[evlatests] Low-frequency rolloff absent at L-band
Rick Perley
rperley at nrao.edu
Sun Aug 25 13:22:05 EDT 2013
Two EVLA memos have discussed the issue of a 'low-frequency bandpass
rolloff' in the 8-bit path. (The subject is in quotes, because it's
seen at the high frequency side for X-band observations). In the first
memo (#152), Emmanuel and I showed that L-band showed a steep loss in
sensitivity in the lower ~150 MHz, and attributed this to the receiver
response. In the second (#154), Keith and Emmanuel showed that a
rolloff is seen at all bands, and discussed its origin in detail.
In the most recent VLA Tests Results meeting, I showed new bandpass
plots, made of recent data at L-band, which clearly showed a near
complete absence of any low-frequency rolloff at L-band. There was some
discussion on how such a dramatic change could have happened by any new
tuning setup -- especially since no change in LO setups have been known
to have taken place since late in 2010.
Since the meeting, I have reviewed the bandpasses of all VLA
antennas from data taken in March 2011 and July 2013, at L, S, C, and X
bands. The two observations taken at each band utilized identical
tuning setups.
The results of this are quite clear. I summarize the results
immediately below. For those who want to know more, the details from
L-band are given further down.
Result 1) There is no essential change in bandpass shapes between
the two dates at S, C, and X bands. The 'low-edge' rolloff addressed in
Memo 154 are present, in both IF pairs, in both epochs, to the same
level.
Result 2) There is a dramatic change in bandpass shape at L-band,
in *nearly* all antennas. (The asterisks are important -- read on if
you want to know the details!). The 'rolloff' noted in memo 152 is
essentially absent on all but three or four antennas now, whereas it was
present in all but five antennas in March 2011.
An important addition is that the 'rolloff', evident in 2011, was
only present on the A and C IFs, and absent on the B and D IFs. This
latter pair was tuned to the upper half of L-band, the A and C pairs
were tuned to the lower half of the band. (This is a clue!)
Conclusion: By comparing the presence, or absence, of the low-end
L-band rolloff to the dates on which the modern, wide-band L-band
receivers were installed, it is proven beyond any doubt whatever, the
the dramatic improvement in L-band bandpass function is entirely due to
the implementation of the new receivers.
What I don't understand is why the 'true rolloff', clearly visible
at S, C, and X bands, is entirely absent at L-band.
-----------------------------------------------------------------------------------
Details, for those so inclined.
The 'true rolloff' -- easily seen in S, C and X bands, and addressed
in Memo #154, reduces the bandpass response by about 8 to 10 dB over a
frequency span of ~ 150 MHz. I make this suppression measure from
where the anti-aliasing filter cuts off (about 20 MHz from the lower (or
upper, for X-band) frequency edge) to the middle of the third subband
(sixth, for X-band).
I provide below a table, showing this 'suppression response' for
L-band. Given is the reduction, in dB, between the bandpass function at
~1000 MHz to that at 1150 MHz. This value is provided for both
polarizations for the two dates: March 2011 and July 2013. The
bandpasses were derived from observations of 3C147. These bandpasses
were 'absolute' (i.e., not normalized), and adjustments for the known
spectral index of 3C147 were made.
March 2011 Comments July 2013
Comments
------------------------------------------------------------------------------------------------
Ant. R L
R L
01 17 19
5 5
02 19 20
9 5
03 16 16
3 5
04 17 18
8 4
05 18 19
8 4
06 8 8 New style 8 11
07 15 9
13 8
08 18 15
5 6
09 12 7
10 8
10 8 13 New style 8 8
11 15 15
15 16 Old style
12 8 8 New style 8 10
13 out out
4 4
14 13 10
12 13 Old style
15 22 22
7 8
16 out out
7 4
17 out out
17 16 Old style
18 14 18
4 4
19 15 15
7 8
20 6 10 New style 7 6
21 16 18
out out
22 11 8
10 8
23 13 14
7 4
24 8 5
4 7
25 16 17
4 6
26 8 9 New style 6 6
27 16 15
4 5
28 18 17
6 5
------------------------------------------------------------------------------------------------------------------
Comments:
1) The antennas listed as 'New style' in 2011 were the first 5 with the
modern full-band receivers. All have minimal bandpass suppression. My
notes don't show ea24 as being 'new style', but it must surely have
been, judging by its response, and by our clear intention that ea24 be
the first to get updated electronics.
2) The antennas lsited as 'Old Style' in 2013 are the three remaining
antennas which do not now have modern wide-band receivers. Two have
bandpass suppression unchanged since 2011 (and for the third there were
no data taken in 2011).
3) Nearly all antennas now equipped with modern L-band receivers show
minimal bandpass suppression -- about half of 6 dB or less. For the
B,D pair (tuned to a center frequency of 1756 MHz, so the low frequency
edge is at 1500 MHz), the suppression is 3 dB or less for *all*
antennas. From this it is clear that the real rolloff mechanism
discussed in EVLA Memo 154 does not operate at L-band. The L-band
rolloff seen now -- typically 4 to 8 dB -- must come from the
receiver. In support of this statement is the observation that the
high-frequency side of the bandpasses shows about the same rolloff as
the low-frequency side.
So to my mind, the only mystery left is why the Memo154 rolloff
mechanism does not show up at L-band.
More information about the evlatests
mailing list