[34S] battery warning indicator / question

[matthiaspaul]

(03-31-2015 09:02 AM)Bernd Grubert Wrote:  What about adding a gold cap as a back-up power source for demanding calculation?

My experiences with super capacitors are limited to double-layer capacitors (EDLCs). In general, they are great to deliver the relatively high-energy peak currents observed f.e. in conjunction with GSM wireless modems, however, I have found them to be not without their own set of shortcomings as well.

For example, they are extremely sensitive to over-temperature and over-voltage (both no problems in a calculator, though). Compared to ceramic capacitors, they have higher leak currents reaching the low micro-amperes range (not a problem in many applications, but certainly an issue in a calculator without a hard power switch and powered by small capacity coin cells).

Further, for a supply voltage of ca. 3V two capacitor cells would have to be used in series (as the maximum allowed cell voltage is typically in the range of 0.9-2.8V only), which would require the cells to be balanced. Passive balancing would increase the "leak" current by a factor of 10 or more, but even with active balancing the leak current could not be reduced below about 2 micro-ampere.

Another potential issue is the fact, that the charge current must be current limited, as an uncharged capacitor represents almost a short circuit to the power supply - and depending on how much current the supply can deliver, it may take seconds to charge the capacitor. Unless power planes are decoupled from each other, this can easily cause brown outs / resets in other devices connected to the same supply. Parts or PCB traces could be permanently damaged due to over-current. Without some sort of current limiter, the batteries could overheat (and in the worst case explode) when inserting fresh batteries. Much less of a problem in a coin-cell powered calculator, as the batteries would most probably be empty before bursting, nevertheless, this is an issue that needs to be taken into account.

One possible solution might be to add some circuitry to disconnect the super capacitor when the calculator is powered off (to eliminate the leak current when it hurts) or the battery voltage exceed the allowed capacitor voltage (even as little as 0.1V too much will damage the capacitor mid-term), to slowly (in order to avoid resets) charge it once the calculator is on and running at low speed, and to directly connect it to the power plane once it is fully charged *and* the calculator wants to run at a higher speed. With the super capacitor under firmware control, unless the battery voltage is above a certain threshold value, the calculator would not attempt to run at high speed until the capacitor is fully charged. Depending on if the capacitor is connected or not, the firmware would use different thresholds to determine when it needs to switch back to slow speed or has to power down. An approach like this might be suitable to get the most out of those little coin cells, but it's a bit more than simply dropping a capacitor into the existing design.

However, I think, the first and primary goal should be to find out, if the WP 34S suffers from similar high current spikes beyond the normal peak current (which we must assume as unavoidable under normal operation) as observed in the HP 15c LE, and if so, to eliminate them.
Even very short spikes would cause the calculator to reset, so if they'd exist and we could avoid them, we would already gain alot of headroom until the voltage under normal peak load drops low enough to reset the calculator. Such spikes may be too short to even measure them (with the calculator's hardware, that is), so it would be next to impossible to take them into account in any possibly more sophisticated power management schemes, except for by adding some empirically derived and undesirably large safety factor. As demonstrated above, a suitable ceramic capacitor may be enough to accomplish this goal already.
Most important is low leak current (since the capacitor remains connected at all times, and even a low leak current would eventually exhaust the batteries), followed by low ESR (so it can stabilize the power plane enough), followed by capacity (larger is better, but to eliminate the current spikes, something between 10-100uF is already enough, and typically, the other parameters weaken with increasing capacity).
Perhaps this would already stabilize the power plane long enough for the voltage detection code to get more "deterministic" readouts and thereby make the low battery indicator more reliably.

Greetings,

Matthias