Simple discrete two transistor boost converter.

[Refrigerator]

I like my circuits discrete and as a bit of a challenge for myself i came up with a DC-DC boost converter that uses only two transistors.

Yes, i know, that's a MOSFET, not a BJT but it's a transistor nonetheless, a Field Effect Transistor, if you will.
The MOSFET could easily be replaced with a BJT, you would only need to increase R1 to 1K to do so, but the current output would be much lower.
Also this circuit is rather sensitive to the input voltage, the particular circuit above is designed to boost from a lithium cell to many voltages set by a zener diode D2.
Lowering R3 increases the frequency, same goes for C1, but you can't make them too low otherwise the circuit won't work.
To make the circuit work with higher input voltage increase R1.

I did make this circuit on some perf board but i couldn't find where i put my 2N3904s so i used a BC547, which worked out fine.
Load regulation is not bad, with some random zener i found the circuit gives me 5.65V no load and 5.5V at 1A, and operates at about 100kHz.
MOSFET is not too important as long as it has a gate voltage of 3V or less it will be fine.

I'll attach the LTspice file if anyone wants to play with this circuit.

[Zero999]

I've cleaned up the schematic. No one can read that spider's web.

See attached.


The simulation gives an efficiency of around 87%. In real life it will be poorer. Did you do any measurements/calculations?

I think it MOSFET is more important than you realise. It needs to have a low gate charge, otherwise the efficiency will fall.

[Richard Head]

How does the MOSFET switch off?

[Zero999]

How does the MOSFET switch off?

When Q2 turns on, it shorts the gate to the source.

[Jay_Diddy_B]

Hi,

I don't think this circuit will start reliably.

If I add some resistance in series with the source V1, it doesn't start to oscillate.

I think you need to try on the bench.

Regards,

Jay_Diddy_B

[Refrigerator]

Hi,

I don't think this circuit will start reliably.

If I add some resistance in series with the source V1, it doesn't start to oscillate.

I think you need to try on the bench.

Regards,

Jay_Diddy_B

I did try it on the bench and it had problems starting at low voltages. It would start boosting at 2V but with very low efficiency up to 3.3V or so, from 3.3V to 4.2V it works without any issues.
I did put a 33µF tantalum on the input near the inductor.
Might be a good time to fire up my scope and make some screenshots. 

[Richard Head]

When Q2 turns on, it shorts the gate to the source.

Agreed but before the BJT can turn on the MOSFET has to switch off to allow the inductor to fly.
Self oscillating topologies are normally implemented using a saturating core or some mechanism that switches the MOSFET off when a certain current has been reached.
Also, another winding is normally used to switch the MOSFET on and off faster.
 

[Richard Head]

This is an isolated flyback but a boost converter can be implemented easily I'm sure.

[Refrigerator]

I did some testing on my circuit.
Idle it uses 51mA, but it has a green clear LED on the output with a 220 series resistor that uses some current.
The MOSFET used was salvaged from a dell server motherboard, here's the datasheet: http://pdf.datasheetcatalog.com/datasheet/fairchild/ISL9N312AD3.pdf
In my circuit R3 = 100k, C1 = 470p.
To test the output on higher load i put a 6.8 resistor across the output, with an 18650 cell on the output through a 0.22 current shunt because my meter fuse was blown  .
Input current was 1.55A @ 3.23V input voltage.
Output current was  0.808A @ 5.5V output voltage.
Efficiency comes out to be about 88.8%

The simulation gives an efficiency of around 87%. In real life it will be poorer. Did you do any measurements/calculations?

Then i must be dreaming 

I've attached two screenshots of the DC-DC at idle on the drain of the MOSFET, you can see how it regulates and you can also take a look at how nice the the waveform is  .
I forgot to take screenshots of the loaded DC-DC but they didn't look that different just that the duty cycle was higher.

EDIT: looks like the first screenshot got chopped off but there's nothing too important just a few more cycles and some slight ringing.

[Berni]

I had my own go at it and this is what i came up with



Its similar to the "jule theif" cirucit or the photoflash boost circuit. But i used the other transistor to give it a regulated output so that it can power an actual circuit with varying current consumption.

Here are the simulation results of it with the input voltage swept from 3V down to 1V and it keeps on working fine, providing roughly 5V at 100mA from it.



The efficiency is probably not great tho but it does offer a lot of features for a tiny component count.

[Refrigerator]

I made some more screenshots of my circuit under load.
First screenshot is of the circuit at idle. I included this since the other one didn't have the full set of pulses visible.
Second one is on the drain of the MOSFET with the 6.8 resistor on the output.
The third one is the voltage on the output with the 6.8 load.
Also here are some pics of my prototype.



Now i have to find where to put this thing i made. I'd quite fancy myself a power-bank. 

[T3sl4co1l]

If you allow a coupled inductor, it is possible to do in one transistor.  This is a traditional blocking oscillator, with rectifiers handling the power, and zeners controlling off time (usually).

