Two transistor non-inverting Ćuk converter.

[Refrigerator]

The recent GreatScott video reminded me of the Ćuk converter and pointed out that any boost converter can be converted into a Ćuk converter.
So as an exercise i decided to take my two transistor boost converter and modify it into a Ćuk converter in LTspice.
Haven't played around in LTspice for a while so it took some getting up to speed, but i managed to get the converter going.
My two transistor boost converter is only good for a narrow input voltage range, so it's not the most practical application for the Ćuk converter, IMO.
But it does work and when boosting gets about 60% efficiency at a 10W output without any optimizations.
My boost converter circuit was optimized for a single cell lithium battery as input and the same input was used in this simulation also.
I noticed that in the output voltage range of 0 to double the input voltage it works pretty well but anything above and it starts to struggle and needs more tuning to work.
But there's only so much you can achieve by tuning the circuit any further.
Voltage control is achieved through a zener, but i also tried using an NPN transistor to control the output voltage.
When using the NPN transistor i noticed that, when regulating the voltage, the efficiency would drop massively because the circuit would struggle to oscillate efficiently.
Using a zener diode to control the voltage works a little better.

Anyways, it can be done and that's all i wanted to find out. 

[T3sl4co1l]

Better known as SEPIC.  Cuk is specifically the inverting kind.

Having played around a lot, years ago, I don't think there are any real great two-transistor no-transformer converters; with transformer, you can have a blocking oscillator with peak current mode control, which can be zener controlled for somewhat mediocre regulation.  Add a 3rd transistor (as opto) for isolation.

I guess you've been feeling out those sorts of problems; the narrow voltage range and low efficiency are fairly typical in that respect, I think.  Though doing it at fairly low voltage here, may be hitting efficiency more than my experiments did (mostly in the 12V range), so maybe it's not too bad, that considered.

For reference, I'd made a 1W "joule thief" with around 60% efficiency from a single 1.25V cell; which requries a coupled inductor, of course.

Speaking of coupled inductors, SEPIC/Cuk of course perform better with them; it's not required, but it helps.  Not a problem for efficiency per se, you can always use bigger/better caps and inductors to account for that.  It's just bulkier, so a matter of cost/size optimization at same efficiency.  Which, if your components are good enough then you can discount them as loss sources, and focus on just the switching aspect, its losses.

Around 7 transistors, you can get behavior closer to something like UC3843: free-running oscillator, maybe frequency modulated, or some hybrid of fixed-on/off-time and current mode.  And you can generate sharper edges, and can afford drivers for MOSFETs, netting higher efficiency.

I'm also fond of using TL431 or relatives, since, that's just one single very fancy transistor...right? RIGHT?!

Tim

[Zero999]

There are three transistors in that circuit. Where does Q2's collector go? Presumably it should be connected to M1's gate and isn't being used as a diode?

Adding an emitter follower, with a diode between the  base and emitter could help reduce M1's turn-on time.

[Refrigerator]

Better known as SEPIC.  Cuk is specifically the inverting kind.

Having played around a lot, years ago, I don't think there are any real great two-transistor no-transformer converters; with transformer, you can have a blocking oscillator with peak current mode control, which can be zener controlled for somewhat mediocre regulation.  Add a 3rd transistor (as opto) for isolation.

I guess you've been feeling out those sorts of problems; the narrow voltage range and low efficiency are fairly typical in that respect, I think.  Though doing it at fairly low voltage here, may be hitting efficiency more than my experiments did (mostly in the 12V range), so maybe it's not too bad, that considered.

For reference, I'd made a 1W "joule thief" with around 60% efficiency from a single 1.25V cell; which requries a coupled inductor, of course.

Speaking of coupled inductors, SEPIC/Cuk of course perform better with them; it's not required, but it helps.  Not a problem for efficiency per se, you can always use bigger/better caps and inductors to account for that.  It's just bulkier, so a matter of cost/size optimization at same efficiency.  Which, if your components are good enough then you can discount them as loss sources, and focus on just the switching aspect, its losses.

Around 7 transistors, you can get behavior closer to something like UC3843: free-running oscillator, maybe frequency modulated, or some hybrid of fixed-on/off-time and current mode.  And you can generate sharper edges, and can afford drivers for MOSFETs, netting higher efficiency.

I'm also fond of using TL431 or relatives, since, that's just one single very fancy transistor...right? RIGHT?!

