I didn’t expect Gaoyun’s FPGA to be so durable
[Copy link]
I was recently designing a board using Gaoyun's GW1N FPGA. There were many components on the board, so I used a lot of 0402-packaged resistors and capacitors for layout convenience. I had hand-soldered 0402 components before, but this time there were more of them, and the multi-layer board dissipated heat too quickly, so I just used solder paste by hand. I didn't make a steel mesh because I needed to make changes to the board after adjustment, and I regretted it so much.
At the beginning, there was too much solder paste. When welding, the 0402 components floated on the board along with the solder oil, and finally gathered into a pile. The other chips were all soldered. After removing the 0402 components, I re-applied solder and arranged the components. After finishing, I heated them again and put the components that were stuck back in place. After repeated attempts, they were finally soldered. At this time, the FPGA and other chips on the board had been heated repeatedly many times, and it felt like they were hanging. After measuring that each capacitor was not short-circuited to the ground, I first measured the output voltage of the power chip and it was normal. After the power jumper was turned on, it was found that the adjustable power supply was overcurrent protected. Turning on the thermal imaging, the power supply stopped jumping after the current was increased, and the temperature of Gao Yun's FPGA instantly rose to more than 90 degrees. After powering off, we measured and found that the FPGA's core voltage 1.2V was disconnected from the 3.3V during welding, which means that the 1.2V core of the FPGA was supplied with 3.3V power. After the short circuit was disconnected, the FPGA could work normally when powered on. However, an STM32 chip on the board was not so lucky and stopped working after flashing the program several times. After replacing an STM32, it worked normally.
At this time, the whole board can work normally, because the FPGA with integrated PSRAM is used, and the BANK voltage shared with PSRAM can only be powered by 1.8V. The VDDIO of another acquisition chip is also powered by 1.8V, and the data is sent to the 1.8V i/o of the FPGA. During the test, it was found that the voltage that should have been 1.8V is now 2.6V. It is normal for the power supply to output 1.8V when it is no-loaded before, because the output voltage is greater than the set voltage. At this time, the switch power chip is not working. Remove the magnetic beads and measure the FPGA side separately. There is no voltage. 2.6V flows back from the acquisition chip side, because 2.6V and 3.3V are exactly the voltage drop of a diode. Careful inspection found that D0 of the acquisition chip is connected to the 1.8V bank i/o of the FPGA. This I/O is disconnected from 3.3V on the FPGA side, so 3.3V flows along D0 to the acquisition chip, and then flows to the VDDIO of the acquisition chip through the clamping diode of the acquisition chip. The FPGA's 1.8V bank power supply does not have 2.6V, which means that the FPGA pin should not have a clamping diode. The short circuit problem between FPGA 1.8V pin and 3.3V pin was solved and the 1.8V voltage returned to normal.
Why is Gaoyun's FPGA so durable? First, it has been heated for many times, and then it has been powered by 3.3V to power the 1.2V core, and then 2.6V to power the 1.8V psram bank. It can work normally even when powered by 2.6V. Finally, the level conversion used by the emulator did not work well when I was debugging. The measurement showed that there should be a problem with the level conversion chip. Later, I directly used the 3.3VI/O of the emulator to connect to the 1.8V of the FPGA, and it also worked normally.
| 
提升卡
变色卡
千斤顶
京公网安备 11010802033920号
