Maker Pro
Maker Pro

A/D Converter Problem AD7890-4

D

Dave

Jan 1, 1970
0
We're using an Analog Devices AD7890-4 in an existing design that has
been working just fine for several years. This design has been
"transplanted" into a new/different enclosure/chassis and the pilot
build of these new instrument is having problems with the A/D
converter, "very specific problems". My first guess is that the
AD7890-4 is not happy in new environment and is acting up, I'm trying
to figure out what is causing its odd behavior.

ODD BEHAVIOR:
For this analysis I've divided the AD7890-4 into two basic chunks, 1)
the front end consisting of the input circuits (signal scaling and
mux) and 2) the back end that consists of the sample/hold and A/D
functions. When the chip goes goofy the front end (input pin to mux
out) developes a negative DC offset that makes the voltage at mux out
lower that it should be. The DC offset has been measured as small as
10 mV and as large as 1 volt or so. It remains constant from A/D
sample to sample and is not consistant across all eight channels.
Also while it is in this mode the back end (A/D conv) works just fine.
It converts the voltage at SHA IN to the correct digital value over
and over again, but the value is wrong because the front end is
feeding the back end the wrong voltage. The mux out and sha in are
simple connected together is our application. Once the part enters
this goofy mode it stays like this until power is cycled, then it
works correctly once again.

So far the only known way to get the chip into this mode is to run the
production burn in which lasts 24 hours and does numerous operations.
I'm scratching around trying to determine exactly what is going on
when the A/D goes silly. Currently it is not reproducible enough to
pin down the cause.

The power supply (5V) is set to 5.20 volts which is within spec but on
the high side. This elevated 5 volts is necessary to compensate for
voltage drops feeding other circuits in the system. All other
voltages and signals on the other pins are nominal. The device is
operating at 2.5 MHZ. I've also sniffed around the chip using a near
field E-probe looking for the presents of some rouge RF field and
found nothing unexpected.

Since the problems goes away with a power cycle I keep thinking about
some kind of substrate latch SCR condition that resets with a power
cycle.

I have a call into Analog Devices support and hope to hear from them
in the next day or so.

THE QUESTION:
Has anyone experienced anything similiar to this problem with a A/D
chip?

I feel kind of lucky this device has the mux out voltage actually
connected to a pin so I can measure it. Otherwise I probably would be
connecting a logic analyzer to the data in/out pins to see if the part
is telling the F/W bogus information or not.

Any insight welcome...

Thanks,
Dave
 
Top