ESP32 life-time

[svenbieg]

Regarding "erased flash too often", are we talking >100,000 erase cycles here?

Yes, i did save the temperature-curves every minute erasing flash-sectors. The chip did still work, only the curves were gone after about half a year, that means about 100.000 write-cycles like in the specs. This was my fault, like i said. I'm going to use the DevkitC this year, i can post an update when it stops working.

[ESP_Minatel]

I'm using a voltage divider with the 3.3V supply-voltage, the relais are driven with 5V by an ULN2003. I just wanted to know if this a known problem of the ESP32. I'm going to try the DevkitC, maybe the chip got too hot when i was soldering it on the adapter.

Hi,

Are you using the chip itself (i.e ESP32-D0WD) or a module (i.e ESP32-WROM)? If it's the chip, did you reflow it manually with a hot air station for a long time in high temperature?

For this kind of problem, we can't help a lot without any detailed information about your circuit and the environment that it's connected.

Things that you can check:

• Power supply quality/noise

• Voltage spikes on the analog side

• Relay protection/noise (should measure the 3V3 rails while switching the relays)

It's to way complicated to give you a solution, but the ESP32 is very reliable and I've designed several industrial grade solutions using ESP32 without any problem. For any embedded system in rough environment, you'll need to design some protections in your circuit, sometimes using TVS, inductors, ferrite beads, PPTC, diodes, etc.

[svenbieg]

I don't get spikes on the analog-pins, it is a simple voltage divider with 3.3V supply voltage. And the relais are driven with 5V. If the supply voltage would drop the chip would restart immediatly. I think the soldering maybe the problem, i put a WROOM-SoC on an adapter with a soldering Iron. Thank You for Your reply, good to know that it is my fault!

[svenbieg]

I'm now using the DevKitC v4 but the ESP32 still doesn't work over half a year. I don't think i got over-voltage because the chip wouldn't work that long then. I'm using a usual 5V smart-phone charger. Maybe there is a difference where i order the DevKit? Is there anybody having the same problem? The ferrite bead seems to be a good idea. The original controller had problems too, maybe it's the 230V power supply. I'm going to try this next!

[mbratch]

I'm using a voltage divider with the 3.3V supply-voltage, the relais are driven with 5V by an ULN2003. I just wanted to know if this a known problem of the ESP32

Not a known problem. Are these custom boards that you hand solder the ESP32 on yourself?

I'm going to try the DevkitC, maybe the chip got too hot when i was soldering it on the adapter.

Sounds like a good plan for a good reason.

[Kaisha]

I have an esp8266 that's been running for 4y straight now, no issues. I don't think it's the chip failing. Much more likely you're doing something incorrect.

[pinkeen]

Regarding "erased flash too often", are we talking >100,000 erase cycles here?

Yes, i did save the temperature-curves every minute erasing flash-sectors. The chip did still work, only the curves were gone after about half a year, that means about 100.000 write-cycles like in the specs. This was my fault, like i said. I'm going to use the DevkitC this year, i can post an update when it stops working.

Increase the flash lifetime by using some form of wear leveling - either automatically by using SPIFFS/FAT32 or implement a simple flash-based circular buffer.

[svenbieg]

My program was running in an endless-loop with no delay, so the CPU-load was 100% all the time. Maybe this caused the problem.

[vanBassum]

I'm using a voltage divider with the 3.3V supply-voltage, ....

Wait what? If I understand you corectly:
You have 5V supply, and you use a voltage divider to supply the ESP with 3.3V?
That would be an absolute NO-GO.
But then again, maybee I understand you incorrectly.

As others have noted, a schematic would help us to help you.

If your hardware is OKE, it should be inpossible to kill the chip using software. The thing should be able to handle 100% cpu load for a very long time.

[svenbieg]

You did not read correctly. I'm using the voltage-divider for the Pt1000. As supply i have a standard contstant voltage-regulator. I had the same problem when using the DevKitC, looks like it is the software.