Results from First Trial of 8 Joule Smashers are in.

I ran a trial with 8 Joule Smasher Led Flashers powered from brand new AG3 cells. I discarded three sets of results for the following reasons:

1. One stopped working for no reason despite cell having a voltage of 1.2V. After removing and re-inserting battery it ran properly until cell was exhausted - bad solder joint suspected
   
2. Another was stopped after led appeared to remain on. Cause unknown - again a bad solder joint was suspected.
3. Battery was displaced after last measurement and not noticed till morning.

So I was left with 5 trial results. The results were revealing:

1. molded choke was again very inefficient
2. boost control appeared to offer no real world improvement in life
3. dropping boost voltage and adjusting led resistor to maintain 10mA of led current was most efficient configuration.
4. Average rate of depletion at 1v battery voltage was 240mV/day. This is about 8 times faster than the initial trials with the same inductor with an on time almost 10 times smaller.
   

I replaced the molded choke with a better inductor and changed the boost IC to one with 3V output. That gives me a selection of output voltages and the next trial will run 9 Joule Smasher Led Flashers at 10mA for 3ms. The reduction in on time should at least double the battery life.

In addition I reprogrammed the Flat AA and Exhausted AAA cells I was testing to be on for 10ms, instead of about 1ms with the incorrect fuse setting. I left the inefficient molded choke in place. It is obvious in the following chart when this happened. While the AAA cell collapsed, the AA cell bottomed out and kept running at 0.74V. I will leave the exhausted AA to keep running with a 10ms on period.

 

The chart reinforces the previous test results that extrapolating from an AG3 battery, say 30mAh, to a AA battery, say 3000mAh, is not  as simple as multiplying AG3 life by 100. The AG3 batteries basically stop at 0.9V whereas the AA cell keeps powering on well below this.

And one last thing: datasheets for these boost convertors are typically awkward to compare and questionable in meaning. I measured the input current for a 4uA current through a resistor (ie micro sleeping) as 30uA. Take away the 4uA load at 80% , or 11uA, and I have a quiescent current of 29uA. This looks too low given the the typical value in the datasheet of 45uA for no load. 

This is compounded by the conditions under which efficiency is calculated being quite different between manufacturers. This has made it very difficult to confirm by calculation the results obtained. Iteration, rather than calculation, seems to be the only way to optimise the outcome.

Tomorrow morning another trial of 9 Joule Smasher Led Flashers with AG3 batteries begins. Hopefully the soldering has been corrected and everything runs to termination.