[evlatests] 16B DTS problems

[Mike Revnell]

The problem with 16B is RFI combined with the change we put in to guard 
against the digitizer demux coming out of reset in a bad state. The 
problem is actually present in all DTS modules presently in antennas but 
is most manifest in 16B because of how that particular digitizer happens 
to come out of reset.

Improper reset manifests itself samples at maximum (-128 or 127) values. 
Logic in the formatter detects this and resets the digitizer. Most 
digitizers come out of reset very quickly (a few nanoseconds) in the 
correct state. 16B aparrently doesn't. So it's easily visible only in that 
one.

The bad assumption I made is that 8 bits is a lot of dynamic range. It is 
but aparrently not enough. When a strong RFI is present it causes full 
scale samples which causes the logic to reset the digitizer.

We have implemented a change to the reset logic that disables the 
resetting after 10 seconds of good data (no over ranges) after a reset. 
Once the digitizers are up and running properly they seem to stay that 
way.

We will install a module with this change into 16B on Tuesday and watch it 
for a while before changing the rest of them.

Until then when you see 16B acting that way it means some undesired signal 
is causing max scale samples. It could be regarded as an RFI alarm.

On a different level this has implications for level setting policy. There 
will obviously be a trade off between how much spectral damage to live 
with and active bits in correlation.

Mike Revnell
National Radio Astronomy Observatory
Socorro, New Mexico
mrevnell at nrao.edu
revnell at ieee.org