I got the T-962 reflow oven so that I could perform SMT soldering for a larger project:
This reflow oven seems to be the most popular consumer reflow oven, because it is cheap. However it is very poorly designed (according to the internet) and requires some upgrades to work better. I only upgraded 3 aspects that I deemed necessary for this project:
When the reflow oven arrived, I was alarmed to find white cotton like fibers visible at various corners of the reflow oven. Worried that it was asbestos, I emailed the manufacturer, and they said it wasn't asbestos. In one YouTube video of someone replacing the masking tape, they also said that it was fiberglass. I can only hope they are correct and feel slightly more reassured.
This was one of the biggest complaint about the T-962 I read about online, as the masking tape stinks when the oven heats up. This upgrade was pretty straight-forward. I simply had to open the upper body of the reflow oven, tear out all the visible masking tape on top of the insulating material, and replace them with Kapton tape. There was one piece of hidden masking tape below the insulating material however at the front of the oven. If it weren't for existing YouTube videos, I wouldn't have known it was there.
Many, many people online suggested to upgrade the firmware by following this GitHub repo. I don't know much about how USB to serial stuff works yet, so I just connected the specified pins on the reflow oven controller board to the specified pins on the USB to serial adapter.
I first tried to follow the steps by reading the GitHub repo and watching YouTube videos with an old (and cheap) USB to serial adapter I got from AliExpress but it didn't work, so I got another one from HobbyPCB and tried again. The firmware successfully updated with the chip from HobbyPCB FTDI breakout board, but I'm not sure if it's because of the chip or because I was doing it wrong before. For example, I didn't know I was supposed to connect TX to RX and vice versa at first. Either way, it was working now and I didn't bother to try again with the old chip.
The five ISP pins on the controller board were in order from left to right (pin name on LPC2134 in parentheses):
The six pins from top to bottom of the FTDI breakout board was:
The microcontroller in the T-962 was a LPC2134, and I updated the firmware using the LPC2000 Flash Utility v2.2.3 with frequency 11.059MHz and baud rate 1440.
I had a hard time figuring out what cold junction compensation meant. Basically, the connection end of the thermocouple (wire that measures temperature) to the controller board is the cold junction, and the measuring end of the thermocouple is the hot junction. In the old firmware, the calculation of the temperature measured assumes that the cold junction is at room temperature, which it is not, considering it's on top of the oven. Therefore, a thermometer is needed to measure the temperature at the cold junction to properly calculate the temperature.
The instructions for cold junction compensation on the GitHub page is very opague. Thankfully, I found a YouTube video that explained what I needed to do in detail. A lot of the instruction on the GitHub page was shown in the picture, leaving the reader to deduce what to do from the picture, which was insufficient for people like me who has very little experience with electronics. I needed to buy a new temperature sensor form Mouser for this part.
Here's a breakdown of what this picture means (as explained in the YouTube video):
You may be wondering, how is the thermometer being powered? I looked at the datasheet for the DS18B20+, and it turns out it has a parasitic power mode that only requires 2 pins for operation (DQ and GND). Amazing stuff!
The new firmware from Unified Engineering gave users the option to calibrate the oven using two settings (gain and offset) for each thermocouple (left and right). The details of how these values are used to calibrate the measured temperature is detailed on this page in the GitHub repo: https://github.com/UnifiedEngineering/T-962-improvements/wiki/calibration. Basically, the measured value is multiplied by the gain and an offset is added to get the final value. The "actual" temperature displayed on the screen would then be the average of the two.
First thing first, I had to confirm which thermocouple was left and which was right. I couldn't tell just by looking at the connection to the circuit board since there were no labels, and the software could have conceivably labeled either as left or right. To find out, I disconnected one of them from the board and read the measured value on the screen while turning up the heat. The side that didn't change temperature would be the disconnected one, and from there I could deduce which was left and which was right. It turns out if you were facing the front of the oven (where the screen is), the side on the left is... left, and the side on the right is right. Surprising, I know!
Below is an image of the data collection process using a 2-channel thermocouple thermometer
To find out how bad the oven measures temperature, I bought a separate 2-channel thermocouple thermometer and placed the two terminals at approximately the position where the built in thermocouple terminals were. I set the oven temperature in 5 degree increments from 40C to 150C, and plotted the measured vs actual temperature on both left and right using Excel. Measured temperature refers to what the oven measures, and the actual temperature refers to what the external (more reliable) thermometer measures.
Note that since the temperature that the oven measures is very inaccurate, the actual temperature ended up being pretty far from the set point. The average error on the left was -25.17C, and the average error on the right was -13.02C.
It looked like the measured temperatures on the left side had the right slope, but was shifted up from the actual temperature by a large margin, so I set the L offset to -25.17. The value was selected by taking the average error on the left. The right side had the opposite problem. The measured temperatures seemed to be in the right ballpark, but was tilted upwards a little, so I set R gain to 0.86. The value was selected by eyeball... (Keep adjusting the slope until the line for the adjusted values and actual values overlapped pretty well).
To confirm that the calibration was good, I did another round of measurements by setting the oven temperature in 25 degree increments from 50C to 250C, and plotted the measured vs actual temperature on both left and right (again) using Excel.
Notice how after calibration, the actual temperatures are much closer to the set points because now that the oven can read the temperatures accurately, it could heat and cool accordingly with more accurate data. The average error on the left was -7.76C, and the average error on the right was 1.29C. A huge improvement from -25.17C and -13.02C!
With the reflow oven calibrated, I can now use it with the confidence that its temperature won't overshoot and burn my PCBs, nor would it undershoot and not properly reflow the solder paste.