high voltage LED blinker

No, the top resistor is 2M from 120 vdc, so the current is more like

50 uA.

On a sunny day (Fri, 27 May 2022 17:08:44 -0700) it happened John Larkin <jlarkin@highland_atwork_technology.com> wrote in snipped-for-privacy@4ax.com:

How about a neon bulb with a series resistor and capacitor in parallel? Or if it must be a semi-conductor a UJT with a LED in the discharge.. not sure that b1 b2 resistance is high enough, multivibrator with 2 transistors perhap?

No spice here atm

This is one good option, a neon bulb is another. But are you sure you want your discharge current *low*? Mine is set up to 5mA, as it is the DMOS SOA should something go wrong and the discharge circuit got constantly on.

Best regards, Piotr

One of my first circuits, as a kid, was a neon blinker. But I want a lower trip voltage and a surface-mount indicator.

Popular Electronics had a fun circuit: build a NE2 relaxation oscillator with two neons in parallel. The lower breakdown bulb will usually flash, but tiny disurbances - light, efields, radiation - will shift to the other.

For some reason I have a bag full of NE2's.

Diacs are cool. I never used one before. I can imagine off-label uses. LT Spice has a diac in the parts library.

I built that last-thing Friday, and left when it blinked. I might study it more next week.

The test circuit on the SODDB3T data sheet has it discharging into

20K, so it has some low shutoff current. My blink looked much brighter with the 1K resistor in series with the LED.

Of course it's brutal inefficient to discharge 32 volts into a 2-volt LED. An inductor or transformer would help, at some hassle. There are cute multi-junction leds around.

I don't know how I could simulate the diac function with transistors and stuff. There is some semiconductor magic going on inside a diac.

These strobe trigger transformers are cheap n cheerful, I think the tend to be about 1:10 with a couple mH on the secondary.

Just a series L might work, make a sort of single-cycle switching power supply.

It's basically an SCR with a breakdown that triggers it; two transistors makes the SCR, and it triggers by collector-base breakdown

There are variations; GE made a SBS (silicon bilateral switch) with mismatched positive/negative breakdowns, and of course there's BIG CURRENT sidactors

A diiac should just be a triac with a set trigger voltage which shouldn't be too difficult to model. I think that there are Spice/LTspice models of a diac that can be used.

I have a product that uses two neon lights but it is from a HV supply over 200V.

I wanted to do something like what you did but as you did, built the blinky neon lights many years ago.

Too bad that the NE-1/NE-2 isn't available in an SMT package though.

boB

At 1400V, that would waste 7 watts continuously.

Cheers

Phil Hobbs

Last time I looked at a diac spec sheet (many years ago) I noticed the trigger voltage is not exact, has a tolerance like +/-3 volts. Not a surprise I think. But I would up your cap's voltage rating, just to have more safety margin. The cost of 50V over 35V in an Aluminum electro is trivial. If youre going to use Tantalum, thats a more serious price jump. I would also feel better putting a 39V or 47V zener or across that cap, just in case a part in the LED string fails open, and the cap never gets discharged. Exploding caps, even just one, are bad for business ;-) regards, RS

IIRC, John mentioned it was a discharge circuit, so it should act fast. That's correct: 7W would be dissipated if the DMOS shutdown circuit fails and no one notices that, hence me urging to mind the SOA. But the power should be essentially 0 watts under normal circumstances, assuming there is some indicator of normal operation (an intrusion detection switch, low voltage rail presence indicator, you name it).

This is how my discharge circuit works. Basically, a normally-on relay minus the contact reliability problems and coil losses. If it were supposed to be constantly on, a simple bleeding resistor would do.

Best regards, Piotr

Namely, in addition to the regular DMOS current source with a source resistor, there is a diode connected to the gate and then to a tiny 6mm OD pulse transformer with <1% duty cycle, powered directly from the MCU pins. No pulses (a controller-side failure or intrusion detection) => discharge starts automatically. "Shut up" pulses present => DMOS "disappears". 5kV isolation and extreme simplicity thanks to a tiny core and 20cm of TIW wire.

Best regards, Piotr

Glad you tried a diac, have you tried replacing the diac with bipolar transistor (reverse connected and with base open, sometimes called negistor)? It is similar effect I think but with breakover about 9V.

Remember back in the 1960s when diac symbol was like a BJT with two emitters and no base connection - that gives a clue to internals.

I think the semiconductor magic is a combination of emitter-base reverse breakdown, reverse beta and impact ionization? Most diacs are 3 layer devices so the negative resistance is not from transistor-transistor feedback as in a schockley diode.

piglet

The 32v diac is ideal for this use. It is not specified for holding current (stays-on) but that should be OK with my roughly 50 uA current-limited source. The academic challenge is to use the energy in the cap more efficiently to get a brighter blink on less supply current.

Shockley was obsessed with developing a 4-layer device, rather than evolving the transistor. That was one factor in the split:

Microamps won't explode this cap. Al and tant and even most ceramic caps just leak when the voltage gets high. I have seen one brand of polymer elec that failed hard without warning at under 2x rated voltage, but most seem to just leak more as voltage goes up.

Small surface-mount tants are 10 cents in quantity. Ceramics lose c as voltage goes up, so don't store the expected energy.

A small Al cap would be OK, more like 5 cents, but might be tricky over temperature. In our business, parts cost for passives doesn't matter, not when we're lucky to get an FPGA for $200.

The blinking LED is the indicator. It blinks until the HV is below safe levels, and works even when power is off.

A resistor has a very long exponential delay. The depletion fet discharge is linear, much faster, and it blinks!

IIRC MnO2 tantalums are some of the longest-life caps there are once they make it out of the first drop in the bathtub curve, the rise at "EOL" is very shallow when they're treated right.

They have a self-healing mechanism, that however can become a self-destroying mechanism when they're fed by a low impedance supply and

2 Meg likely doesn't qualify.

Or rather large dv/dt plus low impedance source is bad news for tantalums.