I am tinkering around with shift registers and I want to power an IR sensor from one of the output pins on the shift register. The register can push or pull 8mA/pin(recommended) or 20mA/pin(max) with a max of 50mA through the entire register. The IR sensor draws 5mA on average and 70mA during short bursts. I am going to add a 10uF cap across the power and ground of the IR sensor. Will this work, or will it destroy my register in the long run?

Would having capacitors on the shift register output pins destroy the shift register, since they would require brief bursts of current to charge up every time?

  • 3
    \$\begingroup\$ In general, if you exceed the manufacturer's absolute maximum specifications, for any device, all bets are off. Use the shift register to switch a transistor instead. \$\endgroup\$ Jan 16 '17 at 5:52
  • \$\begingroup\$ @Roger: Your comment suggest that within the maximum specifications everything is fine, which is not the case! That is what the normal operating conditions are for. The maximum specifications are the conditions beyond which the device can be permanenly damaged. \$\endgroup\$ Jan 16 '17 at 7:56
  • 1
    \$\begingroup\$ @WoutervanOoijen My comment suggests nothing of the kind. You've read into it an assumption that I neither made nor enunciated. \$\endgroup\$ Jan 16 '17 at 7:58
  • \$\begingroup\$ @Roger: by stating "if you exceed the absolute maximum specifications" instead of "if you exceed the normal operating conditions" you didn't say anything about range between the two. In that range the working of the device is not guaranteed, so all bet's are off too. By limited your "all bet's are off" to the range beyond the maximum specifications you make your statement unneccesarily weak. \$\endgroup\$ Jan 16 '17 at 16:34

What are absolute maximum ratings there for? No, what are they really there for?

They are to protect the manufacturer's reputation. Engineers talk to each other. I used this ACME part and it failed. I used this NXP/TI/Analog/Maxim part, and it kept working! The reason it kept working is because the data sheet accurately reflected what the part would do, and warned me against trying to make it do things that, in the long term, would cause sufficient numbers of devices to fail, that the manufacturer would get a bad name for selling things that didn't work.

So, can you draw 50mA from a pin that's rated 20mA absolute max? Certainly. For how long? Maybe a few uS, maybe a month, you do the tests. The manufacturer has already done the tests, thousands of hours of them, and decided that something about the part, the bond wire fusing, or electromigration in the diffusions perhaps, means that they expect that running at that current will change something vital about the device, like how well it works, or whether it works.

As you are anticipating pulsing the diode, putting a capacitor across the diode is the worst thing you could do, as it increases the inrush current when you pulse the diode on.

Use the limited current from your S/R pin to switch a driver that is specified to handle the IR diode current.

  • 1
    \$\begingroup\$ Many parts' specifications don't contain enough information to make them usable without some level of "how things generally work" conjecture. If one went purely by data sheet specs, one couldn't even connect a typical CMOS output to a typical CMOS input without some kind of series resistor, since nothing in the typical specs would guarantee that the input wouldn't behave as a 1pf cap with a 1ohm ESR, that the output wouldn't switch at a rate of 1,000 volts per nanosecond, and that the part wouldn't be destroyed by the resulting 1 amp spike. \$\endgroup\$
    – supercat
    Jan 16 '17 at 18:47

When the SR output switches from low to high, it will be driving a nearly/completely-empty capacitor that will instantaneously act like a short circuit and then start to charge. It has a similar problem with discharging the capacitance when the output switches from high to low. This will be putting stress on your SR's output stage beyond that in the manufacturer's data sheets and that may well diminish its reliability.

It's a cheap and simple step to drive a switching transistor stage with the output. (You could use a MOSFET but also look at a ProFET, they might be available at whatever your voltage is.) Then you have completely remove that possibility and any consequences from your design.


Putting capacitors on the output may not destroy your shift register, but it's certainly not a good design.
Why can't you buffer your register outputs with a FET device for a single channel that will sink 70+ mA. Perhaps something like this: http://www.diodes.com/_files/datasheets/ds30734.pdf

Or if you need multiple outputs you could even consider the fabulous DMOS FET equivalent of the ULN2003: http://www.ti.com/lit/ds/symlink/tpic2701.pdf

You could even consider a 'Power' shift register like the http://www.ti.com/lit/ds/symlink/tpic6c595.pdf which includes everything in the one chip.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.