search

eez-open / modular-psu-firmware

Firmware for EEZ Bench Box 3 (BB3)
https://www.envox.eu/eez-bb3/
GNU General Public License v3.0
94 stars 42 forks source link

DCP module faults after running on function generator #231

Open fietser28 opened 3 years ago

fietser28 commented 3 years ago

Running FW 1.7-RC1.

Running the function generator on Channel 1 (DCP405) I get a module fault. I tried this a couple of times it happens between a few seconds after starting and 2-3 minutes. The BB3 and other modules keep working. A software restart solves the issue: Module eror2

The offending function generator setting: Module eror2-func

The output of the module looks like (nothing strange is happening): Module eror2-scope

No load (except for scope probes) was attached. Down programmer was on, current range was best. Trying this with a 100ms period I couldn't reproduce the cash within 15 minutes. Ethernet connection was active. DCP405 is a R2B11.

prasimix commented 3 years ago

Ok, that is a clearly result of newly introduced protection #202. You don't have enabled neither SW or HW OVP. What is the connected load?

prasimix commented 3 years ago

Hm, if load not connected, and you didn't move probe, somehow ~3% overvoltage is detected. But signal on the scope looks fine. Please try to move lower level above 1 V, and see what will happen.

prasimix commented 3 years ago

I managed to reproduce this on my side. It seems like a bug in #202. When you have the opportunity, just try to increase the min. value from 1 to 1.1 V and see how it behaves. I think it incorrectly detects overvoltage at a lower value.

I've conducted more tests. It seems that actually everything below ~1.5 V will generate such error more often. Question is also do we really need to put module in error mode or just switch its output off.

fietser28 commented 3 years ago

I tried several variations: increasing/decreasing the lower limit, increasing/decreasing the amplitude, adding a capacitor and or resistor as load, but I can't find a consistent pattern when it happens more frequently. To really find it requires repeating (and writing down) but that might take some time. All without OVP.

On the scope the timing of the fault (output going to zero) always has the same timing pattern: around 5 ms after falling edge.

Looking at the code introduced in #202 I'm wondering if noise plays a role here: A single sample might be off and at lower voltages this is larger in % and more likely to trigger (at 1V, 3% equals 30mV)?