| Electronics > Projects, Designs, and Technical Stuff |
| Protecting against Voltage Regulator failure? |
| (1/4) > >> |
| kizmit99:
Hi All, I was tinkering today on an ongoing project that uses one of these Cyclone-II dev boards: (They are cheap and widely available, so I have no idea who made it, or frankly even who I got it from at this point.) I was testing a new PCB I had made which uses one of these as a 'daughter board'. The board had been running for about a day without any issues, then while trying to debug a portion of the circuit which really doesn't involve the FPGA much at all I noticed everything had crashed. After doing a post-mortem it appears that the 3.3V regulator on the dev board has failed with a short from Vin to Vout. The regulator is an AMS1085CT-3.3 which is basically a 3 pin 3.3v regulator - Vin max should be 15V, max current 3A. I was powering it with a 9VDC (1.5A) wall wart (basically something sold as an Arduino/RasPi wart). I suspect the failure fed 9V to the FPGA (and the rest of my board) and fried all those 3.3v parts. As to the cause - it is certainly possible that I shorted 3.3V to Gnd with a scope probe on one of the chips. I was debugging an audio out amp, and these two pins are very close to one another. I did not notice a spark, pop or magic smoke so I can't say for certain that I shorted them, but certainly a possibility. However, since the regulator is rated to supply 3A and the 9V supply was rated at only 1.5A I'm a little suspect that a brief short should have killed the regulator - but again, always possible I suppose. The only other idea I have is that the regulator just failed on it's own, so I'm willing to chalk the failure up to me shorting the output... What I'm not happy with is the failure mode - shorting Vin to Vout seems like the worst possible failure mode. And, from what I can tell, it does seem to have taken out all of the 3.3v chips on both boards. So, given this kind of failure, is there a reasonable way to protect against your voltage regulator crapping out and over-volting the entire board? Also, I'm curious if anyone else has seen something similar - cheap regulators failing in Vin shorted to Vout mode? I've used a different copy of this board on the previous iteration of the main board for many months without any kind of similar mishap. And I'm probably just as likely to have shorted out that board given I was using it for so long. I had been powering that one from a 5VDC (2A) wart instead of the 9V one though. My new design has another 5V regulator that needs to be powered from this Vin, that's why the change to 9V on this one. Anyway, I'm open to any thoughts or suggestions on how to protect against this failure mode... I don't relish the thought of toasting everything with a small slip of a scope probe... |
| Ian.M:
Ringing on the input (input capacitance resonating with the cable inductance) can transiently double the input voltage. This normally only happens if you connect the PSU to the board with the power already on or if the PSU has a switch in series with its DC output, but briefly shorting the regulator output could also cause it. You didn't have any problems wth 5V in because the ringing never exceeded 10V Therefore any time you use a LDO with an input voltage greater than half of its abs. max. rating, and there is a significant length of wiring between it and the PSU, you need to make sure that any ringing is damped or clamped at well under the regulator's abs. max input rating. Typically one adds an aluminum electrolytic capacitor of ten times the total value of any ceramic input capacitors used so its ESR damps the ringing. A beefy 12V Zener across the input to clamp the ringing would also work. To protect against the consequences of regulator failure due to other causes you need a crowbar circuit that will short the output to ground if the rail voltage becomes excessive. It must latch on so it holds the short till the supply fuse blows or over current protection trips, or if the supply is current limited, hold the short until power is removed. This is easy for rails 5V and upwards - see the classic TL431 + TRIAC crowbar circuit in the full TL431 datasheet - but with only a 3.3V rail, a TL431 wont have enough headroom to trigger a TRIAC reliably. MOSFET based crowbar circuits are possible, but you need a really beefy MOSFET to guarantee it wont fail before the fuse, and plenty of gate drive from a maintained supply so it continues to hold the MOSFET hard on. |
| georges80:
Possibly overheated/died. 9V in and 3.3V out, so 9-3.3 = 5.7V (could be worse if the wallwart is outputting higher voltage) that has to be dropped (linear regulator). No idea what the 3.3V side is drawing, but say it was 300mA. Then you have 5.7 x 0.3 = 1.7W dissipated in the regulator... Your 5V operating one is only having to drop 5-3.3 = 1.7V... cheers, george. |
| kizmit99:
--- Quote from: Ian.M on March 04, 2019, 11:33:13 pm ---Therefore any time you use a LDO with an input voltage greater than half of its abs. max. rating, and there is a significant length of wiring between it and the PSU, you need to make sure that any ringing is damped or clamped at well under the regulator's abs. max input rating. --- End quote --- Very interesting - that's the first I've heard of this effect. (To be fair, my background was in 1990 era digital, and we had analog guys for the power supplies :P) The variant of this regulator that Digikey stocks shows a max Vin of 30V, so perhaps replacing these parts on my remaining dev boards might be prudent... (I'm quite certain the folks making these boards are using the cheapest, sketchiest parts available) --- Quote from: Ian.M on March 04, 2019, 11:33:13 pm ---Typically one adds an aluminum electrolytic capacitor of ten times the total value of any ceramic input capacitors used so its ESR damps the ringing. --- End quote --- That very well may be why I've not experienced it on other projects in the past. I usually drop a bulky cap on both the inputs and outputs of regulators, just for good measure. I usually use tantalum caps (because I have a supply) - you specifically mentioned aluminum, do you think there would there be a difference in this application? --- Quote from: georges80 on March 04, 2019, 11:56:16 pm ---Possibly overheated/died. --- End quote --- Certainly another possibility. I'm not sure what the board was pulling, but I suspect it is in the sub 100mA range. But as you point out that would still be a half-Watt the little package would need to dissipate (with no heat sink of course), compared to just around 100mW at 5V... I didn't notice the regulator getting hot prior to the meltdown, but I may just not have noticed. Although, now that I look, this part is supposed to have an internal thermal shutdown, so I would have thought that a slow-burn would probably just shut it off, not melt it down. In fact they don't even spec the max power-dissipation, just temperature related numbers. My gut now says this isn't the culprit. I should also probably also point out that the specs on the board itself do specify 5VDC input. But looking at the schematic, verified with inspection of the board itself, the input power only goes to this Vreg (and its input cap) nowhere else. So I was thinking, since the part was spec'd for 15V input I should be ok with the 9V input (really only need about 7V in for the other regulator, but 9V is what I had lying around). I did figure the additional power waste would be less than it probably is, and I definitely didn't consider input voltage ringing... |
| Doctorandus_P:
It is a part of the price you pay for cheaping out. Putting too much voltage on circuits often does not destroy them immediately, but causes too much current flow, which heat up the components too much and they release the magic smoke. This is why Lab power supplies with adjustable current limit are invented. If you set it to a voltage and current slightly above what your board needs then if it fails, on over current, the output voltage collapses and your circuit has a chance to survive. 3V3 circuits tend to be more sensitive to abuse than the older higher voltage stuff. Alternatively you can add a "Crowbar" circuit. This is most often a TL431, Thyristor, (poly)Fuse and some passives. The Idea is that if the TL431 detets an over voltage, it fires the Thyristor, and the thyristor shorts Vcc to GND, which protects the circuit. Deliberaty shorting the power will of course cause too much current to flow, and the fuse wil blow. (Or use a decent Lab Power supply). |
| Navigation |
| Message Index |
| Next page |