Hi,
We've designed a TPS562201 into an existing product, which is about to go into production.The TPS562201 is powering an ESP32-Wroom32 module, from a 12Vdc powersupply (Meanwell IRM-30-12).
Our initial prototype worked fine. We bought prototype components through Mouser, and the boards worked as expected, 3.3V output from TPS562201 looked good. Then, we purschased a large batch TPS562201 from the TI store, and built 9 boards in a pilot run before running full assembly of 1.000 boards. Now, with the new (9) boards, the ripple out of the TPS562201 is almost 500mVpp. With the ESP32-Wroom32 is running (drawing around 100-150mA), the ripple is complete sawtooth shaped, with a frequency of around 17kHz (see screenshot below).
We did not make any board changes between the prototype and pilot run. The only difference is the TPS562201 coming from 2 different sources (Mouser / TIstore). They're both marked 2201, but the lines on the SOT23 packages (which I suspect are used for revision) are a bit different from the 2 sources.
If the ESP32 module is in standby (drawing less than a mA), the ripple frequency is around 80Hz, still with a sawtooth ripple of around 500mVpp. Output capacitors are 2x22uF, 10V, X5R. The ripple is almost the same shape no matter the amount of current we draw (1mA og 100mA).
What can explain the high ripple? What I find interesting is that the low voltage on the ripple is actually the setpoint (3.3V), while the high voltage (upper peak) is much higher, about 3.7-3.8V. So it seems there's a massive current boost that charges the output capacitors way too high (up to 3.8V), and then the regulator doesn't add any current (charge to the output cap) until it's down to 3.3V again.
This also explains the frequency changing when the current is higher or lower (lower current = lower frequency). We have about 60-70uF on the output. The frequency is about 17kHz or 60us. This means that we would see a drop of 1V pr. 1A, or 400mA = 400mV given the timeframe (60uF vs 60us).. So when we go into standby on the module, we see the drop taking about 12,5ms, which corresponds to a few mA. So the theory of capacitor discharge is proven to work..., but what causes the TPS562201 to overregulate and make the high voltage/current boost to begin with?
We're ofcourse worried about this, and we have a very tight timeframe to solve the problem, massproduction will start soon.
I have attached a snippet of the schematic (partial). C32+C10 are very close on the PCB, so we have a total of 44uF + the internal capacitance in the ESP32-Wroom32, on the output of the TPS562201). The ripple is measured across C32.
Thank you for any pointers or help.
Jacob
(This is showing the ripple voltage on pilot run boards (total of 9 boards, all same behaviour)):
(This one showing the ripple on our initlal prototype board):
(And finally a partial schematic:)
