Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uwm.edu!ogicse!zephyr.ens.tek.com!gvgpsa!gold!grege
From: grege@gold.GVG.TEK.COM (Gregory Ebert)
Newsgroups: sci.electronics
Subject: Re: Z-80 NMI
Message-ID: <850@gold.GVG.TEK.COM>
Date: 13 Mar 90 20:06:35 GMT
References: <5539@bgsuvax.UUCP>
Distribution: usa
Organization: Grass Valley Group, Grass Valley, CA
Lines: 22

In article <5539@bgsuvax.UUCP> nandan@bgsuvax.UUCP (rajesh Nandan) writes:
>Just a baffling situation-help me solve the problem
>
>If the Z-80 is executing "executable" codes the NMI works & forces a jump to
>0066H but when I make it "HANG" by jumping to a "nonexecutable" part of code 
>the NMI DOESN'T work even though the RTC(58167) chip connected pulls the pin 
>low. ie:- the NMI doesn't work if it has hung. 

	The Z-80 might be executing a block instruction with a huge block
	count (ie LDIR w/ BC=0000). NMI will not break into an instruction.
	Maybe your SP got trashed - NMI routine executed OK, but it can't
	return to the right place; this might 'look' like NMI isn't working.
	The last thing to check is the NMI pulse width. NMI is falling-edge
	triggered, but I'm sure that the Z-80 needs to sample NMI low at the
	end of the current instruction; 'glitching' NMI won't work. If NMI
	is stuck-low, the routine at 0066h will get executed. Even if you
	do a RETN, the next NMI won't occur until NMI has gone high. I know
	from experience that NMI works perfectly when all of the laws
	are obeyed.

	For what it's worth, I've been using the Z-80 for 10+ years and I
	always have/always will love this critter.