The Joule Smasher LED Flasher - 6 to 8 years from a single AA

First please excuse me for a heading that sounds like click-bait. I'm just very excited at the results to data and wanted to share progress with you.

It is important to remember that during the entire life of the battery there is no change in the rate or the intensity of the flashing. No feeble blinking allowed in my projects!

The Joule Thief was a wonderful circuit for getting light loads to operate from batteries that were essentially flat. I built many of them and was amazed at the results. However, I stumbled across boost convertor IC's which cost cents and delivered even better results for no engineering effort on my part.

Recently I tried a variety of circuits for flashing a LED. My need was to be able to see if the gate to our property had been closed at night without walking all the way there. I didn't want to turn on the driveway lights because one of the benefits of going outside with the dog before bed is the night sky. The driveway lights detract from that. And where's the fun in painting a black gate white!

Testing showed that using a micro to flash a LED was way more efficient that discrete versions of LED flasher circuits.  Building on this example I made some code changes to enable pull-ups on unused inputs and adjusted the timing to suit my needs.

Instead of powering this from a 3V battery I used a AA battery and a boost convertor. It was obvious from the start of testing that this was going to be a step change in how long a AA battery could be made to flash a LED. Several years looked possible so I used an AG3 battery to test the concept.

The quoted capacity of an AG3 cell varies from 25 to 35 mAh. Based on earlier measurements I expected an average current draw of roughly 100uA, say 250 hours or 10 days. I used a brand new cell that was several years old from an unknown Chinese battery manufacturer. Given my contempt for most things purchased via AliExpress I would have been satisfied with 5 days given my expectation of 10 days.

After 15 days the battery voltage was still 1.205 volts. It appears from extrapolating the measurements that the AG3 will will last around 25 days.

I have a Joule Smasher Led Flasher running on a brand new AA battery. That battery still reads 1.555 volts after 15 days. It's a wild guess but if the 25-35mAh capacity of the AG3 is comparable to the 3000mAh of an alkaline AA then the Joule Smasher should flash the LED for 6 to 8 years. That exceeds my wildest expectations and is far longer than I expect to be living at this property. 

There is still room for improvement. I'm boosting the battery to 3.3V because at the time of building that was the part I had to hand. So I'm wasting a significant amount of energy in the current limiting resistor in series with the led. That might represent an increase of 20% in battery life.

Sometime next year I'll build a few more of these to confirm results and check the benefit of boosting to 3V or 2.7V.

Li Ion Battery Charging - Preliminary Test Results

Recently I blogged some guidelines for Li Ion battery use. I have ignored the "Store at 50% charge" folklore I read about everywhere because I store in the charger at a terminal voltage around 4v per cell. No research appears to exist on how harmful this is. But then I couldn't find research that supports storing at 50% compared with say 90%. 

Page 10 in this document has graphs which suggest charging to 4.0V represents a battery capacity of around 80% and a substantial increase in the number of charge discharge cycles available.

Still, I wondered if there was any degradation. To see if this could actually be tested I ran a trial on a few rubbish cells to see if anything happened. The trial is not as rigorous as an academic paper, but reflects my real world use of these batteries.

The method is:

• Each morning remove a battery and discharge to a fixed voltage. 
• Record the capacity
• Put battery back into charger in the same slot.
• No attempt to control temperature or other variables. 
• Some batteries do not get tested every day simply due to time constraints. These sit in the charger at the nominal terminal voltage of 4V.

I say 4.0V but I actually adjusted the charger circuit to be 3.95V for Battery A and 4.05V for the Battery B.

As I said - rubbish cells. Brand new 18650 from China but had been sitting in storage a while. Over the course of the trials I decided to reduced the discharge current since 450ma was more than the capacity of the cells being tested. I also learned there is relatively little energy between 3.1 and 3.0 volts so I adjusted the threshold to 3.1V after 18 days.

 

Battery B initially appeared defective. It seemed to terminate discharge early and I twice had to re-start the trial on that day. With the drop in discharge current the battery magically had reduced variation in each trial and each test ran to the terminal voltage.

Since I haven't seen evidence of degradation after 24 days I will in future check less often. In between trials the battery will remain in the CC/CV charger at the nominal 4.0V.

Led Flasher - Could it really be 4 or more years from a single AA batery?

So I built the ATiny13A led flasher and achieved an average current draw of 30uA over several hours by measuring the discharge of a 0.78F capacitor. This is around 25% of the discrete version.

