Increasing the isolation in the downstairs 260 MHz monitor ports.

     On 23apr03 the isolation for the IF2 port1a monitor port was increased. After the increase it was still possible to affect the correlator bandpass by changing the cables on the monitor port. On 04sep03 the isolation was increased for IF2 monitor port 1a. The configuration for the ports were:
 

Port	Configuration	Loss
port 1A before change	10dbPad-24dbamp(30dbRevIsol)-3db pad	19db
port 1A after sep03 change	10dbPad-24dbAmp(30dbRevIsol)-15dbPad-11dbAmp(16dbRevIsol)	36db
port 2A	11dbAmp(16DbRevIsol)	5db

    To test the new isolation two experiments were done.
    The first did a 900 second on/off with the correlator. The correlator was set to 25 Mhz bandwidth (over 1024 channels). After acquisition the data was smoothed by 9 channels to increase the signal to noise. The On was with a cable going from the monitor port to a 5 Mhz filter. The off was with nothing connected to the monitor port. This emulates someone connecting/disconnecting something on the monitor port during a normal observation. This was done to the modified monitor port (1A) and to an unmodified monitor port (2A). Without sufficient isolation, reflections from the 5 Mhz filter would cause ripples in the bandpass.
    The second test injected a sine wave into the monitor port and measured how much of it ended up in the correlator bandpass. This test used a 390 Khz correlator bandwidth with 1024 channels. The data was then hanning smoothed. This resulted in a resolution  of 760 Hz.
    The test results are shown in the plots:
 

• Fig 1 is the filter test. The plots are (900secFilterIn)/(900secFilterOut) - 1. The black lines are polA (which had the filter change).  The red lines are polB (whose monitor inputs were not changed). The top figure is monitor port 1a (after the extra isolation is added). The lower plot is monitor port 2A which has the original setup of just the 11 db amp. After 900 seconds and smoothing by 9, you can not see any ripple in the spectra in sbc1A. The vertical scales of the two plots are not the same. The rms noise for monitor port 1a, 1b match the expected value of .0001 Tsys (  sqrt(2)(..on/off)/(25e6(..bw)/1024.(..chn)*1.2(..chnShape)*9(smooth).*900.(integrate))
• Fig 2 shows the sine wave injected into the monitor port. The top plot is monitor port 1A after the change. There are 4 traces corresponding to a sine wave power of -20dbm,-10dbm,0dbm, and +5 dbm. The sine wave became visible above the noise floor when the sine wave was 0 dbm. The bottom plot is a -20 dbm sine wave injected into monitor port 2a. It sits 14.8 db above the tsys noise floor.

The loss equation is (let T be the noise power in 1 channel,A be the amplitude of sinewave,L be the loss from the components above, X be the loss through the splitters back to the correlator, R be the measured ratio of sinePeak to T)
(A1*L1*X + T)/T=R1=.25db=1.07   Then  L1*X=T*(R1-1)/A1
(A2*L2*X + T)/T=R2=14.8db=30.2  Then L2*X=T*(R2-1)/A2

L2/L1=A1/A2 * (R2-1)/(R1-1)=100*(30.2-1)/(1.07-1)= -46db the expected difference was -55db so we're off by 10db somewhere...

Conclusions:

 home_~phil