Keith's Electronics Blog

Last Saturday I went shopping for lunch with a tape measure.

Delicious Spam™, fried up nice and crisp and served with Grannie’s “homemade” [what exactly does that word mean?] mustard. Yummmm!

The behavior of my CupCake during the first week of testing and parts-building had made me think that the nozzle was actually colder than the extruder controller believed it was — barely able to push plastic at allegedly 220°C and much happier at 230°C with not much scorching.

I had previously cross-referenced the extruder’s reported temperature with my infrared thermometer’s reading at room temperature (which I no longer trust) and had dipped the heated nozzle into a small pool of water to try to find the boiling point.

Condition	Thermometer Measured	Thermometer Converted	Thermistor Reported
room temperature	68°F	20°C	16°C
dipping nozzle into water and adjusting set temperature until water boils			105°C

In the 100-115°C range, the water would sizzle when the nozzle heater was on and stop boiling when it was off. It boiled most evenly between heating and cooling at a reported 105°C, so I was guessing that it thought 100°C was really 105°C.

This turned out to be incorrect; but it convinced me to perform a more proper thermal calibration, which is what counts.

Reference Temperatures

I believe I’m not the first to calibrate the RepRap / CupCake Plastruder temperature this way, but I haven’t seen detailed documentation on the procedure (nor have I tried particularly hard to find any).

The first step was establishing known reference temperatures at which to measure the thermistor’s resistance. Because I’ve come to suspect that my infrared thermometer is inaccurate — in some ranges, wildly so — the only reliable reference temperatures I had were the freezing and boiling points of water.

The Spam™ can was just the right size to fit between the support dinos and hold the water. (Yes, I went to the store specifically thinking I would buy Spam™ for the can. Yes, I really ate the Spam™. Yes, I really enjoyed it, although it was saltier than even I like.)

The PID-controlled hotplate is a red herring — its setpoint has nothing to do with the calibration. It’s just a handy way to heat water with relatively little risk of burning or melting my supports and in a convenient, controlled environment with no inquisitive feline assistance.

Tip: Mash the bottom of the can flat from the inside so there’s more surface area in contact with the heater. The first time I boiled this, it took forever.

Tip: Pine, not so much the best choice for supports when I have to crank the hotplate temperature so high.

Although the measurement I needed was the resistance of the thermistor out of circuit, I was curious and let ReplicatorG take a reading at boiling also, just to confirm that I would really be making a difference. Seeing 97°C boiling was an affirmation that I should be able to improve things.

I put the Spam™ can in the freezer until a significant amount of the water was frozen, broke the surface ice, and repeated the measurements.

Condition	Nominal Temperature	Reported Temperature	Thermistor Resistance
ice water	0°C	4°C	268KΩ
boiling water at 1500′ elevation	100°C	97°C	6.86KΩ

I actually don’t know the boiling point at 1500′, and I understand it’s more dependent on barometric pressure than on raw altitude. Please some chemist, physicist, meteorologist, or politician with good truthiness tell me it won’t have made much difference.

The extruder was reporting 4°C at freezing and 97°C at boiling. I hoped I could improve these readings — especially since not only were the temperatures off but the scale of temperatures was off. It was reading only a 93°C difference between freezing and boiling whereas it should have read 100°C; so at ~220°C operating temperatures it might be reading a full 10°C low.

Note that this is the opposite of what I had suspected — I thought it was reading high. Regardless of whether my suspicion was right or wrong, I wanted an accurately calibrated temperature before starting to tune Skeinforge to my CupCake’s needs.

Also note that the thermistor reading must be taken out of circuit. The rest of the extruder board becomes parallel resistance via the power supply leads while it’s in-circuit (yes, with the power off) and the readings are orders of magnitude lower and useless. (Tip: Don’t waste time finding this out yourself, like I, uh, did.)

Calculating Thermistor Beta

Next I needed to calculate the thermistor’s beta, which nophead describes in detail as a simple model of the nonlinearity of the thermistor’s resistance as a function of temperature — essential to know in order to correct for the nonlinearity and calculate the actual temperature.

I used the RepRap project’s document on measuring and calculating thermistor beta to calculate my beta. Assigning the necessary constants to my measured values:

T₀ = 0°C
R₀ = 268KΩ
T = 100°C
R = 6.86KΩ

And then plugging into their formula gives:

β = ln(R / R0) / ((1 / (T + 273.15)) – (1 / (T0 + 273.15)))
= ln(6.86KΩ / 268KΩ) / ((1 / (100°C + 273.15)) – (1 / (0°C + 273.15)))
≈ 3736

Which matches the answer that their calculator form generates.

I wouldn’t have bothered to do the math myself except the formula is higher in the page than the calculator and I had already done the work before I found the automatic version. Drat. Well, it was a good validation.

Creating a New Temperature Lookup Table

The extruder controller uses a temperature lookup table to cross-reference the A/D thermistor reading to temperature values and interpolate between nearby points.

