Sampling rate of ADC for Thermistor input for Loadcell

SanderLPFRG

Hi all

We are trying to develop a custom board with built-in loadcell support. We are now looking into the options and found many articles stating the HX711 sensor is too slow for accurate Z sensing with an 80Hz sampling rate (Resulting in 0.025mm accuracy at 2mm/s probing speed)

@dc42 advised in 2021 (Re: Duet3 6HC + load cell + amp) to use an op-amp structure connected to the thermistor/ADC input, which would result in higher sampling rate, but at what frequency is the thermistor input running for RRF?

Also, we found the HX717 sensor, which is using 320Hz sampling rate, which would be more sufficient for probing Z (0.006mm accuracy at 2mm/s probing)speed. What are your thoughts on that?

gloomyandy

@SanderLPFRG If you have not already seen it this Klipper discourse thread provides some interesting reading on both sample rates and some of the issues of dealing with noise and other problems https://klipper.discourse.group/t/strain-gauge-load-cell-based-endstops/2134

dc42

@SanderLPFRG what do you intend to use the load cell for? That dictates whether you need low drift or can tolerate high drift, what resolution you need, and what sampling rate you need.

For Z probing you need only detect when the load changes quickly. So you can filter out the low frequency drift, filter out high frequency noise, and look for the difference between the current load cell output and the average output in the recent past. This is what we do in the Duet3D Smart Effector. We use a low-noise op amp to amplify the strain gauge signal and an 8-bit microcontroller to compute the average and do the comparison, with the sensitivity being programmable. If you didn't need programmable sensitivity then you could use a comparator instead of the microcontroller.

If you also want to use the load cell to monitor the filament force applied by the extruder then you need low drift and high resolution. Perhaps the HC717 or HC71708 would be sufficient, depending in part on the sensitivity of the strain gauge. I haven't found a non-Chinese full datasheet for these devices, so I don't know whether the full 24-bit resolution is available at 320 samples/sec, or how high the drift is.

SanderLPFRG

@dc42 We are planning to use it for a Z probing configuration, X and Y offset measuring, and later crash detection.

We now have drawn a schematic based on the structure you mentioned in a previous post but using the INA818 op-amp with a 1V offset voltage. Testing and simulations with our prototype custom load cell, it outputs a range of roughly 0.4 to 1.6V on the ADC input on the ATSAME, of an expected scale of -150 to 350 grams. We currently plan to configure the ADC into a linear analog signal and set the sensitivity with the "Z probe trigger threshold" in RRF. Still, the 8-bit microcontroller/comparator also sounds very interesting. Is that the purpose of the "ATtiny 44A-MUR" that I see in the schematic?

It is an interesting approach but I guess the sensitivity is only programmable via the programming pins, not via Firmware, right?

As for the chips, I think you mean "HX" instead of "HC", right? We have looked at those, but the sample rate is expected to be too low for our wishes in Z and crash sensing.

Thanks again for your answer

SanderLPFRG

@gloomyandy really interesting and a very good article! I have not seen it before.

Funny to see they first chose the ADS1263 but later moved back to the HX sensor. It does seem effective, but probing at 1mm/s is less than we aim for (we have a bed size of 600x500 with a probing grid of 11x11, (if we probe the entire bed)

Therefore we now plan to read the signal directly at the microchip