RustyBolt.Info/wordpress

I found that my modified Joule Thief circuits were running at the frequency of 550 kHz, or at the bottom of the AM broadcast band.  This is the highest frequency that I have settled on as a satisfactory frequency for a Joule Thief.  I have built ones that have gone as high as a few MHz, but they were just to see how high the JT could go and still function.

What is important is that when the frequency gets above a few hundred kHz, then the circuit becomes more dependent on the stray resistances, capacitances and inductances of the wiring and PC board traces.

People often assemble a JT on a proto board or with clip leads.  The stray capacitance between the contacts inside of the board will cause the circuit to have a different operation than if it was tack soldered together.  And using the clip leads to assemble the circuit adds an excessive amount of stray capacitance and inductance so that at the higher frequencies, the circuit may become unstable so that it operates erratically or quits altogether when the clip leads are moved.  The proto board also has more stray capacitance than a PC board, and the circuit may work differently than if it is on a PC board or tack soldered together.

The most damaging of these to a Joule Thief is the stray resistance.  The voltage is so low that any stray resistance becomes much more detrimental to the circuit.  When a circuit is running at 12 volts, a small resistance might be tolerated.  But when a circuit is operating at 1.2 volts, that same small resistance becomes ten times more influential on the circuit and ten times more detrimental – it can ruin the performance.  The solder joint is a very low resistance connection; the contact in a proto board and the clip lead are not low, their resistance affects the circuit much more.  And this is especially true if the proto board’s contacts are old and dirty.

Some circuits (not the conventional JT) require a tuned circuit with a capacitor and inductor in parallel.  These two components have a very high circulating current when they are working properly.  The hams would use silver plating to reduce the resistance and increase the Q or quality factor.  Using a proto board to hold the cap and coil adds resistance to the circuit, which reduces the Q or quality factor, and this makes the tuned circuit’s performance lower than if the cap and coil were soldered together.  The penalty is even greater with clip leads (see  note below).

If the experimenter is going to build any circuit where the frequencies are more than a few hundred kilohertz, then he or she should learn how to use good RF construction practices.  Keep leads short.  Make good mechanical connections to the leads before soldering.  Make good solder joints.  And use good quality components for the critical circuits where RF is highest.

One can download this 256 page manual on RF and microwave grounding.

Note: I bought a hundred mini alligator clips online, in red and black.  I made my own clip leads with wire that is only four inches or 100mm long, so the clip leads have a minimum impact on the circuit.  They are supposed to be temporary, and not the way to build something.  If you want to hold leads together, you should buy a hundred Fahnestock clips to hold the component leads together with no extra wire.  At least they don’t have so much capacitance as the proto boards have.

Update: This circuit is much like the Supercharged JT, see the end.

A while back I built a Joule Thief flasher, like the one I had seen somewhere on the web, most likely on Youtube.  The schematic shows the different arrangement of the feedback winding and base bias resistor.  Normally the feedback winding is in series with the 1k resistor to the positive supply.  But in this arrangement, one lead of the feedback winding is connected to negative, and the other lead is connected to the timing capacitor.  In my first build, I had used a 120k ohm resistor for the base bias from positive to the base, and a 33 uF ‘lytic for the timing capacitor, but the flashes were so fast that I had to put another 220 uF capacitor across the 33 uF.  Even with this large capacitor, the flash rate was still too fast.  And the flash was very weak, not really visible.  The supply current was very low, only a milliamp or less.

So at this time, I decided that if I was going to use an even bigger timing capacitor, I would have to use a smaller timing resistor to give the transistor more base bias and help brighten the LED (I used an amber LED).  I put a 33k resistor across the 120k, which was then effectively 25.8 k.  The flash rate sped up, but the LED didn’t get much brighter.  And at this time, I figured that if I lowered the timing resistor down to where the flash was much brighter, the timing capacitor would have to be a thousand microfarads or more.  Nix to that!  My Supercharged Joule Thief Flasher does a much better job with a capacitor of a few uF.

So now what should I do?  I’ll change it back to a Joule Thief, but I will just leave it like it is, with the feedback winding to negative.  So I removed the capacitor and put a 0.1 uF capacitor in its place.  Well, of course the flash rate sped up so fast that the LED looked continuously on to my eye.  But it was still too dim.  The frequency was 4567 kHz.

I removed the 0.1 uF capacitor and put a .01 uF capacitor in its place.  The LED was brighter and the narrow pulses were getting closer together, as I saw them on my ‘scope.  With the scope on the collector, the waveform was a narrow low positive voltage spike, followed by a much more negative voltage spike.   The frequency was 46 kHz, but there was still a lot of ‘dead time’ between spikes.  I figured that if I decreased the timing capacitor even further, the spikes would get even closer together and less dead time would mean a brighter LED.  So I replaced the .01 uF with a .001 uF or 1 nF.  The LED became brighter and the spikes were closer together.

