# efficient low-voltage ac/dc conversion for bicycle battery charger

Discussion in 'Electronic Design' started by Tom (at tomsweb.net), Jun 9, 2005.

1. ### Tom (at tomsweb.net)Guest

Everyone,

I want to construct a device that takes the AC power generated by my
bicycle dynamo and uses it to charge 4 AA NiMH cells in series (the
batteries may at the same time power a load (GPS) through a low-drop
regulator).

Here are the specs:

input power (dynamo measurements):
* biking speed 10 km/h: 5.6VRMS unloaded, drops to 2.5VRMS at 0.5 amp
* biking speed 20 km/h: 9.9VRMS unloaded, drops to ca. 4VRMS at 0.5 amp
* did not measure the AC frequency but I expect it to be in the range
10..200 Hz, and pretty much sinusoidal

output voltage (for charging batteries): 5.6 V (1.4V/cell)
output current: whatever the source can give and the batteries (+ load,
if any) will take (expected ca 0.3 amp)

Since I have limited power available and need nearly all of it, I am
looking for the most efficient solution (would like 85-90% efficiency).

I see basically two major options:
1) rectify + filter AC to DC and use a switching converter to the fixed
output voltage
2) use a circuit which switches the AC directly into a DC output
voltage

for 1)
- what rectifier circuit to use? a bridge rectifier with schottky
diodes seems the most straightforward, but still costs two schottky
drops of wasted power, not negligible for such low input voltages;
perhaps some sort of active rectifier circuit with MOSFETs or so is in
order?
- will the typical short conduction angle of such rectifiers, where a
peak current is drawn to recharge the filter capacitor, negatively
impact efficiency? if so, how to avoid this?
- since the desired output voltage of ca. 5.6V can be higher or lower
than the input voltage (input expected to vary between 2VRMS and going
upto 20VRMS at high speeds), some sort of buck-boost regulator seems
needed; I was hoping to be able to use a simple IC switching regulator
but most seem to be either buck or boost, rarely both
- perhaps a voltage doubling rectifier can get the voltage high enough
so a simple buck regulator can be used; will this cost me efficiency?

for 2)
- intuitively I would say this would allow higher efficiencies, but all
switching IC's I looked at take in DC, and I would expect a
configuration like in an off-line switching power supply (with
transformer) to be very inefficient at these low voltages, power levels
and frequencies; any ideas?

Thanks for any suggestions!

greetings,
Tom

PS I am aware I could use a solar panel but that is not convenient to
add, and the dynamo is there anyway

2. ### John SmithGuest

Run the dynamo through a bridge rectifier, perhaps with a 100 uf or so
cap across the + and - of the output of the bridge...

You can construct the bridge out of 4 single diodes, 1 amp should be
more than enough, for instance a 1N4001 or better...

John

3. ### Ian StirlingGuest

Random thoughts:

The most efficient way would be a AC-DC converter that presented an
impedance optimised for frequency, to extract maximum power.
Constant impedance is probably the easiest to do though.
It should convert the AC waveform into pulsating DC current which is
fed to the batteries. (as you mention, resistive drops are bad)

I'd start with something like a standard boost converter - with the
wrinkle that the power switching device is two complementary FETs back
to back, so it has no problem switching AC.

This switches the voltage up to the 6V range, where it's rectified by
a shocketty or active bridge into the batteries.
To get the input impedance right, you'd want the control circuitry to
mimic a resistor, so the feedback loop wouldn't be to keep it at a certain
output voltage, but a comparison of the current with the absolute value
of the voltage...

Anyway, this is probably a bit hard.

I suspect you'd get a fair portion of the way there with a correctly picked
maxim/... switcher designed to run from low startup voltages.
4 biggish shockety diodes (sized for minimum Vf) feeding the switcher chip
directly, with a small decoupling capacitor.

The switcher is set to output 6V (or whatever), with a small resistor to the
batteries.
Ideally, you'd probably want to feed some fraction of the input supply voltage
in to the regulator, so it tries to raise the output voltage a little when
the input rises.

That's my first thoughts, though I am really a bit tired.
Anyone else contribute?
voltage rises.

4. ### GPGGuest

Have a look at linear's LTC3780,

5. ### John SmithGuest

.... once you exceed the forward voltage of the diodes (silicon--about
..62 per volt or 1.4 in the bridge circuit) the voltage out of the
diode/cap circuit with be about 1.4 times the ac in p-p voltage. At 5V
ac in, this will be ~7V dc out (unloaded.)
You can spend a lot of time trying for more "efficiency"--I think if you
try the "occams razor" approach you will be just as well off...

