Choose a compatible DIP conversion board

The LTC3600 comes in a 12-pin MSOP package. The pin pitch of the 12-MSOP is 0.45mm, which is much smaller than the pitch of a universal board, so a DIP conversion board is required to mount it. The conversion board adjusts the pitch to match that of a universal board. Thinking that this was a convenient product, I tried to purchase a compatible DIP conversion board on an e-commerce site, only to find that it was "discontinued." Reluctantly, I resorted to a brute force strategy of using a 20-pin conversion board that the company had with 8 extra pins.

​DIP board with extra 8 pins

As a result, the wiring was stretched unnecessarily on the conversion board, the distance between the IC pins and the peripheral components became too far, and it was not possible to output a good output voltage. No matter how many peripheral components are placed directly below the pins of the DIP conversion board, it is pointless if the wiring on the conversion board becomes too long. At first glance, it may seem like bigger is better, but you need to be very careful with DIP conversion boards.

Also, because the LTC3600 used this time is a current feedback type, I think it was more sensitive to noise and susceptible to influence than other types. Trying to take the easy way out backfired. It was a good lesson that we should not be obsessed with short-term profits.

Voltage feedback method (FB resistor omitted)

Current feedback method (FB resistor omitted)

Too thin GND

There are various possible causes of the damage, but we believe that the first three were definitely caused by stretching the GND wiring to a long, thin shape.

This time, I used a cheap universal board, which had a major flaw: when removing a component that had been attached, the via plating would come off along with it, making it impossible to re-solder. So, to hedge the risk of damaging the IC, I used pin sockets to make the DIP conversion board removable. I thought to myself, "What a great idea!" But then I ended up destroying three power supply ICs because of these pin sockets.

Too thin GND

At the time, I didn't know what I was doing wrong, so I clutched the board and went to a veteran colleague for advice. He told me with a big smile, "This is going to break. No matter how you look at it, the GND wiring is too thin." Thinner wiring increases parasitic inductance, which can cause voltage spikes and destroy the IC.

When designing a power supply, make sure to use thick and short GND wiring.

Be careful when you get used to it

Based on advice from my senior colleague, I designed the GND to be thicker and shorter, redesigned the board, and succeeded in outputting 5V with no load. However, I had an issue where the output voltage dropped to 0V when I drew a load of about 700mA. Fearing that I might damage the IC, I disconnected all the probes and recorded the results. To find the cause, I reconnected the probes to check the waveforms from other pins, and I heard a "snap" and the IC became very hot.

I wondered what was going on, and when I looked at the wiring, I realized that I had mistakenly connected the electronic load input to the GND check pin. Even though I had succeeded in outputting 5V for the first time, destroying the IC due to human error was a huge blow not only to the IC but also to my mental health. From this point on, I couldn't get power supply design out of my head, and eventually I even started designing power supplies in my dreams. It was essentially a 24-hour workday. I felt terrible at the time.

I vowed to myself that I would never again feel this way, and to always make sure to check things even when doing tasks I was familiar with.

Soldering iron temperature

Although I had artificially destroyed the IC, I was able to output 5V, so if I create a circuit in the same way, I think this time I'll finally be able to complete the power supply circuit! So I changed my mind and created the power supply circuit again from scratch. This time, I wasn't able to draw a sufficient load, so I decided that since I was going to rebuild the circuit anyway, I might as well change the capacitors, inductors, and resistors to chip types and make the layout smaller! I made some changes to the circuit. This should allow me to draw a load, so I thought I'd finally be able to complete the power supply! However, the reality was not so simple.

The remaining two​ ​ICs were not switching at all, and there was an issue of not outputting 5V. This is thought to be due to damage to the SW pin, but the exact cause of the damage is not clear. However, one possible cause is that the soldering iron temperature was too high, exceeding the rated temperature.

Thinking that a higher soldering iron temperature would make the circuit easier to solder and melt the material faster, I set the temperature to around 400 °C. When I checked the datasheet, the absolute maximum rating listed 300 °C for up to 10 seconds. This is way over the rating. I believe that heat exceeding the rating caused accumulated damage to each pin, which led to destruction as the IC was operated.

I actually left the circuit unchanged, but changed the soldering iron temperature to 250 °C and created the power supply again, and was able to confirm that it was switching. However, 5V was not being generated at the ISET pin, and when I checked the VOUT pin, only about 2V was being output. The cause of this is not clear, but as mentioned in the DIP conversion board section, I think it may be due to noise.

Finally considering changing IC

After making repeated improvements and creating seven power supply circuits, we were still unable to create a power supply that could output a stable 5V, which was affecting the schedule for completing the project, so we changed the model number to the LT8609, which uses a voltage feedback system and a compatible DIP conversion board was available.

By utilizing what I learned from designing the LTC3600 power supply, I made the GND wiring thick and short, and designed the layout to be as small as possible, and was able to complete a 5V output power supply without damaging a single​ ​IC.