1
\$\begingroup\$

I have a device which has a 16 bit D/A converter which through an analog linear circuit generates an output voltage between -1V and 1V. I have 32 of such channels on my device.

schematic

simulate this circuit – Schematic created using CircuitLab

I need to design a circuit to be used for factory calibration. I need to find the gain through the analog part of the circuit so that I know which digital value val to write to the D/A converter to achieve a DC voltage output of, say Vx=800mV.

I need to calibrate it well enough that I can achieve an output voltage of accuracy better than 0.5mV, preferably 0.25mV. The calibration does not have to be very fast. Anything up to two minutes will do.

I have been thinking about just using a D/A converter. Another way could be to use a comparator and a precision voltage reference and using bisection to determine the gain, but it seems to be hard to find one that has an input offset voltage that is low enough.

How would you go about going this?

Conclusion:

I am going to use two 16 channel ADC's and a LTC655BHMS8-1.25 voltage reference which has +-0.025% accuracy at 2 ppm/C. The ADCs will be LTC2439-1IGN which are slow but accurate. Using a small microcontroller I will control the ADCs and use averaging to get rid of noise.

I think this should meet my requirements without breaking my budget.

\$\endgroup\$

3 Answers 3

1
\$\begingroup\$

I think you will find chopper stabilised amplifiers with low enough offsets.

And if you can guarantee linearity in the analog stage better than 0.025% (1 part in 4000) you can calibrate at 0, +1V, -1V and the rest will follow.

I think the tricky bits will be :

  • sourcing exactly 1V within 0.25mv and keeping it there across temperature.
  • verifying that all sources of error combined do not exceed your error budget.
\$\endgroup\$
2
  • \$\begingroup\$ You are right, calibrating just a few points and relying on linearity to solve the rest is good enough for my application. I had not heard about chopper stabilised amplifiers before. They seem like they would work for this. You are right that finding an appropriate voltage reference is not so straight forward - nothing off the shelf seems to be available. \$\endgroup\$
    – martinhans
    Commented Dec 4, 2013 at 12:48
  • \$\begingroup\$ I'm accepting this answer not because it is what I decided to do but because it helped me in my thinking process. \$\endgroup\$
    – martinhans
    Commented Dec 4, 2013 at 13:19
1
\$\begingroup\$

This is a bit out-of-the-box, and may be too expensive, but I think it could be quicker to implement. The basic concept is that you use a multimeter as your voltage sensing to get you an accurate voltage reading, and then either calibrate each channel individually, or use a multiplexer to switch between inputs.

With a multiplexer, I think you could meet your 2 minute limit. It would also save some pain and suffering involved with making your own "high accuracy" meter and calibration issues between channels, and also (probably) eliminate the need for a PCB design.

\$\endgroup\$
1
  • \$\begingroup\$ If OP has any budget at all I'd definitely go this way. Agilent has a 6-1/2 digit benchtop meter listing as $1100. If you buy this, the problem is solved. If you build your own, your problems are just beginning (like do your customers want a traceable cal on your production test equipment, ...). \$\endgroup\$
    – The Photon
    Commented Dec 3, 2013 at 22:22
1
\$\begingroup\$

Depends on your budget. I would probably go for an Agilent DAQ box, personally. If you're going in to production, the cost of buying or renting a DAQ unit is not very significant. The 34970A and 34972A units both have built in multimeters and can take up to 120 channels worth of relay mux cards. The 34970A is programmable via GPIB and RS232 while the 34972A is programmable via LXI (Ethernet). It would not be very difficult to write some scripts that can interface with the DAQ box and your device and perform the calibration. The DAQ unit is also very fast at scanning its inputs - the 16 channel mux card can be used to acquire at 250 channels/second. The 34970A and the mux cards can be found on ebay for a very significant discount if you're on a budget.

\$\endgroup\$
1
  • \$\begingroup\$ Looks like a good device, but I don't have the budget for this. I need to build a handful of calibration stations. \$\endgroup\$
    – martinhans
    Commented Dec 4, 2013 at 12:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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