EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: helge on October 19, 2014, 10:38:27 am
-
Yesterday I built a few test boards featuring the MAX1968 TEC controller (see picture). It may have been the leaded but non-inductive 0.1R shunt resistor that I soldered to the board (in a second and also failed attempt I replaced it with a 0.05R one) or the lack of a small cap over the current sense terminals but that bloody chip just was unstable (creating audible noise) and destroyed itself after some time.
It appears that some of the internal switches got damaged during start-up or normal operation. But why? Is there a fundamental flaw in the layout?
Here are my test scenarios:
2x 22µH 3A inductors, 0.1R shunt resistor (current limited to +/- 1.5A)
supply: 5.5V
switching frequency: 500kHz
peltier device: 12V 6A cheap chinese peltier labeled "TEC1-12706" mounted to a CPU heat sink
power-up with peltier, setpoint voltage = internal reference (no nominal peltier current)
observations: current draw with and without peltier is around 0.02A and subsequently +/- 1.2A can be reached by attaching a trimmer to the supply rails and the control voltage input.
After playing around with it, mostly to test whether different compensation caps and more output capacitance across the peltier would eliminate instabilities a standby current draw of ~0.3A appeared no matter whether the peltier was attached or not.
And it got worse: the minimum peltier current setpoint shifted away from the 1.5V of the reference to 1.688V, indicating internal damage.
Different setup: 4.7µH 6A, 0.1R
after power-up, current consumption was found to be ~0.1A, changing from positive to negative peltier currents again exposed some audible instabilities. That test ended up in a fried MAX1968 drawing ~0.8A even without a peltier cooler attached (I should mention that voltage sensing and bypass caps ought have prevented excess voltage).
So what have I learnt? I'm not sure. Unfortunately I wasn't serious enough about testing the board with the scope attached because.. well.. it's just a kind of dual synchronous buck converter with X7R caps close to everything and I expected it to work - or at least not destroy itself on power-up.
The exposed pad TSSOP packages were soldered using a hot plate and hot air so I'm rather sure I wasn't too hard on them w.r.t. thermal gradients and heating / cooling rates. The layout is a bit strange, with the VDD polygon coming in from one side on the bottom layer but that's what you end up with when the power switch pins are scattered all over the package. I've filled all the vias with solder to minimize resistance.
Maybe someone can enlighten me with what's wrong with this circuit.
documents:
datasheet http://datasheets.maximintegrated.com/en/ds/MAX1968-MAX1969.pdf (http://datasheets.maximintegrated.com/en/ds/MAX1968-MAX1969.pdf)
evaluation board http://datasheets.maximintegrated.com/en/ds/MAX1968EVKIT.pdf (http://datasheets.maximintegrated.com/en/ds/MAX1968EVKIT.pdf)
reliability report http://www.maximintegrated.com/en/reliability/maxim/MAX1968EUI.pdf (http://www.maximintegrated.com/en/reliability/maxim/MAX1968EUI.pdf)
edit: attached eagle .brd file