2013-02-09 Flashers Bowden Style

I have built several of the “Single Cell Flasher With Two Year Battery Life” flashers found at Pigeonsnest.co.uk and they flash for what seems like eternity – they’re real misers when it comes to battery current.  But the flash is somewhat anemic and I had never tried this type flasher with any color LED other than red.  When I saw Bill Bowden’s similar flasher, I decided to try building one to see how well it worked.  The Pigeonsnest flasher used 10k and 27k resistors to supply current to the LED part, so I figured that since Bowden’s flasher uses 330 ohm resistors, it would be much brighter.  I was happy with giving up the long battery lifetime of the Pigeonsnest flasher to get a brighter LED flash.

For my first attempt at building one I used a 3/4 inch by 1-1/2 inch piece of birch plywood that I got from Michaels crafts/hobby store.  They’re about 0.08 inch (2mm) thick and come in a bag of assorted sizes for $4 or $5.  I made the holes with a small drill in a pin vise turned by hand.  The wood is thin and easy to work with.  Assembly was easy.  In the picture, I show my first build in the upper left corner, the second in the lower left, a third build lower right and the blank board.

I followed the schematic with two changes.  Instead of 1k, I used 1.1k, which was just ten percent higher so it should not make any difference,  I used a S8050 NPN high current transistor for the LED driver transistor, in place of the 2N3904.  I thought that this would be an improvement, given that the Pigeonsnest flasher recommends that this transistor should be a high current transistor.  The other two transistors were as shown in the schematic.

Performance Of First Build

I finished the assembly and connected it to a fresh AAA cell.  I observed a few sporadic flashes, but it refused to flash steadily.  I double checked the wiring to make sure it conformed to the schematic.  I checked a few points.  I tried a 22k resistor across the base to emitter junction of the S8050, but it would still not flash steadily.  I have built and experimented with many of Bowden’s circuits and I know that the circuits shown on that website have met Bowden’s expectations, one of which is that they are good working designs.  So I wasn’t sure what the problem was with my first build.  It could be a bad part or it could be that one of the changes I made had adversely affected the circuit.  I decided that I should give up on this one for now and try another one and get it working.

Performance Of Second Build

I got the parts together and two blank boards.  I used the exact parts, no changes or substitutions, even though I think the 2N3904 is not the best choice for the driver transistor.  I also used a red high brightness LED.  I clamped two boards together, one on  top of the other with a spring clamp.  I had the first build as a pattern so I eyeballed where the parts were to go and drilled both boards at the same time.  This way, I can build one of the boards and keep the other as a pattern.  I assembled the second board – everything went together smoothly, and soldered the red and black wires on.  I connected it up to a AAA cell, and it worked.  The flash rate was a bit on the fast side – 10 flashes in 7 to 8 seconds.

I connected it up to a power supply along with a 100 ohm resistor in series with the positive lead and 60 thousand microfarads of capacitors across the red and black leads.  This filtered out the current spikes caused by the flashing, and at 1.5V, my DMM measured 115 millivolts across the 100 ohm, which equals 1.15 milliamps average  current.

Thinking and Sorting It Out

Now I have two flashers, one that’s dysfunctional and one that’s working fine.  The one that works has the exact parts that are shown in the schematic.  The non-working one has some different parts, but in my judgment the differences should not cause it to stop working.  I needed to prove that I could make changes to the working one that would make it like the non-working one but without causing it to stop working.  But instead of making changes, I decided to build the third one with the changes, to see if it would work.  I examined the schematic and believed it was a valid design, and that those changes would not stop it from working.  Onward.

The Third one

I had planned ahead and made the third board so it was quick and easy to stuff it with the parts.  Just like the second except I included some changes,  One was the driver transistor: I used a 2N4401 – it’s capable of higher current.  I decided to use a yellow LED, which has a higher forward voltage, but should not have an effect on its operation.  Since the second one flashed faster than what I wanted, I changed the 1.5 meg to 2.2 meg.  This third flasher went together quickly and powered up with bright yellow flashes without a problem.

Back to the problem

By now I was convinced that the reason for the dysfunction was not in the changes I had made.  I had three of them working, and I decided to try fixing the first one.  I changed the output transistor from the S8050 to a BC338, which wasn’t much of a change.  Not a help; it still flashed sporadically.  I replaced the 2N3906 with another 2N3906, but it still flashed sporadically or not at all.  Feb 10 morning…  I replaced the first (left most) 2N3904 with another 2N3904.  Wow.  It started to flash regularly without the sporadic problem.  I think I fixed it, but I don’t understand why the original 2N3904 was not working.  I’m using new parts, never before used.  They should not be defective.  I suspect that the circuit is sensitive to some parameter that varies from part to part.  This may be the gain or leakage of the first transistor.

I decided that it would be a good idea to coat the boards with some clear acrylic spray.  I tried to get the spray can to spray, but apparently it had been sitting on the shelf for so many years that the propellant slowly leaked until it would no longer spray.  Instead I gave two of the boards a coating of clear nail hardener, something I got at the dollar store and was still liquid enough to brush on.

While they were drying I decided to change the 2.2 Meg resistor to 1.8 meg.  This was board #3, which was flashing at about 10 flashes every 12 seconds.  I wanted it to be about 1 second per flash, so I unsoldered the 2.2 Meg and put a 1.8 Meg in its place.  I powered it up and it was acting just like the first one: sporadic flashes once in awhile, maybe ten seconds or more.  This third build was working okay before.  Something  in the circuit design seems to be causing this instability.

