Possible Ways to Brick?

[Nick]

Hello,

I have a board with an RFDuino on it which uses a Nordic nRF51822 internally. We are seeing a few of them auto-brick when the input voltage dips too low.

My first guess is that it is doing an EEPROM write even when the voltage dips or something, but I am just wondering what can brick these? Any known errata for this?

Basically it is running off of a big battery and as long as it stays above 2.9-ish volts, it works forever it seems. When it goes to 2.8v or so, sometimes it stops responding and then never turns on again. I haven't written the software but I am trying to get the code to look at.

[Øyvind Karlsen]

Hi,

What do you mean by brick, do these devices not turn on again when powered by lab power supplies?

Are you sure that the battery can take the approx 15 mA current spikes (radio TX + cpu active)? You can alleviate this somewhat by adding capacitor reservoirs on the VDD line.

What is probably happening is that your battery is being pulled down from 2.8V by a radio TX current draw, the battery voltage will drop due to not being able to supply the instantaneous current. At 2.8V a coin cell battery is typically at the end of its life.

Best regards,

Øyvind

[Nick]

Hello,

Correct that they do not turn on again. No signs of life.

The battery can take 30W draw. Yes 30W, not mW, it is a gigantic Li-Po battery. On previous designs the RFDuino (Nordic part) was powered via a 3.3v SMPS buck capable of 1.5A continuous. We saw 3 parts die in this manner and it was suggested we power it from 3.0v. The SMPS has tighter than 1% regulation as measured by a high end Agilent scope, there were no spikes, very minimal ripple even under full load. Now with the LDO, it is an exact 3.000v with 4.7uF hanging off the LDO output and another 4.7uF right next to the RF chip.

The battery is nominally 7.2v (2 series lithium). When the battery gets very low, down to 2.8v after a total deep discharge sometimes the RF chip never recovers. The battery charges back up to 8.4v no problems. But the RFDuino is no longer discoverable. An LED hooked up to blink, never blinks.

[Nick]

And yes I have tried with my lab supply and they do not turn on again. Out of about 20 parts, we have had 4 fail in this way which is very concerning. As far as I know, I am not doing anything that the datasheet says it cannot handle.

Yes by "brick" I mean non-recoverable fault. I have received the code and there are no writes to the flash memory at all. So back to no idea why these would fail.

Are there any signals that are probe-capable on the output that will show it completed a POR? Like measuring a node for something internal to connect? A small change in an output voltage when a programmed trim value gets loaded, etc? I would like to prove that the RFDuino or Nordic device is powering up and completing a POR or not and bricked. All other aspects of the circuit work perfectly.

[Øyvind Karlsen]

The Li-Po voltage is 2.8 V? Then that Li-Po is dead.

What voltage range is your LDO rated for, do you have a product sheet for reference?

[Nick]

Current LDO is a Microchip MCP1802T-3002I. Good for 10V max input, 300mA max output. The ONLY things hooked up to the +3v supply is the RFDuino and a 4-IO I2C expander (PCA9536TK,118). Previously when this happened everything was supplied by the +3.3v 1.5A SMPS.

Yes the LiPo goes down to 2.8v and is dead, correct. That is exactly what is being tested, dead batteries. However I have completely 100% disconnected the battery and it is being powered by a Keithley 2602B SMU ($10K USD). With the input voltage at 8.2v or any voltage for that matter, the RFDuino is completely unresponsive.

All other circuitry is operational. There is an additional 5v LDO and an LTC3883 with full telemetry capabilities, all are operational.