I am building a custom microscope, PhD in biomedical engineering. I would like to have a scaling amplifier, with programmable offset and gain, that takes 0..10V signal X (from piezo element) and turns it into output -1..1V Y signal (into galvo mirror) such that Y=A*(X-B) where A~0.1-0.5 and B~-1..+1V are adjustable within 2-3 digits. In fact, currently we use A=0.18, offset B=+0.119

The frequency bandwidth should be around 1kHz or higher. The input signal is usually a sawtooth with 1Hz repetition rate. Input impedance should be > 10kOhm, devices outputs into 20kOhm impedance. The delay between input and output should be as short as possible, but 1ms is OK. I think we can tolerate output noise of 0.5-1mV.

Currently we are using rather expensive 1MHz scaling amp, so I want a copy of that with lower quality and much lower price tag. The problem is that we'll need roughly 4-5 of these, at around $1000 a piece. I feel like there is a way to make these devices 10x cheaper.

Is that a feasible project for an EE undergrad? I have access professor with much more EE experience than I, for double-checking the final design.

I am considering custom PCB as a solution, or something arduino-based for programmable part.

  • \$\begingroup\$ Do you know the impedance of the 0..10 source and the load on the -1..1? Also 3 digits of adjustment/precision is possible. However, 3 digits of accuracy is beyond the tolerances of regular electronics and can only be done with calibration. Have you ruled out 16b USB IO? Is there a latency requirement? \$\endgroup\$ – scorpdaddy May 4 '20 at 16:15
  • \$\begingroup\$ @scorpdaddy re: accuracy. In effect i will do the calibration during alignment. I don't care if device says "B=0.001" but outputs "B=0.005" as long as these numbers are stable over 12 hours, does that make sense? \$\endgroup\$ – aaaaa says reinstate Monica May 4 '20 at 16:22
  • \$\begingroup\$ Can you be clearer about what A can be and what B can be numerically. Also, is your input signal DC biased such that signal minus B doesn't go negative? What sort of noise output is acceptable for the design and how stable do the offsets and gain need to be drift wise and against temperature. \$\endgroup\$ – Andy aka May 4 '20 at 16:23
  • \$\begingroup\$ @Andyaka thanks for the questions, I added more details. The temperature we can assume to be stable ~RT. \$\endgroup\$ – aaaaa says reinstate Monica May 4 '20 at 16:39
  • \$\begingroup\$ As currently written, you state that the bandwidth should be 100Hz or higher, and the input to output delay should be of order a microsecond. Are these both correct? Because they don't seem to fit together. \$\endgroup\$ – Jack B May 4 '20 at 17:25

For signal fidelity (SNR), here are concerns

(1) plan on using a steel box, unless you can guarantee no magnetic fields (only steel handles the 60Hertz power and rectifier surge peak fields)

(2) plan on a Ground Plane; I'd default to 4" by 4" area, plus Power Supply; plan on a steel metallic strip between Power Supply and your analog circuits.

(3) assuming a goal of 0.1 millivolt PeakPeak random + deterministic "noise", with half of that for the Random, in 100Hertz bandwidth, the OpAmp noise density is

50uVPP / [ 6.2 sigma * sqrt(100Hertz) ] = 50uV/62


about 800 nanoVolt noise Density.

(4) assuming the output stage is gain-of-10, you are allowed 80 nanoVolt noise Density referred-to-input; that is total noise contributors of 400,000 ohms.

(5) to not be burdened by 1/F noise, the 1/F noise plot corner needs to be 50Hz or lower

(6) If this is knob-controlled, then use 10-turn pots with verniers.

If this is computer-controlled, then consider CMOS 10-bit or 12-bit multiplying DACs or DigiPots.

(7) Topology: do the math up front, followed by Gain_of_10X amplifier including a Low Pass Filter to set the noise (random and deterministic) floor.

(8) plan on +- 15 volt supplies, so life is easy. You only need about 1 watt total power, so the Ground plane easily removes heat. Maybe as low as 0.1 watt.

(9) stability over 12 hours: have heat paths from the Ground Plane (metal mounting posts that contact the Steel case and contact the Ground Plane, with metal screws). The thermal time constant of 1 cubic meter of copper is 11,400 seconds. The thermal time constant of 0.1 cubic meter (4") is 100X faster at 114 seconds. Having a central metal mounting post speeds that up by (2*2) to 32 seconds. That's probably acceptable.

(10) pick a good opamp: ensure it is spec'd by a plot for 1/F noise (I may have gone overboard on the achievable noise floor; your call); ensure opamp handles +- 15 volt rails (IMHO)

As with any free suggests, your mileage may vary.

(11) black brick power supplies (switching Regulators) will be a problem. You want to avoid Ground Loops (charges such as from the black_brick that MUST find a return path).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.