I've been tasked to design a circuit which measures the peak impact force of a crimping press on a terminal. I am using a load cell under the press as a sensor. The load cell outputs 20mV (differential) at full scale (4000 kg force). The impact is quite fast - an entire crimp is over in about 20-30 ms!

The approach I've come up with is as follows: feed the load cell's differential output to an instrumentation amplifier (gain ~ 200) which in turns a peak detector and that in turn is fed to a 10 bit ADC on my AVR. I've tried this approach and while it works quite well it's not repeatable enough. The noise in the system makes repeatable measurements difficult. Please note: I've only tried this on a veroboard and haven't yet tried it out on a PCB. I intend to do that soon.

The 2nd method which occurred to me was that I could use a differential ADC with built-in gain of, say, 200x+ and directly obtain the result that way. The only downside is that I'm not reading the peak value - but I can always do some processing and find the peak value. Is the second approach inherently better?

I am also open to other ideas. The goal is just to measure the peak force and then compare that with a programmed range in the MCU to see if the force was too high or low.


A 20mV output should not be too noisy so I reckon it's the layout issues you have and a proper PCB should cure that. I'd definitely be wanting to analyse the peaks inside the MCU as it gives you many more options.

Whether you use an ADC or instrumentation amplifier you should consider low-pass filtering the signal before it goes into an ADC input to improve the noise situation as much as you can. You'll also need to consider putting an anti-alias filter inline too. Both filters are easily combined.

If you do sample at a much higher rate than what you actually need you should also consider an averaging filter in software too.

What load-cell excitation voltage are you currently applying? Maybe this can be increased to get a bigger signal?

  • \$\begingroup\$ I am using the recommend 10V currently. Maybe I should switch to a constant current excitation? Would that be better? \$\endgroup\$
    – Saad
    Sep 18 '13 at 13:08
  • \$\begingroup\$ @Saad No, constant current is not needed on full-bridge devices - it helps when you have quarter bridge devices where only one element from the 4 is active i.e. it provides better linearity but not on a full-bridge load cell. Maybe you can check that there are four active elements in the bridge. If so stay with constant voltage. Maybe you can operate it at higher excitation voltages to reduce noise - that's worth considering. \$\endgroup\$
    – Andy aka
    Sep 18 '13 at 13:14

Are you not source loading your load cell if you place a filter directly on its output. I understand that ADC have buffered inputs, making it possible to place a filter at its inputs. However, doesn't this same filter cause problem at the output of the load cell.


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.