Tuning the heater temperature control

This document is relevant to: all Duet boards

Firmware versions: RepRapFirmware 1.15 and later

Each heating controller in RepRapFirmware 1.15 and later can work either in bang-bang mode or in PID mode. When operating in PID mode, it can get its PID parameters in either of two ways:

• Use two sets of PID parameters computed from a first-order + delay time (FOPDT) model of the heating system. The FOPDT model parameters may be derived using auto-tuning, or configured manually using M307.

• Use one set of legacy PID parameters configured manually using M301. This provides backwards-compatibility with previous versions, and provides a fallback if the behaviour of your heating system cannot be approximated well enough by a FOPDT model.

PID parameters are set in config.g manually, or in config_override.g when saved with M500. If there are no heater definitions or parameters in config.g:

• In RepRapFirmware 3.x, no heaters are defined by default; set up the temperature sensor and heater with M308 and M950 respectively, then set PID parameters with M307.

• In RepRapFirmware 2.x and earlier, the bed heater controller defaults to bang-bang mode and the extruder heaters use the default PID parameters.

If you use the M301 command to set the legacy PID parameters, the controller for that heater switches over to use them. If you run a successful auto tune or you set the model parameters manually using the M307, the controller switches back to using model-derived PID parameters.

To see which set of parameters is being used, run M307 H# where H is the heater number, and that will tell you either that the model is in use (so model-based PID parameters are being used for that heater), or not in use (so legacy PID parameters are being used).

Each heating controller performs temperature monitoring to try to detect fault conditions such as a heater of thermistor falling out of an extruder heating block. The temperature monitor relies on the model parameters to decide what is reasonable behaviour. Therefore, even if you decide to use legacy PID parameters or bang-bang control, you should still define an approximate model of the heater to provide good heater monitoring without false alarms.

Auto tuning is initiated by the M303 command:

M303 H1 S240 ; auto tune heater 1, default PWM (100%), 240C target

M303 ; report the auto-tune status or last result

[invalid guide link]

RepRapFirmware 3.2 introduced a new heater tuning algorithm. This algorithm is more accurate than the old one (especially in measuring the dead time), often completes more quickly than the old algorithm, and is more portable to expansion and tool board firmware (auto-tuning of heaters connected to Duet 3 expansion and tool boards is implemented in RRF 3.3). It also allows for tuning a tool, rather than just a heater, which tunes the heater with related fans both off and on. The purpose of this is to allow the heater control to implement feedforward, which monitors fan PWM changes and adjusts the heater power, before the PID algorithm recognises that something has changed.

M303 T0 S205 ; tune the primary heater of tool 0

Additionally, the new algorithm allows for the setting of the ambient temperature (A parameter) for tuning a heater that has been on and has not cooled down to ambient temperature. The tuning cycle hysteresis (Y parameter) and Fan PWM (F parameter) can also be set.

When tuning, auto tuning will perform a number of phases (heating, settling, fan off, fan on), cycling between the target temperature and 5C below it. It measures the heating and cooling rates on the last 3C each cycle, and accumulates the mean and standard deviation. Typically it will cycle heating/cooling 8 times, but can be as many as 30, if the readings are not consistent (noisy thermistor, or a large thermal mass). You can use the Y parameter to alter the hysteresis.

If you already have working values for M307 from using the old versions of autotune, there is no particular need to rerun this new version of auto tune (though it may improve temperature stability). Just use your existing M307 values.

The S parameter is the temperature to heat up to. The default value depends on which heater you are tuning. During auto tuning, the heater will be run at the specified power until this temperature is reached and then it will be switched off; so the temperature will overshoot the target somewhat

There is also an optional P parameter, which is the PWM value to use. Unless your heater is greatly over-powered, use the default PWM value of 1.0.

Warning! During auto tuning there is minimal protection against heating faults. Therefore you should not leave the printer unattended during auto tuning.

Warning! Tuning will always overshoot the target temperature. You need to be aware of this when choosing the target temperature. The heater itself is switched off when the target temperature is reached, so the overshoot is purely due to excess thermal mass in the heater and poor coupling to the thermistor.

Only one heater may be auto tuned at a time.

