2013-10-27 Low Power Joule Thief With Photocell

One Joule Thief I have not built (at least not in the same JT) is the low power JT with the CdS photocell.  Almost all of my JTs with daylight shutoff used a LED as a light sensing diode, and then most were flashers.  I have a bunch of CdS photocells but I haven’t really tried to use them for shutoff because they are linear and will cause the circuit to shut off gradually, instead of abruptly.  They work better if they are used with a circuit that gives abrupt shutoff, such as a comparator or Schmitt trigger.  This time I just connected the photocell across the base and emitter, and it works okay, with some dimming, but it shuts off completely.  It is very sensitive; it doesn’t take much light to shut it off.

The rest of the circuit is a conventional Joule Thief with a 68k resistor in place of the usual 1k resistor.  This reduces the base current to the point where the battery drain at 1.5V is less than 10 mA.  That’s about 1/8 of the usual JT current.  For the coil I used a Fair-Rite 2643002402 core with about 9 turns which gave an inductance of about 100 uH.  The transistor was a SS9014D (D means it’s the highest gain range, a very high gain 400 to 1000). It’s a low noise audio transistor, with a max current of only 100 mA.  I reckoned that it would work okay since the battery current was less than 1/10 of that.  The 2N3904 doesn’t have high enough gain, only 300 or so; with the much higher gain SS9014D, the resistor can be much higher and when the CdS shorts the base during daylight, the battery current will be only 22 microamps, which is so low that the battery will last a very long time when it’s shut off. Remember that if the circuit is left indoors where there are lights on much of the time, it will shut down during the lit periods.

I assembled it on a 9/16 by 1-1/16 inch piece of thin birch plywood.  I probably would have had it together and working in less than an hour, if I hadn’t goofed and crossed the wires of one winding.  After I got that straightened out, it probably took me another ten or so minutes.  The only thing it needs is a spot of silicone or hot glue to hold the toroid firmly in place.  A coating of acrylic spray would be better to seal the wood to prevent it from absorbing moisture.  This is not critical since the impedances are only moderate; the 68k resistor is the highest.  But if it’s outdoors, it would be best to waterproof the circuit.

I had trouble getting a photo of it with the LED lit.  When the light was high enough to get a good shot, the LED would shut off.  And when the LED was lit, it was too dark for the camera, so I decided it would be shown without light.  I can say that the circuit meets my expectations, and if I want to make it less sensitive, I can enclose the photocell in something dark to cut down  the light and reduce the sensitivity.  The LED current is not all that much but it’s visible and can light up a small area.  I haven’t tried it at night but it may make a good night light.  At 8 or 9 milliamps, the AA cell will not last all that long, maybe a few weeks or a month.  I also have another idea that might work.

Well it’s now dark, and I measured the frequency while lit.  It measured 284 kHz, which is much higher than I had expected.  That’s 4 or 5 times higher than the typical JT, and indicates to me that the coil needs more turns.  There is enough room for 12 or maybe more turns, and if the feedback (purple) winding is changed to a finer wire, it could hold even more turns.  Perhaps the reason why it’s running so fast is the high value of the resistor.  With the base current severely limited, the transistor doesn’t get switched on and off fully, and it’s running more in the linear region, instead of saturating or cutting off.  This reduces the efficiency, but then it’s using very little power and the light output is adequate.  A small capacitor across the resistor might help to slow it down, but I haven’t got the inclination to try it since it’s working okay now.

The next day I put the LPJT near a small night light and I found that it’s sensitive enough to shut off when the dim light from the nightlight hits the photocell.  Now if I want it on, I just place a small object in front of the photocell and it comes on.  I used a pop bottle cap and  put the photocell inside so it’s surrounded by most of the cap.  Even like this, when the room light is on it still goes out.  Same for daylight.  I bought the night light in Feb of ’07 and less than a year later the cheap white LED in it had dimmed to the point where it was useless.  So I removed  the white LED and put in a yellow one, and it’s been running bright for more than 8 years.

One other thing it can do.  As long as the night light is on, the LPJT will stay off.  Then when a power failure occurs, the night light goes off, the LPJT comes on and acts as an emergency light.  The problem is that it’s okay for a dim night light, but for an emergency light it needs to be much brighter.  I think if I could get a regular power JT to shut off with a photocell and draw only a few dozen microamps, the cell would last long enough for it to work okay after a year or so.  But the normal 1k resistor would drain the cell in only a month or so.  It will probably need another transistor to reduce the photocell current down to the point where the cell will last a year or more.