PINDA v2 temperature sensor / compensation

[tompe-proj]

Requirements

• Heated bed with temperature sensor
• Z-probe with temperature sensor (e.g. P.I.N.D.A V2)
• Spare input for the probe thermistor

Description

I mainly print ABS with enclosure and experienced inconsistent first layer. Therefore I added my spare P.I.N.D.A V2 probe (from original MK3S) to the system. Thermistor of probe is simply connected where usually the thermistor for a second extruder goes (Note: for tests one can also zip-tie a spare thermistor to a standard probe).
Therefore, first changes were to add temperature readings for a probe thermistor.
When running G29, after each probe the temperatures will be read and according compensation values will be added to the measuread z-value. There are three tables (bed/hotend/probe), whereas each holds compensation values for a certain range of temperature.
In order to find these values I initially planned to simply copy the Prusa MK3 P.I.N.D.A calibration process. However, this process does not even work properly on my original MK3S and I honestly think Prusa does not care too much about this feature so I came up with my own but similar process. Also I found that different bed temperatures significantly effect first layer so I included it in the calibration process. For the sake of completeness I also added a table for hotend temperatures (which have to be set manually). (calibrate: G76, modify values: M871).

For calibration it is important that while one component gets calibrated, the others temperature is held constant. E.g. when calibrating probe, bed temperature is held constant (e.g. 100°C). First z-probing is always the base value (e.g. probe at 30°C), following probes will then calculate the delta (i.e. the relative error due to temperature change of probe). E.g. second probing (probe at 35°C) will measure a delta of 5um to base value, therefore if I ever do a bed leveling and my probe measures 35°C I would have to subtract this value from the measured_z. This relative errors are the compensation values that will then be stored in a table.
When calibrating bed temperatures, we first park the print-head in a position where the probe can cool down below 30°C, then move it to the probe point close to the bed and wait until it reaches 30°C, finally do the probing. Doing this over and over again allows to measure different bed temperatures by constant 30°C probe temperature.

Probe calibration:
Note: make sure fans are not blowing at probe or it will not heat up enough, we should at least reach 45-50°C to get enough data for a linear regression and extrapolation to be somewhat accurate

• Move probe to cooldown point.
• Heat up bed to 100°C.
• Move probe to probing point (1mm above heatbed).
• Wait until probe reaches target temperature (30°C).
• Do a z-probing (=base value) and increases probe target temperature by 5°C.
• Wait until probe reaches increased target temperature.
• Do a z-probing (delta to base value will be a compensation value) and increases probe target temperature by 5°C again.
• Repeats last two steps until max. temperature reached or timeout (i.e. probe does not heat up any further).
• Missing compensation values of higher temperatures will be extrapolated (using linear regression first). Note: While this is not exact by any means it is still better than simply using the compensation value of highest temperature in table.

Bed calibration:
Note: heating up the probe to 30°C should be no problem, even at a bed temp. of 60°C (otherwise shield probe from fans again)

• Move probe to cooldown point.
• Heat up bed to 60°C.
• Move probe to probing point (1mm above heatbed).
• Wait until probe reaches target temperature (30°C).
• Do a z-probing (=base value) and increases bed temperature by 5°C.
• Moves probe to cooldown point.
• Wait until probe is below 30°C and bed has reached target temperature.
• Move probe to probing point and waits until it reaches target temperature (30°C).
• Do a z-probing (delta to base value will be a compensation value) and increases bed temperature by 5°C.
• Repeat last four points until max. bed temperature reached (e.g. 110°C) or timeout.
• Missing compensation values of higher temperatures will be extrapolated (using linear regression first). Note: While this is not exact by any means it is still better than simply using the compensation value of highest temperature in table.

I´m not a professional when it comes to Marlin, in fact this was my first time changing more than just the config files. But since it was clear that this is not a feature to be implemented soon, I thought I would give it a shot. There might be errors in it or simply malfunction due to a different setup that I had not thought about. However, I run this code on an "almost clone" of a Prusa MK3S and had no problems so far.
In case there are any questions feel free to contact me.

Benefits

Adding support for z-probes with a thermistor (e.g. P.I.N.D.A V2) and a way to calibrate first layer temperature compensation values (will then be used with G29).

Related Issues

More consistent first layers at various bed and probe temperatures.
P.I.N.D.A V2 support in Marlin was already requested as a feature before.

[boelle]

now you just need to implement the prusa style of probing the mk42 bed and prusa printers can almost run marlin without missing important features :-D

of course all the selftests prusa has will be missing

[tompe-proj]

The probing of bed calibration points they do on setup is their way of skew correction, right?
I actually have a genuine MK52 on my Marlin printer so I might play around with it some day.

[boelle]

yes they do a zig zag search over each point, just high enough over the bed that it will only trigger when right over the center of each point

and they do multiple passes over each point so i gues on each pass they go a bit up

and they know where the points should be so they can compare them to where they are and figure XY skew and at the same time they also have done probing for height/level

or so i think it is, i'm not good at coding but might be worth looking in to their copy of marlin (old version based)

[tompe-proj]

Right, I can confirm that they in fact do change their z while running zig zag, I realized that when running my own setup. I think they start rather high and then go lower while running zig zag, this way you should get x, y, and z for each calibration point. Assuming the coordinate system is fixed to the frame, one can then find the bed plane in it... or rather the orientation (I think that is all you need for skew correction). However, I have not really thought about this in depth. I think it would make sense to take a peek at the Prusa sources first and then think about how to feed this data into the already existing skew correction in Marlin.

