### Author Topic: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?  (Read 14013 times)

0 Members and 1 Guest are viewing this topic.

#### NotionalLabs

• Supporter
• Posts: 23
• Country:
##### Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« on: July 13, 2016, 11:31:04 am »
Hey everyone,

I've been playing with attempting to make a regulated joule thief to learn more about boost techniques and to see if it's possible to run a low-power micro from an AA or supercap for an extended period, etc. I've encountered some behaviour when experimenting with different supply voltages via my bench PSU that I'm not sure I understand, so I'd appreciate your thoughts.

I have a Tenma 72-2540 30V/5A PSU that I've been using to test the voltages. The joule thief circuit is based on this instructable: http://www.instructables.com/id/Cheap-Voltage-Regulated-Joule-Thief-power-supplych/?ALLSTEPS:

The only modifications to this are a 1K resistor to the transistor base as in BigClive's joule thief youtube video, a 16v 470uF cap (what I had lying around) for the output regulation, and I'm missing the output regulating zener (it's in the mail). I'm using a 2N2222 transistor and I got about 17 paired turns of 26AWG enameled magnet wire for the toroid.

The basic joule thief appears to work, it oscillates at about 210kHz and with a single red LED load on the output I see a 1.5v AA boost to 2.1V and the LED lights.

The confusion comes when trying to characterise how this circuit will behave under different source conditions.

Setting the PSU to:
1V 100mA = the LED lights (boosts to about 1.9V), the PSU is in constant voltage mode, about 1V is drawn at 55mA.
2V 100mA = the LED lights (boosts to about 2V), the PSU is in constant current mode, 1.5V is drawn at 100mA.
2V 200mA = the LED dimly lights (output drops to 1.85V), the PSU is in constant current mode, 2V is drawn at 200mA, and the transistor gets very hot .

My questions are:

- Why does the PSU switch to constant current mode when I set the PSU to 2V at 100/200mA? My current theory is that the winding to the CE path of the transistor will appear to be very low resistance when the transistor is fully saturated, triggering the PSU's switch over to Constant Current mode. Is it possible to force it to stay in Constant Voltage mode (although I imagine the PSU is switching to CC mode in the assumption that if the load is very low resistance you're either going to hit the OCP limit immediately or set something on fire/blow it up)?

- How does the resistor on the base of the transformer affect the current draw of the joule thief? Is it solely for protection? If I increase this, will it affect how the PSU's CV/CC detection circuit perceives the circuit load?

Ultimately, I was thinking of seeing if I could trickle-charge a supercap (likely 2.7V) using a small solar panel, and using it to power a very low-power micro (ATTINY45V) for an extended period (overnight) using a joule thief or some other boost circuit to meet the voltage requirements of the micro for the maximum possible time (the ATTINY45V operates as low as 1.8V, I believe). I know there are probably way more efficient ways to go about this (which is likely to be the topic of another post in this forum fairly soon...), but I'm currently interested in learning more about the theory of operation of the joule thief. As the supercapacitor can discharge huge currents very quickly I'm concerned that not being able to regulate the current draw to something sane means this idea is a no-go.

Thanks!

#### MosherIV

• Super Contributor
• Posts: 1530
• Country:
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #1 on: July 13, 2016, 12:07:53 pm »
Quote
Why does the PSU switch to constant current mode when I set the PSU to 2V at 100/200mA?
At 1V there is just enough voltage to enable the transistor to operate, it takes 2 x 0.7V to make a silicone transistor operate correctly. At 2V the transistor can now fully be turned on. When the transistor is fully on, the only load the PSU sees is the resistance of the coil. Hence the PSU cannot maintain the Voltage and switches into constant current mode. Since this happens at 210KHz, the PSU looks like it is in CC mode but if you look with an oscilloscope, you should see it either going back to CV (V goes back upto 2V) or there is not enough time for it to recover into CV mode.

