[sdiy] Roland DCOs

[Ingo Debus]

Am 25.02.2008 um 22:34 schrieb Tom Wiltshire:

>>>
>>> What I wanted to know is if anyone has built a DCO along these  
>>> lines, and what they discovered. I've wondered about doing it  
>>> myself since the 82C54 triple counter chips are only a few pounds
>>
>> AARGH! The nastiest chip I ever came across!!
>
> I'd be interested to hear why. It looks like a pretty straight- 
> forward IC from the datasheet.

Didn't think I'd ever touch that datasheet again, but here it is:
http://download.intel.com/design/archives/periphrl/docs/23124406.pdf

Here's the crucial sentence, page 16, note 3:
"Low-going glitches that violate tPWH, tPWL may cause errors  
requiring counter reprogramming."

Yes, that's very true, I found it out the hard way. This chip can  
forget what it was programmed to do. It simply stops working and has  
to be re-programmed. Doesn't happen very often (seldom enough that I  
missed that bug in the first prototypes), but it does happen.

I couldn't ever catch any short glitches in the clock signal (only a  
100MHz scope back then). But note there's also a tWC and tWG  
specified. So not to violate that you'd have to watch the clock or  
gate when accessing the '54 from the microprocessor. That's almost  
impossible to do. I suspected that I didn't violate the tPWH, tPWL  
specs but rather the tWC and tWG which crashed the chip.

Ingo