LED driver

There are some kilovolt SOT223s around, too! But I hate to have all silicon between 5KV and ground.

I may be able to run the HV at 2400 volts. If I use a multi-junction LED, 24 volts maybe, it just about works with just a 3M resistor, nothing else.

I'd still like to come up with a good series blinker circuit, something that runs at microamps, has good controlled fire and release voltages (max power transfer!) and no hangup states. Just for fun, maybe.

Yeah. I got one green LED to make barely visible photons at 800 pA. The white Cree is blinding at 10 mA. I replaced the 20 old LEDs on my Mantis with 8 Crees.

Science experiment: use photodiode and lock-in to see how little current it takes for an LED to make any light.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Since no one keyed up on the obviousness of the solution, I will attempt to compress the width... ;-)

Synchronous bridge drive around the HV transformer. Connects +Vin to

+Vout as a DC transformer.

Discharge +Vin to 0.0V and +Vout is also ~0V.

Boom, done.

So, you need some regular transistors on the primary side, some high voltage transistors on the secondary side, a drive transformer with enough HV isolated secondaries for all the gates (or enough transformers, plural, to do it), and a chopper driver, which can just be a dumb 50% square wave. For soft start, you could further use a synchronous buck to go between system supply and primary side bus.

The controller or chopper or whatever has to keep running, so it needs a delayed supply, or auxiliary supply, or whatever.

Tim

-- Seven Transistor Labs Electrical Engineering Consultation Website:

From the few led's I've looked at, it stops being linear with I at ~100uA or so. you might fit it to some power law below that, I haven't looked below ~ 10nA.

George H.

The simplest way would probably be a very sensitive SCR or, if the gate current is too high, a helper circuit that acts like an SCR. A zener in series with the gate which sets the trip voltage, minus Vf of the LED stack. 3M resistor charges up cap, 48V is reached, SCR or SCR-style circuit triggers, LED which is in series with the whole chebang comes on, voltage across cap drops so the hold current vanishes, LED extinguishes, everything waits for next cycle when the cap is recharged enough.

Of course, the blink frequency will increase with the level of your HV source. Which could be a good thing, telling the service folks "Now is really not a good time to touch this!". If that's not desired there'd have to be another circuit that sense and shunt-regulates the voltage rise time across the cap.

For a really simple experiment it probably suffices to blink it at a low but odd frequency, maybe 135Hz or so. Then hang a large enough photo diode to the sound input of a laptop run on its battery and use a narrowband level meter or spectrum analyzer like in Daqarta. That software offers incredible noise immunity.

--
Regards, Joerg 

http://www.analogconsultants.com/

It depends on AC feedback and can be made to hang. I don't like sircuits with hang states.

Sure, but what happens when the power connector is unplugged? Or a fuse blows?

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation

C-multipliers don't store much energy. The usual circuit can't stand over a few hundred mV input ripple.

Given that, I don't need a blinker!

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation

I've looked into that. Sensitive-gate SCRs and sidac type things need 10s of uA trigger current.

One thing that might work is a classic two-mosfet astable multivibrator, powered by a 5 volt zener, underneath the entire stack. That can be made to oscillate at

1 uA supply current. It can turn on another mosfet that does the actual blinking. Done right, it always oscillates with no hang states. Even if it stalls the LED still gets the available supply current.

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation

"Lighting"

what you

parallel.

Sure. Also at about 30 times the space. How do you light pipe that out to your externally visible indicator port? Remember that?

10 to 1 time concentration will help a lot here, 96 mW led pulses will be quite usefully bright. 10-chip

Sure, but you have to control them in a way that keeps them all functioning. Have you solved that one yet?

BTW how high of Vds can you find depletion mode (IG)FETs?

?-)

It is too dumb o understand the question..

Can you pot the entire HV assembly and make it impossible for human contact? Like the HV PDA of a classic Tek scope, including the connector.

