An actual circuit

The jfet astable is cute.

I had a high-voltage supply and wanted to blink an LED when there was potentially dangerous voltage. I used a Supertex depletion fet to charge a cap, and a diac to dump into the LED at about 1 Hz. Five parts, including the LED.

The classic NPN astable circuit can hang up, with both transistors saturated. I wonder if he jfet circuit can hang too, with Idss grounding both drains and not enough gain to oscillate out of that state.

Even when they have a hang state, luck usually kicks them off into oscillation. Your source resistors and asymmetric drain resistors help it start up. Try making both drain resistors 3.3K.

If you make the source resistors lower, it will hang up.

...

Yes I noticed both points when I was designing it. I wanted to have it start up by itself, preferably without a kickstart capacitor. So I had a complicated circuit with two more diodes and a transistor in the hope that I could detect the hang state and force it off balance. I couldn't get that to work Then I accidentally made R2 3,3k and R6 3.3k and I didn't see how it could start so quickly with no other help. Eventually I noticed 3,3k which maybe LTSpice takes as 3k.

If R2 and R6 are both 3.3k then LTSpice says it slowly drifts into operation after 40 seconds. But why does it go one way and not the other? Is that an artefact of asymmetry in the simulation? Or is there some hidden asymmetry in the circuit I'm not seeing when R2 is

Yes, it is mysterious. Set both drain resistors to 3.3K and zoom way up on the two source voltages for the first 20 seconds.

LT Spice may have some hidden asymmetry. It has to do math sequentially, maybe.

In real life, parts will be plenty asymmetric. Especially jfets.

There are bunch of variants of this circuit, I call it the "Forest Mims Oscillator" cuz in one of his radio shack circuit books was where I first remember seeing it as a teenager. Variants with an inductor in Q2 collector are often used in those cheap solar garden lights.

It can be a pain to get to oscillate in LTSpice but works pretty good IRL.

I'm pretty sure he didn't invent it though and that I've seen it in books from the 60s also, it was probably invented about 5 minutes after the PNP.

On 5/24/24 17:59, Edward Rawde wrote:

Below are a pair of astable circuits. The left one is like yours, with a hangup state. I start it by specifying an initial condition. The right one will start all by itself.

