Increasing the isolation in the downstairs 260 MHz monitor ports.


The bandpass change (on/off-1) caused by changing the FP load.
Injecting a -20dbm tone into the FP ports.

     The downstairs if/lo has an 8 way splitter/buffer amplifiers for each of the 8 sub bands that can be generated (4 for polA, 4 for polB). The frequency range is 0 to 500 Mhz and can be sent to  7 back end devices. The eighth channel is a front panel monitor port that is normally used for the spectrum analyzer. It is also used to take data with the radar interface (ri) a/d system. The front panel output has caused ripples in the spectra of the correlator if a cable/device was connected or disconnected during an observation.
    Lisa Wray modified the front panel output on 23apr03 to improve the isolation. The before and after setup is shown below...

The power for a transmitted and then 100% reflected signal was 11db - 17db = 6db.
After the change the power was : -10db+24-3-3-30-10=-32 db. The difference is 26 db (twice the attenuators added since the amps have the same ratio of gain to reverse isolation).
The correlator is connected to the output adjacent to the front panel (FP) output.

On/Off measurement changing the load on the monitor port.

To measure the improvement, data was taken using the correlator and a 25 Mhz bandwidth on buffer amplifiers 1 and 2 using both polarization's. Band1A FP had its amplifier modified. The other 3 bands (band2a, band1b,band2b) used the old setup. The  isolation for  band1A and band2a were measured by placing two  unterminated 5 Mhz BW, 260 Mhz cfr filters on their FP ports.  band1b and band2b FP ports were left open. Data was taken for 300 seconds on the 4 bands (call this the on). The filters were then removed from bands1a,2a and  another 300 seconds of data was taken (call this the off). The average of the 300 second integrations were computed and then on/off-1 was computed.
 The plot shows the change in the bandpass shape (on/off-1) caused by changine the FP load. On/off-1 is plotted (units become fractions of Tsys). The black traces had the inputs to the FP ports changed. The red traces were left open the entire time.     You can easily see the ringing from the filter in the bottom plot (before the change). The maximum ripple is about .04 Tsys. The black top plot with lisa's change still has the ripple but at a reduced level. The maximum amplitude is now .0025Tsys. The maximum ripple has decreased by 12db.

    The decrease is not as large as the change in the power for a signal that traversed the path and then had a100% reflection.

    addendum.. jon hagen pointed out that you have to follow the voltage through the path and then add this to the volaage that got sent directly to the correlator. The power difference along the path was 32 db. The voltage difference would have been half of this in dbs. Call this deltaV. If you add this to the value that went directly to the correlator (using units of Tsys=1) and then square it to get the voltage you get:

powerRipple(units Tsys)= (1+deltaV)^2= 1 + 2*deltaV + deltaV^2 .

Since deltaV is small you can ignore delatV^2.  The ripple size will be 2*deltaV. The ratio of before and after should then be the 2*deltaV1/2*deltaV2  . This is 13 db (or one pass through the attenuation added). The measured value was about 12.

Injecting a -20dbm tone into the front panel port.

    A second test was done by injecting a -20dbm tone into the FP port. The correlator was set to 300 hz resolution. The plots show how far the tone stood up above the noise floor. The noise floor of the 8 bands were probably within a few db of each other.
processing: x101/030423/