Before commencing auto tuning, the heater to be tuned must be at or near room temperature and its temperature reading should be stable. The exception to this is if you use the A parameter in M303 (RRF 3.2 and later).

[RRF 3.2 and later] When tuning a hot end heater, move the nozzle close to the bed, i.e. within 1mm. The tuning will then compensate for the effect of the hot end fans (hot end cooling fan, and part cooling fan if tuning the heater as a tool).

Send M303 with an H parameter (and optionally P and S parameters) to start the auto tune process. A message will be generated when auto tuning is completed or abandoned. You can also run M303 with no parameters to see the current status of auto tuning.

If auto tuning is successful, new model parameters are set (but not saved) and the PID parameters computed from them are used. You can see these parameters by running the M307 H# command, where # is the heater number M307 will also indicate that the model is in use, meaning that the PID parameters displayed by M307 are used, not the PID parameters displayed by M301.

Tuning a hot end heater typically takes between five and ten minutes. Tuning a bed heater may take more than half an hour, depending on the thermal capacity of the bed. You can cancel tuning by sending M0.

After you have run auto tuning and checked that the heater control is working well, save the heater parameters. There are two ways to do this:

• If you have a M501 command at or near the end of config.g to run the config-override.g file, run M500 to save the heater parameters in config-override.g.
• Otherwise, copy the M307 command that was suggested when auto tuning completed into config.g, replacing any existing M307 command for the same heater.

Make sure there is no M301 command for the same heater after the M307 command, otherwise the model will not be used.

Following a successful tune with M303, you may still get this error when heating up the bed.

• Make sure you have copied the settings of the successful M303 heater tune to config.g, or they have been saved to config-override.g with M500
• Copy the M307 line from config.g or config-override.g and paste it into the command line of DWC. e.g.

M307 H0 R0.300 K0.572:0.000 D8.00 E1.35 S1.00 B0

• Reduce the value of the R parameter and increase the value of the D parameter a little, e.g.

M307 H0 R0.200 K0.572:0.000 D10.00 E1.35 S1.00 B0

• Press enter. Copy the new setting back to config.g, or send M500 to save changes to config-override.g
• Check behaviour of bed heating, and adjust again if you still have problems.

In RRF3 there is no longer a default heater model assumption used for the bed. Therefore, you must perform a PID auto tuning to give the firmware a model for bed heater performance. Failure to do so will result in an error message that the heater temperature is not rising fast enough and therefore triggering thermal protection. Running the PID auto tune for the bed and saving the results should resolve the error.

• If auto tuning fails with a message that the temperature is not rising fast enough, this indicates that either you are using too low a P value in the M303 command, or the dead time in the existing model is much too low for your heater. You can increase the dead time using the M307 command. For example, sending M307 H0 D30 will increase the dead time to 30 seconds. The actual dead time will be known after a successful auto tuning.

Model parameters can be changed and reported using the M307 command. Examples:

M307 H1 A350 C139 D5.5 B0 ; set model parameters for heater 1 and use PID mode

M307 H1 ; report model parameters for heater 1, and whether the model is being used

The H parameter specifies the heater.

The A parameter is the model gain, which is the ultimate temperature rise divided by the PWM fraction. For example, a gain of 350 means that at a constant 50% PWM, the temperature would eventually reach ambient temperature plus 350 * 0.5 degC.

In RRF 3.3 and later, the A parameter is replaced by the R parameter, which is the heating rate in degC/sec at full power when the heater temperature is close to ambient.

The D parameter is the dead time, which is the delay between a change in PWM and an appreciable effect on the rate of temperature change.

The C parameter is the model time constant. This may be thought of as the time taken for the temperature rise to reach 63.2% of its ultimate value after the heater is switched on from cold at constant PWM, less the dead time.

You can limit the PWM by adding a suitable S parameter, for example S0.8 will limit the PWM to 80%.

To use bang-bang mode instead of PID, change B0 to B1. In bang-bang mode, the S parameter is still used to limit the PWM when the heater is turned on.

