[evlatests] T304 Power linearity tests

[Rick Perley]

    This is a report on the results of a test, run yesterday evening, 
which sought to find if, and how, non-linear responses can be generated 
within our electronics.  The report given yesterday was based on varying 
the output power of the T304, and examining the response of the digital 
system.  The current test modified the input power to the T304, while 
keeping fixed (where possible) the output power. 

    For the purposes of this test, the T304 module consists of three 
components:  An input attenuator, a bunch of analog electronics with 
gains and losses, and an output attenuator.  Both attenuators have a 
full range of 31 dB, accessible in 1 dB steps.   The purpose of the 
input attenuator is to prevent signal saturation within the T304, and to 
prevent T304 noise power from degrading sensitivity.  The purpose of the 
output attenuator is to provide the optimum signal level to the samplers. 

    The observations were made of the strong calibrator 3C147 at the 
default OSRO C-band frequencies, 4872 and 5000 MHz, with 128 MHz BW and 
64 channels. 

    The input attenuators for all A and C T304s were stepped in 2dB 
increments, starting at their maximum attenuation of 31 dB, and ending 
at 1 dB.  For each input attenuator level, the output attenuators then 
adjusted their value so as to place the output power at the desired 
level for the samplers.  Depending on the actual input power level and 
(presumably) the gain within the T304, it was not always possible to 
place the output voltage at the desired level.
    The B and D IFs were left in the normal (optimum?) mode, with no 
changes. 

    From the data, we can monitor the switched power sum and difference, 
their ratio (= 'Tsys'), and the fringe power amplitude, both with and 
without the switched power corrections. 

    Some Results. 

    Perhaps the most surprising result is the wide range of behavior 
seen.  I had expected some similarities amongst all the antennas...

    1) Output Power. 

    This is easily monitored through the switched power monitors.  There 
is a wide range of behavior, which I'll illustrate by noting the extrema:

    A)  For antenna 4C (LCP), there is *no* change in output power over 
the entire range of the input attenuators.  Apparently, the input power 
level and intervening gain are such that the power supplied to the 
output attenuator is 'just right' so as to permit an output attenuator 
setting to match every one of the input attenuators.

    B)  For antenna 24A (RCP), a very different scenario is seen.  The 
input attenuation changes were perfectly seen in the output power as 2 
dB steps of 2 dB until the input attenuators were decreased to the 9 dB 
level.  Only after that did the output power level stabilize (i.e., was 
at a point within the range of the output attenuators).  Hence, either 
the input power level to the T304A module is extremely low, OR, the 
intervening gain within the module is extremely low. 

    All other antennas have behavior between these.  The median input 
attenuation for which the output attenuators start to come into play is 
about 20 dB, but with a very wide spread -- nearly a uniform 
distribution between 31 dB (for antenna 4C) and 9 dB (for 24A). 

    2) Measured system temperature. 

    We would expect that linear performance will provide a uniform 
measurement of Tsys over some reasonable range of input attenuation.  
This is met for nearly all antennas -- in fact, the input attenuation 
range over which there is little change in measured Tsys is typically 20 
dB!  This is good.  Only a very few antennas show significant 
degradation at one end of the other (never both), and they all 
correspond to extrema in the input power (or T304 gain). 

    A) For antennas like 4C, for which the input power is very high (or 
the T304 gain is high), Tsys degrades notably when the input attenuation 
is low.  For 4C, Tsys rises from its 'good' value of 35K (from 31 to 16 
dB input attenuation) to about 47K as the attenuator decreases from 16 
to 0 dB. 

    B) For antennas like 24A, where the input power appears to be 
extremely low (or the T304 gain low), the opposite is seen:  Tsys is 60K 
at maximum (31 dB) attenuation, dropping to its stable value of 35K 
after the attenuation is decreased to about 12 dB. 

    All other antennas are in between these.  But most are remarkably 
stable in Tsys, with nearly all demonstrating less than 1K change over a 
20dB range in input attenuation. 

    3) Fringe Power. 

    In an ideal system, all necessary changes in attenuation in the T304 
would be invisible to the user -- the corresponding changes in fringe 
power would be offset by the switched power monitor, and the sensitivity 
would be the same. 
  
    We're not far from that ideal, in fact.  I again start with the extrema:

    A)  For antenna 4C, (high input power), the corrected visibilities 
becomes significantly wrong (are too high) when the input attenuation is 
down to about 15 dB.  Presumably, we are seeing compression within the 
module electronics, which reduces the PDif values below their correct 
values. 

    B) For antenna 24A, (low input power), the inverse is seen:  the 
corrected visibilities are too small when the input power is attenuated 
to a very low value. 

    The good news is that nearly all antennas don't look like either of 
these two extrema, and show a very wide range -- exceeding 20 dB in 
input power range, over which little change in visibility amplitude is 
seen. 


    4)  Based on these results, antennas can be categorized into three 
groups, based on the Tsys performance:

    A) High Input Power (like 4C):  These include:  1A, 3C, 4A, 4C, 15A, 
15C, 25A, 25C.  Only 4C is extreme. 

    B) Low Input Power (Like 24A):  These include:  10A, 11A, 12A, 12C, 
13A, 13C, 14A, 14C, 16A, 17A, 17C, 19A, 19C, 21A, 24A, 24C, 26A, 26C.  
Most of these also show variation in visibility amplitudes. 

    C) The rest show essentially no variation at all accross the full 30 
dB range of this test. 

    5) Relation with the PDif compression. 

    In short, there is none!  The list of antennas which show 
significant PDif compression when we observe a source of different power 
(like Cyg A) is not related to the antennas listed above. 
    Furthermore, the range of power offered by Cygnus A at this band is 
only 3dB -- this is a far, far smaller increment than implemented in 
this test.  If the T304s were non-linear over a 3 dB range due to Cygnus 
A, the effect on switched power in the test described above would have 
been quite obvious. 

    So what is the difference in these tests? 

    The T304 and digitizer tests that we have run in the past two days 
affect only the downstream electronics -- no change was made to the 
input signal.  Based on what is reported above, the origin of the 
compression problem cannot lie within or after the T304 module.  It's 
clean. 
    The PDif compression problem apparently is induced by a change in 
input power, or at least a change in power prior to the T304. 

    I have some data which might shed light on this -- the solar data, 
where we inserted a 20 dB attenuation within the T302 module while on a 
strong calibrator.