LED Flasher goes ON-dim-ON-dim

Thread Starter

Roderick Young

Joined Feb 22, 2015
408
flasher.gif
Here's an interesting puzzle for you. I have this LED flasher as part of a larger design. It is flashing a series string of 8 high intensity LEDs on a bicycle trailer. Some design requirements:
  • Must self-start, and not lock up.
  • Must tolerate a 50V supply, 24V being nominal.
But when this was hooked up, the LEDs flashed ON...dim...ON...dim... instead of ON...off...ON...off

It's already committed to a PC board, but I found a way to get close to the desired behavior without adding additional components. Maybe someone has a better idea?

p.s. Don't worry about the flash rate or brightness for now.
 

Alec_t

Joined Sep 17, 2013
14,280
The diodes in series with the bases of the Darlingtons are preventing them turning off fully. Remove them, so that the capacitors can pull the bases negative: or else, if you want to limit (clamp) the negative excursion, leave their cathodes connected to the bases but connect the anodes to ground.
 

ian field

Joined Oct 27, 2012
6,536
View attachment 96581
Here's an interesting puzzle for you. I have this LED flasher as part of a larger design. It is flashing a series string of 8 high intensity LEDs on a bicycle trailer. Some design requirements:
  • Must self-start, and not lock up.
  • Must tolerate a 50V supply, 24V being nominal.
But when this was hooked up, the LEDs flashed ON...dim...ON...dim... instead of ON...off...ON...off

It's already committed to a PC board, but I found a way to get close to the desired behavior without adding additional components. Maybe someone has a better idea?

p.s. Don't worry about the flash rate or brightness for now.
If it was 50V steady - I'd just use a resistor to charge a 47uF cap and a diac to dump it into a bank of power LEDs. The DB3 diac can handle a peak pulse of 2A.

The diac is bidirectional, but since you're using DC you could use a small TO92 SCR with a zener diode to trigger it at some voltage less than the minimum supply.

You could also use the SCR equivalent made with a complementary pair of BJTs - then you have access to both "gates", and can take advantage of any programmable unijunction circuit.

All these options produce narrow intense pulses depending on the capacitor size, repetition rate depends on the supply voltage and charging resistor.
 

Thread Starter

Roderick Young

Joined Feb 22, 2015
408
Dont see the point of the diodes, just use a normal multivib circuit with two caps and the base drive resistors.
Since the circuit might be exposed to higher voltages, I was scared that I would violate the minimum base voltage of -5V (might have been -8V, I'd have to look it up) on the transistors, during part of the cycle, causing them to break down. That's the idea behind D8 and D10. If my supply was just 6 volts or something, I wouldn't have such fears. D7 and D9 are to prevent a lockup state where both transistors stay on. I found on the previous prototype that the multivibrator would lock up after about a day of operation. It's solar powered, and the supply ramps slowly in the morning.
The diodes in series with the bases of the Darlingtons are preventing them turning off fully. Remove them, so that the capacitors can pull the bases negative: or else, if you want to limit (clamp) the negative excursion, leave their cathodes connected to the bases but connect the anodes to ground.
Interesting idea. I don't want to clamp the bases at ground, because the base going negative is part of the timing. However, I could put a high value resistor between the base and ground, to bleed off any leakage current. I should try that and see what happens.
The diac is bidirectional, but since you're using DC you could use a small TO92 SCR with a zener diode to trigger it at some voltage less than the minimum supply.
Should have done that in the first place. I don't really need control of the pulse width, just need some bright flashes to tell people that there is something present. With that scheme, the thing would flash faster at higher voltages, but that's ok. What I absolutely don't want is for it to go dark, or to lock up with the LEDs on dimly - a concern at minimum supply voltage. Could be done, I think. That would possibly be a good revision if I ever spin the PC board.
 

DickCappels

Joined Aug 21, 2008
10,152
I see the point of the diodes on the bases, to keep the emitter-base junction from breaking down, but it is possible that since the diode blocks current from the base that leakage current times the effective beta at that current is enough to keep the LEDs dimly lit?

You could check that by putting a from the base of Q4 to its emitter.
 

Picbuster

Joined Dec 2, 2013
1,047
current runs via les diode 316K( can't readcleearly) to base of darlington.
Try to measure base voltage and calculate resistor to force it low when needed.( in the off cycle)
 

Thread Starter

Roderick Young

Joined Feb 22, 2015
408
current runs via les diode 316K( can't readcleearly) to base of darlington.
Try to measure base voltage and calculate resistor to force it low when needed.( in the off cycle)
This is exactly what was happening. In later experiments, I found that there is a current path even when the LED is supposed to be off, through LED3, R43-R46, D9, R41, D8, and into the base of Q3 to ground. With the high efficiency LEDs I'm using, these few microamps are enough to give a visible glow. I blame the auto-start diodes D7 and D9. I was able to remove D9, and tie the top end of one of R41 or R42 to the positive supply, and still guarantee an auto-start. It was not entirely clear which resistor to tie to positive (only one can be tied high). Anyone care to guess which one it was?
 

ian field

Joined Oct 27, 2012
6,536
Since the circuit might be exposed to higher voltages, I was scared that I would violate the minimum base voltage of -5V (might have been -8V, I'd have to look it up) on the transistors, during part of the cycle, causing them to break down. That's the idea behind D8 and D10. If my supply was just 6 volts or something, I wouldn't have such fears. D7 and D9 are to prevent a lockup state where both transistors stay on. I found on the previous prototype that the multivibrator would lock up after about a day of operation. It's solar powered, and the supply ramps slowly in the morning.

Interesting idea. I don't want to clamp the bases at ground, because the base going negative is part of the timing. However, I could put a high value resistor between the base and ground, to bleed off any leakage current. I should try that and see what happens.

Should have done that in the first place. I don't really need control of the pulse width, just need some bright flashes to tell people that there is something present. With that scheme, the thing would flash faster at higher voltages, but that's ok. What I absolutely don't want is for it to go dark, or to lock up with the LEDs on dimly - a concern at minimum supply voltage. Could be done, I think. That would possibly be a good revision if I ever spin the PC board.
Another real easy dodge is possible because of the complementary asymmetry of the 4-layer device that could just as easily be a SCR instead of a PUT.

You can basically turn the textbook PUT pulse generator upside down and use an SCR instead - you could probably go up to about 12A rating before gate trigger current became a problem.
 

Picbuster

Joined Dec 2, 2013
1,047
This is exactly what was happening. In later experiments, I found that there is a current path even when the LED is supposed to be off, through LED3, R43-R46, D9, R41, D8, and into the base of Q3 to ground. With the high efficiency LEDs I'm using, these few microamps are enough to give a visible glow. I blame the auto-start diodes D7 and D9. I was able to remove D9, and tie the top end of one of R41 or R42 to the positive supply, and still guarantee an auto-start. It was not entirely clear which resistor to tie to positive (only one can be tied high). Anyone care to guess which one it was?
The current= (mainout-Vdiodes) /(20+316) all in mA this will flow into base. now what you want is <2 x 0,6V in off lets say .5V.
R(base-emitter) = .5/(mainout-Vdiodes) /(20+316)) now you hit the mainout. R is depended on mainout. this makes it off state instable.
 
Top