Probably someone on this list have experience of this chip?

I«ve been struggeling to find bugs in my firmware for a week now, and this weekend found that the bug really seem to be in the #Û*@!%Û£** MAX110!
I should have demonstrated the system for the customer today...

Everything is working just perfect until I restart it with reset.

(Yes, I of course want it ro reboot properly in case of brown out reset at temporary supply problems, or (in case of program failure) a watchdog reset)
This is industrial tool application.


MAX110
This is a "low cost" auto calibrating 14 bit differential two channel A/D.
It communicates trough SPI.

PROBLEM
after reset (to the PIC) I sometimes find MAX110 babbeling on the SPI line without the processor having selected it using Chip Select!
This destroys all SPI comminication. (except MAX110 is useless then)

I believe that it is caused like this:
If resetting in the middle of a SPI transaction, and then when teh PIC is starting new transaction after reset, the MAX110 is misinterpreting the communication as some undocumented test mode command.

I can accept that it enters test mode.  The manual clearly states "dont set theese command bits".  The broblem is i can«t find a easy way to detect nor get out of it before it destroys other communication.
Especially as the MAX110 is also signaling its Busy line, my program nicely leaves it alone (i can change that if necessary, adding a busy timeout).
I would expect MAX110 to eventyually finish the test command, as it is supplied with sampling clock (and also SPI clock through other chip transactions)
But what is really bad, is the MAX keeping transmitting though it«s Chip select is inactive!!

Yes I am scoping the clocks, busy, CS, data in, data out, and having also put 470R resistors on the Data out to see which chip that did talk without permission.

The chip have no reset input, nor reset command, but an idea is to when detecting busy active signal for too long time, I will override the busy signal and push any command (except a "NOP" which is not even transferred for decoding internally) to the MAX110 anyway?

The MAX110 has internal power up reset, but then i have to add switches to +5V, reference voltage, and probably -5V, and at the same time set all SPI low a while.  But the switches will degrade performance. (and increase cost)

But the biggest problem is that at that point it have already destroyed communication between the PIC and other devices, as it is transmitting without active CS signal.  This is the bottom line.

Only solution seem to use software spi on separate pins for MAX110, but I have not very much pins nor CPU power left.  (using the f877)
Maybe a bus switch IC?
Another solution could be to add a 1k series resistor to MAX110 data out, so any other device can override MAX110 output, but i am already using that trick to connect a HC165 which have no output enable...

I have to work on other things today, and my two MAXIM FAE AD experts are both out today too.  

(BTW there is a bug in the datasheet too, minor about number of conversion cycles; page 18&19: "81240" should be "81920", max FAE have confirmed, if you wanna know... maybe other errors too?? - somebody knows?)

Excuse for the long post
Thanks in advance
/Morgan

Morgans Reglerteknik, HŠllekŒs, 277 35 KIVIK, SWEDEN
   tel +46(0)414-446620, fax -70331,   mrt@iname.com