Switch Mode Power supply - finding controllers

1. Good old UC3843 covers most of that range. But it stops at ~5V.

3-18 is an awfully wide range to expect from a single circuit (6:1!). That deserves a multistage or transformer (flyback/forward) circuit to handle the 17A input. You realize that's 17A input at 3V, right?!

1. is easily done with the old MC34063. Or put a tertiary winding on #1.

Tim

-- Deep Friar: a very philosophical monk. Website:

How about a comparator followed by a MOSFET driver?

One thing I like to do is use an LM339 quad open-collector comparator - using "wired AND" to have the comparator used for MOSFET current sensing ANDed with one used to sense output voltage, and any others used to sense conditions requiring lockout. Connect the outputs together, and connect from the outputs to B+ a "pullup resistor". The value of this resistor should be a little higher than that which passes the minimum current sinkable by one comparator output, at maximum input voltage.

(If input voltage range is inconveniently wide, one could add a regulator of some sort or another to limit the voltage used by the control circuit.)

The comparator combined output feeds a MOSFET driver IC, which is a simple buffer made for that job. One that I like comes in a SOT23-5 package, which I have found to be hand-solderable. In the past when those were more expensive, I have used 555s, also 4049s with most or all of their individual inverters paralleled.

However, I am not getting a good feeling about using input voltage less than about 4 or 4.5 volts at this rate... For one thing, MOSFETs that reliably pass 17 amps or whatever with gate drive of 3 volts, maybe minus a diode drop, sound to me not all that common. For another, a bipolar MOSFET driver (whether marketed as one or otherwise) will have a voltage drop around a diode drop or maybe somewhat less when passing the high current needed to quickly charge and discharge the MOSFET's likely-large gate capacitance. And anything CMOS that will source and sink that sort of current at 3 volts will likely need to have its supply voltage regulated to not exceed less than 10 volts, maybe 5 or 6 volts.

-------------------

One idea I have used before, but not in anything put into production: Use a complimentary pair of smaller power MOSFETs, that can reliably pass an amp or a couple amps or whatever with 3 volt gate drive. Build a bigger version of a 4049 or 4069 stage. This requires its gate signal to not exceed some reasonable limits of amount of time or percentage of time more than .5 or 1 volt or whatever from both supply rails, otherwise much time with both MOSFETs conducting at the same time is a bad thing. Also, during switching, such a "4069-zilla" will draw a serious current spike through its supply rails.

--
 - Don Klipstein (don@misty.com)

2. >

That one looks good! I don't know how I missed that one before...

The datasheet for it implies it has an undervoltage cutoff at about 8 volts...

My use case for this is a device which uses variable amounts of power. It could pulse to 2 amps at 25 volts for 0.5 - 1 secs, but most of the time, it'll be using more like 0.1 amps at 25 volts. Annoyingly these timespans don't allow me to "cover" with a big capacitor, but they probably would allow me to underspec wire thickness and heatsinks, since hopefully things aren't going to overheat in 1 second. The reason for my wide voltage range is I'd like to power it from batteries, but the main concern is if the batteries can meet the pulse power requirements. When at maximum output power, the impedance of the input side of the power supply should match the internal impedance of the battery, so the terminal voltage at the battery will be half the open circuit voltage (with 50% efficiency)., so if my battery is nominally 9 volts, the power supply must work down to 4.5 volts...

My experiments show that a pair of 9v PP3 batteries could probably supply the 50w required for half a second

#1.

Thats a very hostile and unsafe way to treat batteries, not to mention hardly functional.

Why dont you make your own controller from the ground up?

NT

Oh? Thought it was around 5V max falling threshold. Make sure you aren't looking at the 42 or 44, which is the "high" voltage model...

Sure they do. C = 2A * 1s / 5V = 0.4F. Ten 4.7F in series will take that up nicely.

I don't know how picky supercaps are about balancing. Resistors would be contrary to battery operation. You could use zeners to enforce 2.5V across them.

These look possible. 0.3 ohm ESR, or 3 ohm total, which is kind of a lot at 2A. Hehe, 300mA max discharge current...

$30 for the caps is about what you'll spend on the batteries and converter to do it, so you're paying either way...

Holy shit, where did you find those? Are you sure they don't have plutonium in them?!

These only claim 22.5 ohms ESR for light loads, a couple amperes short circuit.

If you were measuring alkalines, well, you certainly weren't, according to the curves they give..

Bleh, batteries... nonlinear, long-time-constant, rapidly aging temperature sensors!

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

Ya know... they make BJTs in that range. I've used KTC5001 before, which is good to 7A at < 0.2Vce(sat), at hFE >= 80. These BJTs are almost better than MOSFETs, certainly by transconductance, which is a big limit at this point. If you don't mind building discrete, you can implement an UC3842 without the UVLO and with lower operating voltage.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

- t

ss

t

to

o,

For a moment I vaguely wondered about 2 parallel circuits, one deisgned to cover most of the voltage range but not the bottom end, and the other optimised for the low end. When the higher V one runs it fully disables the low v one.

NT