Warmest regards,
John

6. ### Roger LascellesGuest

1. I suggest MOSFET for the rectification part, not diodes. You will have
to check diode specs, but I suspect this alone will give the biggest payoff.
There are MOSFETs which are well on at 3V gate drive. Perhaps schotkeys in
parallel will get it started, then the MOSFETs can take over when there is
DC available.

2. You could use the configuration used in PFC (Power Factor Correction)
circuits. This IC probably won't be suitable, but if you grab the LT1249
datasheet from lt.com (Linear) you will see the idea. The circuit uses a
single inductor and a mosfet switch. A small cap across the output of the
rectifier makes the raw DC look like a low impedance source at the 100KHz
switching frequency. You don't have to do the whole PFC thing, in accurate
detail.

3. Personally, I would just rectify into a capacitor, to make it simple.
The peaky current does reduce efficiency, but numbers I have calculated for
mains rectifiers tells me that it is not too bad - certainly not the
disaster you might think intuitively. Using a smaller capacitor to extend
the conduction angle has so little effect on efficiency, that it is better
to use a big enough cap to get the ripple down.

4. You might want a "pause" button which stops charging on hills !

Roger Lascelles

7. ### martin griffithGuest

On Thu, 9 Jun 2005 13:05:18 +1000, in sci.electronics.design "Roger

snip
going up or going down?

martin

8. ### Tom (at tomsweb.net)Guest

Have a look at linear's LTC3780,

That looks like a useful chip for the switching part, even if normally
intended for higher powers (although its minimum Vin of 4V may be a bit
high for my application). Unfortunately Farnell etc. don't seem to
carry it. I tried ordering samples from the Linear website... too bad
LTSpice does not seem to have a model for the LTC3780.

Thanks for the suggestion.

greetings,
Tom

9. ### Tom (at tomsweb.net)Guest

John,
Unfortunately I have no room to "keep it simple" else I would. My
available input power is 1.5W and I need at least 0.2A x 5.5V = 1.1W
into my load, so there is not much left for inefficiencies.

The input voltage under load (which is what matters) drops to 2.5VRMS,
or ca. 3.5Vpp. If I would use ordinary silicon diodes I would lose
twice 0.65V leaving me with 1.6Vpp, or a measly 45% efficiency just in
the rectifier.
Schottkys are better, but the MOSFET circuit also sounds interesting.

greetings,
Tom

10. ### Tom (at tomsweb.net)Guest

Roger,
OK, exactly how would I drive the gates of these MOSFETs? Do they need
a square wave in phase (and/or opposite phase) with the incoming power
frequency, which is variable? If so, do you know a convenient way to
generate that?

Is there no problem with MOSFETs having a parasitic reverse diode which
would inhibit their use as a rectifier? Or do lower-power models not
exhibit this parasitic diode?
I had a quick look at this datasheet and must admit I don't understand
much of it or how it applies to my situation - can you maybe clarify a
bit?
Will it not be a problem that the source may not be able to provide
these high peak currents and go to its knees? If so, would a (bipolar)
capacitor on the input to handle the surge currents help?
Well, uphill the power available will just be too low, and the battery
will not charge (I may need another diode there if the switching
circuit's reverse leakage when off is too high and would discharge the
battery).

Downhill, I would clamp the input voltage to below the max. input of
the switcher chip with a big Zener diode or so.

To have a rough idea of the battery status I plan to add a simple
pushbutton-activated voltmeter that I can press when bicycle is
standing still, i.e. charger circuit is switched off (not very
accurate, I know, but should give enough information).

greetings,
Tom

11. ### Tom (at tomsweb.net)Guest

Ian,
That's an idea, the output does not really need to be "proper" DC. Will
NiMH batteries charge happily on some sort of pulsating voltage?
Yes, and a simple boost converter is probably not sufficient, because
at higher cycling speeds buck operation is needed.
Looks the most feasible way. So you suggest a SMALL capacitor, to
maintain a large conduction angle? Someone else suggested this hardly
improves efficiency vs. a big cap.

Perhaps I should consider the possibility of having TWO switchers:
- a low-voltage 1..5.5V or so Maxim/... step-up switcher for the low
voltage range
- a 5.5-36V or so step-down switcher for the higher voltage range

if both have a shutdown, I could use a comparator (or 2) to select one
of them depending on the current input voltage.

That approach would probably not be able to switch fast enough to work
good with pulsating AC, and a big filter cap might be in order (and
hysteresis on the comparator).

