[Back to SMPS Page]


Negative "buck-boost" regulator; +12v in, -5.4v out

App a06; Bo Nielsen's 3-transistor negative voltage generator
12th December 2011

This is not my design!

This nice 3 transistor SMPS is not my design but was designed by Bo Nielsen of Denmark. He contacted me in an email saying he had built an SMPS similar to my designs and I have included it here at his request as my page did not have a negative voltage regulator. (All photos and and diagrams and files on this page were provided by Bo.)

I have not tested Bo's design myself but it does look fine on paper. Bo has nicely included the LTSpice files so you can tune this design to suit your own needs and he built it in hardware as final proof that it works as the spice simulator suggested.

A 3-transistor +12v to -5.4v SMPS regulator.

The circuit accepts +12v input and generates a -5.4v output. With minor changes it will also work as a +5v input and -5v output SMPS.

Circuit operation (as I see it)

The power oscillator section is similar to my 5v to 13v converter and operates in the same way. Q1 is turned on at the start and current rises through the inductor and the current sense resistor R4.

When the current is high enough the voltage on R4 is >0.6v and this turns on Q3 (Trip) and turns off Q1, so the power oscillator oscillates around a fixed L1 peak current of about 180mA (0.6v / 3.3 ohms).

The power oscillator will produce a negative output voltage through the output rectifier diode D1 (1N4148). This will charge the output cap C3.

Output voltage regulation was cleverly added by Q2 and a zener diode. When the negative output voltage is enough to turn on the zener diode (about -5.5v) this turns off Q2, and Q2 is needed to provide the base current for Q1. So when the output voltage is high enough it becomes impossible for Q1 to turn on, so the output voltage is regulated in operation as duty cycle will be reduced.

So the input peak current is regulated by Q3 and R4, and the output voltage is regulated by the zener and Q2.

Recommendations for the circuit

Bo says; I tweaked the design a little hoping to push a little more juice out of the somewhat weak small signal pnp-transitor. It gets rather hot when going above the rated 100mA collector current.

First recommendation must certainly be to go with a little beefier pnp such as the BC327 - unfortunately I do not have one so you must settle for my humble data collected from the old and trusty BC557B.

Three snaps of my incredible old scope
(in the zip file) - the first is shows the collector of Q1, the second shows the basis of Q2 and the third the basis of Q3. All three snaps was taken with 12.5 volt in, 5.4 volt out, 100 ohm load. The fourth snap is the circuit under test and the fifth snap is the simulation of said circuit.

I suspect most people would want higher efficiency or more power at low input voltages - I do not have a simple way to achieve good efficiency in such a low input voltage range.

More current could be realized with a more suitable pnp-transistor, the efficiency could be improved a little by switching to a schottky diode - a schottky diode would also provide a significant boost of the output voltage in the 3-4-5 volt input scenarios with medium to high current output due to it steals less energy during the inductor ramp down cycle.

But the real troublemaker is the current sense resistor, it is stealing 0.7 volt from the inductor both during ramp up as well as ramp down, which in dissipation terms amounts to around 20-40% of the total input power. Q3 needs to be a germanium type transistor;
if the low input voltage scenarios is ever going to look any better. Or a fourth transitor to clamp the current sense resistor during inductor ramp down. But that road leads away from the corner stones of the RB-regulators: simple and yet very useful.

Another tip, if the regulator is not going to be used with low input voltages then R2 can be lowered with favourably impact to all the parameters in which most people would optimize for. Try out 470R in the simulation and notice the wast improvement to the switch-off performance. But alas, 470R in combination with 5 volt input will cause the transistor to exit saturation fairly quickly with increasing collector current.

It's always a puzzle to get the right values.

And yes, the current limiting I mentioned is with respect to the peak inductor current (where it should always be I might ad) and in every boost/flyback topology that will translate to input current limiting - and for every buck/forward converter it will provide output current limiting.

My comments; I agree completely about changing Q1 to a BC327 and D1 to a schottky like a 1N5819. Those 2 parts are responsible for much of the "high efficiency" of the other designs here.

Also I like to use an inductor of about 470uH as the increased inductance gives reduced current ripple, reduced peak current and reduced switching frequency (for a given current output). All three factors will improve performance in these simple circuits where the Q1 switching and drive is never really ideal.

I can see what you mean about R2, from your graph Q1 turnoff seems a little slow, costing efficiency there. It's a shame there's no easy way to drive the base of Q1 high to turn it off quicker. Like you said changing Q1 to a BC327 which has higher gain and a lower Vce sat voltage should require less base current to turn on well, and the resistors could then be optimised for much faster turn off. I tuned the BC327 in a similar way on my regulators.

Bo's ZIP FILE can be downloaded below, it contains the LTspice files and his photos.
BoNielsen-buckboost.zip (439kb)

- end -

[Back to Home Page]