Joule Smasher Led Flasher - Interim Results from Second Trial

 

Overview

I have conducted a number of trials using up to 9 Joule Smasher Led Flashers running off AG3 cells. The following is some preliminary findings from the second trial of 9 Joule Smasher led Flashers. At present 5 boards are still running and I expect they will cease in the next day or so.

AG3 Battery Life

Boost Control - Ineffective

Having re-worked the math several times now I kept coming back to a possible 30% improvement in life. During the 32ms after the Led turns off the micro sleeps and the boost control keeps running to finish topping up the capacitor.  This brings the voltage up over 3.25V on the digital volt meter which I felt was close enough given the refresh rate of the digital volt meter being used. I left the boost control version running with the extra 32ms of "on" time. 

Surprisingly, the boost control delivered no improvement. In addition to no improvement in life there were also instances where the micro appeared to freeze or lock-up. One of the poor performing boards with boost control which had flattened the AG3 battery was converted to be always on and a new battery was inserted. At the time of writing the always on version appears to be lasting around 30% longer.

While some of the improvement in this example can be attributed to a better battery, it is clear across the comparisons done that the boost control did not deliver the expected improvement.

Going forward I am dispensing with boost control. The boost control boards are being progressively converted back to always on with a fresh battery fitted. This will delay the end of the second trial and is not ideal given the different temperature conditions. However, 3 of the 4 boost control versions proved undesirable from a battery life and lock-up perspective. 

Trial 2 Results for AG3 Batteries to date

The best 5 batteries are still running. 

The only surprise here was it took so long for the 3V boost control versions to emerge as contenders for longest life.  And the special diodes I purchased don't appear to have added any meaningful improvement in this application.

"ON" Time : 10ms versus 3ms 

From the preliminary trials when the On time was 1ms the actual life was 50% longer than the calculated life. Having completed the first proper trial with incorrect fuse settings now allows me to compare a 10ms On time with a 3ms On time.

My measurements of current in hte On and Off states result in the 10ms versions having an average current consumption of 190uA compared to 81uA for 3ms of On time. If I scale the time axis for the second trial by the ratio of average current I can compare the two trials. 

Here is an extreme outcome from this:

 

In all cases where I did this comparison the battery life for 3ms was significantly longer than the ratio of average current would suggest. 

The Voltage Cliff

Reviewing the results to date from AG1, AG3, AAA and AA batteries suggests the smaller the battery capacity the higher the voltage at which the cell is exhausted. 

AG1    1.1 Volts
AG3    1.0 Volts
AAA    0.9 Volts
AA       0.6 Volts

While the energy available from a AA battery between 1 volt and 0.6 volts is unknown it appears substantial.

Implications for AA Life

I have measured the actual battery current required in both the On and Off states. However, using these currents and the duty cycle to estimate life is misleading. As Panasonic stated in their Alkaline Handbook:

"Actual testing is needed to determine the amount of additional service expected in pulse applications since there is no simple equation to accurately calculate the impact of duty cycle on service. "

The fresh AA battery I started testing some months ago was programmed for a 1ms flash until day 89 when I discovered the fuse bits were wrong. Since then it has been using 10ms of On time in an attempt to accelerate the discharge. It will take another 6 months to confirm the outcome but early signs are promising. Even with the much longer On time it appears the AA battery will last much longer than a life based on extrapolating capacity would suggest.

From my measurements using smaller calls and a range of On times it appears the life for a A battery will be around 10 years. 

Conclusions

The current boost convertor IC I'm using is remarkable at getting that last amount of energy out of a battery. I tested a CR2032 battery removed from a motherboard because the bios said it was flat for 5 days. Then it just stopped with 0.2V measured across the terminals.

Trying to estimate battery life with the Joule Smasher Led Flasher discharge characteristics from datasheets is difficult and error prone.

If I waited until the AA battery was exhausted before reporting results the devices I use would be unavailable due to obsolescence.

The test results could be applicable to battery powered sensors and wearable devices.

My testing supports the hypotheses that the larger the battery the greater the effective capacity. The capacity quoted for a AA battery becomes meaningless for a Joule Smasher Led Flasher given the minuscule current and duty cycle used. 

I found in testing that small batteries appear to discharge in relative terms faster than larger batteries and total collapse happens at a higher voltage for the smaller batteries.

Moving from small AG3 cells (25-35mAh) to a AA cell represents a change of over 100 fold which makes predicting life uncertain. It would be preferable to wait many years for trials running with fresh AA batteries to end. At present that appears a decade away. The first of those is on my desk nearby. The second, with the better inductor is now on the gate and is expected to run far longer than I will live here. In coming weeks further AA powered Joule Smasher Led Flashers will be built for extended testing.

For this project that about wraps it up. I have settled on a circuit which I believe will deliver the 10 years of life and revised the PCB so it is about the same diameter as a AA battery. That will allow battery and Joule Smasher Led Flasher to be slid into a tube end-on-end. I'll publish a photo in a few weeks when assembled with the final results from Trial 2.