Breakthrough #1

[Tom Droege <tdroege@veriomail.com>]

I think I have finally tracked the bad data problem to Windows.

This evening I was setting up TOM2.  In a hurry to take measurements, I 
interrupted the program while Downfits was writing data to the disk.  This 
gave me a windows error.  Program fouled up, must terminate program.

After this action, the system wrote bad data.

I then restarted Windows without turning off power.  The system recovered.

It is my theory that Windows can have an odd byte sitting in an I/O 
channel.  When you say from Downfits, give me a byte it first gives you the 
odd byte, then gets the next one from the I/O.  The result is that the data 
is one byte off, and with this data, the high and low bytes are 
interchanged.  Surely something to give one bad data.

Note that the memory board has a four byte conversion.  It reads four bytes 
in parallel from the memory and then outputs them one at a time to the I/O 
bus when asked by an I/O operation.  I suspected that this was not getting 
reset properly.

Note that once this problem starts, it can persist for days if power is not 
turned off.

Before I was not sure if it was something that the Memory board was doing, 
or that it was Windows.
Now I am pretty sure it is windows.

Does someone know how to write a "FlushTheI/O.exe" program that I can run 
that would confirm this?  Next time I get bad data I could run it and see 
if the bad data went away.

Tom Droege