2
\$\begingroup\$

How do you properly monitor supply voltages for digital circuits (microcontrollers, FPGAs, RAM)? (I stumbled upon this question when working on safety critical systems)

What I have seen a lot is to have ADCs which measure the supply voltage regularly, but these kinds of circuits perform weakly against glitches as most of the time the ADC is not measuring at the right moment. Glitches can lead to memory corruption, which can be catastrophic. Furthermore, having a look at http://www.analog.com/media/en/technical-documentation/product-selector-card/Supervisorsfd.pdf it seems like some resilience against detecting glitches seems even desired. What makes sense, because as long as a glitch does not have a certain width or depth, it has no impact.

But I have also stumbled across some ICs (an RTC for example) which besides their normal brown-out circuitry have some edge detection to detect sudden rises or falls on the supply voltage.

To summarise, isn't it desirable to have beside the normal "static" voltage monitoring (with things like ADCs or comparators) some circuitry which can detect glitches (of a big enough width of depth to have an impact)?

\$\endgroup\$
  • 2
    \$\begingroup\$ Ask yourself what is it that you are trying to achieve and does a simple "product" have some special need for power supply self-monitoring. What equipment is required to have self-monitoring of power supplies and why? What type of product might your question be aimed at? And, when you do detect a glitch, what do you do about it or what should you do about it? \$\endgroup\$ – Andy aka May 25 '18 at 9:44
  • \$\begingroup\$ Many electronic systems do not have any form of supply monitoring and just rely on the power supply to supply the correct voltage. Most of these systems work fine. Having a power supply monitoring system in place does not guarantee safety critical operation either. You should avoid falling into trap of learning about a possible issue, do everything to avoid it but missing other (more critical) issues that your system might have. \$\endgroup\$ – Bimpelrekkie May 25 '18 at 9:57
  • \$\begingroup\$ @Andyaka No, a simple product does not need this usually. The case of the RTC is special as glitches can have a bad impact on the accuracy of the clock. On the other hand, safety critical systems need this, but just there I was surprised that often there is just an ADC which IMHO does not properly protect digital circuitry. If a sufficiently large glitch is detected, the system is unsafe and should take measures to get back into a safe state (reset or reload corresponding memory). \$\endgroup\$ – njg May 25 '18 at 10:00
0
\$\begingroup\$

But if there is a 'brown out' and the voltage dips below a certain value then having an ADC to monitor the situation isn't going to help.

An 'electronic fuse' is something that I've used in the past for current faults. The TPS2420 also has UVLO but it is not set-able, on some other electronic fuses they are. These type of parts have a fault pin and you can set the fault current.

enter image description here

If your looking to monitor the voltage for 'dips' this is a good circuit and is found in many power IC's, or you could build one with discrete components:

enter image description here

The schmitt trigger comparator ensures that the output doesn't flucuate and has hysteresis so if there is a dip and it's rapidly switching it doesn't switch on fast oscillations. The resistive bridge determines the voltage that the circuit detects (which should be matched with the comparator reference voltage). The current source ensures that the comparator still operates even under low voltage conditions (and you should have a capacitor on the linear supply that should keep the circuit running in bad conditions). You could run the output of this circuit into a S/R latch or the interrupt pin for a microprocessor.

|improve this answer|||||
\$\endgroup\$
0
\$\begingroup\$

The problem is that once your supply is not stable, all bets are off. Any glitch that takes Vcc below the minimum threshold in a processor circuit or similar is really bad news.

If I suspected that this was the case but was not able to catch the event with a scope (perhaps because it was very infrequent) I would consider designing an analogue latch circuit to indicate in some way the fact that the supply had fallen below limits. But this would still be to confirm the problem so that I could set about trying to fix it at source. I would not be trying to make a processor detect when its own supplies were flaky (although things like brown-out detection exist, but that is a bit different).

|improve this answer|||||
\$\endgroup\$
0
\$\begingroup\$

There are some different ways to monitor supply voltage. One simple and effective solution is a power on reset chip such as the DS1233AZ-15+.

https://www.digikey.com/products/en?keywords=DS1233AZ-15%2B-ND

Read through the data sheet to understand how it works. You could make your own supervisor circuit using a high speed comparator and a voltage reference. Sampling of the supply voltage using an ADC probably isn't necessary unless you need a log of the data. Most likely you just need to know if something happened and act accordingly.

|improve this answer|||||
\$\endgroup\$

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.