AVR Freaks

Do you need to distinguish "low voltage" from "no voltage at all" ?

Your diagram shows rectified signals - so why not just smooth with a capacitor, and read the voltage ?

Probably add a suitable "bleed" resistor to discharge the cap when the mains fails ...

Each input is used to lock a digital phase-locked-loop oscillator running at something like 16 times your input frequency. The output of those is used to internally generate a sinewave of the same frequency and phase as your input signal. You sample your incoming signal at 16 times its frequency and use a window comparator to check the incoming signal against the expected signal out of your internal sinewave generator.

I wonder if that's a real requirement ... ?

Perhaps some more information about the actual goal would help ...

http://www.catb.org/~esr/faqs/smart-questions.html#goal

A MID400 will monitor input AC and give a logic level output when the input faults (is missing at the next ZC).

https://www.onsemi.com/pub/Colla...

Jim

already bitten - in #10

(also #6 & #7)

Who cares if it is 5ms or 100ms?  You will see the led the same either way.

The line is full of transients, so you will probably need to wait a full cycle or two anyhow.

how long does the led stay on?  forever?  until the next valid cycle?

Come back and restate your specifications & be realistic.

You could possibly use reverse logic & light the led if it has not seen a voltage greater than 10 adc counts for more than 5ms...that just finds it has stayed very close to zero for >5 ms.

However, it will not be very effective working with transients/spikes

clear timer

loop:

  if voltage>10 counts

     led off   (if not using led pulse, below)

     clear timer

 else

  if timer>5ms

    light led   (or start an led 100ms pulse)

   clear timer

Note this is detecting if/when you leave the valley...it ignores how much voltage (how high) you eventually get.

And don't get me started on those where the hardware has been chosen and PCBs have been made before any R&D has been done.

I was called in on a project where they had a dialup modem module on a PCB they developed & were finding severe connectivity issues & needed to fix ASAP.  They gave me a prototype board & I soon found issues with their interfacing & presented a relatively simple fix.  Then was told they could not change the board layout at all, since they had already made & populated 50000 boards!

Slope = derivative. ..... Easy and uncomplicated

Maybe...probably not....since the line voltage is rather noisy/spikey you will get plenty of realworld large slopes when trying a derivative (an inherently noisy process)....so it, as a minimum, needs some sophistication  (maybe filtering, rejecting outliers, etc).

If you go through all that trouble of matching to known derivatives of the wave, you may as well just match to the wave itself as a template & avoid the noise sensitivity.

What interrupts?

You need to be monitoring both inputs "simultaneously"

How about:

1. set up a suitable sampling rate - twice the rate that you'd need for a single input
2. on the "odd" ticks, sample input 1 
3. on the "even" ticks, sample input 2.

From the samples, you can do zero detection and matching against your "nominal" sinewave

(this is basically how old 2-trace analogue scopes used to work in 'alternate' mode)

There were many posts made since your last posting over a month ago (at post 5)  & you have not yet even replied to those...why should anyone bother with you now?

You get interrupt A, you also get interrupt B, but B will be pending.  The moment A is finished, B will be handled.   So B might be delayed a few microseconds,  (waiting for A to finish) which won't matter to you.

At 60 Hz, 5 us  is only 0.108 degrees....big deal.  And you only need to detect whether zero crossings are occuring or not.

in case of AC fault, turn on a led in a time less than 5 ms for each input

Please explain why...is someone going to be staring at this led 24/7, waiting for it to come on?  Will this person have specially trained high-speed eyes? 

Please write your specs in a meaningful way, otherwise a lot of effort can be wasted.

SNAP !

edit

oh - too late.

So it looks like #6, #34, #35 would solve your issue ?

If so, see Tip #5 in my signature (below).