Tim

My two transistor boost circuit had a real world efficiency of 86% (or was it 88 i can't remember) at around 5W so i wouldn't call it inefficient.
But it's a dumb converter because there's no protection so when it blows it goes kablamo. 
The input voltage range is limited because the entire control circuit is just a self-biasing inverting BJT amplifier and swings in the supply voltage offset the bias current.
But more importantly it offsets the idle voltage (let's call it) on the gate of the MOSFET, which is the voltage you would get without the feedback capacitor.
Because this voltage is directly responsible for "how hard" the converter wants to work, that's actually how i get it tuned.
If it doesn't work, most of the time it's because it's turned on too hard or not enough, it's easy to adjust it to work when you know this voltage.
That and the gate drive voltage comes directly from the supply voltage, so it's limited to that as well.
An emitter follower gate drive transistor with a zener diode reference to limit the gate voltage would make the circuit immune to input voltage changes.
But then it wouldn't be 2 transistors anymore 
I have used my 2 transistor boost converter to boost the voltage from a single cell lithium battery to 32V to power a 20W COB LED module for a DIY work light.
And it worked quite well but in reality it struggled to supply any more than 5W of power to the LED.
Although the brightness was just right for the kind of work i used it for.

Personally i like to consider the TL431 as a perfect MOSFET of sort, because it's voltage controlled and has a near infinitely steep I-V curve.
I've even used it as a MOSFET to control a string of LED's when i didn't feel like looking for BLT's small enough to not clutter up my breadboard.

[Refrigerator]

There are three transistors in that circuit. Where does Q2's collector go? Presumably it should be connected to M1's gate and isn't being used as a diode?

Adding an emitter follower, with a diode between the  base and emitter could help reduce M1's turn-on time.

It's in the OP:

Voltage control is achieved through a zener, but i also tried using an NPN transistor to control the output voltage.
When using the NPN transistor i noticed that, when regulating the voltage, the efficiency would drop massively because the circuit would struggle to oscillate efficiently.
Using a zener diode to control the voltage works a little better.

Edit: oh yeah i didn't say that it just shunts the gate of the MOSFET to turn it off.

[T3sl4co1l]

Oh yeah it was actually unconnected... that explains that.

Also, what's D3-D4 doing?

I recall a similar conversation with Roman Black; http://www.romanblack.com/smps/smps.htm he was especially concerned with doing it in just the two transistors, with any practical considerations almost irrelevant.  So, different priorities...

Tim

[Refrigerator]

Oh yeah it was actually unconnected... that explains that.

Also, what's D3-D4 doing?

I recall a similar conversation with Roman Black; http://www.romanblack.com/smps/smps.htm he was especially concerned with doing it in just the two transistors, with any practical considerations almost irrelevant.  So, different priorities...

Tim

D3-D4 shunt the current from the feedback capacitor because the peak current can be quite high.
It also helps improve the switching speed a little bit by not saturating the base of the BJT so much.
I think one diode might work as well but i put two in series to not interfere with the biasing of the BJT.

I remember the black converter because i accidentally recreated it by trying to make a two transistor buck converter.
I started at a two transistor multivibrator and worked towards a buck converter that worked and in the end what i had was something like the black converter just with three transistors.
I've since significantly improved on the original black converter and made one with a P-MOS transistor.
IIRC i achieved some 90%+ efficiency with my design at some 40W. I have like 20 variations of the design in my old laptop, including a constant current high power switching laser driver.

[Refrigerator]

So i mentioned the emitter follower gate driver and how it would make the design immune to input voltage variations previously and decided to just build it in LTspice to show what i mean.
And by doing so i proved myself wrong, because it does not make the converter completely immune to supply voltage variations but does help improve the input voltage range.
It also seems to have improved the stability of the output voltage but my previous circuit wasn't really optimized so it could be just a coincidence.
Also the input voltage is different now at 8V but the output stays the same.
The screenshot is a bit hard to read but D9 at the base of the gate drive transistor is what limits the voltage on the base of the MOSFET.

Edit: actually i calculated the efficiency of this particular circuit and it's 92.6%  I'll include the spice sim also.

[Refrigerator]

Looks like i grabbed a bad sample because the real world efficiency is lower.
It gets 84% around 6V input but doesn't manage to maintain a 10V output with a 10 resistor.
At 8V input it gets close to 10V output but efficiency drops to 78%.
So it's not bad, might be some room for improvement also.

Included is a screenshot of efficiency calculations with 10V output and input stepped from 2V up to 20V in 2V increment and a 10 load resistor.

[Refrigerator]

Some more testing results:
Changing R1 does not affect the efficiency much, i tried changing it from 2k to 20K and currently settled at 2k just because it works at 2V.
Changing the load resistor does not affect the output voltage much. I tried 5, 10 and 20 Ohms and they behave mostly the same and output reaches 10V only when input is 8V.
Increasing the load increases the efficiency, i noticed higher efficiency with the 5 load than the 20 load.

So some interesting finds, looks like the circuit might be useful for something but it's not really great at boosting.
But it's nice that it's sort of immune to short circuits and with the addition of the third transistor also seems to handle input voltage swings well.

Edit: this is for the 3 transistor version

[T3sl4co1l]

Try a resistor in series with C1, those diodes won't be needed then.

Tim

[Refrigerator]

Try a resistor in series with C1, those diodes won't be needed then.

Tim

As it turns out removing those diodes actually increased the efficiency a little bit.
But i did test out many resistor values to see if it has any effect on the efficiency and seems like it doesn't.
Increasing the resistance in the feedback loop only brings instability in the output voltage.