Quote
How does the resistor on the base of the transformer affect the current draw of the joule thief?
Transistors are controlled (turned on) by current flowing into the base junction. Changing the resistor is adjusting the base current and therefore changing the bias (how turned on) the transistor is

At least that is my understanding of what is going on

#### NotionalLabs

• Supporter
• Posts: 23
• Country:
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #2 on: July 13, 2016, 01:13:13 pm »
Transistors are controlled (turned on) by current flowing into the base junction. Changing the resistor is adjusting the base current and therefore changing the bias (how turned on) the transistor is

From my understanding of the joule thief principle, doesn't the secondary CE winding cause a chain-reaction increase in current to the base of the transistor by inducing current in the primary base winding (which in turn opens the CE gate more, resulting in more induced current, etc...) until the transistor is saturated? I imagine changing the resistor to a lower value would result in a higher frequency oscillation (the transistor would saturate more quickly each cycle) at the cost of drawing more current from the source?

Is my understanding correct?

• Contributor
• Posts: 45
• Country:
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #3 on: July 13, 2016, 01:18:01 pm »
Yes I believe that is basically how it works apart from the fact that it is the inductor which saturates and then causes the base current flow to stop and the field collapses which provides the current to the output. This is where you can tweak things as the frequency tends to rise with input voltage. You can increase the value of the base resistor to make the circuit work better at higher voltages.

#### MosherIV

• Super Contributor
• Posts: 1530
• Country:
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #4 on: July 13, 2016, 01:34:21 pm »
Quote
From my understanding of the joule thief principle, doesn't the secondary CE winding cause a chain-reaction increase in current to the base of the transistor by inducing current in the primary base winding (which in turn opens the CE gate more, resulting in more induced current, etc...) until the transistor is saturated? I imagine changing the resistor to a lower value would result in a higher frequency oscillation (the transistor would saturate more quickly each cycle) at the cost of drawing more current from the source?

Is my understanding correct?
Yes, someone else beat me to it

Does the explanation of why the PSU go into constant current still satisfy you?

#### SteveyG

• Supporter
• Posts: 993
• Country:
• Soldering Equipment Guru
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #5 on: July 13, 2016, 02:08:45 pm »
What are the characteristics of the ferrite material used? Is the transformer saturating?
Use code: “SDG5” to get 5% off JBC Equipment at Kaisertech

#### Ian.M

• Super Contributor
• Posts: 12977
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #6 on: July 13, 2016, 02:49:20 pm »
One way of  regulating a Joule thief without wasting (much) power is to shut off the base drive of the switching transistor once the desired output voltage is reached.   To do this easily you need another transistor to short the switching transistor base to ground. The feedback transistor's base can be driven from the output via a potential divider (crappy regulation) or via a Zener + current limiting resistor (much better regulation).

#### NotionalLabs

• Supporter
• Posts: 23
• Country:
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #7 on: July 13, 2016, 04:04:26 pm »
One way of  regulating a Joule thief without wasting (much) power is to shut off the base drive of the switching transistor once the desired output voltage is reached.   To do this easily you need another transistor to short the switching transistor base to ground. The feedback transistor's base can be driven from the output via a potential divider (crappy regulation) or via a Zener + current limiting resistor (much better regulation).

This sounds very interesting - I have just received the current limiting Zener diodes (3.3v) I was going to experiment with, so I'd like to try this approach out. I've attached an updated schematic of how I *think* it would look, do you mind taking a quick look to see if I'm on the right track?

EDIT: I just realised after posting it that I think I've put the Zener on the wrong side of the connection to the regulation transistor base and the +ve rail!

What are the characteristics of the ferrite material used? Is the transformer saturating?

This is a good question. I salvaged a spare ferrite toroid from an ancient scrap graphics card that I was practising hot-air rework on and used that - its very possible that the ferrite core isn't the right spec. Having said that, here's a single-shot of the joule-thief oscillations (test points are between the collector and emitter of the transistor) in case you can judge it's effectiveness from this:

I'm not sure how to tell if the toroid is magnetically saturated - do you have any tips?

Does the explanation of why the PSU go into constant current still satisfy you?

Yes - many thanks to you and the other posters, I feel like I'm starting to understand the parameters of the circuit better now. Incidentally, I noticed that if I enable OCP then the PSU is trusting enough to not automatically switch into CC mode, so that's one way to force it (albeit at the risk of allowing the coil/transistor to pull enough current to blow themselves up if you set the OCP limit too high!).
« Last Edit: July 13, 2016, 04:06:54 pm by NotionalLabs »

#### Ian.M

• Super Contributor
• Posts: 12977
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #8 on: July 13, 2016, 04:36:01 pm »
Yep the Zener's in the wrong place. It should be K to Vout, A to Feedback TR base.  You may want a 1K resistor in series with the Zener to limit the base current.
Also you *MUST* add a reservoir cap to the output - try 10uF.

