killed my onyx rev2 bed?

[gschora]

So I was upgrading to 24v today:
got my meanwell psu with 24V/27A (dialed it down to 20V)
a crydom 100A ssr
wired everything with 2*15AWG wires.

So everything worked fine, heated from 23°C to 95°C in about 3 minutes with bangbang-mode enabled, and therefore I did the pid autotune to switch to pid.
Problem was that my psu has a fan that is activated either when the psu gets over 50°C or when the load is above 35%. So the fan was always on and really loud when the bed was heating.
I did play around with the "Bed Pid Max Value" and dialed it down to find the sweetspot where the fan isn't always activated.
Suddenly the fan of the psu wasn't activated anymore and the bed had trouble maintaining 85°C.
I first thought I'd killed the ssr or the psu, but voltage measurement seemed right.
So I disconnected the heated bed to measure the resistance and it had 3.2 Ohm.
I then took everything apart to measure resistance directly on the solder-pads and it has 2.6 Ohm, which would be a bit high IHMO.

So is it possible that I've killed my onyx somehow?
Any suggestions?

[gschora]

So I hooked everything up again and dialed the psu back up to 24V to see if it works with 3Ohm also.
The bed was heating, but the temperature in the middle where the thermistor is wasn't rising.
When I checked I found that the bed had a cold spot in the middle with a diameter of approx. 10cm and on the outside it was getting pretty hot.

So this looks like that somehow the pcb-traces in the middle are broken.
Am I right?

[jdurand]

It's all one circle so if it was broken it wouldn't heat anywhere. What's probably happened is the board delaminated. Putting 576 Watts into a board that size gives you about 6 Watts/square inch and that can EASILY pop the traces loose. The pc board material including glue is rated for 140C continuous so it's not too hard to damage.

Reason: There's a lag time for the heat from the main part of the board to work its way from the copper down to the thermistor meaning the copper can be a lot hotter than you think. The trace expands rapidly with the temperature rise while the PC board doesn't move quite so much. This puts a large strain on the glue that holds the copper to the board and as the glue is probably already over temperature at this point, it lets go.

Now you have an air gap between the board and the copper. This removes the heat sink of the board from the bottom face of the copper so the delaminated bit gets even hotter.

To prevent this the PID should have two stages, one has a large D component to make the surface heat SLOWLY so it all gets there at the same time. Once in the regulation band it would then switch to the normal D parameter.

A product I'm designing now has a heater in it that could damage the item being heated if it overshoots so I'm doing the PID much the same way. Intentionally slowing down the initial heating so there is no spike. Have also done that with liquid valves, I designed a control system for a factory and the 10" proportional valves could open/shut in a small fraction of a second. If you know what water hammer is, you understand the problem of overshooting in either direction!

[gschora]

Thanks for the info! Really appreciate it.

So in short I need a new bed?