Fix Progress

[Tom Droege <tdroege@veriomail.com>]

OK programmers, the following is guaranteed to put you to sleep.  But some 
of you might still be interested in hardware and it's problems.

Besides worrying about data set 18 and taking data, I have also been 
worrying about why multiplexer chips burn up.  I am beginning to understand 
the problem.  It seems there are 406 series multiplexer chips (Typical part 
number is DG406) which you can get, and 506 series chips that you 
cannot.  I just received 25 of the 406 series chips so that I am back in 
business to continue checking out Mark IVs.

This has been a problem from the beginning.  But it has not been at the top 
of the list.  Arne may recall that I had this problem when checking out 
ARNE.  (This is related to the problem of reading an additional temperature 
sensor.) OK, as I understand it, I have connected two of the power supplies 
directly to the input of the mux chip to monitor them.  This is the +5 and 
the -5.  In the event that the -15 supply trips out for some reason, this 
makes the -15 pin zero and the -5 attached to the input pin is now more 
negative than the -15 pin.  This is a no-no for some CMOS devices, and I 
think this is one.  This causes the chip to look like an SCR and it just 
shorts the -5 pin to ground.  Or it tries to, in most cases it burn up a 
bond wire trying.

There are several solutions.

1)  Make sure the +/-15 supplies are always there.  Of course this can't be 
guaranteed.  My recent problem with a camera which shorted the -15 to 
ground is an example.  No doubt a sick capacitor, I just have not taken it 
apart yet for diagnosis.

2) Limit the current on an input pin to a value that cannot trigger the 
built in SCR.  This is apparently what the 506 series does.   I can do this 
by cutting and pasting in a series resistor outside the chip.  I will try 
this on the test stand.  Another cure is to add a diode between the -5 and 
the -15 so that the -15 pin is always at least as negative (within a diode 
drop) of the -5.

3) Buy a chip with built in protection.  I have ordered some 506 series 
chips.  Delivery was quoted as November.

The plan is to do 2) with the chips I have.  Then when the new chips I have 
ordered come in we can change out the mux chips.

Some of you may be familiar with this problem.  I would appreciate advice 
as to what conditions trigger the burn up.

Note that there are too many pins to protect by the cut and paste 
method.  All the DACs are going to the mux, and there are time when the op 
amps that buffer the DACs get quite hot.  Looking at the specs, they can 
put out more current than the MUX can stand.  Sigh!   I use the particular 
amp because I had several tubes of them and they were quite high 
performance.  So the best solution is 3).  But 2) will work most of the 
time, and it seems that only the power supply puts out enough current to 
burn up the mux chip beyond recovery.

I notice in reading the data sheets that the highly sophisticated solution 
for the 506 over the 406 was just to insert a 1000 ohm resistor in series 
with each input.  OK, we have the technology to do a really clever thing, 
put in a series resistor.  ;^)  Of course if you are trying to do high 
speed switching you cannot afford the 1k resistor.

I am actually pretty pleased to find this problem and to begin to 
understand it.  Slowly I am getting rid of the production problems.

Tom Droege