Does that sound at all feasible?

Also an interesting idea, so the demanded battery current depends on
available voltage. But then the battery state, current demanded by the

The frequency of the AC also has information about the available energy
of course (how fast I cycle), but that's getting a bit complicated
then...

greetings,
Tom

12. ### Fred BartoliGuest

Your pb can be view as a PM motor that you want to brake.

A full MOSFETs bridge will do:

.-----------------+--------------.
| | |
||-+ +-|| ---
Q2 ||<- ->|| Q3 -
-||-+ +-||- |
| _ | ---
| ___ / \ | -
+----UUU--( ~ )---+ |
| \_/ | ---
| | -
||-+ +-|| |
Q1 ||<- ->|| Q4 ---
-||-+ +-||- -
| | |
'-----------------+--------------'

(created by AACircuit v1.28 beta 10/06/04 www.tech-chat.de)

By having on one half cycle the lower mosfet of one branch permanently on
(say Q4) and the other branch switching, you'll boost the dynamo voltage to
the NIMH voltage (Q1 will play the role of the boost the switch, Q2 will be
the synchronous rectifier).

This will grant you for the highest efficiency, but driving the upper
mosfets may not be very practical.

A slightly lower efficiency, but much easier, solution would be to change
the upper mosfets for schottkies :

.-----------------+--------------.
| | |
| | ---
D1 - - D2 -
^ ^ |
| _ | ---
| ___ / \ | -
+----UUU--( ~ )---+ |
| \_/ | ---
| | -
||-+ +-|| |
Q1 ||<- ->|| Q2 ---
-||-+ +-||- -
| | |
'-----------------+--------------'

The 0.5V schottky drop, weighted against the 4.8V NIMH voltage would only
cost you a 10% efficiency penalty.
The driving circuit will be easy to power from the NIMH.

13. ### Tom (at tomsweb.net)Guest

Fred,
Looks like a very elegant solution!

Could you maybe show me the control waveforms applied to the gates of
the transistors in both cases, w.r.t. the incoming sine waves?

I guess in such a setup I would anyway need a small micro to send out
the correct pulses, in phase with the incoming AC. So perhaps doing the
full bridge configuration will not need much extra hardware vs. the
schottky (apart from maybe a gate driver if the micro's outputs are to
feeble to properly switch the MOSFETs with a decent risetime).

greetings,
Tom

14. ### SpajkyGuest

rectify the current with (graetz) bridge constructed with 4 schottky
diodes (+followed by maybe a el.capacitor of few hundreds uF in
paralell) & than thru a resistor of min. 1W/100ohm or more in series
to batteries ... IMHO cheap & effective solution ....

15. ### John SmithGuest

Absolutely true, Schottkys ARE better.... one real improvement would be
if a method could be introduced to increase generator (dynamo) output at
low bike speed...
I have often wondered about replacing the cheap dynamo of higher output
with a voltage reg circuit... but seldom ride at night these days and
haven't been able to find the motivation--but certainly, surplus, a guy
could get a hold of a better generator--somewhere...

Warmest regards,
John

16. ### Jim ThompsonGuest

[snip]

I was just struck with a thought... isn't the "pedal rate" relatively
constant, as opposed to the wheel rate which depends on gear?

Maybe some gearing directly from the chain to run a dynamo?

Or magnetize the links alternating and pass through a coil ;-)

...Jim Thompson

17. ### John SmithGuest

That is an excellent take on it too, directly off the chain, and speed
regulation would get magnitudes better...
non-magnetic sprockets and sufficient distance from magnetic
metals--there would also be the great danger of the chain attracting
bits and pieces of magnetic metals and causing increased wear and rider
injury...

Warmest regards,
John

18. ### Ian StirlingGuest

They are perfectly happy with pulsating current to charge them.
Missed that bit, I'll respond again when I look up the high range voltage.
I dunno, it was a thought.
The idea was to basically just use the capacitor to provide negligable
impedance at the switching frequency, not actually smooth the voltage.
The converter starts up each time the voltage hits 1.2V or so.
I'll respond more fully a bit later, after I've had time to read the

19. ### BobGGuest

Make the generator pully larger diam to slow it down and REDUCE the
output voltage. Use the mosfet sybchronous rectifier discussed
previously to charge a 5V supercap.. essentially a D cell. Now get a
COTS DC to DC converter to go from supercap to battery charger or lamps
or whatever.

20. ### Jim YanikGuest

Then you don't get any power generation when coasting.