• If necessary, use M307 to set some parameters that let you achieve a steady temperature. The default ones may work.
• Set the heater to a temperature you typically use and wait until the temperature is stable. It doesn't matter if it is still creeping up slowly, but it must not be oscillating (if it oscillates, increase the M307 D parameter).
• Send M573 P# (where # is the heater number) to report the average PWM.
• The gain is: (actual_heater_temperature - ambient_temperature)/average_pwm.

Turn the heater on from cold, wait a few seconds for the temperature to start rising, then time how long it takes for the temperature to rise by a further amount e.g. 10C (for a slow bed heater you might wish to use a smaller amount e.g. 5C). Divide that temperature rise by the time in seconds to get the heating rate.

With the heater hot at a steady temperature Tstart, calculate the following target temperature:

Ttarget = Tstart * 0.37 + Tambient * 0.63

Turn the heater off and time how many seconds it takes for the temperature to drop to Ttarget. That is the time constant.

If necessary you can make manual adjustments to the M307 model parameters, as follows:

• If during initial heating the temperature undershoots the target, reduce the A parameter ((RRF 3.2 and earlier) or increase the R parameter (RRF 3.3 and later). If it overshoots the target, increase the A parameter or reduce the R parameter. Try increasing/reducing it by 5% or 10%.

• If the response to temperature changes caused externally (e.g. the print cooling fan turning on) is too slow, reduce the D parameter. If the temperature is not stable but oscillates around the target temperature even when the print head is stationary, increase the D parameter. Try increasing/decreasing it by 30%.

This mode is intended as a backup, for use if model-based tuning is not working well enough. The parameters are configured using the M301 command. Example:

M301 H1 P10 I0.2 D50 T0.3

The H parameter is the heater number. Usually, 0 is the bed heater, 1 is the extruder 0 heater, and so on.

P, I and D are the standard proportional. integral and differential coefficients, scaled by 255 for compatibility with older firmware. A negative P value means use bang-bang control.

Previous firmwares also had B, S, T and W parameters in the M301 command. These are no longer used.

Even when using legacy PID parameters, the heater model set by M307 is still used by firmware to detect whether the heater is behaving as expected and to raise a heater fault if not.

If the firmware detects a temperature anomaly, it will turn off the corresponding heater and put it in the Fault state. This can be seen in DuetWebControl, which will show the status of that heater as "fault" instead of "active", "standby" or "off".

You can reset a heater that is in the fault state using the command M562 P# where # is the heater number.

When a heater goes into the fault state, an error message is generated giving the reason. You can view the message in the Console page of DuetWebControl or PanelDue. The most common reasons are:

• Temperature reading was too low for several consecutive readings, suggesting that the thermistor has become disconnected
• Temperature reading was above the limit set by M143 for several consecutive readings
• A thermocouple or PT100 interface chip reported an error for several consecutive readings
• During initial heating, the temperature was not rising as fast as expected. This can occur if you have a slow or low-powered heater and you have not tuned the heater model, so that your heater has a greater dead time or a lower gain than the default model
• The heater reached the target temperature, but the temperature subsequently varied by more than +/-10C. This can occur if you are using unsuitable PID parameters, or if you turn on a very strong print cooling fan that cools the nozzle excessively.

There are three ways of using two thermistors to control one heater:

1. Connect them in series. The temperature readings will be averaged with a bias in favor of the colder one. If any wire breaks, you will get the usual -273C reading and the heater will fault.
2. Connect them in parallel. The temperature readings will be averaged with a bias in favor of the hotter one. If one of the thermistor wires breaks, the temperature will under-read, so you will end up heating too much.
3. Use one to control the temperature, and the other as an over-temperature cutout, configured using M143.

Note: if connecting 2 in series, the resistance of the thermistors needs to be combined in config.g. So if each thermistor is 100k at room temperature, they need to be set as 200k when declaring the sensor with M308. And if connecting in parallel, the resistances need to be half their combined average (or just half of one if both are the same).

However, when connecting thermistors in series or in parallel, although the B value remains the same, the effective C value changes. So it's easier to adjust the R parameter instead. For example, on a Duet 3 the normal R value is 2200. If you connect two identical thermistors in parallel, you can specify the usual T, B and C parameters for the thermistors, but double R to 4400. Likewise, if connecting two thermistors in series, halve the R value.