When the chance arrived to buy a mechanical keyboard for not a lot, naturally, [Hales] jumped at it. Then it started having odd intermittent problems with some keys appearing stuck, which led to a teardown and some fault finding. The culprit was a white LED — but why this was the case is a fascinating story.
Stripping it down there didn’t seem to be an obvious culprit, but eventually, the trail led to a lack of diodes in the matrix. This keyboard had an extremely clever yet rather cursed design in which it used LEDs as both illumination and as a diode in the keyboard matrix circuit, and the faulty LED had a reverse breakdown condition that could be triggered under certain operational conditions.
More unexpectedly, it would sometimes hold on to its reverse breakdown state even after power off. Just when you think you understand a component’s properties, there’s always room for surprise. And we can safely say we’ve learned something about the design of cheaper keyboards in reading the account. It’s clear that when it comes to ‘boards, it’s best to take no chances.
>LEDs are not meant to be used as diodes.
leds are diodes, its right in the name
> LEDs are almost always only ever used in the forward direction and every circuit I’ve seen that potentially subjects them to reverse bias puts a standard diode in series with them.
bi color 2-pin LED say hi
Bi-color LEDs are two LEDs in reverse parallel, so they always have a diode going the other way, and neither LED is reverse-biased.
And I have been wondering recently about using LEDs as diodes, since I’ve ended up with so many from old Christmas light strings. (Which, by the way, use alternating sub-strings in parallel to handle both phases of AC.) Now I have some ideas of what to watch out for.
They are reverse biased, but only up to the forward voltage of the other LED.
LEDs tend to have fairly high leakage current compared to other diodes, and rather low maximum reverse bias voltages, but if you take that into account you can use them as diodes. I’ve also used LEDs as makeshift reference diodes in the past.
Looks like he was reading spikes of 7v, add that to any photon based voltage and it could easily become a photocell.
> I’ve also used LEDs as makeshift reference diodes in the past.
Hahaha I did this for some BJT amplifier biasing once.
Some red SMD leds I had were the perfect voltage when fed with a 10K resistor. I added a cap to get rid of noise, but still laughed at the thought that this amplifier probably had a bad light rejection ratio and maybe even amplifiter-to-amplifier coupling if placed near a white object.
Even more fun: I was able to adjust the bias point of all these amps in the circuit by changing VCC slightly. My LEDs did not have as flat of a V vs I curve as I had expected.
Whilst not the best technical solution, I still think it’s cool to have your bias circuits glow.
Harman tried that once. The LEDs were machine-placed into the PCB and the leads were stressed during insertion. The stress caused some LEDs to fail open and the amp overheated and failed.
I think I read somewhere that the light sensor used in the original Lego mindstorms sets used its LED as a voltage reference in addition to functioning as a light source.
Oh, wow… digging a little more, I have found that a reverse-biased LED can be photo-sensitive and even act as a “single-photon avalanche diode”. That may explain why the keyboard would suddenly fail for no apparent reason. Perhaps ambient light could trigger the “failure”.
See https://www.teachspin.com/newsletters/ and look at May 2013.
That was also my first thought. I used LEDs as photosensitve devices already at many occasions, mostly, when I want to investigate some light effects with an oscilloscope and the first opto-part from the parts bin is an LED not a real photodiode. I did not know about the avalanche-didoe effect.
Thanks for the link.
I did try shining a torch on the LEDs of the keyboard to see if that triggered the problem. It was difficult to correlate cause and effect (because sometimes it would take hours before the problem started surfacing again). The only thing that got the keyboard into a marginal state (sometimes) was changing the brightness of the LEDs on the whole keyboard; but again perhaps a torch would have done it too if I had tried for longer.
Entertainingly, not only is every LED a diode, but technically every diode is also a LED. Just not always a useful color.
I suppose they’re also all solar panels, but that gets a little weird.
Being an indirect band gap semiconductor, silicon has a negligible probability of radiative recombination (basically, it’s impossible to conserve momentum while emitting a photon), so silicon diodes aren’t LEDs. There do exist exotic silicon microstructures that can be coaxed into emitting photons; the electronic band structure in those materials becomes tweaked enough by the physical configuration that light emission is possible.
>>LEDs are not meant to be used as diodes.
>leds are diodes, its right in the name
Hm, that seems to lack a bit of reading between the lines. I’m normally a stickler for precision in writing, but not to an obsessive extent. I’m pretty sure the intended meaning was “LEDs are not meant to be rectifiers” (where “rectifier” is meant as “allowing current to run only in one direction”).
>> LEDs are almost always only ever used in the forward direction and every circuit I’ve seen that potentially subjects them to reverse bias puts a standard diode in series with them.
>bi color 2-pin LED say hi
Again, the implied meaning is that LEDs are not normally exposed to anything exceeding their reverse voltage limits, which includes bi-colour LEDs.
Bi-colour LEDs work without causing trouble as the forward current drop of the LED emitting light (i.e. conducting) is less than any voltage that may cause malfunction of the one not emitting light (i.e. blocking) — and definitely less than the actual breakdown voltage (forward voltage drop of an LED is typically about 1.2V and the breakdown voltage around 5V, 10 times less than that of low-voltage, “normal” diodes at around 50V).
Argh, too quick to fire off, it should have been: “LEDs are not meant to be used as rectifiers”
When i repaired mine i accidentally damaged one diode and soldered jumper into it, while waiting for propper part. It would freak out, pressing neighbour keys.
This is very silly. The anti-ghosting diodes I use in my keyboards cost me a stunning 40c. Forty cents. For the whole keyboard. Come on.
Yep :) Not to mention that they had to use lots of zero-ohm resistors in this design to jump traces anyway.
I suspect it might have been done as an experiment into “can we do this in a product” in the hope that maybe some future design might somehow cost benefit. Perhaps once you add 3 colours of LED it becomes much more cost effective?
If lots of other units suffered the same fate as mine from the factory (bad reverse led) then perhaps they might scrap this design, but if this fault only appears later after purchase they might not notice.
I think I remember reading on this site (or somewhere else) about a radio telescope that was taken down by a faulty indicator LED in the (custom) main Z-axis motor drive.
Single point of failure for the fail 😞
There’s also the case of a particular floppy drive that would fail when the status led burned out.