[Khaase]

An update: I got the Tag Connect socket and it works fine on my Pololu Atmel ICSP programmer, and using it burning bootloaders and program Mayflys is fast and fun.

I am still sorting through my test results and will probably have to do a new battery of tests with an external datalogger because the internal ADC gets confused by low voltages, but regardless of that data, I don’t think that setting the brownout fuse to 1.8V is a solution to many problems one might encounter. In particular to the Mayfly, it does seem not help with the long-wire input capacitance problem that can cause reset loops.

[Khaase]

So far, your super-capacitor idea is holding up pretty well for me. A 100 uF capacitor buffered voltage sag to a degree, but the 10,000 uF supercap (1/10 the voltage rating and 1/2 the physical size) gets our setup able to use any given battery voltage measurement without having to consider if the sensors are on or not. (I want to write “Battery voltage measurement fairly close to the aliasing limit of the ADC” but I average a few hundred measurements, so one would expect the optimum result to be nearly invariant. Now we are aliasing across five 0.019V levels instead of ten.). I am curious how the capacitor buffers our pulses when the LiSOCL2 cells are nearing depletion, we’ll tune our deep-sleep cutoff voltage based on that.

[Khaase]

I ordered the cable. FWIW The folks that run Tag-Connect are very friendly!

[Khaase]

I’ve been testing with Mayfly 1.1 and LiSOCl2 D batteries. Our package is using a Mayfly 1.1, and SD card, and a sensor on the switch power that pulls ~350 mA in pulses when its on.

I initially tried using 1S4P configuration with 95sq015 protection diodes with the intent of preventing heating from cell imbalances. That puts the Vin at about 3.4 V. At temperatures down to about -10C it works fine, but below that the voltage would drop to the point that the 1284p will brownout. The internal battery measurement shows less than 3.1 V, but as the powersupply is not a buck/boost, I doubt this is accurate and I think the voltage is sagging to 2.7V and triggering the BOD (herm, note to self: try measuring battery voltage using internal 2.7 V VREF). After reading some papers and datasheets about LiSOCL2,  I wrote the company that makes the LiSOCL2 (Xeno) I’m using, they said for the use case where the cells are the primary power source and replaced as a group, the diodes aren’t needed.  I removed them, and, at least in the short term these are able to operate at -20C.

Something else I noticed was that the LiSOCL2 cells have an activation period: they couldn’t power the Mayfly for a few moments after I first connected them (reboot loop) after a few minutes it started working. Several datasheets mention that the lithium electrode gets deactivated when the cells are not used and that some current draw is required to reactivate it.

[Khaase]

Thank you for the link. I was googling for pogo pin grids, what a nifty widget.

[Khaase]

I probed the traces to the vertical uSD socket and FTDI and determined that with the isolation capacitor and DTR, it’s not possible to update the bootloader from uSD/J7. Of course, Right there on the circuit diagram there is an ICSP port (H4), which is a set of pads for what I believe are 1.94mm pogo pins by the uSD slot. I wonder if there is a part number or diagram for this? It should be easy to design my own, but an engineered part might survive better if I have to reflash many of these.

Though so far I am finding that the source of power-on resets is simply having the uA meter in the power loop.

[Khaase]

That is useful, and along the lines of my thinking: moving to LiSOCL2. Though I am having trouble getting supplies to test. The internal impedance of those, especially when low or cold, is a big question to resolve. My system can use all the deep-sleep tricks, but when measuring the load averages ~50 mA with spikes up to ~350 mA for fractions of a second, and I wonder if this will cause a brown-out with the LiSOCL2 cells. Additionally, if I use these cells, we’ll need to use an array of 4, behind protection diodes, which drop the input voltage a bit, but also could provide more current during operation.

I had not thought to try using a bigger buffer capacitor in front of the Mayfly. I got some more stability with 1000uF with my ammeter in the circuit, but some configurations still got resets: I figured that it was confusing the voltage regulator somehow, as I could not resolve a deep voltage sag on the inputs when measuring. I will look at the datasheet again and meditate.

But if I could just set the brownout to 1.8V instead of 2.7V that would still be a nice test/check. I tried hooking up through the SD card vertical socket and the DTR pin of the FTDI connector, but wasn’t successful. I am wondering if the pin order in J7 is correct in the schematic, I can’t see the traces under the SD card slot on the PCB, so hoping that I can get clarity here, though I will probably just try to probe out the pins soon.

[Author]

Posts