Marcel duchamp <marcel.duchamp <at> sbcglobal.net> writes:
 
> William Chops Westfield wrote:
> > On Mar 6, 2007, at 1:38 AM, Alan B. Pearce wrote:
> > 
> >> So the data sheet is correct, and you are trying to
> >> operate the chip outside the limits.
> > 
> > I dunno.  If the chip behavior has changed and the datasheet
> > HASN'T changed, that seems like a microchip-side problem to me.
> > The inability to adequately clamp inputs with ordinary silicon
> > diodes seems like a pretty serious issue. 
> 
> The PIC chip itself is made from, as you call it, "ordinary silicon" 
> Bill.  Adding an external "ordinary silicon" clamp diode simply puts two 
>   "ordinary silicon" diodes in parallel.
> 
> Which one will turn on first?

The one that is more strongly doped. In particular, the 'reverse diodes' in
MOSFETs such as those in PICs are poor diodes as diodes go and the the external
diodes should turn on first. But with die shrinks doping can become violently
high and exceed that of external diodes. I tried to compare 16F54 and 16C54 bulk
diodes using a DVM on diode scale. Results are not very conclusive but the 16F84
seems to be different enough. Testing was with the DVM + at pin 5 (Vdd) and DVM
- at pin 6 (PORTB.0 - the pin that first gave me trouble). I also compared this
with 1N4148. I can't say that this is conclusive.

> Flip a coin. In fact, since the PIC is powered up and dissipating energy 
> as heat, it will be (likely) warmer and have a lower Vf than the 
> external one so the external one will tend not to help.
> 
> Schottky clamps are "de rigeur" for clamping SILICON microcontrollers.

I strongly disagree. All commercial tranzorbs clamp at -0.6V (silicon Vf). 
There are specialty tranzorbs that clamp lower. I accept that the voltage may
not go below -0.6V iff that can be ensured with an ordinary clamp. I know how 
to make high speed accurate clamps that clamp to 50mV of the rails, but I don't 
want to use such things in a project that uses a $0.5 micro. Even the two 
external Schottky diode clamps I have to add for this project increase the 
price by 3 cents, lus labor, plus board, plus plus.

And yes, I tested with 'virgin' chips and with chips that have seen the current 
drain. The problem is the same. I did this fairly methodically, first w/o 
diodes, then with 1N4148, then with 1N6263 and then BAT42. A 16C54(C?)JW works 
even without diodes, ditto 16F84. 16F54 does not work w/o diodes, does not work 
with 1N4147 (Vf is about 0.6V), and works with any Schottky.

One more thing: The OSCOUT amplitude on these chips is low, about 2Vpp, and
rides on GND. This is unusual. I use XT mode with ZTT series 3.58 MHz crystal
filters. With a different chip (16C54, 16F84) the OSC amplitude is healthy and
over 4Vpp. All this gives me a bad feeling about long term reliability. The
programmer seems to like the chips and they verify at high and low voltage but I
must say that they behave differently from most PICs I've used over the past ~13
years.

Peter P.


-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist