开云体育

I'm pretty sure I understood what Raymond was saying.
Raymond always writes very clearly:

1) Scope plugged in, but turned *off*.

2) A5 board installed in scope.

3) Desoldering, and resoldering NVRAM.

4) Worrying about soldering iron tip grounding /WE, /CE,
or /OE, and activating the NVRAM to overwrite its contents.

.: Presents scheme to connect soldering iron's tip grounding
to Vcc, to prevent /WE, /CE, and /OE from being driven low
when desoldering/soldering them...

(Note: Since any reasonable soldering/desoldering station
has its tip hard wired to earth ground, bringing its tip's
ground to VCC is the same as grounding VCC)

What I am saying is, if the Dallas NVRAM's controller cannot
handle the grounding of /WE, /CE, or /OE, without writing to
the contents of the NVRAM, it cannot handle being powered off.
That would be a major fail for Dallas. A fail that I have never
witnessed. I was one of Dallas's early adopters, having used
some of their preproduction NVRAMs to fix my own EEPROM design
fail.

I believe the problem is that Raymond thinks the NVRAM is passing
the /WE, /CE, and /OE signals straight through to the internal
CMOS RAM, and the lithium cell is simply keeping VCC up to save
the RAM contents. I believe his thinking is that grounding out
the control signals may activate the CMOS RAM to write over some
random cell's contents.

The problem with that belief is the Dallas NVRAM's whole purpose
in being is to prevent that scenario.

One of the capabilities of the NVRAM that Dallas featured was
the designer's ability to preprogram the NVRAM, and then wave
solder the programmed part into the circuit board. Something
that I caused done thousands of times.

I first came to use Dallas NVRAM's as a young engineer, when I
believed that EEPROMS, such as the X2816 worked in the way their
data sheets implied. The early data sheets didn't emphasize that
/WE, /CE, and /OE remained functional until VCC was down to, I
think, 1.5V... This created a situation where the address, data,
and control lines for the bus all were in a race to the bottom
as the power supplies came crashing down during power off. If /WE
hit active LOW before VCC hit 1.5V, the EEPROM could get written to...
just what Raymond fears might happen with the NVRAM.

I feel badly about making that mistake, but I take some slight
solace in knowing that I wasn't the only one. I have seen X2816's
misused in dozens of products by dozens of different manufacturers.

Analyzing some of those products has shown creative software fixes
for the random data lost due to power down. Most seemed to settle
on writing the data in multiple places and voting on the rightness
of the data... before cleaning up the record.

Meanwhile, Dallas exploited that failure to take the market from
the X2816.

-Chuck Harris

Steve Hendrix wrote: