2 Transistor Buck LED Driver -> Voltage Output

The control loop has to have enough gain/hysteresis to "snap".

You could run on current-control until correct output voltage is reached, then shut off... what do they call that, "burst mode"? ...Jim Thompson

Yes, that was what I meant, bang bang control, so burst of energy delivered to the load

Cheers

Klaus

The immediate problem is that you're trying to run the switcher off the

Another is that your sim's inductor is 220 attohenries. :-)

Another is that the 28V load voltage is so close to the 30V input voltage that the load never draws enough current to trigger the current-mode switcher's current limiter. This style converter will stay 'on' until the current limit is reached, then coasts, then kicks 'on' again.

In general the circuit works over a wide range. If it fails to oscillate, there's not enough drive to the switch, or (more often) insufficient positive feedback.

All that said, your sim file works for me if the supply voltage is raised to 35v instead of 30v.

HTH,

James Arthur

The original thread explains the version Klaus chose has a higher efficiency, but also a dropout voltage than the other version.

---quote--- John Lark> Doesn't that have a hang state?

I don't think there's a hang state.

As long as there's enough headroom Q1-Q2 are linear with positive feedback, and there's no start-up bias problem. If there's not enough headroom it may fail to oscillate, and Q1-2 will form a crappy linear current source, at less than designed value. I'd call that a graceful failure mechanism, not a hang.

---/quote---

Fig. 2's headroom trade-off was part of the original circuit description--

---quote--- R3 accounts for a fair amount of dissipation. If dropout voltage doesn't matter, most of that can be recovered by returning R3 to ground through the load, LED1.

Fig. 2

------ R1 Q1 _______ 5.1 2n4403 .-.-.-.

---/quote---

You can certainly close a voltage loop around it, after all, that's how voltage-output current-mode switchers work. A some point, though, you've added enough parts that you might as well use an IC.

Cheers, James Arthur

Yep. Most of these "simple" discrete lash-ups are hacks guaranteed to fail.

Except for high speed switchers you can hack quite nicely with LM339's... all the GenRad portable testers were built that way... durable, reliable, no field failures. ...Jim Thompson

The ASC file was copied from the old thread and some characters were wrong, and I apparently missed that one. Thanks :-)

Cheers

Klaus

Sure. Your version, with the inductor value corrected, works at Vin=30V and two fewer LEDs in series, to give the switcher some headroom.

The original was for a 21Vf Nichia array.

Cheers, James Arthur

Anyway, it's a nice circuit. I've been playing with one around a schmitt trigger, but this one is simpler and cheaper :-)

Cheers

Klaus

LED drivers are constant current and constant voltage so the simple hacks work very well.

Not the circuit mentioned, though. The current sense is on the wrong end so it will be unstable as hell. Maybe it will be a good candle simulator at just the right voltage.

Kevin won't see this, but he's wrong. This circuit's switching current limit varies with input voltage to produce quite decent line regulation / constant current into a given, designed load.

Cheers, James Arthur

I havent looked at your sim however, using a simulator means the components are equally matched, at least as far as the Sim is concerned.

Offset one of the components and set the power supply at 0 on start up.

Jamie

Klaus' difficulties were from a 220aH inductor and 2V input-output differential. The version he chose runs off the input-output difference, and needs more than 2V to operate. There's another version with a lower drop-out voltage in the OP.

Restoring the original 220uH inductor and ~9V input-output differential corrects Klaus' woes.

Cheers, James Arthur

A modified circuit based on James Arthurs circuit. Added voltage mode control

A TL431 could be used instead of the zener and BJT (not directly due to the needed cathode current, but via a BJT)

Cheers

Klaus

Random thoughts....:

Of coarse this converter does not provide shutdown feature if it hangs in any way. One option could be to sense the voltage at the junction of R1 and R2, if the voltage is lower than 2 times a VBE junction than VCC, restart the thingie

Could be done by a microcontroller, to facilitate shutdown and fine tuning of the output voltage

Cheers

Klaus

That was leveraged around constant current, here is a buck for constant voltage if you absolutely have to use just two transistors:

Making a proper diff-amp comparator much improves performance, as in:

If your load can run differently ground referenced then an inverting flyback is also simple:

piglet

You should drive LEDs with constant current, not constant voltage!

Thanks Rob. My interpretation of Klaus's OP was he asked about constant voltage output - I guess he intended the load to not be LEDs.

piglet

I will try those out, they look very interesting :-)

The other one I could finetune by increasing R3 and using a higher beta BJT for Q1

Cheers

Klaus

Correct. I intend to use it for a power supply for low power circuitry

Cheers

Klaus