 

Which is crazy compared to what I was expecting. After poking and prodding I determined that some of the difference was due to not allowing for the internal resistance of the pin driver on the ATTiny13A. Instead of getting 20mA I was getting 7mA of current through the LED. However, this did not explain all of the difference between what I measured and what I expected.

With some more tweaking of software and hardware I achieved a LED that flashes for 10ms every 2 seconds drawing 10mA. That's the rated maximum for a single pin. I was thinking of using two pins in parallel but it seems bright enough for my purposes.

As a sanity check I put a new but several year old AG3 battery onto one of these Tiny flashers on 9:30am on Dec 6th. I believe the battery was rated at 25mAhr when new. I expected this battery would  run the led for 10 days.  Note the makeshift button cell holder below.

 

However, after 10 days of operation it appears there is still substantial life left in the battery. 

   Day  Voltage   
    6        1.35           
    8        1.32           
    9        1.31
   10       1.30    
   11       1.28

I can't explain the observed result which is way better than I expected. What was supposed to be a sanity check has sent me back into my spreadsheet looking for an error. Wherever the error lies I am amazed that the AG3 cell looks like it will run the Tiny Flasher for 20 days.

It would seem that based on the range of outcomes at this point the Tiny Flasher will run for around 4 years from a single AA battery.

 

The only way to be sure is to monitor the one running on a fresh AA battery to see what happens. 

More on this in 2027?

Some notes on Li-ion battery charging

I've used lithium ion batteries in many projects for several years without mishap. I never realised how much mis-information exists in relation to these batteries. Claims such as "You have to cycle the battery 5 times before it reaches it's rated capacity" do nothing but demonstrate how stupid the person is making such a claim.

Many years ago I sighted research which showed that if you can keep a lithium ion battery between 80% and 20% of it's capacity it will last essentially forever. Now I question if that was real research or folklore. Consider that one of my laptops a decade ago actually had a bios setting that allowed you to only charge the battery to 80%. Like all "information" in this field I don't know if this is still, or even was, valid.

Guidelines

However, I have a few guidelines that have not let me down:

• Never fully charge 
• Never discharge below 3.1V (revised from 3.0V after measurement)
  
• charge at C/10 or less.
  

BRCL4054CME

When I have to package the battery and charging circuit into a small enclosure I reach for the BRCL4054CME. Simple, reliable and effective. However, like every charging IC I have looked at it is cumbersome to terminate the charging voltage at less than the designed limit of 4.2V. So when I use this IC I generally don't satisfy the first guideline. 

The second guideline is satisfied by shutting down the circuit using voltage detection and third by selecting the appropriate resistor.  Simple but not quite the complete solution. And it limits you to charging from a 5V source.

Preferred Solution

Where space permits the following circuit is the basis of my preferred approach.

 

Any DC source from 8V to 30V can be used and it has survived mishaps. In my multiple single cell charger I now include a mosfet to isolate each battery during a power outage. That stops discharge through the series resistors that set the voltage. It also means where fitted I can simply switch off the power when leaving the workshop.

My longer term goal is to have enough batteries sitting charged that I only need to switch the chargers on when a cell has been used in a torch or project and needs recharging. I plan to write up the finished project once my battery testing is finished. Preliminary findings will be blogged in a few days or so.

I never had to consider extended power outages before but on reflection of the increasing natural disasters in the world it seems prudent to keep as much charge in the battery  as possible without doing harm until it is needed.

Led Flasher - Some Preliminary Results

 To test the boosted AA flasher concept I used the following methodology:

1. Charge capacitor to around 1.5V
2. Connect to a 3.3V boost circuit
3. Load 3.3V output with a 68k resistor (simulates the 48uA average current draw of concept)
   
4. Measure voltage at periodic intervals
5. Calculate average current drawn

I had previously estimated the average current drawn from the AA battery as 140uA when boosting to 3v.

The results I obtained were quite consistent with this until the voltage dropped below 1V. The divergence after 48 minutes is likely to be the fixed assumption for efficiency. As the input voltage drops more current is required and the losses increase. 

Still, not bad for a thumb suck.

The planned 4ua sleep versus brief 20mA flash may give a different result due to the boost circuits efficiency varying with load. 

It appears this approach will deliver a flashing led with around 3 times the life of the discrete version I built. The next step is to build this prior to Christmas.