If not, then two transistors are required.  One inverts the other, so that, as is necessary for all oscillators, negative and positive feedback can be applied in the circuit.  The negative feedback is at DC, so that the amplifiers can always settle into a bias where the loop gain is large, and thus begin oscillating.  Positive feedback is at AC, so the oscillation grows over time.

Note that the MOSFET circuit breaks this rule, which is why it is unreliable: there is no feedback mechanism to adjust the first transistor's collector current to around the MOSFET's gate threshold voltage (where the loop will have AC gain, and be able to take off).

Using a BJT, this is okay, because the first transistor is biased such that its collector sits at a bit more than 1*Vbe.  It acts as a crude current mirror, where both collector currents are set by Vbe (and some by hFE), and their area and doping ratios.

Note that these circuits cannot possibly incorporate useful control or protective features, like undervoltage lockout, auto-restart in case the oscillation is quenched, or current limiting.  (The blocking oscillator can be current-limited by adding a transistor in the emitter circuit, which turns off the main switching transistor when its emitter current rises too high.)

Tim

[T3sl4co1l]

I've played this game myself; this is an example I'm pretty comfortable with:



My goal was to ask: if we add a minimum of transistors, how much functionality can be introduced as a result?

This circuit behaves similar to the venerable UC3842, incorporates variable frequency control, though lacks a wide supply operating range (due to the lack of current sources/sinks, which would gobble up many transistors, quickly!), or UVLO.

It uses a separate oscillator core, so the frequency does not depend upon the output transistor's switching waveform.

The part I'm most disappointed in, is the pulse inverter (bottom left), which is necessary to retrigger the timer.  There's probably a better way to solve that.

Note: I don't consider TL431 an IC.  I prefer to think of it as an ideal BJT.

Tim

[Refrigerator]

Cool circuit. 
In my case i wanted to make a high power DC-DC converter, preferably with high efficiency while using as few components as possible.
I might put together another prototype that would look a little cleaner and maybe be slimmer for a power-bank project. I do have a few TP4056 modules with under-voltage protection that i would use together with this boost converter.
Might be a good time to make a constant current load because the 6.8 resistor gets hot in a hurry , makes testing this circuit cumbersome.

[Zero999]

When Q2 turns on, it shorts the gate to the source.

Agreed but before the BJT can turn on the MOSFET has to switch off to allow the inductor to fly.
Self oscillating topologies are normally implemented using a saturating core or some mechanism that switches the MOSFET off when a certain current has been reached.
Also, another winding is normally used to switch the MOSFET on and off faster.
 

Look at C1. There are other ways for getting the circuit to oscillate than the blocking oscillator topology you're probably familiar with.

[Refrigerator]

When Q2 turns on, it shorts the gate to the source.

Agreed but before the BJT can turn on the MOSFET has to switch off to allow the inductor to fly.
Self oscillating topologies are normally implemented using a saturating core or some mechanism that switches the MOSFET off when a certain current has been reached.
Also, another winding is normally used to switch the MOSFET on and off faster.
 

The MOSFET is not fully on at startup, both Q2 and M1 are floating.
Firstly, the MOSFET lets a little bit of current flow thus reducing the voltage at the drain, which then couples to the base of Q2 through C1 and decreases current through Q2, thus increasing the voltage at the gate of M1, this then decreases the voltage at the drain of M1 even more and the cycle goes on until the MOSFET is fully on.
When the voltage at the gate of M1 stops decreasing both Q1 and M1 start floating, this then causes the voltage at the drain of M1 to increase, this voltage couples through C1 to the base of Q2 and increases the current through Q2, which then decreases the current through M1 and causes the voltage on the drain of M1 to increase, this goes on until the MOSFET is off, causing a very sharp voltage spike at the drain of M1.
That current spike goes through D1 and charges C2. 

[BrianHG]

My simplest boost converter was an 8 pin dip pic connected to a logic level mosfet, then to a power inductor.  Basically your first circuit with the MCU in place of the 2N3904 & it's tuned timing circuit and no longer needs the zener diode as I used the MCU's ADC to regulate the output.  Ok, this had that 1 50cent MCU, but I was able can program the thing to a fixed frequency select output voltage via software & it worked from 5v to 2.5v.  However, your design doesn't have any software to deal with...

As for a 2 transistor boost device, I've only ever played with xenon flash tube devices which was a simpler version of T3sl4co1l design but had a 2nd separate transformer for the HV trigger.

For 1 transistor solutions, I've only seen these in battery powered fluorescent flashlights, which also had transformers with a return coil to drive the base.

[SingedFingers]

This is a low current design I have stolen a couple of times which can be built on:

http://lea.hamradio.si/~s53mv/spectana/sa/sa13.gif

From: http://lea.hamradio.si/~s53mv/spectana/sa.html

I've used it, post regulated as the sweep voltage drive for a varactor tuned oscillator. Nice and quiet for a switcher.

[technix]

For the lowest component count, you can connect the power through an inductor to an CMOS MCU's PWM pin, and pull the output from its power pin, with filtering caps there.