If you don't have a decoupling cap across the supply input, the lead inductance back to the PSU  can cause problems.  Joule thiefs only get away without one if they are connected via short wires directly to a battery.
« Last Edit: July 13, 2016, 04:51:39 pm by Ian.M »

#### T3sl4co1l

• Super Contributor
• Posts: 21974
• Country:
• Expert, Analog Electronics, PCB Layout, EMC
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #9 on: July 13, 2016, 05:03:16 pm »
You want a circuit like this, which controls base bias:

Note that, when the left PNP is fully on, it's just the same thing, 22 ohms into the base (well, 22 is more current than 47 ohms, but that just means more available output power).  Full throttle.  But as the output voltage rises, the right PNP starts to turn on, which pulls off the left PNP, reducing base bias, and reducing the switching frequency.

Don't use a ferrite toroid!

The right kinds of cores are identifiable:
- Powdered iron toroids: enameled or painted.  Two-color cores are usually a www.micrometals.com designation (one or two digits).  These are generally suitable -- but, the most common grades are also the most lossy, which makes them really poor for this kind of circuit.
- Solid color toroids: various manufacturers.  Includes ferrites, so you'll have to test these to be sure.
- Bare, charcoal-gray to black, toroids: ferrite -- unsuitable, does not store much energy.
- Gapped ferrite: rods, dumbbells and other cut core shapes ('C', 'E', etc.).  These store energy (in the air gap), and have low losses.  Excellent for this.

If you have bare, ungapped cut cores (i.e., the parts fit together flush, with no air gap), you can customize the air gap by adding a spacer between the pieces.  The required gap, and number of turns, can all be calculated.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!

The following users thanked this post: oPossum

#### NotionalLabs

• Supporter
• Posts: 23
• Country:
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #10 on: July 14, 2016, 02:27:19 pm »
Don't use a ferrite toroid!

The right kinds of cores are identifiable:
- Powdered iron toroids: enameled or painted.  Two-color cores are usually a www.micrometals.com designation (one or two digits).  These are generally suitable -- but, the most common grades are also the most lossy, which makes them really poor for this kind of circuit.
- Solid color toroids: various manufacturers.  Includes ferrites, so you'll have to test these to be sure.
- Bare, charcoal-gray to black, toroids: ferrite -- unsuitable, does not store much energy.
- Gapped ferrite: rods, dumbbells and other cut core shapes ('C', 'E', etc.).  These store energy (in the air gap), and have low losses.  Excellent for this.

That's a great rundown of magnetic cores, thanks! I took a look at my toroid and I think it might be a powdered iron variety. It is enameled, has one side painted yellow, has a yellow dot on one side, and is attracted to magnets:

The left photo is the current state of my joule thief with my hand-wound toroid in place. Is there any way to confirm if it is a powdered iron toroid or a ferrite?

You want a circuit like this, which controls base bias:

Note that, when the left PNP is fully on, it's just the same thing, 22 ohms into the base (well, 22 is more current than 47 ohms, but that just means more available output power).  Full throttle.  But as the output voltage rises, the right PNP starts to turn on, which pulls off the left PNP, reducing base bias, and reducing the switching frequency.

This is very interesting, I will definitely give this a go - would a couple of 2N3906's be sufficient for the left and right PNPs? I'm curious, why is the Zener's cathode connected to the base of the right PNP rather than the Vout directly?

Yep the Zener's in the wrong place. It should be K to Vout, A to Feedback TR base.  You may want a 1K resistor in series with the Zener to limit the base current.
Also you *MUST* add a reservoir cap to the output - try 10uF.

If you don't have a decoupling cap across the supply input, the lead inductance back to the PSU  can cause problems.  Joule thiefs only get away without one if they are connected via short wires directly to a battery.

This is very good to know - I was wondering why some of the example circuits I've seen had decoupling caps on the input.

