I've been working on an old power amplifier (SAE MK3CM) that has a problem with one channel. It works OK for a short time and then the positive side of the output waveform starts to clip, somewhere around 5V. Most of the voltages measure reasonably close to what the SM says they should, but one clue is my nose. I can smell something getting hot. So I decided to make myself a thermal IR camera to determine what's getting hot w/o burning myself or potentially connecting my body to up to 150V (the supply rails are +/- 75V).
I found a relatively inexpensive IR camera that has a 32 x 24 pixel array, the MLX90640. Sparkfun and Adafruit will sell you a board with one on it for about $70 -- and that doesn't include the display and Arduino-ish board needed to process and display the thermal image. However, I already had a Teensy 4.0 and a 320 x 240 TFT display so I just bought the camera chip from Digikey. The camera only has 4 pins -- Vcc, ground, SDA and SCL -- so I just soldered it down to a piece of perfboard.
Adafruit has some example programs for their version so I downloaded them and modified one for my particular display (it has about 4X the number of pixels than the one they use). And it works surprisingly well. So I cobbled together a setup to view the component side of my power amplifier and did find significant differences between the two channels. So here are a few photos showing my admittedly crude setup:
The above photo shows the display. The camera is attached to the tripod using a "C" clamp with some scrap 3D printing failure I had lying around. The amp is shown in the lower right side.
Above you can see the Teensy 4.0 and a better shot of the camera. The camera chip actually is on the bottom side of the board, because I installed it upside-down....oops...
The above image shows what the "good" side of the amp looks like, thermally speaking. The two hot spots are two cascode-connected transistors whose inputs are the front-end LTP's. In normal operation they should dissipate about 76mW.
This is a photo showing the faulty channel. At least one of the transistors -- which happens to be associated with the positive side of the output waveform -- clearly is running hotter. So I have narrowed-down my search by quite a bit. BTW, my hand, which is around 80F, only generates a yellow color so the hot spot is pretty warm.
One initially-challenging problem was figuring out just exactly WHERE those hot spots are. I tried using my finger to point at the hot spot but it was just too big. So I made a hot spot "wand". Just a 120 ohm 1/4 watt resistor taped onto the end of a wood dowel. I adjusted the power supply driving it so it doesn't overload the camera. And here's my high-tech "hot spot":
I found a relatively inexpensive IR camera that has a 32 x 24 pixel array, the MLX90640. Sparkfun and Adafruit will sell you a board with one on it for about $70 -- and that doesn't include the display and Arduino-ish board needed to process and display the thermal image. However, I already had a Teensy 4.0 and a 320 x 240 TFT display so I just bought the camera chip from Digikey. The camera only has 4 pins -- Vcc, ground, SDA and SCL -- so I just soldered it down to a piece of perfboard.
Adafruit has some example programs for their version so I downloaded them and modified one for my particular display (it has about 4X the number of pixels than the one they use). And it works surprisingly well. So I cobbled together a setup to view the component side of my power amplifier and did find significant differences between the two channels. So here are a few photos showing my admittedly crude setup:
The above photo shows the display. The camera is attached to the tripod using a "C" clamp with some scrap 3D printing failure I had lying around. The amp is shown in the lower right side.
Above you can see the Teensy 4.0 and a better shot of the camera. The camera chip actually is on the bottom side of the board, because I installed it upside-down....oops...
The above image shows what the "good" side of the amp looks like, thermally speaking. The two hot spots are two cascode-connected transistors whose inputs are the front-end LTP's. In normal operation they should dissipate about 76mW.
This is a photo showing the faulty channel. At least one of the transistors -- which happens to be associated with the positive side of the output waveform -- clearly is running hotter. So I have narrowed-down my search by quite a bit. BTW, my hand, which is around 80F, only generates a yellow color so the hot spot is pretty warm.
One initially-challenging problem was figuring out just exactly WHERE those hot spots are. I tried using my finger to point at the hot spot but it was just too big. So I made a hot spot "wand". Just a 120 ohm 1/4 watt resistor taped onto the end of a wood dowel. I adjusted the power supply driving it so it doesn't overload the camera. And here's my high-tech "hot spot":
Well done sir. Nice DIY!
I bought a Topdon unit, primarily to find high resistance connections in marine wiring, but quickly found out how useful it can be on the test bench and elsewhere. It's become a tool that I would replace immediately if it were lost or broken.
I bought a Topdon unit, primarily to find high resistance connections in marine wiring, but quickly found out how useful it can be on the test bench and elsewhere. It's become a tool that I would replace immediately if it were lost or broken.
As a follow-up, it appears that the hot spot was a symptom of the failure, the transistor was running hotter due to a resistive cold solder joint elsewhere in the amp. The transistor was running hot because the feedback loop was doing its darndest to follow the signal input.
Nice job. Remember exploring diy MLX90640 option for debugging few years back during COVI* but were all out of stock everywhere due to high demand. Finally settled with readily available seek thermal with higher resolution.