Jeroen Belleman

========================================= Version 4 SHEET 1 1192 680 WIRE -80 -96 -272 -96 WIRE 96 -96 -80 -96 WIRE 240 -96 96 -96 WIRE 416 -96 240 -96 WIRE 864 -96 416 -96 WIRE 1088 -96 864 -96 WIRE -80 -64 -80 -96 WIRE 96 -64 96 -96 WIRE 240 -64 240 -96 WIRE 416 -64 416 -96 WIRE 864 -64 864 -96 WIRE 1088 -64 1088 -96 WIRE -272 32 -272 -96 WIRE 96 48 96 16 WIRE 112 48 96 48 WIRE 128 48 112 48 WIRE 240 48 240 16 WIRE 240 48 192 48 WIRE 768 64 704 64 WIRE 864 64 864 16 WIRE 864 64 848 64 WIRE 704 112 704 64 WIRE 944 112 704 112 WIRE 1040 112 1008 112 WIRE 1088 112 1088 16 WIRE 1088 112 1040 112 WIRE -80 128 -80 16 WIRE 128 128 -80 128 WIRE 384 128 192 128 WIRE 416 128 416 16 WIRE 416 128 384 128 WIRE 96 160 96 48 WIRE 416 160 416 128 WIRE 864 160 864 64 WIRE 992 160 864 160 WIRE 864 176 864 160 WIRE 1088 176 1088 112 WIRE -80 240 -80 128 WIRE -32 240 -80 240 WIRE 0 240 -32 240 WIRE 48 240 0 240 WIRE 240 240 240 48 WIRE 320 240 240 240 WIRE 368 240 320 240 WIRE 704 256 704 112 WIRE 752 256 704 256 WIRE 816 256 752 256 WIRE 992 256 992 160 WIRE 1040 256 992 256 WIRE -272 288 -272 112 WIRE 96 288 96 256 WIRE 416 288 416 256 WIRE 864 320 864 272 WIRE 1088 320 1088 272 WIRE 0 336 0 304 WIRE 0 336 -272 336 WIRE 320 336 320 304 WIRE 320 336 0 336 WIRE -272 352 -272 336 WIRE -272 448 -272 432 FLAG 96 288 0 FLAG -272 288 0 FLAG 416 288 0 FLAG -32 240 G1 FLAG 320 240 G2 FLAG 112 48 D1 FLAG 384 128 D2 FLAG -272 448 0 FLAG 1088 320 0 FLAG 864 320 0 FLAG 1040 112 D4 FLAG 752 256 G3 SYMBOL nmos 48 160 R0 SYMATTR InstName M1 SYMATTR Value 2N7002 SYMBOL res 80 -80 R0 SYMATTR InstName R1 SYMATTR Value 1k SYMBOL res -96 -80 R0 SYMATTR InstName R2 SYMATTR Value 100k SYMBOL voltage -272 16 R0 SYMATTR InstName V1 SYMATTR Value 10 SYMBOL nmos 368 160 R0 SYMATTR InstName M2 SYMATTR Value 2N7002 SYMBOL res 400 -80 R0 SYMATTR InstName R3 SYMATTR Value 1k SYMBOL res 224 -80 R0 SYMATTR InstName R4 SYMATTR Value 100k SYMBOL cap 192 112 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C1 SYMATTR Value 100n SYMBOL cap 192 32 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C2 SYMATTR Value 100n SYMBOL voltage -272 336 R0 SYMATTR InstName V2 SYMATTR Value 3 SYMBOL diode 304 240 R0 WINDOW 0 -33 66 Left 2 SYMATTR InstName D2 SYMBOL diode -16 240 R0 WINDOW 0 -40 61 Left 2 SYMATTR InstName D3 SYMBOL nmos 1040 176 R0 SYMATTR InstName M3 SYMATTR Value 2N7002 SYMBOL nmos 816 176 R0 SYMATTR InstName M4 SYMATTR Value 2N7002 SYMBOL res 1072 -80 R0 SYMATTR InstName R5 SYMATTR Value 1k SYMBOL res 848 -80 R0 SYMATTR InstName R6 SYMATTR Value 1k SYMBOL res 864 80 M270 WINDOW 0 32 56 VTop 2 WINDOW 3 0 56 VBottom 2 SYMATTR InstName R7 SYMATTR Value 100k SYMBOL cap 1008 96 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C3 SYMATTR Value 33n TEXT 88 376 Left 2 !.ic v(G1)=0 TEXT 48 -136 Left 2 !.tran 100m

Rufus P Turner's FET circuits is probably where I got the doing it with FETs idea from. It's on page 50 and easy to find a pdf. It also says the RC circuit should matched to within 1% but not sure why.

I've not previously seen it with source resistors. The idea of not connecting the sources to ground came about because I didn't want the LED drive to affect the RC circuit. So in my teens I just connected the source directly to the transistor base with no other component and put the LED current limiting in the collector circuit.

Instead of nailing the top end of the gate resistors to V+, one can connect each to its own drain. That makes the fets go linear if they are not oscillating.

Your right circuit sort of does that.

Thanks for that. I guess I'm biased towards components I could easily get in the 70s. And also through hole components I can easily build a real circuit with.

...

I like to use surface-mount parts on a Dremeled PCB.

Thanks I probably do have some copper clad board I could do that with.

Use a round-end carbide dental burr. They are cheap on ebay.

It's cool, kind of an art form. Takes some practice.

I have a few square feet of gold-plated copperclad FR4, which is even nicer. Regular copperclad looks grungy after a few months.

I’m not nearly as posh as you are, so for my purposes a spritz of Krylon clear acrylic on a clean board is a ticket.

You can solder right through it, and it keeps the copper looking nice for a long time.

Cheers

Phil Hobbs

Not sure why that made me laugh.

I seem to remember, a long long time ago using copper clad board, some clear sticky tape and a very sharp knife. Followed by Ferric Chloride. An etch resist pen could also be used if the tape removal didn't go exactly as planned.

But yes it's true that after removing the etch resist you might want something to keep the copper looking nice and shiny.

Oh, I haven’t etched a PCB in forty-odd years. This is strictly dead-bug stuff.

Cheers

Phil Hobbs

Soft Scrub. It's basically an optical polish.

Insect cruelty. Plus you have to count their tiny feets backwards.

I generally use Barkeeper’s Friend on a damp paper towel. (That’s what’s under the bathroom sink at the lab. )

The only good bugs is dead bugs. ;)

Cheers

Phil Hobbs