Results
The hot plate heats up fine but stays hot even after turned off completely, making it impossible to cool down on its own at a reasonable rate. This is due to the thermal mass of the plate which makes the plate retain heat even when the power is cut. Below shows a reflow curve that was collected using the Excel HID keyboard feature in the Reflowduino Demo code. The starting temperature was a bit high because I hadn't waited for the plate to cool down (yes, it takes a very long time!) after a previous use. The preheat temperature was set to 50°C, soak set to 80°C, and reflow set to 100°C for testing purposes. The rates of change were set to 2°C/s for preheat, 0.7°C/s for preheat, 2°C/s for reflow, and -4°C/s for cooldown.
As seen from the data, the temperature actually overshot past 100°C to about 113°C. The rate of temperature increase seemed to be a mere 0.46°C/s increase all the way up to the reflow temperature. The rate of temperature decrease was a teeny tiny 0.08°C/s (very slow!!!). Additionally, the following graph of the hot plate turned to HIGH and directly plugged in (without being controlled by the Reflowduino) indicates that this 0.46°C/s increase is the best the hot plate can do:
(As you can see, I didn't bother waiting a day for it to cool back down...) This data proves that this particular hot plate is not suitable for reflow because of its inability to heat up quickly and cool down quickly due to high thermal inertia. For cooking ramen noodles this might be OK, but for us this is a big no no. However, if you end up with a hot plate that exhibits similar thermal characteristics, one thing you might consider doing is to factor in the 13°C overshoot and subtract it from the desired reflow temperature. Although it takes a while to heat up and cool down, you would likely still get a decent reflow job, just much slower (especially the cooling part). For example, after reflow you might want to consider using a pair of tweezers to carefully move the PCB onto another surface to cool instead of waiting an hour for the plate to cool.
The toaster oven results were much better than the hot plate, mainly because of more responsive heating and cooling compared to the laggy hot plate with extremely high thermal inertia. The thermal profile below was for the test run from the videos in the previous page of this tutorial with the following parameters:
- T_preheat = 50°C at 2°C/s
- T_soak = 80°C at 0.7°C/s
- T_reflow = 100°C at 2°C/s
- Preheat PID gains: 100, 0.025, 20
- Soak PID gains: 300, 0.05, 250
- Reflow PID gains: 300, 0.05, 350
The kink in the graph was because I opened the toaster oven door and the thermocouple moved off the tray (but I put the thermocouple back on; hence, you can see the temperature warm back up shortly after.) I found that opening the door was the best way to cool the oven down at a reasonable rate. The results indicate that this toaster oven isn't capable of heating up the board at 2°C/s with the current configuration. However, so far I'm perfectly happy with its performance, because you can't go wrong with lower temperature change rates! I have yet to try insulating the oven (I probably won't) and moving the drip tray up one notch to the highest position (very close to the top heating filament!) as well as adjusting the PID gains to adjust things like the max overshoot. The max overshoot in this case was 5.75°C which I will try to lower by adjusting the PID gains.
In order to test the effects of the gains, I wrote the PID Test sketch to set a desired temperature value of 75°C and see the response for varying PID gains. The results are shown below:
Using lower P and high D gains surprisingly yielded more overshoot:
Using a lower P gain decreased overshoot but increased the settling time as expected:
From these results one may conclude that it might just be easier and more effective to simply compensate for the overshoot by subtracting a constant since the reflow temperature is at the very "tip" of the reflow curve.
Below are a couple closeups (sorry, not very good resolution!) of a board I reflowed using the Reflowduino and some low-temperature solder paste with the reflow temperature set to 155°C (this particular solder paste seems to reflow at a lower temperature than the typical 165°C recommended, and I observed that reflow movement started to occur at around 125°C rather than at the advertised 138°C "melting temperature"). The solder joints were perfect, and I couldn't have asked for anything better!
The results outlined above indicate that hot plates may not be very suitable for accurate tempertaure control, or at least require testing beforehand to determine the proper compensation amounts and PID gains. Since single-knob hot plates require absolutely no internal tweaking they could be a very cost-effective, compact, and easy solution for reflow control! All you need is some metal object (like a wrench socket) to press the thermocouple against the plate.
Toaster ovens are a much better solution when larger size, accuracy, and heating capabilities (rate of temperature increase) are required. They are also quite simple to tweak but require a little more experience and patience. The main goal is to disconnect everything from the neutral and live wires from the power cord and only connect the heater filaments directly to the live wire! Opening the door of the toaster oven after the reflow temperature has been reached is probably necessary to reduce the temperature at a reasonable rate. Insulation might increase the heating performance but the gains may not be worth the trouble! Also, I find it best to simply define a subtraction constant to compensate for overshoot during the reflow phase which will be subtracted from the target reflow temperature value. This code is included in the Reflowduino Demo sketch.
If you have any inputs, questions, or suggestions, feel free to ask them here on Github! Just open an "issue" and I'll take a look at it as soon as I can. Enjoy and share your projects with everyone!