[evlatests] 3-bit Friday

[Vivek Dhawan]

Here is a quick summary, we can discuss the gory entrails (alas)
on Monday.

1. 3-bit data from C-band tests. The best 3-bit modules, like
    the previous versions, are ~10% worse than 8-bit.

Only bits 1-2-3 were used in this test. There are still some
system issues before we test other variations, see below.

RFI on the 3-bit  (not seen in 8-bit because of restricted
bandwidth) was a problem as well.

ea12 3-bit did not have the right flavor of T304 to drive it
properly (right Keith, Michael?) and it performs the worst.
Excluding it and antenna 7 which had other woes, the best
3-bit paths are about 8-10% worse than the 8-bit.

Bandpasses for 6C1 and 12C1 changed by ~5% over 3 days.
The rest were stable to better than 1%. Other than RFI,
things look pretty stable over the 1-hour test.


2. Dropouts and system oddities:

The dropouts that have appeared recently are more pronounced
(~50%) in short-dump data. They are on specific antenna/IF
paths, and seem to be related to delay models because of
the 10sec periodicity, but more is not known.

In the 3-bit data, ea07C1 is very strange with lots of time
variable drops. ea07C2, ea12C1,C2, ea14C1 have 10sec drops.
The rest of the 3-bit paths and all BD 8-bit paths show no
drops. This pattern has repeated now on about 5 datasets over
several days. I do not think it affects all 3-bit data, but
it would be nice to remove this thorn before proceeding.

There were also other periodicities in the fast-dump data,
I hope they are triggered by fast dumps and not a new problem
in regular 1-sec data.

We will get (or already have) new antennas for 3-bit next week.
X-band should be tried I guess.

Michael Rupen wrote:
> Hi Vivek --
> 
>   please have a look at two 3-bit data sets from today:
> 
> * _50msec: 50msec dumps -- what do our drop-outs look like?
> 
> * _offset: this is 3C286 + offset blank field, run for an hour.
>   How does the noise look? and how stable is everything?
> 
> Would also be interesting to compare bandpasses with Thursday's
> data.
> 
> Thanks --
> 
>      Michael
>