Passband Gain Flatness and Calibration
Question from Clint and Bill...
The new system technical requirement for gain flatness is 5.7 dB p-p. EVLA has a similar requirement and Matt Morgan at CDL is planning equalizers for the EVLA receivers in order to comply with the spec. Rick Perley says that phase effects of the equalizers can be removed in
calibration, but will ALMA have sufficient calibrators to do likewise if ALMA were to use equalizers? Also, gain equalizers will flatten the amplitude vs. frequency response
(passband) of the astronomical data, but will add some to the deviation
from linear phase across the passband, which will be largish even without gain equalizers.
Both amplitude and phase responses should be stable relative to the atmosphere.
First, some clarification of the numbers...
Dick's "ALMA System Technical Requirements" (ALMA-80.04.00.00-005-A-SPE) lists the gain flatness requirement as 5.4 dB p-p with 2.7 dB p-p allocated equally each to FE and BE. I will assume that Dick's value of 5.4 dB is the right one to use.
It seems to me that we need answers to the following questions:
- Is the signal-to-noise requirement for a phase calibrator which can be used to calibrate the phase effects induced by these equalizers different from the other calibration requirements for phase calibration?
- How much will the gain equalizer add to the deviation from linear phase across the passband?
- 21 Jul 2004
The precipitating email by BillBrundage
Original Message --------
Subject: [Fwd: EVLA Memo 80 Released] EVLA gain slope correction
Date: Wed, 14 Jul 2004 08:24:25 -0600
From: Bill Brundage <email@example.com>
To: Dick Sramek <firstname.lastname@example.org>, Larry D'Addario <email@example.com>, Clint Janes <firstname.lastname@example.org>, "hrudolf" <"hrudolf"@eso.org>, Bill Holmes <email@example.com>, Alain Baudry <firstname.lastname@example.org>, Charles Cunningham <Charles.Cunningham@nrc.ca>, Al Wootten <email@example.com>, Darrel Emerson <firstname.lastname@example.org>, Brian Glendenning <email@example.com>
Because both FE IPT and BE IPT are doubtful about achieving the ALMA
system technical requirements 271 and 272 for 5.4 dB max pk-pk gain
variation acros any 2 GHz channel (2.7 dB each to FE and BE), we should
join the EVLA effort to achieve gain slope correction. Their proposal
is referenced below.
Note that EVLA has allocated their 5 dB requirement equally among 3
subsystems, whereas ALMA has allocated theirs equally between 2
subsystems. Note also that the BE allocation includes both the IF
processor and the digitizer (DG). Also note that EVLA selection of the
correcting slopes are derived via software from the autocorrelation
Of course, this assumes that passbands can be significantly improved by
slope correction in spite of the presence of ripple.
I suggest that SYSE IPT evaluate this scheme and determine where in the
signal path it (or an ALMA equivalent) would be optimally placed, if used.
The EVLA scheme is described in the following EVLA Memo:
EVLA Memo # 80
A Gain Slope Correction Scheme for the EVLA Receiver System
R. Hayward (AOC), M. Morgan (NTC) and K. Saini (NTC)
National Radio Astronomy Observatory
Abstract : The EVLA Project Book specifies that the entire RF/IF
signal chain must have less than 5 dB of gain slope and ripple across
the 2 GHz input bandwidth provided to the Sampler. The Front-
Ends are allowed slightly less than 2 dB of this gain flatness
specification while the LSC and UX Frequency Converters and the final
Downconverter are each allotted a similar amount. This is a very
challenging requirement. The gain flatness of the first K and Q-Band
receivers to be installed on the EVLA Test Antenna have been analyzed.
It is evident that these wideband receivers will be unable
to meet the 2 dB gain slope allotment, let alone the entire 5 dB spec
at the band edges, without gain slope correction. A fixed gain slope
equalizer will not be adequate since the receiver gain slope
changes in both magnitude and direction depending on which part of the
RF band has been selected.
Also, the gain response will vary depending on which EVLA frequency
band is being used and each of the 27 individual receivers in a
particular band on the Array will be different. Thus a dynamic
equalizer will be required. One possible solution to the problem is
presented, based on a custom programmable equalizer which provides ±15
dB gain slope corrections in fixed 2 dB increments. The preferred
location for this new equalizer, consisting of surface mount
components on a microstrip circuit board, would be inside the T304
Downconverter module. The autocorrelation spectrum provided by the
WIDAR Back-End would be used to determine the required gain slope
View a pdf version of EVLA Memo #80 at:
- 25 Jul 2004
--Comment by BillBrundage
- 23 July 2004
Refer also to ALMA Memos 452 and 407 on effects of passband shape.
Stepped passband slope equalizers would only be implemented in each 2-4 GHz baseband channel because, among other reasons, the downconverter frequency mixers create different slopes among channels and tunings. BE IPT has no baseband equalizers at this time to test for phase variation across the passband. However, tests of 4-12 GHz equalizers of 3 dB and 6 dB slopes across that passband have parabolic phase vs. frequency, with 6 to 10 degree pk=pk variation. Maxima are near band center and minima at band edges. The tested equalizer to equalizer variation indicates that variation among stepped (switched) equalizers over a range of slopes could be a few (~< 4?) degrees.
-- Comment by BillBrundage
- 23 August 2004
EVLA Memo 80 (see hyperlink above) discusses phase response of the switched (programmable) equalizers for EVLA, ALMA and GBT on pages 12 and 19. To move this question to a resolution for ALMA, I will have to submit a CRE (soon).