[weblacky]

Well...Well...I may owe Jan-Jaap and others an apology. They may have been right about the DS1780...Maybe. Here is what I know right now.

I was reassembling my Fuel from the bottom panel tear out and mainboard fan header replacement procedure, when I stopped because it was going to be the time to put a PIMM in...I said to myself...well either I do more testing now or I likely don't get to it (now of course I DO NEED to assemble my Fuel as it was to show that it still works after total disassembly! Then I'd "play" with these other parts. But I'd have to place my working PIMM into it...and I need it on my bench for comparison testing (so not in the Fuel, right now). So, I decided to stop assembly and do more work.

I just finished up another testing go-around with my tools. I've come to an interesting finding which MAY suggest that the on-board DS1780 is slightly damaged, but functional. I'm not 100% sure until I take it out and compare it out of circuit with another DS1780 IC to know for sure!

Here's what I did find:

Using my Huntron 3200S (yeah, I'm going to keep plugging it, it's my cheater-tool) I compared points around BOTH my working 700Mhz PIMM and this unon-working 600Mhz PIMM's DS1780. Layout appears identical in all respects I'd care about for voltage. I probed the 5V line and 12V line sensor positions on the DS1780, on both PIMMs.

For those that aren't aware, part of the magic of newer huntron curve tracers (versus the competition) is their ability to go to EXTREMELY low test voltages. One reason you want to go very low is both A. to be gentle on low-voltage circuits and B. to be able to separate "composite signatures" related to semiconductors.

I'll explain B: Curve tracers produce a deflection graph of voltage/current for a set AC test voltage setting, AC Hz setting, and internal (and display) impedance/resistance. However, most semiconductors need a minimum threshold of voltage to actually "turn on" or conduct ANYTHING. Most of the time (not all) that threshold is ~500mV or 0.5V. It's at this level and higher that semiconductors actually turn on (or have completely turned on...remember threshold). Modern Huntron curve tracers can do 200mV, 300mV, 500mv, and then steps in 100mV increments until 1V, then it goes 1v-20V (I have mine limited to 10v... but in reality I never use past 3V!).

When you have the voltage equal to or higher than semiconductor threshold voltages, your scan (graph) is going to include ANY passives (coils, resistors, cap, AND semiconductors in the path of your probes (unless they are diodes, then I'll see that as a change in the graph's deflection).

But what happens if you're scanning in-circuit (power off of course, power is off using this method) and you go from PCB ground to say the leg of a semiconductor. You may end up getting a graph that has characteristics of BOTH a capacitor, a resistor, a diode, and a semiconductor all in one graph because you went through all those things. Well if I want to STOP at the semiconductor (assuming there is another trace to ground that doesn't HAVE TO go through that semiconductor, somewhere around the cap or resistor (to compete the probing circuit), I can turn down the test voltage to BELOW the semiconductor's threshold which will cause it not to conduct the test signal, thereby removing it's effect/influence on the deflection graph by its "removal" from test (while still being in-circuit). You can also do something similar to remove the effect of capacitors by changing the frequency of your test voltage to diminish or erase their response on the graph to reveal other signature attributes (I almost never do this though).

So, using this little trick, I decided to probe both PIMMs, using on-screen (human visual) comparison, at once by overlay compare. Things were looking pretty similar, until I got to the 12V sensor leg (Pin 15) and its neighbor the 5V sensor leg (Pin 16) on the PIMM's DS1780 IC. These appear to be the only only real variations between the two PIMMs (in-circuit of course).

In BOTH instances (but varied values) the working PIMM has greater impedance (resistance) then the non-working PIMM on BOTH these lines!

While a curve tracer is very poor display of exact impedance, the difference was so great that it was a change of like 15+ degrees on the signatures, very telling!!

Now the question is...is that IN the chip, or outside the chip!? After all a problem between the two rails will of course show up on any points along those rails to the DS1780's legs. So, I just used the threshold trick again to attempt to "exclude" the DS1780 from its own measurement.

When I lowered the test voltage to 300mV, I got (roughly, not exact but close enough in real life) the same signatures for PCB ground to leg 15 and PCB ground to leg 16 on BOTH PIMMs (so similar leg to ground connections on BOTH PIMMs). I forgot what I did between the two legs (directly) but I think was something I thought was fine. When I performed the SAME test again with 800mV test voltage I once again observed the noticeable resistance different between the two.

Now if this was some other resistance along the voltage rail, that I was seeing, it SHOULD have shown up on ultra-low testing! A shorted semiconductor (static short) is now a resistor and looks like one too. So, any shorted external component SHOULD have shown itself on BOTH tests (ultra-low and normal tests). It didn't, so I can only conclude that the test is saying the change in impedance is INSIDE THE DS1780 IC. While it's possible it's a damaged but not shorted semiconductor on the same rails as the DS1780 sensor legs, I cannot be 100% until I remove the DS1780 and scan it out of circuit. If the scan still shows as different from another same (new) IC...then it's the chip.

So, I'd say there is an 80% chance the DS1780 suffered small internal damage which changes its internal impedance on the two sensor inputs but did not short it out (thereby altering measurements!). However, I really did find a bad diode in the VRM signaling though, so it's still possible these legs were exposed to a reverse polarity spike (though PCB ground) due to lack of diode clamping?

Anyway, now the plan is to remove the DS1780 from the PIMM and scan it out of circuit, if it tests okay then it is something external on the voltage rails I haven't located yet. If it does test different, then it really is the DS1780 and I'll replace it and we'll see if that solves the two voltage rail issues in ENV (and the cache failure issues I say before any of my repairs).