#### T3sl4co1l

• Super Contributor
• Posts: 21974
• Country:
• Expert, Analog Electronics, PCB Layout, EMC
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #11 on: July 15, 2016, 12:00:04 am »
That's a great rundown of magnetic cores, thanks! I took a look at my toroid and I think it might be a powdered iron variety. It is enameled, has one side painted yellow, has a yellow dot on one side, and is attracted to magnets:
...
The left photo is the current state of my joule thief with my hand-wound toroid in place. Is there any way to confirm if it is a powdered iron toroid or a ferrite?

Hmm, looks like gray and black to me, which would put it as a #38 (permeability 85).  Yellow-gray would be #35, yellow-clear would be #6, and white-clear, #7.

Do you have an o-scope and signal generator?

Quote
This is very interesting, I will definitely give this a go - would a couple of 2N3906's be sufficient for the left and right PNPs? I'm curious, why is the Zener's cathode connected to the base of the right PNP rather than the Vout directly?

Yes, and the NPN should be something a little beefy, like a ZTXxxx.  My 1W flashlight,

uses a #35 toroid and a PBSS303NX.  With max base bias (100 ohms; there's a bypass capacitor after it, which performs better than the big-series-base-resistor circuit), it runs something like 500kHz, and 70% efficiency at 1.5V supply.

A '2222 won't be good for much more than 300mA or so, peak, or on the order of 75-100mA average input (and proportionally less at the output!).  So keep that in mind if you're after a lot of power.

Quote
This is very good to know - I was wondering why some of the example circuits I've seen had decoupling caps on the input.

Single cells aren't too bad, but ESR eats into performance especially as they expire.  Pictured above, I have a nice thick 22uF 6.3V ceramic cap across the supply, which does a fine job at the switching frequency.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!

#### NotionalLabs

• Supporter
• Posts: 23
• Country:
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #12 on: July 15, 2016, 02:09:47 pm »
Hmm, looks like gray and black to me, which would put it as a #38 (permeability 85).  Yellow-gray would be #35, yellow-clear would be #6, and white-clear, #7.

Do you have an o-scope and signal generator?

I have a Rigol 1054Z (100MHz bandwidth upgrade installed) oscilloscope, but I'm afraid I don't have a signal generator.

Yes, and the NPN should be something a little beefy, like a ZTXxxx.  My 1W flashlight,

uses a #35 toroid and a PBSS303NX.  With max base bias (100 ohms; there's a bypass capacitor after it, which performs better than the big-series-base-resistor circuit), it runs something like 500kHz, and 70% efficiency at 1.5V supply.

A '2222 won't be good for much more than 300mA or so, peak, or on the order of 75-100mA average input (and proportionally less at the output!).  So keep that in mind if you're after a lot of power.

Nice flashlight! I've taken a stab at your proposed alternative:

I don't really have any NPN's laying around much beefier than the 2N2222, unfortunately. I also realised that I should probably be using a 10,000pF ceramic on the input decoupling rather than the rubbish 1uF electrolytic I had lying around. Still, the circuit is substantially better than before, with much cleaner behaviour and better efficiency. I can squeeze about 1.8V out of it from as low as 0.6V at 34mA input draw before it bombs out.

I am having some issues with the expected output though - my Vref zener has a Zv of 3.3V (1N746A), yet I only generally see a regulated output of up to 2.6V-ish. There seems to be a threshold where if I increase the voltage past 2.9V, it jumps to the 3.3V regulation. The output exhibits an interesting waveform when it's under the 2.9V threshold:

And an even more interesting one above 2.9V input:

My immediate theories are:
• My coil is insufficient due to the number of turns (about 22) or the core material (or both), to store and discharge enough energy to raise the output above the Zener voltage.
• The electrolytic I'm using for the output reservoir cap is a piece of crap and probably has ridiculous ESR

Any ideas what might be the cause of this behaviour?

#### bktemp

• Super Contributor
• Posts: 1616
• Country:
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #13 on: July 15, 2016, 02:31:05 pm »
• The electrolytic I'm using for the output reservoir cap is a piece of crap and probably has ridiculous ESR
That would be my guess. You could try adding a 10uF MLCC.