The CMOS ESD protection diodes will provide the MCU with initial power. The PWM gives it the required oscillation for boost operation. If your MCU offers an internal voltage reference for the ADC, you can implement feedback in software by measuring this reference voltage against supply voltage (which is the output of the boost converter.)

You need to make sure your MCU can start with the input voltage minus a diode drop, the switched current does not exceed the MCU's I/O current handling ability, and the output voltage does not exceed the MCU's maximum supply voltage.

You can use an external totem pole MOSFET pair (or NMOS + diode) to increase the power handling capability, this time connect the PWM output to the gates of the totem pole.

[Refrigerator]

For the lowest component count, you can connect the power through an inductor to an CMOS MCU's PWM pin, and pull the output from its power pin, with filtering caps there.

The CMOS ESD protection diodes will provide the MCU with initial power. The PWM gives it the required oscillation for boost operation. If your MCU offers an internal voltage reference for the ADC, you can implement feedback in software by measuring this reference voltage against supply voltage (which is the output of the boost converter.)

You need to make sure your MCU can start with the input voltage minus a diode drop, the switched current does not exceed the MCU's I/O current handling ability, and the output voltage does not exceed the MCU's maximum supply voltage.

You can use an external totem pole MOSFET pair (or NMOS + diode) to increase the power handling capability, this time connect the PWM output to the gates of the totem pole.

That's way too much hassle for just one simple boost converter.
MCUs cost more, then i need to protect the MCU, for high efficiency i need fast edges so i would need a totem pole output buffer for the MCU.
Then comes the code and many calculations to get it working just right.
I'd rather just open LTspice and half hour later have a boost converter that works just as well.

[T3sl4co1l]

Nah, you can do quite a lot with an MCU -- and they're quite cheap (what, have you never been shopping for 8051s before?).

As long as you don't need much current, of course.  Ports are only rated for low 10's of mA.

What's nice, at least, is you get free synchronous rectification: instead of toggling the port between active-low and Hi-Z, toggle active-high and low.  This also means it works both ways, so you can boost up the supply, or buck it down.

The main downside is the same as any other overly simple method: no current control.

Tim

[technix]

For the lowest component count, you can connect the power through an inductor to an CMOS MCU's PWM pin, and pull the output from its power pin, with filtering caps there.

The CMOS ESD protection diodes will provide the MCU with initial power. The PWM gives it the required oscillation for boost operation. If your MCU offers an internal voltage reference for the ADC, you can implement feedback in software by measuring this reference voltage against supply voltage (which is the output of the boost converter.)

You need to make sure your MCU can start with the input voltage minus a diode drop, the switched current does not exceed the MCU's I/O current handling ability, and the output voltage does not exceed the MCU's maximum supply voltage.

You can use an external totem pole MOSFET pair (or NMOS + diode) to increase the power handling capability, this time connect the PWM output to the gates of the totem pole.

That's way too much hassle for just one simple boost converter.
MCUs cost more, then i need to protect the MCU, for high efficiency i need fast edges so i would need a totem pole output buffer for the MCU.
Then comes the code and many calculations to get it working just right.
I'd rather just open LTspice and half hour later have a boost converter that works just as well.

STC15W401AS is a 8051-compatible MCU with built-in PWM and ADC, sadly no built in reference or high current handling. It costs 20 US cents each. It comes with internal RC oscillator and POR so it runs on its own. It can with from 2.7V up to 5.5V. You can clock it up to 35MHz. It's minimal package is SO-16.

TL431 is your run of the mill bandgap reference chip. Hook one to STC15W401AS's ADC pin with external pull-up. You now got the reference voltage. STC15W401AS always reference its SAR ADC to its supply. Here goes the digital feedback path. It's minimal package is SOT-23

AO4606 is a MOSFET complimentary pair in SO-8 package. It can switch 5A current. Connect the gates together to one PWM pin of STC15W401AS. Connect the drains together to one end of the power inductor. Ground the source pin of the NMOS. Tie the source pin of the PMOS to the power pin of STC15W401AS and the reservoir cap.

You have only used 2 pins out of 14 on the MCU. You can hook the other side of the inductor to another ADC pin (through a resistor) for a crude gas gauge. The MCU also have hardware UART so you can make this module speak to another device.

This circuit contains 6 packages, costs less than a dollar. Yet you get a synchronous rectifying bidirectional DC-DC converter, battery gas guage, with digital interface and runs your own code out of it. You can even use the remaining pins to implement other features for your system.

[SingedFingers]

Those STC15W401AS's look pretty nice at that price point.

Can you recommend a basic dev kit for these at all?

[Refrigerator]

@technix
That does look like a pretty inexpensive MCU, i might order some to play around with.
But for now i have to make myself a constant current load to test the circuit i made. 

[BrianHG]

I like the PIC12F1571, or newer, 1.8v thru 5.5v operation on one 8 pin IC, internal bandgap voltage reference, AD/PWM/internal OSC, available in DIP8, SO8, 39 cents US...