LM3909 replacement

[fernando_g]

This thread got me thinking about this project:
http://www.picaxeforum.co.uk/showthread.php?t=14374

Back in the 70s, the LM3909 was introduced. This wonderful circuit would flash an LED from a single cell battery for many, many months. The key for its skimpy current drain was to pulse the LED with a short, bright pulse, and then wait a couple of seconds. Thus the average current would be quite low.

But times have changed, the LM3909 is no longer available, low cost PICAXEs are here, and there are supercaps. Could they be used as a LM3909 replacement.

Yes indeed!. By using the 08M's low voltage capability, and pulsing the LED briefly and immediately going to sleep with DISABLEBOD, the average current can be made real small.

I employed an 08M, which would feed a small high brightness red LED via 330 ohm resistor and used the following ultra-simple, almost trivial, code:

#picaxe 08m

main:
pulsout 1, 600 
'selected 6 msecs since that was the pulse width from the LM3909
disablebod
sleep 1
enablebod
goto main

I took a 1000 uF capacitor (actual 950 uF, measured with a Peak Atlas) charged it to exactly 5.00 volts and allowed to power the circuit while I timed it with a stop watch.
252 seconds later, when the cap voltage reached 1.88 volts, the LED finally stopped flashing.
Some quick calculations yield an average current of ~11.8 micro amps.
Extrapolating these values, a 1 F cap should be able to power this circuit for 254,000 seconds or almost 3 days!

Of course, components have tolerances and your mileage may vary. I suspect that the supercaps (especially the cheaper ones) may have substantial self leakage which would shorten the actual time.

And one could use a small coin lithium battery and power the circuit for months. The Duracell CR2032 is rated for 180 mAH, and thus could last almost two years.

 

[Dippy]

Well tested fernando.

This was exactly what I suggested to someone recently who wanted a flashing light on their boxes of lenses.
Thankyou for proving the point.

 

[hippy]

Do you need the ENABLEBOD ?

Putting a long sequence of PULSOUT then SLEEP until memory is full will save a tad more current when running as it won't execute the DO-LOOP everytime which sucks a bit of current.

You can also experiment with under-clocking the PICAXE and using NAP instead of SLEEP with correspondingly decreased PULSOUT or switching back to 4MHz after NAP.

You could also experiment with lower resistors and shorter pulse lengths which should give comparable brightness and perhaps keep flashing visible for longer as voltage drops - blowing up your chip because of exceeding output current is the risk you'll have to decide upon taking. I'd suggest running on permanent 5V and start with a tiny PULSOUT and increase it until LED seems satisfactory with average current still being < 25mA.

Setting unused pins output low or high, or tying inputs to 0V or +V might also have an effect. When looking for ultra-low power every little bit gained adds up.

Good work on trying it, and for reporting back.

 

[lbenson]

I'd be interested in hippy's suggestion of omitting enablebod. Just put disablebod before main. This should allow the 08M to run at a lower voltage without cutting out because of brownout. But perhaps the cutout at 1.88 volts is not because the picaxe stopped, but because below that there isn't enough juice to flash the LED.

Nice experiment.

 

[fernando_g]

This should allow the 08M to run at a lower voltage without cutting out because of brownout. But perhaps the cutout at 1.88 volts is not because the picaxe stopped, but because below that there isn't enough juice to flash the LED.

Nice experiment.

My thought exactly. That is the reason one should use a red LED, as it has a lower Vf than other colors.
Of course, Infrared LEDs have a lower still Vf, but they would appear to be permanently dark

The reason I chose 6 msecs is to emulate the LM3909's actual pulse width. I did try shorter pulse widths, down to 1 msec, and the flash is still visible in subdued light rooms. The run time did not go up sixfold however, so the PICAXE actually running the code (and drawing current) does have an impact.

Hippy; interesting thoughts on enablebod and lowering the clock frequency. I'll run those experiments as time allows.

 

[Marcwolf]

Many thanks for this.
Although I would prefer to use a LM3909 (still available on the internet) this could suit me.

The copany I work for make spectical lenses and we needed a simple way to mark a job a "Priority"

So a Flashing LED would be perfect. The battery length is also good...

Dave