Hi, I posted a couple weeks ago about a project I was thinking about; a large array of alphanumeric LED displays.

I got them today and took some current measurements:

With a 1.2 kHz refresh rate and 1/32 duty cycle, I can get a decently bright display with a 56 ohm resistor per segment. Current per segment (strobed) is about 2.5 mA. This is quite reasonable. But... what if the microprocessor froze, leaving one character stuck on? If all 18 segments are lit (worst case), then the current through the character's common cathode would be

(1.6V segment forward voltage / 56Ω) * 18 segments = 514.286 mA

If I have a 4-row display, with the rows multiplexed simultaneously and the 4 cathodes of a column tied together, the current draw would be over 2 amps!

So, should I use a power transistor to drive each column (to be safe), or just stick with general-purpose transistors and replace them if they ever get toasted?

If it's better to err on the side of caution, I'm kinda reluctant to use TO-220 MOSFETs because they're huge and would stick out like sore thumbs. Do transistors that can handle that much current come in smaller packages?

My guess is that it's fine to just use transistors rated for 500mA, and if I blow one, the flow of current will stop and no damage will be done to the display. I know it's nitpicky, but I'm being super-careful since these displays are over 20 years old, very rare, and very expensive.

 

If you use the proper current limiting resistor per segment there is no problem to leave all the time on. Be aware of not shorting current limiting resistors (solder splash or whatever), by fault, and destroy a segment(s).

You can use a TIP 110 to drive the segments. It is in a TO-220 package but you can place them horizontally if don't like them to appear tall compared to the displays.

 

The TIPP31 (NPN) and TIPP32 (PNP) are just made for your job.

TO92 package
2Amp max current
800mw max power.

There are also ZTX transistors which would do.

Surely you would also fuse against this fault condition so the transistor would not have to support the excess current for very long.

 

The TIPP31 (NPN) and TIPP32 (PNP) are just made for your job.

TO92 package
2Amp max current
800mw max power.

There are also ZTX transistors which would do.

Surely you would also fuse against this fault condition so the transistor would not have to support the excess current for very long.

Thanks! It looks like no one stocks the TIPP31 anymore, but a Zetex transistor looks like it would get the job done.

 

Thanks! It looks like no one stocks the TIPP31 anymore

I do. I've still got a couple of hundred of each in the spares box.

I am always amazed to see how older, common or garden components seem unavailable to Americans these days. Used to be the other way round.

 

It seems like the display is what you really want to protect. If this is the case, I would leave the transistors small and if the program gets stuck, the small transistor will act like a fuse before the LED blows because of the extended application of the peak current value. The transistors are cheap and easy to replace.

Also, the multiplexer should probably be running off of a timer interrupt. If your main program gets stuck, the interrupt routine will keep running, its just that the values will stop updating.

The last resort... if your interrupt routine does not keep kicking the watchdog timer, the response should be to disable the outputs.

 

If you use the proper current limiting resistor per segment there is no problem to leave all the time on.

Actually, this isn't necessarily so in a multiplexed display. The average current might be one value, and the peak current be a significantly higher value. This means that if you are lighting the LED continuously, you can only use the "average" value, but if you multiplex it (he mentioned a 1/32 duty cycle), your peak current could be much greater than the average rating, as long as it stays below the peak rating (and the average current is also less than the average rating).

The eye sees this blinking as the peak brightness. So if you strobe an LED with the peak current, the eye will ignore the off time and see the light as if it were continuously shining at the peak level, much brighter than if not strobed. This is a reason for strobing LEDs even if they aren't being multiplexed.

Lots of the new automobile tail lights are strobed like this. You can see it by darting your eyes back and forth while looking at one. The lights will seem to leave a dashed pattern when you do this. This is so that the lights will appear brighter than if they were lit continuously at a lower current.

 

My suggestion, if these displays are not found anymore in the market, to buy new ones because if one fails then you won't be able to replace it and the whole circuit will be useless.

 

I think I'm going to use 2N4401 transistors (max Ic = 600mA), since I have a bunch of them on hand already, and a low-valued fuse, maybe around 500mA.

I'm also adding circuitry that makes sure the displays are always off at power-on, as well as a manual cutoff SPST switch in-line with ground and the transistors' emitters. This is for when I'm programming the microcontroller with ISP. During programming, the circuit has power but the microcontroller is halted; the shift registers maintain their state and the uC's output pins can change state randomly. Therefore, I think it's a good idea to cut off the displays from the rest of the circuit when they're not being continuously strobed.

 

Nothing like belt and braces!

 

You can't rely on a transistor as a fuse because you can't guarantee that it will not short when it fails. Since there is a pulse train for the strobing, why not use a 555 as a missing pulse detector and drive a pass transistor for the LEDs? If it doesn't see a strobe the 555 will shut the pass transistor off.

 

and a low-valued fuse, maybe around 500mA.

You can't rely on a transistor as a fuse

I think he was going to rely on a fuse as a fuse. Quaint old custom.