The Leakage Problem

I thought about it, and the one thing I had not replaced before was the 1 uF electrolytic capacitor.  Lytics tend to be more leaky, and in this case I was not sure if the polarity was correctly shown in the schematic.  I decided to solve both by replacing the electrolytic with a 1 uF ceramic capacitor.  That didn’t help the problem at all.

One other concern I thought might be occurring was the leakage problem again.  The first transistor’s collector is connected directly into the second transistor, which is connected directly into the third transistor.  If each of the transistors has a current gain of 200, then the total gains is 200 times 200 times 200.  Just 200 times 200 is 40 thousand, which is so much that when a hundredth of a microamp (10 nanoamps) of leakage leaks into the base of the first transistor, it gets amplified up to nearly a half milliamp (0.00000001A times 40000 = 0.0004A or 0.4 mA).  But this has three transistors, each with a gain of maybe 200.  So 200 times 200 times 200 is 8 Million!  Even a thousandth of a microamp gets amplified up to 8 milliamps!  To bleed some of that leakage away, I put a 10k resistor from the base to emitter of the 2N3906 PNP transistor (second transistor from the left).  That was enough to stop the leakage current from the first transistor from being amplified by the second and third transistors.  And 10k is high enough to not be a problem when the transistors are switched on during the flash.  But it did have one slight effect: with the 1.5 Meg resistor the flash rate was a bit high, about 10 flashes every ten seconds, and with the 10k resistor added, the flash rate slowed down to about ten flashes every 10 to 12 seconds.  That was closer to what I wanted.

Now that I had found the Achilles Heel of this circuit, I drilled holes in the other boards and put the 10k resistors on them, too.  They all slowed down a bit.  I changed one of the resistors from 1.8 Meg to 1.3 Meg and it flashed just about exactly once a second.

The 10k value is just a convenient one; any value from a few k up to 100k may do the job.  There was one last thing that I wanted to add: a daylight sensor to turn off the blinking during the daytime.  I thought about using a CdS photocell from the base to emitter of the first transistor.  The problem was that the 1.5 Meg resistor supplies so tiny a current to the base that a CdS photocell would shunt most of it away even in a dark room.  Instead, I used the same circuit that I added to my Blue Blinky.  I connected an LED between the base and emitter of a PN2222A transistor.  I then connected the emitter to the negative and the collector to the base of the first transistor.  The tiny current generated by light on the LSD (light sensing diode) is enough to turn on the transistor and shunt the 1.5 Meg’s current to negative.  This simple circuit is not very sensitive; it will allow the board to blink in normal light, but when the LSD is held close to a bright light or outside in daylight, the blinking is stopped.  With the Blue Blinky, I changed the PN2222A to a BC547C, which has higher gain and will turn off the blinking in a normally lit room.

Conclusion to end of confusion

I now have four stable boards that brightly blink a red, orange or yellow LED down to about 1 volt, with no signs of instability.  This circuit is now brighter and better than the Pigeonsnest circuit.  I have added the daylight sensor, which is doubling the battery life.  If I had my old desktop running, I could have designed a PC board with ExpressPCB and submitted it to the company to get PC boards made.  But I haven’t had it running for months, and I’m not sure if it will start up after sitting so long.  I have backups, but I can’t do any designing with ExpressPCB since I’m running Linux, not Windoze.  But I still have my Blue Blinky PC boards, which work with any color LED.

I uploaded a short Youtube video of a comparison of the Pigeonsnest 1.5V flasher on the left with the modified Bowden’s flasher on the right.  It’s obvious which is brighter.

One change that can be made is to replace the lower 330 ohm resistor with a diode, cathode to negative.  I can’t guesstimate what the benefits are, but it’s better than the resistor as long as the cell voltage is 1.5V or less.  If the cell voltage were to increase, the LED and diode would both be forward biased and excessive current would flow, with possible overheating and damage.A 1N4148 regular diode would work, but a 1N5817 Schottky diode would be best.

I built the Supercharged Joule Thief Flasher, which is simpler than this one but has some disadvantages.  The SJT Flasher will run to a lower voltage, but will not turn off, it just stops flashing and glows dimly and uses up the battery.  The Bowden’s Modified Flasher will flash down to about 1 volt, but when it gets too dim to flash, it stops and doesn’t draw any more current.  That allows you to remove the cell and use it in a regular Joule Thief.

I have found that Bowden’s circuit is similar to the one found here (scroll down to the bottom).  It looks like the leakage problem there was solved by the 100 ohm resistor.  Note that the LM3909 circuit to the left of it has long been out of production and unavailable for many years.  Here is another site that has Bowden’s circuit.  Notice that except for the watermark “freeciruitdiagram.com”, it is identical to Bowden’s schematic.  From what I’ve seen on this site, it looks like they have stolen schematics from other websites and use them with adverts to make money.  Perhaps Jobs was right. (I emailed Bowden with the change I made.)  However after the having the above leakage problems, I would say that anyone who attempts to build it will have the same problem I had.  This should be a lesson to the plagiarists.  When they steal someone’s schematics, they eliminate the possibility of getting it updated when a correction is made to the original.  If they had linked to it, any changes would be in theirs, also.

Update Feb 16 – I put the Pigeonsnest flasher on a power supply along with a 1k resistor in series and 40 thousand microfarads of capacitors.  That averaged out the pulsing current, and I measured the average supply current at under 50 microamps at 1.5V.  No wonder the AA cell lasts for years!  As I said I’ve built several of them, and the circuit is a good design (compared to the above).  But it does need a serious injection of LED current to make it brighter.  Instead of 50 microamps, the cell current could be 200 to 250 microamps, and the cell would last for many months.

For more modifications on this flasher, see my 2013-02-17 blog.  Back to experimenting…