Thanks again to Jan for the correction on the operation of the PIC PORTs.  I
made this change (to tri-state the PIC pins instead of forcing them low) and
still the original problem exists.  That is - if I sample every 1s (2s is
borderline - sometimes works, sometimes doesn't (right now I tried and it
worked at least for 15 samples)) I miss every other sample because the
reference voltage doesn't come on properly.  I have taken thousands (I think
it was just over 9k) of consecutive samples with a 5s interval with no
errors.

I noticed when looking at the PIC pin that powers the reference IC with the
sampling interval set to 1s that there are two alternating pulse shapes.
Both quickly rise to 5V (looking at 50ms/div) and hold for about 25ms but
then one will decay gradually from the full voltage with a time constant of
~50-75ms while the next will quickly decay to about half voltage and then
show an exponential decay from that point.

When I set the sample interval to 5s (and everything works properly) I see
the second waveform on the PIC pin that powers the reference (drops to half
voltage then exponential decay from there).

So it seems that the variance in waveforms is just a symptom of the fact
that the reference isn't delivering the proper voltage and thereby is
drawing less current and discharges the cap more slowly.

But why would this happen?  Even with the sample interval set to 1s and
tri-stating the PIC pin that powers the reference rather than forcing it
low, the power to the reference goes to 0 for at least a couple of hundred
ms between samples.  I have checked the LT1790 datasheet (
http://www.linear.com/pdf/1790fa.pdf ) a few times and don't see anything
regarding this issue.

If no one has any ideas by the time I finish dinner I guess I'll try
replacing the reference IC.  This is my first SMD project and I have fixed a
few wacky problems like this by replacing the affected IC.  I am blaming it
on overheating the chips when I solder as the number of problems I am having
seems to diminish as I use more and more caution and develop better
soldering technique.

I would love to hear any other ideas though as I have already replaced the
reference IC a few days ago for a different problem (which was fixed by the
replacement) and I am reasonably confident that I did a good and gentle job
soldering this last time.

The intended application of this unit will likely use sample intervals in
the range of tens of minutes so I could potentially just make 5s the minimum
in software, but I'd rather get this figured out in case it is symptomatic
of a larger problem.

Thanks,
Nick



----- Original Message -----
From: "Spehro Pefhany" <speff@INTERLOG.COM>
To: <PICLIST@MITVMA.MIT.EDU>
Sent: Friday, January 09, 2004 8:50 PM
Subject: Re: [PICLIST] [PIC:] LT1790 Voltage Reference problems


> You can set the port pin low or high with movlw/movwf,
> which doesn't read the pin state, then write to the TRIS
> register.
>
> The main condition I think you want to avoid is actively
> driving it low, after it's been high, because that discharges
> the output capacitor through a sneak path in the LT1790 (diode
> and forward-biased collector-base junction). Just
> a guess as to what's happening.
>
> Bset regards,
>
> Spehro Pefhany --"it's the network..."            "The Journey is the
reward"

--
http://www.piclist.com hint: To leave the PICList
mailto:piclist-unsubscribe-request@mitvma.mit.edu