The attached oscillator works pretty well in SPICE, supply voltage is 19V and the transformer/inductor is wound on a small toroid. In the model I can change pretty much any component by a factor of 10 and it'll still oscillate. Design frequency is around 75 kHz

After building it and several days messing around I still can't get it to take off. I think the ferrite bead I used for my first attempt was just too lossy (looked like 10 ohms series R at 100 kHz but the inductances were right) but an air-core one and one with better ferrite didn't work either.

Tried so far:
Tweaking the turns ratio, inductances, core material.
Reducing the 10k for more bias
Shunting the 100R with a capacitor (more ac gain)
Shunting the 10k with a capacitor to give it more startup kick
Winding the supply up to 30V
Bypassing the input filter for more of a startup kick
Reversing the feedback winding in case I screwed up the phasing
Different transistors
Shouting at it

Any ideas please? It just doesn't have the look of a marginal design in the simulator so I suspect I've missed something obvious.

 

What type of oscillator is that?

 

I don't know what it's called but I can describe how it operates.

R1, R2, R3 set the d.c. conditions with around 1V on the emitter and 1.7V on the base, 10 mA flowing through the collector load.

The tuned collector load (L2 and C2) is excited at switch on and by transformer action (3:1 turns ratio) a small fraction of the voltage across L2 couples into L1 and modulates the base current in anti-phase with the collector voltage and the oscillation builds until limited by the supply voltage with collector swinging between roughly 0V and 2x V_in.

The limiting condition is pretty unkind to the transistor as its collector is pulled below its emitter but the second half of the circuit which is currently unconnected would clamp the the voltage below this limiting point. That section has been removed as I thought the loading might hinder oscillator startup.

When oscillating the average collector current should be equal to that set in the d.c. state and about 10x this circulates in the tuned circuit.

Of course given that it doesn't oscillate I'm not entirely sure what to do next.

 

OK, I missed the transformer coupling between the two inductors.

So how many turns are on the transformer core? How big is the core?

 

Have you tried swapping the leads on either one of the coils? The feedback must be in the proper phase to be regenerative.

 

It's some kind of joultheif but with superfluous components. Or do you have them to be able to use it at higher voltages?

 

The turns ratio is around 3:1 and I've tried it with:

1. L2 20 turns on a large ferrite bead approx 12 mm diameter, 3 mm thick, 8mm long. That worked out about right with L2=400uH and L1=38uH but the core was too lossy - when I put it on the LCR meter I saw around 20 ohms in series with L2 at 100 kHz (in spice the circuit doesn't oscillate with that above around 7 ohms).
2. Similar but with 25 turns on a small toroid 15mm diameter, 5mm thick, 6 mm long.
3. Air core with 300 turns on L2, around 12 mm diameter and 25 mm long. Measured as 100 uH and 2 ohms which is within the range spice predicts to work. I did that in an attempt to eliminate core loss and saturation from the possible problems.

The circuit is related to the joule thief but with the following differences:

• R2 and R3 and the different turns ratio make it more suitable for higher supply voltages. R3 forms a divider with R1 that reduces the voltage on the base to something reasonable, R2 provides feedback to limit the maximum collector current, the 3:1 turns ratio reduces the base drive.
• In the .asc there are some other extra components not on the png schematic that are just supply filtering and output.
• The biggest difference is the presence of C2, It totally changes the operation of the circuit.

In a joule thief the current in the collector inductor builds until either the core saturates or the transistor comes out of saturation (a joule thief can work in either mode and may change between them depending on load, supply voltage, etc). The transistor then turns off hard and there is a big voltage spike on the collector that typically powers an LED, the inductor current drops to zero, the transistor turns back on and the cycle repeats. The voltage on the collector is basically a series of large spikes and the inductor current is roughly triangular in time.

In this circuit when the transistor turns off the current in the inductor flows into C2 and the tuned circuit formed by L2 and C2 rings at around 75 kHz for the component values shown. The feedback mechanism is largely the same. Both collector voltage and inductor current are approximately sinusoidal and when the oscillation has built up to the limit there is a sine wave of approx 2*V_supply on the collector.

The extras in the .asc sniff off a fraction of this and generate an output that can supply a few mA about 5V above V_supply. The sine wave is a lot nicer from an emc perspective than the output of a joule thief and given I only want a modest voltage step up I'd prefer to get this working than to do something similar with a flyback type circuit.

Reversing the feedback winding was the first thing I tried. Given I have a design that's pretty robust in simulation (I can change most values by a factor of 10 and it'll still oscillate) that suggests the problem is some non-ideality not included in the spice model. Inductor saturation and core loss are the obvious ones but then I'd have expected it to go with the air-core version.

 

Well my suggestion is not to rely so much on inductive coupling. Small transformers can easily become split into two different coils, replacing the coupling with a small capacitor.

