Building an Efficient 12V to 5V Regulator

Despite
following a few dead end pathways, the result of a weekend's effort
includes three different options to achieve an efficient 12V to 5V
regulator for output currents up to 250mA.
Introduction
It was one of those weekends.
I started out with a plan to build a new compact version of my
ATmega8-based DDS VFO. It uses one of those ultra-cheap Chinese AD9850
modules, a rotary encoder, a few switches and a two-line 16 character/line LCD display. With a few additions and changes to my BASCOM software, it ends up being a fairly useful and inexpensive VFO.
But, as is often the way, I got side-tracked.
The problem: This Chinese DDS module, and specifically the AD9850 chip
and the 125MHz crystal oscillator on the module, draws about 150mA at
5V. Adding my microprocessor and a few LEDs to the mix, I ended up with
a load current of about 160mA. My prototype version just used a
standard LM7805 linear regulator, but with a typical 12V supply on a
transceiver, this 7805 regulator has to dissipate over 1 watt. As you
might expect, and even with a modest heatsink on it, the device got
very hot. Certainly, this power dissipation is well within the ratings
of the TO-220 package bolted to a reasonable heatsink, but I don't like
that sort of heat in a transceiver if I can avoid it. Besides, it
smacks of poor design. That LM7805 is only about 40% efficient.
So, I decided the first step with the new VFO version was to quickly
build a switched mode regulator to replace the LM7805. If I could
achieve, say, 80% efficiency, then I’d only have 150mW to worry about.
That means that the regulator would be operating at a much cooler
temperature, maybe 35C at worst case.
First Version
My first attempt was to quickly build a Roman Black two-transistor regulator. That design
works exceptionally well and, with the right inductor, 80% efficiency
is readily achievable. But that was the first problem – Despite
searching everywhere in my parts box, I could not find a suitable
inductor. Added to that, I couldn’t find a 5.6V zener diode either.
These are not trivial problems where I live, a very long way from the
parts suppliers.
If you are fortunate enough to have the right parts, this one is definitely worth trying.
Second Version
So, I
looked for an alternative design I could build from the parts I had
available. I have to confess, I had not stopped to think about other
ways to achieve the actual objective i.e. What could I use as a cheap
efficient 12V to 5V regulator? If I had thought about that question
first, I could have saved myself a couple of hours of messing about. On
the other hand, I wouldn’t have tried out this design, so maybe it was
for the best.
Anyway, the second option I decided to try was built around a LM317
regulator chip which was used in a switch mode regulator design
developed (as far as I can determine) by Bob Haver of Motorola and
published in Electronic Design in December 1979. I’ve never seen a copy
of that original article, but the design was summarised briefly in
RSGB’s Technical Topics in May 1980. Similar designs also appear in
some 78xx regulator datasheets although these lack any detail on
specific components or indicate any performance data to show how well
they work.
The original 1979 article described a 5V/5A switch mode regulator
application using a large power transistor and a TO-220 version of the
LM317 regulator. I suspected that a load current of 200mA (just to give
myself some margin in the design) could be achieved with a more compact
BC327 and LM317L solution, both small TO-92 parts.
Here’s my final version of this modified design:

Being
in TO-92 packages, the BC327 and LM317T are quite small and the
efficiency of the design means these parts require no heatsinks. R6
(4k7) was added in parallel with R5 (330 ohms) to bring the output
voltage very close to 5V.
I did manage to find an inductor that made this circuit work really
well (i.e. The design gave the correct output, well regulated, and just
under 80% efficient!), but it was physically very large. The right
component would have been less than 30% of its size, but I just didn’t
have anything suitable. And, golly, how I tried to find something that
would work.
Meanwhile, other tasks were becoming more critical, including having to
glue a broken catch back on to the bathroom cabinet. I was out of epoxy
glue, so a trip down to the hardware store was required. As I was
browsing the shelves in the hardware store, I suddenly realized I’d
been wasting my time with designing my own switching regulators. In
front of me were several very cheap (less than US$5) vehicle USB
chargers for iPads and celphones that performed the self same task.
Duh! An ideal cheap solution, probably quite compact inside their cheap
plastic cigarette plug packages, and I’d completely forgotten all about them.
I quickly bought one and headed back to the house.
Testing the Commercial Version
As soon
as the bathroom cabinet repair was completed, I proceeded to tear the
newly purchased USB car charger apart. Inside, I found the expected
switching regulator built around a Chinese-made MC34063 8-pin IC along
with a mid-sized power transistor and a few miscellaneous components.
I proceeded to quickly test it. The charger was rated at 2A, but since
this was an ultra-cheap device, I suspected that a maximum current of
1A, or possibly even 500mA, would be closer to the real limit. Well,
the unit passed my first test – The output voltage was just a little
over 5V. That’s always a good start, and a vital parameter to check
with these cheap products.
But imagine my surprise to find that, even at my very modest test load
of 230mA, this commercial regulator’s efficiency barely reached 60%!!
This was awful. Some checks revealed that while the transistor was cool
to the touch, the MC34063 was getting really hot – Over 75C – and that was at just over 10% of its rated load! What was going on?
Further checks around the Internet showed that this is a very common
problem. The MC34063 has some internal transistors which are used to
control the drive current to the external pass transistor but many
designs use such low gain (cheap) external pass transistors that the
MC34063 is forced to handle higher than desirable currents. This
reflects back as poor reliability and early failure.
I located the datasheet for the H772 external PNP power transistor used
in the unit I'd pulled apart. It was made by a Chinese company,
Huashan. The device had a much better than average current gain of 60 –
400. It is typically more than 100 with currents up to 2A. That
performance allowed me to modify the bias resistors to better suit my
load current and the Huashan transistor, a solution recommended by
NTC/Philips in one of their application notes discussing this problem,
but of course they recommend the use of one of their devices.
The schematic for the original charger and the changes I made to it are shown below.

With
the two resistor changes noted on the above schematic (R3 is changed to
1k and R4 is changed to 330R), I improved the regulator’s efficiency to
75% with my test load of 230mA. More importantly, the MC34063 was cool
to the touch, as it should be. I did not test it with higher load
currents however.
The modified unit with resistors changed and PCB trimmed slightly, and
with the USB connector sub-board removed, is shown (top side) in the
picture at the top of the page, while the picture here shows the
modified underside of the PCB. The two replacement resistors are in the
centre of the board. (Right-click the image to see it full-size)
Conclusions
All
of this goes to prove a few things. First, cheap products are often
just that – A design produced to a price. A little extra effort may be
required to turn it into something you can use. I would almost be
prepared to guess that the company making this car charger board simply
copied something they saw elsewhere and used parts readily to hand
without thinking about what they could do to improve its performance.
Secondly, while my first two attempts did not result in a workable
solution - the Roman Black design because I couldn’t find the parts I
needed (and it is a really superb performer when you have the right
parts) and the LM317 design because the inductor that I used was too
big for practical use - I did verify that both designs would have
produced viable results with the right parts. I also succeeded in the
end, and the path I followed helped me identify and solve the
overheating and efficiency problems with the car regulator that I will
use with my new DDS VFO. It also resulted in a very compact solution
which will fit nicely on the final VFO board.
Of
course, by the time I’d completed all of this, I’d run out of time to
actually build the new compact version of my DDS VFO! I guess that task
will just have to wait until next weekend.
Want to go back to the main page? Click
here to
return directly.