[tompe-proj]

In current commit of file probe_temp_compensation.h there seems to be a typo after code cleanup (line 43-45):
static constexpr temp_calib_t calib_arr_init[TC_COUNT] = { { 10, 30, 5, 10 + ( 30 * 5) }, { 10, 60, 5, 10 + ( 60 * 5) }, { 5, 180, 20, 5 + (180 * 20) } };
The end temp is calculated 30 + (10 * 5) and so on.

[GMagician]

samd51 needs more code to add probe adc reading...
hal.cpp

#if HAS_TEMP_ADC_PROBE
  #define GET_PROBE_ADC()         PIN_TO_ADC(TEMP_PROBE_PIN)
#else
  #define GET_PROBE_ADC()         -1
#endif

adc_pins array:
in
above // ADC1 pins

#if GET_PROBE_ADC() == 0
  TEMP_PROBE_PIN,
#endif

below // ADC1 pins

#if GET_PROBE_ADC() == 1
  TEMP_PROBE_PIN,
#endif

adc0_dma_regs_list array:

#if GET_PROBE_ADC() == 0
  { PIN_TO_INPUTCTRL(TEMP_PROBE_PIN) },
#endif

adc1_dma_regs_list array:

#if GET_PROBE_ADC() == 1
  { PIN_TO_INPUTCTRL(TEMP_PROBE_PIN) },
#endif

[thinkyhead]

samd51 needs more code to add probe adc reading...

That's outside the scope of this PR and should be submitted separately.

[thinkyhead]

The end temp is calculated 30 + (10 * 5) and so on.

Thanks for reviewing. Now patched!

[thinkyhead]

I tried to preserve all the same functionality with my patches, but there can always be more typos. So, of course, keep us posted and submit fixes for any issues that you find. The new G-codes will need to be documented in the MarlinFirmware/MarlinDocumentation repository before this is considered complete.

[tompe-proj]

I just added some doc files and opened a merge request. However, checks failed so maybe somebody can review my formatting. I don't know what caused it to fail.
https://github.com/MarlinFirmware/MarlinDocumentation/pull/269

[tompe-proj]

Somehow my link does not work when clicking but it does when copy-pasting. If neither does for you just search for request 269 in MarlinDocumentation.

[rado79]

Does G76 need automatic bed leveling activated? or does itworks independently.

[3d-gussner]

Thanks to getting this into Marlin.

Can you also please update RepRap Gcode Wiki that Marlin uses G76 and add M871 as well?

Looking into Prusa Firmware the #define TEMP_SENSOR_PROBE should be 1, maybe a note that PINDAv2 uses the "1 : 100k thermistor - best choice for EPCOS 100k (4.7k pullup)" would be helpful.

Question: How are the pins_xxx.h files updated?

I was missing this new feature especially in pins_EINSY_RAMBO.h and pins_MINIRAMBO.h as these are used in Prusa MK2.5/s MK3/s.

[tompe-proj]

Does G76 need automatic bed leveling activated? or does itworks independently.

G76 itself is standalone. However, currently the compensation values will only be taken into account when you run G29 (ABL).

[rado79]

This is a nice feature, thanks! Would be great to have it without G29, just to adapt z-offset.

[tompe-proj]

Would be great to have it without G29

It should be possible to move the temp-compensation from G29 directly to the function probe_at_point() in probe.cpp. Although the function would need an extra parameter to disable compensation for certain calls -> we also use this function when running G76 and it would mess with us. I'm not a Marlin dev so I don't know if this change could break other parts in code.

[boelle]

now that its fully working do you have any plans to add the mk42 and mk52 bed?

some initial work was made here: #13348 and there is a FR for it here: #6199

[tompe-proj]

@boelle I still do but don't want to guarantee anything. Was quite busy the last days but it got better again. I will start by checking Prusa code to get an idea of what they are doing exactly. Since Marlin already has skew correction it would probably be best to just estimate all values that would normally be measured manually after printing a test object.

[thinkyhead]

There seems to be some confusion in the code about the meaning of the probe / wait-for-heat position. Do you intend for PTC_PROBE_POS_X/Y to indicate the position to put the probe or the position to put the nozzle? It is unclear from the code what the intention is, but it seems to refer to the nozzle position and not to the probe position. See #16995 and #17005.

[npmomchev]

I installed Marlin 2.5.0.3. I put on the printer P.I.N.D.A. v2 sensor. I'm trying to find the temperature dependence of the sensor. After reading the bed temperature offset values, the G76 command moves the sensor outside the bed perimeter and heats the bed to the maximum temperature. The temperature of the probe is about 26 degrees. Then nothing happens. I know from the manual that I have to read the values manually. Can anyone help step by step how to do this.

[tompe-proj]

@npmomchev So if I understand correctly you run G76 and bed calibration completes successfully but when it starts the process for the probe calibration it just keeps sitting there doing nothing.

• When running probe temp-calibration it moves the probe outside the print area to cool it down and heat up the bed to max temperature at the same time. Only when the probe is below probe start temp (usually 30°C) and bed has reached target temp (PTC_MAX_BED_TEMP) it will continue. In Configuration_adv.h there you will find the parameter PTC_MAX_BED_TEMP, make sure your bed can actually reach this temperature. E.g. in case it is set to 110°C and your bed only reaches 100°C max your printer will wait until it runs into timeout and as a consequence fail the calibration process.

• Setting values manually means the following: after a successful calibration process, all values are stored in RAM only, so they are already active but after a reboot they will be lost again. In order to store them permanently run the M500 command to copy them from RAM into EEPROM. It is also good practice to read them first using M871 and check if the values actually make sense (e.g. there might be some outliers). You can also correct/set single values by manually using the M871 command. Once you are satisfied store them using M500.

You can find example values and the whole process of calibration here:
https://marlinfw.org/docs/features/probe_temp_compensation.html

Hope this helps.