And why do the extra work for 300 turns?

I had the idea too, that the purpose is to work it at higher voltages.

For some mA you don't need large ferrites as you mentioned, even very small one's will work. One time I built a joultheif using a small ring originally intended for magnetic core memory. Only a few mm in diameter!

Or maybe use a voltage divider, that gives a voltage just 2V below Vcc, and connect this as ground. Then if required use a seperate winding to draw off the stepped up voltage. This would totally do away any kind of efficiency, however.

I like to see detailed and qualified replies like here in the thread.

 

The 300 turns was more out of desperation - I figured if I could get it working with an air-core then it proved my problem was with the inductor, it actually took less time to wind than the toroids. The larger ferrites are easier to wind and space isn't really a problem here .

Can you explain your capacitative coupling suggestion - I can't see how to get the required phase inversion for the feedback with that method unless I use a second transistor.

A little about the application for those who were wondering - I often find myself wanting a supply a few volts above the main supply of a circuit to run op-amps or drive mosfet gates. In the past I've added an extra winding to mains transformers or capacitively coupled a second rectifier off of the secondary but neither works if you're using an external psu. In this case the project is powered of a 19 V 4.5A laptop brick and I could do with a rail at around 25V. There are other ways of achieving it (I've even been down the LED + solar cell route) but I'd like to come up with a simple, reliable circuit I can use in future designs whenever I want such a supply (or to make a negative rail if the output is capacitively coupled).

 

The 300 turns was more out of desperation - I figured if I could get it working with an air-core then it proved my problem was with the inductor, it actually took less time to wind than the toroids. The larger ferrites are easier to wind and space isn't really a problem here .

Can you explain your capacitative coupling suggestion - I can't see how to get the required phase inversion for the feedback with that method unless I use a second transistor.

A little about the application for those who were wondering - I often find myself wanting a supply a few volts above the main supply of a circuit to run op-amps or drive mosfet gates. In the past I've added an extra winding to mains transformers or capacitively coupled a second rectifier off of the secondary but neither works if you're using an external psu. In this case the project is powered of a 19 V 4.5A laptop brick and I could do with a rail at around 25V. There are other ways of achieving it (I've even been down the LED + solar cell route) but I'd like to come up with a simple, reliable circuit I can use in future designs whenever I want such a supply (or to make a negative rail if the output is capacitively coupled).

OK here you go- I am not so good with theory and formula, even if I have related literature available, and more than 300k pages datasheets. Too much to read it all literally!

I found your explanation interesting to read.

The circuit oscillates inside LTSpice, and it also should be possible to make it work in real world. I have built a similar circuit a while ago. You get a voltage swing even with 200R inlined.

Originally this circuit was a joulethief, I tried to make it work using a small EMI filtering inductor. Was not possible until I added the capacitor! Then I figured out, it also works, even better, with two discrete inductors.

I looked up the theory, and as far as I understand, inductive coupling can indeed become replaced with a capacitor.

 

Thanks for that. I think it'll take some time to get my head around it but it looks promising for the future. It also should make t possible to use the wire ended inductors which would simplify construction quite a bit. Given that I still don't understand why the circuit didn't work I wouldn't entirely trust my explanation but it's probably pretty close.

I have now got it oscillating by re-arranging the feedback network. The attached schematic shows exactly the circuit I now have and with the zener disconnected I get a nice smooth 34V from the output at C5. Input current is around 15 mA, I haven't tried loading the output yet.

The collector waveform is far from sinusoidal though, much more so than predicted by the simulation. Spice predicts it becomes closer to a sine wave but with lower amplitude if I reduce C6. That fits as I would effectively be lowering the excess gain.

 

If only Tesla would have had transistors available for his experiments.

My circuit is unusual a bit there seems to be a large circular current in the inductors, however emitter current is only 1/10 of that. For simulation purposes I have assigned resistance to the inductors. At 2 Ohms nothing oscillates. At 1 Ohms, oscillation builds up rather slowly. It is still a riddle to me if in reality this circuit indeed maintains a large circular current, and what flows through the transistor mainly is the amount that is lost through physical effects.

Anyway, it works in real world.

I normally use small electronic transformers, since they are inexpensive. Never tried an extra voltage rail to drive (N-channel?) MOSFETs.

 

If it falls over at more than an ohm then maybe I can't get away with the little axial inductors - as you can see from my model the 220 uH ones I'm using for input and output filtering have 10 ohms series R.

Mine has large circulating currents too, I think it's characteristic of any oscillator that uses an LC tank circuit.

In this case the extra rail is to supply an op-amp which has to be able to drive the gate of an N-channel MOSFET whose source is tied only a few volts below V_in. With an 8V2 zener my new rail gives me around 27 V which is perfect.

 

If it is using inductive coupling it is an Armstrong oscillator.