2
\$\begingroup\$

I am using a Wheatstone load gauge through an AD8295 op-amp to a SAMD21 uC ADC. The op-amp uses a resistor to set gain. An analog switch is used to change gain. Since the resistors are not exact (G = 49.4k/R + 1) I want measure the overall gain by using the DAC output of the uC. The gains will be stored in EEPROM during programming or memory on startup. Analog.com tech article shows a DAC connected to the +IN of a single ended op-amp to set offset but I am using a differential op-amp, so put a summing junction on the +IN side of the op-amp for the DAC? Should the DAC output be buffered with another op-amp? The offset of the zero load point in the ADC also has to be adjusted to get the maximum resolution. Loads are typically higher in one direction (the application is a shock dyno). By auto changing amplification and offset I can get the maximum resolution of the data. Thanks.

Some additional information - The op-amp used in the Analog.com article is a differential op-amp. The article is here: https://www.analog.com/en/technical-articles/offset-adjust-for-a-differential-op-amp-driving-an-adc.html I am currently using a PGA204 with fixed gain. However, I am changing from a Labjack board to a Feather M0 because of cost and upgrading to WiFi connection to the PC. Ti.com shows using a PGA204 and PGA205 to get programmed gains but the result is not in the ranges that I need. The 1000lb calibrated strain gauge output is 3mV/V with 12V supply giving +/-3.6mV output. The SAMD21 (corrected) works on 3.3V with options for Vcc, 1V or Vref for scaling. For full scale ADC at 3.3V the gain has to be 45.83. Reading full scale at lower forces are multiples of this, hence the analog switch and variable gain op-amp. The resistance of the switch can be taken into account by adjusting the fixed resistor value. However, the gain will not be exact which is why I want to calibrate the amplifier using the DAC. The non-loaded Wheatstone bridge will be connected when the DAC adds the gain calibration offset so the differential op-amp will see this as a signal input. Reading the ADC with the known DAC input should give the actual gain within the resolution of the ADC. The resolution of the ADC can be increased from 12 to 16 bits through oversampling (this cannot be done during runs due to time constraints). Once the actual gains are measured they can be stored in non-volatile memory. Peak loads on a shock are typically ten times higher on compression than rebound. Having a gain for reading +/-500lb to read forces that are +500/-50lb will need to be lower resolution than a gain for 600lb. range. Using the DAC to change the zero load offset will allow higher resolution.

\$\endgroup\$
1
  • \$\begingroup\$ Maybe it would be better to use a PGA rather than trying to DIY with an in-amp which will have errors based on the on-resistance of the analog switches. You might be painting yourself into a corner by sticking to the AD8295 + analog switches. \$\endgroup\$ Commented Aug 29, 2023 at 5:00

2 Answers 2

0
\$\begingroup\$

There is a lot to unpack here.

So if I understand right, you want to measure the gain of your amplifier using a known voltage to correct for the resistance tolerance. That can indeed be done using an analog multiplexer and a DAC. To do it, you must feed your calibration signal (DAC) to the input of your chain. The further you can go, the more you can compensate for. Connecting to the positive input of the amplifier won't do what you are expecting.

What happens if you connect the in+ to the DAC is essentially, you are creating yourself an offset for your amplifier. This can be done if you oversaturate your amplifier but it won't allow to calibrate your amplifier.

The difference between an opamp and a differential amplifier is significant. You can't use one to the other interchangeably. The differential amplifier as the control loop inside. Therefore, it is useless to look how opamp circuit are designed if you have a differential amplifier.

To offset the 0 and the gain to maximize your range is a huge part of the design of your analog front end. If you need to do signal conditioning of the sort, you probably should use op amp. Some differential amplifier do have a biasing voltage. Regarding gain, this is never perfectly accurate. You always need to consider some tolerance on the component and do a linear 3 point calibration to adjust for gain and offset. Calibrating your analog loop is not easily scalable in production. There are technique to do it, but 99% of the time, these are over engineered and don't work well. Your best bet might be to place your differential amplifier with the right gain then offset your signal using an opamp. There are other topologies, but that one might be the simplest for you.

Creating a multirange analog front end isn't easy. There is a lot of software going into that and the hardware gets a lot more complex when you had several ranges. If your application allows it, I would use 2 separate front end. When that's designed, improve it to add the second range, but for start, keep it simple.

Follower amplifier might be required depending on your sampling rate. They allow to have a good impedance at the input of the ADC. Most design don't require them. It really depends on how your amplification stage are designed. Too many follower add uncertainties too, so if you need to be accurate, you might want to minimize them.

\$\endgroup\$
0
\$\begingroup\$

I'm not sure if the MCU-integrated ADC will be good enough for what you want to do. A cheap 24-bit external ADC would do the job and either connect directly to the strain gage, using an internal PGA to amplify the signal, or with a very small external gain (<10).

In most cases, unless you want to sample a rather high bandwidth signal (more than 0.5kHz say), there's little point to variable gain in strain-gage front-ends. They are all designed mostly to have the same full-scale strain, and thus their relative output (in mV/V) is similar for similar strain in the body of the transducer.

Good instrumentation amplifiers are great, but they can be expensive, and recently the availability has been hit-and-miss.

I've been doing lots of resistive bridge data acquisition, and with a 24-bit ADC it's never been necessary to have variable external gain, if all you need is roughly a 16-bit resolution.

Modern high-resolution ADCs perform so well, that it's very hard to maintain their performance with just about anything connected in front of them, other than the signal source itself. A $5 ADC channel may require $20 worth of analog parts in front just so that the front-end doesn't degrade system performance in terms of linearity and noise.

\$\endgroup\$
1
  • \$\begingroup\$ Thanks, I ordered a TI evaluation board for the ADS127L01 ADC. When I first looked at 16 bit could ADCs the prices were much higher and TI's 24 bit ADC is within reason for this application. \$\endgroup\$ Commented Aug 30, 2023 at 22:02

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.