switching current supply

Typically, in a switching regulator, there's a feedback pin that (eventually) connects to the node that you're trying to maintain at a constant voltage (with respect to your circuit common (aka GND)). This feedback pin usually connects through a simple resitor divider because the output supply voltage is usually larger than what this feedback pin wants.

So, rather than using a resistor divider to connect this feedback pin, if you connect it directly to a low value resistor (Vfb/Iout) and the other side of the resistor to GND, and then tie the low side of your load to this feedback point, too, then you'll convert the constant voltage regulator into a constant current regulator.

I've done this and it works.

Bob

Something like this?

The MAX1640 lets you use a wide supply range and has current mode options which are very simple.. Just use the LED string in place of the batter shown as the example..

sOOhixrXnZ2dnUVZ snipped-for-privacy@supernews.com...

es

s 10%

ht

is

nto

=3D

Your scheme works, though it is pretty inefficient if the voltage at the feedback pin is a full bandgap, unless the LED string is pretty long.

You can use current sense with hysteresis to turn your led's on and off. Jim Thompson has a simple two-transistor circuit on his website that fits the bill

You might want to check the Joule Thief thread for some ideas. I made a current regulated version that holds the output current to 11.25 mA +/- 2% at 75% efficiency with input voltage of 1 to 5 VDC, and also a more powerful version of a blocking oscillator with a MOSFET that drives a string of 7 white LEDs at up to 30 watts at 80% efficiency with a nominal

I made a more complex circuit with a PIC and PWM which has up to 90% efficiency at 12-25 watts. It uses an amplifier from the 0.1 ohm current sense resistor so it does not affect efficiency.

For maximum battery life there are LED drivers that work at up to 98% efficiency if you have a voltage source greater than the LED string. You can probably make a boost voltage switcher that is at least 90% efficient. There are different design considerations if you have rechargeable batteries that should not be discharged beyond a certain point, or throw-away batteries that can be milked for the last drop of energy like a Joule Thief.

Paul

It may be satisfactory to simply run a UC3842 full blast (short IN to COMP). It's a current mode boost/flyback converter, so you'll get whatever maximum current the inductor runs. If the inductance is fairly high, current will be continuous and therefore fairly well regulated without regard for supply variation.

Tim

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

The "backward" switchers (flyback, sepic...) are the sources of current by nature of their operation.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

You can use current sense with hysteresis to turn your led's on and off. Jim Thompson has a simple two-transistor circuit on his website that fits the bill

Hmm, try putting a cap across the sense resistor in that circuit, and use logicout to drive the led's. then you'll get a sawtooth and the current will run up and down in the window between the two trip points. I think.

Linear Technology, and others, have a multitude of switching power supplies for LEDs. Check out data sheets and app notes for ideas.

Search on "joule thief".

Good Luck! Rich

The LM317-and-a-resistor current source wastes 2V to bias the regulator, and 1.25V drop across the resistor, so lighting a 2V LED consumes current from a 2+2+1.25 =3D 5.25V (minimum) source. It's not power-efficient to proceed in that way, of course.

So, you want an LED and a current sense resistor (that maybe steals

You also need a (I presume) battery that drives the reference and an amplifier, in addition to standard SMPS circuitry. One battery cell can be 1V (NiCd end-of-life) to 3.6V (lithium); easy recharging can be had for pairs of NiMH, 2 to 2.8V if you put 'em in series.

So, your problem is to do current sensing with op amps and switchmode power switching with lower-than-convenient voltages. One solution is to use more cells (four AA NiMH?).

My power supply is on the order of 6 volts (four NiMH in series) and the charge voltage, should I need it, is on the order of 12 to 20 volts variable as a function of sun angle.

I'll post the final article in ABSE.

Jim

The LM317-and-a-resistor current source wastes 2V to bias the regulator, and 1.25V drop across the resistor, so lighting a 2V LED consumes current from a 2+2+1.25 = 5.25V (minimum) source. It's not power-efficient to proceed in that way, of course.

So, you want an LED and a current sense resistor (that maybe steals

You also need a (I presume) battery that drives the reference and an amplifier, in addition to standard SMPS circuitry. One battery cell can be 1V (NiCd end-of-life) to 3.6V (lithium); easy recharging can be had for pairs of NiMH, 2 to 2.8V if you put 'em in series.

So, your problem is to do current sensing with op amps and switchmode power switching with lower-than-convenient voltages. One solution is to use more cells (four AA NiMH?).