Well, I figured that the spikes needed to be a bit closer, so I changed the timing capacitor to a 330 pF.  Now the LED was getting bright enough to see decently and the pulses were close together.  The frequency was 172 kHz.   Even so, the supply current was only 5 mA, which is less than a tenth of a normal Joule Thief.

Conclusion

Right now, I can say from this experience that this circuit the way it is, is not the optimum for a Joule Thief, and that it is definitely no good for a flasher, even though the circuit does flash dimly.  I will try to increase the base bias current by reducing the value of the resistor below 33k in parallel with 120k.

I have changed R1 to a 5.6k resistor, which was the value that gave about 50 milliamps supply current.  The amber LED brightened considerably.  I put a 0.47 ohm resistor in series with the LED so I could measure the LED current.  The LED current was maxed at 15.5 milliamps when I changed the capacitor to a variable capacitor and tuned it to 101 pF.  So I replaced it with a 100 pF disk capacitor. The frequency was almost 300 kHz and the supply current at 1.5V was about 50 milliamps.  That was very low for a current of 25.5 mA through the amber LED.

The amber LED was brightest when the variable capacitor was broadly tuned to a certain point.  This is much like the Supercharged Joule Thief, where the capacitor interacts with the coil’s feedback winding to give higher efficiency than a conventional Joule Thief.

I removed the amber LED and replaced it with a white LED.  This is so I can make an “apples to apples”comparison with the Supercharged JT.  I further reduced the R1 to 3.9k, to increase the supply current up to the point where the Supercharged JT usually is.  I was getting about 15 mA to the LED with a supply current of 48 mA.  The frequency was 593kHz which is very high, at the bottom of the AM broadcast band.  Some figures for calculating the efficiency:

3.3V * 0.015A = .0495 W or 49.5 milliwatts.

1.5V * .048A = .072W or 72 milliwatts.

49.5 / 72 = 68.75 percent.

So this circuit is on the very high end of efficiency of the conventional JT, but on the low end of the efficiency of the Supercharged JT.  I will have to make another one or two of these to see if the circuit is consistent with the parts I used.

I’m building a second one, so read my blog about it here.

Back to experimenting…

I received a google alert with this Youtube video of a Joule Thief which uses a C9011 transistor.  The notes give a link to the data sheet for this transistor.  The actual designation for this transistor is C9011, not 2SC9011 (see note at end).  As can be seen in the data sheet, the transistor’s absolute maximum collector current is only 40 mA.  Also, the Pc is only 400 mW – most transistors can dissipate 625 mW or more.  Notice that it says it is meant to be used as a converter and IF amplifier.  To me, this means that to get lower capacitance and better RF performance, the makers made the chip and junctions smaller than the typical transistor to reduce Cob for example.  That’s another reason why the current and power are lower.

The typical Joule Thief needs over a hundred milliamps from the battery to give the typical white LED the 20 milliamps that gives true “full brightness”.  This C9011 should not be made to go past 40 mA collector current, but it probably is, and very reluctantly.  It just can’t do it because the transistor is too small, and if you could hear it talk, it would be screaming and complaining about the current demands.  As a result, the circuit is severely limited in its power input and output.

The video claims that it gives “100% bright[nes]s” with a supply current of only 18 mA.  But assuming that it takes about five times as much supply current as the LED current, that means that the LED is getting only 18/5 or about 3.5 milliamps, which is much less than full brightness.  And this assumption that this circuit can put out 1/5 as much current is probably too optimistic; the actual current is probably even less because the transistor is being pushed far beyond its limits.  It is very likely that this circuit is being wasteful and inefficient because of the transistor is unable to handle the higher current.

The proper way to do this is not by using a wimpy transistor, but by using the proper transistor and changing the base bias resistor to a higher value to get the battery current down to 18 mA.  The transistor will then be coasting along and doing a much better job of converting to the higher voltage, with less penalty in efficiency.  And the transistor will not be driven past its limits, protecting the circuit against failure.

Some transistors to use could be the 2N2222A, PN2222A, 2N4401, BC337 or at this lower current, even the 2N3904 or BC547 would do a better job.  If you like those C90xx parts, then use the C9012 or C8050, which can handle more current than the BC337.  But whatever you use, don’t push the transistor beyond its limits.

Note: There was a need for a line of transistors for high volume, very low cost use, in very low prices consumer equipment.  The line was given the designations of 9011 through 9015 and 8050 and 8550.  The numbers are often preceded by a letter such as C or S, depending on the manufacturer.  I’ve seen datasheets for them from many companies such as BEL (India), KEC (Korea), and many Japanese companies.   These transistors are often found in the low end consumer devices, such as transistor radios.

Back to experimenting…

Update May 25, 2013 – I bought a bunch of Fairchild SS9018 VHF transistors, just to make regenerative receivers.  These are supposed to be capable of about 700 MHz, 2 or 3 times what the typical 2N3904 can do.  Mouser had them for really cheap.