I downloaded the createTemperatureLookup.py utility from the RepRap repository. I haven’t looked exhaustively and it may already be available in the RepRap code set, but I didn’t find it there.

The utility is invoked from a command line like so:

./createTemperatureLookup.py --r0=268000 --t0=0 --r1=0 --r2=4700 --beta=3736 --max-adc=1023 > ThermistorTable.h

where r0 is the measured resistance (268KΩ) and t0 is the temperature at the measured resistance (0°C), r1 is the value of the resistor in parallel with the thermistor and r2 is the value of the series resistor (see below), beta is as determined above, and max-adc is the largest reading of the analog-to-digital converter.

The schematic for my v2.1 extruder controller confirms that the circuit has no parallel resistor and that the series resistor is 4.7KΩ. The Python script has code for a special case of r1=0 to indicate the lack of a parallel resistor (rather than having to enter r1=∞). These values for r1 and r2 match the values that were used to generate the default RepRap code for the Plastruder, so I knew I was on the right track.

The official firmware, by the way, uses r0=100000 and t0=25. That sounds awfully like datasheet values to me, which is a perfectly reasonable decision for a default firmware configuration; but the manufacturing variation of the thermistor makes it not a close enough choice for my extruder.

Here’s the output from createTemperatureLookup.py:


// Thermistor lookup table for RepRap Temperature Sensor Boards (http://make.rrrf.org/ts)
// Made with createTemperatureLookup.py (http://svn.reprap.org/trunk/reprap/firmware/Arduino/utilities/createTemperatureLookup.py)
// ./createTemperatureLookup.py --r0=268000 --t0=0 --r1=0 --r2=4700 --beta=3736 --max-adc=1023
// r0: 268000
// t0: 0
// r1: 0
// r2: 4700
// beta: 3736
// max adc: 1023
#define NUMTEMPS 20
short temptable[NUMTEMPS][2] = {
{1, 1108},
{54, 280},
{107, 225},
{160, 196},
{213, 177},
{266, 161},
{319, 149},
{372, 138},
{425, 128},
{478, 120},
{531, 111},
{584, 103},
{637, 95},
{690, 87},
{743, 79},
{796, 70},
{849, 60},
{902, 47},
{955, 31},
{1008, -1}
};


And comparing the original lookup values to the newly-generated ones:

ADC	Old Temp (°C)	New Temp (°C)
1	841	1108
54	255	280
107	209	225
160	184	196
213	166	177
266	153	161
319	142	149
372	132	138
425	124	128
478	116	120
531	108	111
584	101	103
637	93	95
690	86	87
743	78	79
796	70	70
849	61	60
902	50	47
955	34	31
1008	3	-1

The important line is that at an actual (or at least newly-calculated) operating temperature of 225°C, the code previously on my extruder thought the temperature was only 209°C. I was actually way off in thinking that I was running too cold; I was running quite a bit too hot (if the new table is to be believed).

Incorporating the New Temperature Lookup Table

Next I needed to merge the new temperature table into my firmware. I was working on a different computer than the one on which I had originally set up the CupCake, so I downloaded a fresh copy of the RepRap generation 3 firmware for the MakerBot 1.1 – 1.2 and extruder controller 2.2 from the Google Code repository (not the SourceForge repository).

I unpacked the code, went into the ArduinoSlaveExtruder directory, replaced ThermistorTable.h with my new table, loaded up the ArduinoSlaveExtruder program into the Arduino IDE, and compiled and downloaded it to the extruder board (using the Arduino board type — on the CupCake, only the motherboard is a Sanguino).

Retesting

Finally, I ran another set of hot and cold cycles with the Spam™ can.

Condition	Nominal Temperature	Reported Temperature
ice water	0°C	1°C
boiling water	100°C	99°C

Not perfect but much better (improved from 4°C and 97°C) — and since ReplicatorG only appears to report with 1°C precision, pretty darn acceptable.

Dripping Gunk

Addendum:

I forgot to include this part last night.

By the time I was done with the re-test of temperature readings, this sap-like drip had formed on the extruder nozzle, on the edge below the thermistor. I’d had the end of my ceramic insulator “blanket” and kapton tape sitting in the boiling water, and I don’t know whether the water interacted with the thermistor, the insulator, or the tape.

After taking the nozzle out of the water and cleaning off the “sap,” I turned on the heating element and slowly raised the temperature for a while, to heat off and then boil off any water remaining in the assembly. I didn’t see when it happened, but I later found these drips on the build platform.

And the first plastic to extrude from the nozzle after I was done was gross, which makes me think some of the black drippiness may have come from inside the nozzle. Any ideas?

Lesson: Only dip the nozzle tip, and not any other part of the heating assembly, into water for calibration. I think.

This entry was posted on Friday, September 11th, 2009 at 09:48 pm by Keith Neufeld and is filed under Arduino, CNC, Equipment, MakerBot CupCake. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.