inexpensive log amp is simpler and cheaper, and may be MUCH faster, than a high dynamic range (i.e. very high resolution) digital front end.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.