#### MosherIV

• Super Contributor
• Posts: 1530
• Country:
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #14 on: July 15, 2016, 02:48:07 pm »
Quote
I am having some issues with the expected output though - my Vref zener has a Zv of 3.3V (1N746A), yet I only generally see a regulated output of up to 2.6V-ish.
Quote
Any ideas what might be the cause of this behaviour?
Not entirely sure. What happens if there is no load?

What you have is a 'poor man's buck regulator' without the closed loop control.
In other word, the circuit oscillates at some freq based on the feedback between the transformer and the transistor, and the higher voltage is generated by the magnetic field collapse of the inductor/transformer but there is no control of this voltage other than to try and tweek the oscillation frequency.

(FYI It is inefficient because there is always current flowing through the inductor.)

#### T3sl4co1l

• Super Contributor
• Posts: 21974
• Country:
• Expert, Analog Electronics, PCB Layout, EMC
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #15 on: July 15, 2016, 08:08:59 pm »
Beautiful waveforms!  Thanks to the capacitors' high ESR, we can see exactly what's going on!

The rampy sawtooth stuff means the transistor is turning off pretty sharply, and turning back on after a suitable time period.

The hold-off time varies with base bias, so the PNPs are also doing their job.

The zener, too, is doing what it does best; low voltage zeners (under 6V) have terrible leakage, so you should expect poor regulation.

The humpy sawteeth suggest the PNPs aren't quite doing their job, as well as we might hope; in other words, the loop is going unstable under some conditions.  Or, it may be a sign that your base drive is too high, or not bypassed well enough (in the extreme case, >5Vp-p base drive causes E-B breakdown, magnifying the base bias current and the oscillator runs at full power all the time -- destructive!).

You want base drive to be about 1.5-3Vp-p, for whatever the collector voltage waveform is (for a 3.3V output, it should be going between about 0V and 3.9V peak, with 1.5V average inbetween, so a 1:1 winding should be okay here).

To implement loop compensation, try a C or R+C across one of the PNPs, base to collector.  You'll want to check if the oscillator is behaving itself, first.

And yes, you want more like 1uF ceramic (or more) on input and output.  Or more like 100uF+ electrolytic.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!

#### BravoV

• Super Contributor
• Posts: 7547
• Country:
• +++ ATH1
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #16 on: July 15, 2016, 08:15:32 pm »
Don't use a ferrite toroid!

The right kinds of cores are identifiable:
- Powdered iron toroids: enameled or painted.  Two-color cores are usually a www.micrometals.com designation (one or two digits).  These are generally suitable -- but, the most common grades are also the most lossy, which makes them really poor for this kind of circuit.
- Solid color toroids: various manufacturers.  Includes ferrites, so you'll have to test these to be sure.
- Bare, charcoal-gray to black, toroids: ferrite -- unsuitable, does not store much energy.
- Gapped ferrite: rods, dumbbells and other cut core shapes ('C', 'E', etc.).  These store energy (in the air gap), and have low losses.  Excellent for this.

That's a great rundown of magnetic cores, thanks! I took a look at my toroid and I think it might be a powdered iron variety. It is enameled, has one side painted yellow, has a yellow dot on one side, and is attracted to magnets:

The left photo is the current state of my joule thief with my hand-wound toroid in place. Is there any way to confirm if it is a powdered iron toroid or a ferrite?

If you have the right cutting tool, like rotary wheel, cut one like the one at the left below, you will see an increase in the inductor's saturation current capabilty, aka higher energy store as Tim stated above.

See the gaping cut ? Of course, it will need more turns once gapped.

#### alsetalokin4017

• Super Contributor
• Posts: 2055
• Country:
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #17 on: July 17, 2016, 03:59:38 am »
Slightly off-topic... but here's my favorite JT:

(The inductor is a 10mH unit from an old CRT chassis, with a 20 turn primary wound on top of the stock winding. The heat sink on the 2n3055 is just for show.)
The easiest person to fool is yourself. -- Richard Feynman

#### Kilrah

• Supporter
• Posts: 1852
• Country:
##### Re: Bench PSU + Joule Thief Experiment: Constant Current vs Voltage?
« Reply #18 on: July 17, 2016, 01:20:44 pm »
Relevant video I just came across:

Smf