I have a load cell attached to a crimping press (20mV Full-scale differential voltage). The press presses down on the load cell when it's crimping a terminal and the load cell outputs a differential voltage.

enter image description here

It doesn't explicitly state in the screenshot but the pulse lasts for about 35-40 ms. The operator can be fast enough that he is able to crimp two wires in approximately a second. I am considering using an ADC with an integrated PGA but I'm not sure what sampling frequency I should choose. I can see the benefits of sampling this as fast as I can because I can average the results to reduce noise but I don't know where to draw the line - would 2kHz be enough? 10kHz? Perhaps I haven't look hard enough but it seems there aren't many high speed ADCs with integrated PGAs.

An alternative way would be to use a PGA like PGA281 (with excellent drift and error specs) along with a 16-bit or so differential ADC and sample at, say, 200kHz and average 50 points (effective sampling frequency: 4kHz)

The peak detection itself is just a simple if statement. I start sampling as soon as the operator presses the foot pedal and continue sampling for approx. 300 ms (this is the window in which the pulse is bound to appear).

  • \$\begingroup\$ Are you trying to recreate the waveform or capture the maximum? \$\endgroup\$
    – Samuel
    Oct 1, 2013 at 7:16
  • \$\begingroup\$ @Samuel Capture the max. \$\endgroup\$
    – Saad
    Oct 1, 2013 at 7:29
  • \$\begingroup\$ Sampling fast and then having to do maths to smooth it out is a bit laborious, if you sample slower the smoothing is done for you. You want some basic filtering in there to avoid capturing small spikes/transients in the peak-hold circuit. \$\endgroup\$
    – John U
    Oct 1, 2013 at 7:53

2 Answers 2


You have a screen shot - why don't you plug a memory stick into the oscilloscope and take the readings into excel. The data will be high resolution of course but in excel you can average, decimate, filter, decimate and average or do whatever you feel fit to reduce the data until you are starting to lose temporal and amplitude information that is important to the quality of the parameters you wish to know.

When you reach that point you will know what sampling rate is sufficient.

As for PGAs inside a chip with a combined ADC - you can have them separate but still controllable digitally. I use a PGA from Microchip for this - the MCP6S21-I/MS and we measure strain gauges and load cells.


There are a couple easy options.

Just build a peak detection circuit and sample shortly (maybe 100ms) after the foot pedal has been pressed. The peak detection circuit will hold the maximum value until reset or until enough time has passed. I'd recommend a resettable one, like this.

Otherwise, just sample fast like you're already saying. The best way to determine exactly how fast depends on how precise you need the max value to be. You're on the right path splitting the waveform into how many samples would occur at different sampling rates. For instance, if you were sampling at 2kHz, you'd have 40 samples during your 20ms pulse. That's a sample every 0.5ms, so just see if a 1ms window is enough resolution to get an accurate enough maximum. In this case it looks like you'd be quite alright with 2kHz. With an engineering hand wobble I'd say you'd get within 5% of the absolute maximum.


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.