# Self-adjusting analog signal translator

I think this question is best asked on concrete-ish terms, so, here it goes.

Let's say we have 4 wires, all with common ground:

A) ≈9V-20V, varying.

B) ≈3V-6V, varying.

C) A low frequency analog signal (<1kHz), which can vary from 0 to A.

D) In here, I need to output the analog signal from wire C, but translated from range 0-A to range 0-B.

Point of the circuit being that it should self adjust to always translate the signal from range 0-A to range 0-B, even if A, B or both change.

I'm basically stuck, any pointer or a suggestion would be appreciated.

• Wouldn't this basically be an analog multiplier? If you really need this, expect to pay a pretty penny for it. Dec 29, 2020 at 17:50
• What does "low frequency" mean to you? I've worked with folks who hear "10kHz" and say "that's insanely high frequency", and with other folks who hear "10MHz" and say "oh, that's low frequency". Dec 29, 2020 at 17:52
• Define real purpose and components Dec 29, 2020 at 17:53
• And how accurate do you need the translation to be? I assume that you want $V_D = aV_C + b$, with $a$ and $b$ adjusted based on $V_A$ and $V_B$ -- correct? Dec 29, 2020 at 17:53
• B seems to be about 1/3 of A. Is that ratio constant or some random DC Dec 29, 2020 at 17:57

• Get an ADC. Set its analogue reference to be the voltage value on line A.

• Get a DAC. Set its analogue reference to be the voltage value on line B.

• Connect the ADC digital output with the DAC digital input.

• Connect signal "C" to the ADC input.

• Connect line "D" output to the DAC output.

• Set both ADC and DAC for continual conversion (you might need a pulse generator).

If the "A" line range of 0 to 20 volts is too high for a particular ADC, then pot down to make 20 volts become 5 volts (for instance). Ditto the signal line "C". Same ratio of potting down.

If you can't get a suitable DAC that can have a reference input of 6 volts (line "B") then pot it down then, take the DAC output through an op-amp amplifier to restore to the value you need for line "D".

• You'd have to make sure the ADC and DAC support this kind of operation, but this seems like a good solution to me. Avoids mucking about with microcontrollers. Dec 29, 2020 at 19:02
• IF I can find a pair of ADC and DAC which can communicate without an MCU in the middle, this actually seems like as elegant a solution as I've seen to this problem yet. Dec 29, 2020 at 19:14
• Well, it was doable mid 80s when I built a simple 8 bit analogue storage thingy. ADC -> RAM -> DAC different read and write speeds, no MCU. Dec 29, 2020 at 19:46

The solution that uses the least board space is to find the smallest microcontroller you can that has an adequate ADC and an adequate DAC, or perhaps an even smaller microcontroller with better off-board ADC and DAC. At this point you can:

1. Use an ADC with at least three inputs.
1. Attenuate your $$\V_A\$$, $$\V_B\$$ and $$\V_C\$$ so they are in the input range of the ADC.
2. Read $$\V_C\$$ at a sufficiently high rate (not 2kHz -- you probably want 5kHz or 10kHz), and $$\V_A\$$ and $$\V_B\$$ at the same rate, or at whatever slower rate fits with your judgement.
3. Do the obvious math in the microcontroller, pump the answer out to the DAC, and amplify the result sufficiently to hit the correct range. This keeps the circuitry simple.
2. Use a 1-input ADC with a flexible $$\V_{ref}\$$. Ditto the DAC.
1. Attenuate $$\V_A\$$ and $$\V_C\$$ identically (or appropriately, depending on the ADC's $$\V_{ref}\$$ arrangement),
2. drive the normal ADC input with the attunuated $$\V_C\$$ and the ADC $$\V_{ref}\$$ with the attenuated $$\V_A\$$.
3. Attenuate $$\V_B\$$ to match the DAC's $$\V_{ref}\$$ range, then amplify its output by the reciprocal of the attenuation. Then the microprocessor just needs to shuttle the number from the ADC to the DAC (with chips circa 1980's, you could do this without the microprocessor -- oh, this degenerate age when you need a million transistors to do the work of 20...)

OR

Get on your analog high horse and eschew any of this digital nonsense. Make a triangle wave or sawtooth oscillator that operates at some fixed fraction of $$\V_A\$$ (preferably 0V to $$\V_A\$$). You'll need a way-high frequency here -- probably 20kHz at least, if not 200kHz. Feed that and $$\V_C\$$ to a comparator to get a square wave. Use that square wave to switch $$\V_B\$$ on and off. Low-pass that result, and you have $$\V_D\$$. It's functionally the same as the microprocessor solution, it just needs way more way lower integration components and it'll be more of a pain to keep tuned in manufacturing (I'm an analog circuit designer by inclination, but one who's continually disappointed by the reality of digital circuitry).

OR

Use multipliers as suggested -- this will be even more hard to tune than my PWM suggestion above, it'll be driftier, more sensitive to component variations, and as an added bonus, the available stock of analog multipliers is getting ever-more narrow and more expensive. But it's about as pure analog as can be done.

• Wow, talk about a comprehensive answer. I think you've covered all the theoretical prospects excellently. Now I'll roll up my sleeves and start testing each for practical viability in my application. Thank you very much. Dec 29, 2020 at 19:16