Put a divider in for making 100:1 measurements with an external test point.

Then you don't need any indicator.

tm

thinking

redundancy) that

overlay on

officially safe

Cool. Just the same since this is for a unusual hazard indicator RED LEDs are a better choice. Better effective brighteners as well.

?-)

Right, that's the advantage.

You mean the load, or the c-mult? Surely the c-mult's fine with it--just drop a few (or twenty) volts across the pass transistor.

Two of those cascaded and you've knocked ripple to nil, and eliminated the need for an equivalently large bulk filter cap.

[snip ASCII ckt]

1. What you need is irrelevant--WE're having fun.
2. Yes, methinks you'd still need a blinker, otherwise 200uA isn't visible enough.
3. But okay, here's the actual fix:

D1 R1 +5v >--->|----+--/\/\/---. (blinking) | | LED C1 --- V ~~> .47F --- --- | | === | | | ||--' Q1 R2 ||--/\/\/------+---'|--| | | | === .-' ^ Z1 | 20V ===

Seven parts. It's boring.

Cheers, James Arthur

These are cool. Surface mount, 3.5 mm square, come in 12, 24, and 48 volt versions.

24 is ideal for max power transfer from 48 volts with a resistor.

There are lots of tiny multi-chip LEDs around.

That's an ancient problem. You use a string of high-value resistors to set the gate voltages in a cascode-type stack.

Supertex SOT23s go up to 600 volts. But the stack need not be depletion fets. It may be convenient if the bottom one is depletion mode.

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation

Current limiting? Well that is your problem. ;-) I'm just taking care of the isolation.

Since you are generating the high voltage, I don't consider it cheating if you integrate the LED in the circuitry that creates the high voltage.

I meant "what's the quiescent current?"

If the blinkLED needs, for example, 1mA in-between blinks, you're toast.

Cheers, James Arthur

A multivibrator would work but it'll have three transtors and a lot of miscellanea. A more interesting approach would be a UJT:

They still sell inlarge quantities and cost around 10c in bulk. Peak current is in the uA range.

--
Regards, Joerg 

http://www.analogconsultants.com/

Maybe, if total circuit current can be kept down.

I meant something like this...

With the right-most mosfet maybe reconfigured as a current sink. Yeah, that's better.

The best blinker would turn on the LED at, say, 28 volts and turn it off at 22, and run the led constant-current. That would operate a 24 volt LED around the maximum-power-transfer point, from 48 a volt supply. There should be a way to do that.

I still don't really understand the PNP-NPN pseudo-SCR circuit, which the PUT is, specifically how to make one with low trigger current and no latch states. Spare-time project.

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation

It should draw no current until a threshold is reached and this threshold can be programmed. Leakage is in the tens of nanoamps. Then it essentially shorts like an SCR until the current through it (valley current) becomes too low. That valley current is tens of microamps but this should be ok because you want some current in the LED so the flashes can be seen even in daylight. The energy comes out of the capacitor during that time so it doesn't matter.

Peak current is low so I don't see a problem using the programmable UJT as an oscillator in the low single digit microamps.

AFAIR there is a SPICE model for the 2N6027 somewhere in the web. I've always wondered what a UJT would be useful for nowadays and this may just be the perfect application for one. Someone must use gobs of them because Arrow stocks almost 100,000 of these.

The multivibrator is ok but is more complex and does not have quite the range of operation as a UJT. Because it forces a fixed oscillation frequency on the LED and when the HV gets to be too low the 3M resistor can no longer support that. It could start choking and produce a less than pristine zippiness in light output. The UJT solution will just drop in flashing frequency but should still flash nicely.

--
Regards, Joerg 

http://www.analogconsultants.com/

One problem with the PUT is that the side electrode thingie has to be biased up, with a fairly low impedance source. That voltage has to come from somewhere. The higher the impedance there, the lower Iv, and the easier it is to latch it up. It's really just a PNP-NPN pseudo-SCR.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com