INTRODUCTION:

The purpose of these tests was to get a decent estimate of the Noise CAL
power that would need to be inserted at the S- & X-band coupler ports of
the duan-band feed originally (and still currently) installed on the 12m
telescope.  The couplers' coupling coefficients are unknown, as are the
Tsys values at this time.  Also, there is some doubt in my mind as to
whether or not the couplers are installed in the correct orientation.


DESCRIPTION OF TESTS:

Cables (Belden 1673A) were run from one of the two CAL coupler ports on
each polarization for each band, down to the pedestal.  The couplers in
question are each positioned in WG immediately prior to the LNAs in the
signal paths.  The coupled ports, however, are SMA.

The losses in the cables are considerable, especially at X-band, and
were checked for agreement with datasheet values at X-band.  For this
study, interpolated datasheet values will be used since they are precise
enough.

The general idea is to see what S/N results from injecting a moderately
weak CW signal of known level into the coupled port on each band, with
the noise being measured in a specific bandwidth.  The results should
then permit estimation of the noise CAL PSD that should be required at
the coupled ports on each band in order to boost the system floor by a
specified fraction of the base level, without knowledge (yet) of either
the coupling coefficients or the actual Tsys. 

The actual resolution bandwidth and power levels at the injection ports
were chosen to facilitate accurate measurement of the signal peaks'
levels without undue averaging time requirements.  The choices were
unexpectedly constrained, however, in that an excessively high power
level caused the noise floor to increase, presumably due to a rather
high phase noise level somewhere in the system, probably in one or
both of the LOs in the UDC.  High phase noise levels over a very wide
bandwidth were noted in earlier diagnostic work, but the details have
not yet been determined.  For the reported results of the current
tests, the signal level was held low enough that the increase in noise
floor was minimal, a fraction of a dB.

The CW test signals began at an HP83712 generator in the 12m pedestal,
then through a short cable to an HP manual step attenuator, whose
ouput then drove the bottom end the Belden 1673A cables leading up to
the selected CAL coupler.  Prior to moving the sig gen and attenuator
to the pedestal, the output setting of the generator required to yield
0 dBm at the attenuator output (with the attenuator set to zero dB)
was determined with the SA in the lab, at the rough center of each
frequency band.  The required generator settings were recorded and
carefully adhered to during the actual tests. 

Measurements were performed at the 512 MHz - 1024 MHz VLBI outputs of
the UDC with an E4445A spectrum analyzer.  The RESBW settled upon was
1 kHz, and the detector mode was RMS Average.  Signal peaks were read
with the normal marker function, and the noise level was measured off
to one side with the "noise marker" function enabled, with the noise
levels recorded in units of dBm/Hz.


X-BAND:

Cable length:  80 feet, loss taken to be 36.0 dB in 80 ft at 8.60 GHz

Sig Gen level setting = +4.16 dBm to yield 0 dBm at step atten output
when atten setting is 0 dB (established in lab prior to experiment).

Sig gen output level setting used in pedestal = +4.16 dBm.

Power level at coupled port of CAL coupler is (-36.0dBm - atten setting)

Atten setting used was 65 dB; therefore:

Power level at coupled port of CAL coupler was  -101.0 dBm

MEASURED levels at UDC output follow:
Note: Levels reported as (pol A) were measured at UDC output pol A;
      likewise for pol B, with no change to input configuration
      between the two measurements (only the SA connection was moved).

8.1 GHz:
Signal peak = -38.0 dBm (pol A);  -68.0 dBm (pol B)
Noise floor = -89.5 dBm/Hz (pol A);  -85.1 dBm/Hz (pol B)
1-Hz S/N = 51.5 dB (pol A)
Coupled-port-referred PSD = -152.5 dBm/Hz (A): for Tcal = Tsys

8.6 GHz:
Signal peak = -39.5 dBm (pol A);  -?? dBm (pol B)
Noise floor = -89.9 dBm/Hz (pol A);  -84.5 dBm/Hz (pol B)
1-Hz S/N = 50.4 dB (pol A)
Coupled-port-referred PSD = -151.4 dBm/Hz (A): for Tcal = Tsys

9.1 GHz:
Signal peak = -42.8 dBm (pol A);  -78.9 dBm (pol B)
Noise floor = -91.4 dBm/Hz (pol A);  -85.3 dBm/Hz (pol B)
1-Hz S/N = 48.6 dB (pol A)
Coupled-port-referred PSD = -149.6 dBm/Hz (A): for Tcal = Tsys



S-BAND:

Cable length:  80 feet, loss taken to be 16.6 dB in 80 ft at 2.27 GHz
  
Sig Gen level setting = +1.68 dBm to yield 0 dBm at step atten output
when atten setting is 0 dB (established in lab prior to experiment).

Sig gen output level setting used in pedestal = +1.68 dBm.

Power level at coupled port of CAL coupler is (-16.6dBm - atten setting)

Atten setting used was 80 dB; therefore:

Power level at coupled port of CAL coupler was  -96.6 dBm

MEASURED levels at UDC output follow: 
Note: Levels reported as (pol A) were measured at UDC output pol A;
      likewise for pol B, with no change to input configuration
      between the two measurements (only the SA connection was moved).

2.19 GHz:  
Signal peak = -41.3 dBm (pol A);  -66.1 dBm (pol B)
Noise floor = -94.1 dBm/Hz (pol A);  -90.4 dBm/Hz (pol B)
1-Hz S/N = 52.8 dB (pol A)
Coupled-port-referred PSD = -149.4 dBm/Hz (A): for Tcal = Tsys

2.27 GHz:
Signal peak = -41.3 dBm (pol A);  -67.8 dBm (pol B)
Noise floor = -94.7 dBm/Hz (pol A);  -88.7 dBm/Hz (pol B)
1-Hz S/N = 53.4 dB (pol A)
Coupled-port-referred PSD = -150.0 dBm/Hz (A): for Tcal = Tsys

2.364 GHz:
Signal peak = -45.5 dBm (pol A);  -78.5 dBm (pol B)
Noise floor = -97.2 dBm/Hz (pol A);  -93.5 dBm/Hz (pol B)
1-Hz S/N = 51.7 dB (pol A)
Coupled-port-referred PSD = -148.3 dBm/Hz (A): for Tcal = Tsys


DISCUSSION & CONCLUSIONS:

1. As Phil has noted, pol B noise floors are about 5-6 dB higher
   than those on pol A, for both frequency bands.  This is a
   little bit "off", and should be investigated.  Perhaps there's
   another loose/bad connector in the system, or a fiber that did
   not seat correctly.  This cannot be explained by tolerances in
   the two links unless one of them has changed since the two
   were tested at incoming inspection.  It might also be explained
   by something as mundane as tolerance stack up between the two
   pols, either in the UDCs or between the duplexer units in the
   pedestal, or (heaven forbid) by a miscalculation on my part
   somewhere.
   
2. Eureka!  The CAL couples were plainly installed backwards, and
   upon correcting this flaw the measurements reported above make
   pretty good sense.
   
3. CAL levels of ~10% of Tsys should be marginally achievable with
   a standard 15dB ENR noise source.  A little more would perhaps
   be preferred, however, by some users.
   
4. We also made measurements of the CW level required to cause a
   small rise in noise floor due to presumed phase noise in the
   overall system.  These were observed in about 50 kHz SPAN, and
   the rise was fairly flat across this span.
   
   X-band:  Signal of -81.0 dBm at 8.6 GHz raised floor by 2 dB.
   
   S-band:  Signal of -81.6 dBm at 2.19 GHz raised floor by 1 dB.