The xband receiver (7.8 to 10.2) Ghz was
installed on 19dec01 in the gregorian dome. It is a native circular
receiver with setup as: horn, omt, dewar (isolator,
calcoupler,amplifier), then postAmps. It is inserted into port 11 of
the upstairs if/lo. The IF processing uses the 2 to 12 Ghz mixer to
mix down to the 1 to 2 Ghz 1stIF (prior to 3aug03 the receiver
used the 10 Ghz upconverter for the 1st IF mixing). This gives an
instantaneous bandwidth of 1 Ghz. When using the 2 to 12 Ghz mixer
important to use a low side lo below say 8500 Mhz.
04nov16: polA (rcp) (left hand rack downstairs), polB (lcp)
(right hand racks downstairs). .. measured using jpl xband
Installation (prior to 15feb02)
- 04dec12: polB Tsys jumps up be about 4k. Stayed high until
07dec12. denis found a loose rf cable from postamps.
- 04dec12 was the same day that lbw got brought down.. so the
cable may have gotten twisted.
- 16aug12: installed new cal
values that were measred ju12. (backdated to 01may12).
- 04jan10: cooling now on one of the alfa compressors (with
- febmar04: receiver down. new cals installed, some cables
replaced in dewar.
- 08jan04: xband receiver moved to new position on the turret
- 05oct03: xbrcvr to antenna test range, measure cals, rcvtemp
- 03sep03: new xband horn installed, cal cables cleaned
- 03aug03: switched to use the 2 to 12 Ghz mixer rather than the
10 Ghz upconverter for the default first IF
- reshimmed horn to move it up .44 inches vertically, moved
turret position .35 degrees (from survey).
- switched to pointing model 13. offsets still large.
- 27dec01: heater installed on wave guide for condensation.
- 19dec01: receiver installed.
installed new cal values that were measred ju12. (backdated to
12jan06: check xband receiver after
13jun04: xband performance above 10
gain curve using data 05oct03 thru 25jan04.(GAIN
05sep03: New horn
performance: SEFD new horn,old horn.
(SEFDB/SEFDA) and (TsysB/TSYSA)
05sep03: Tsys vs
frequency old and new horns.
gain variations a1704 recomb lines
gain curves mar03 thru sept03 (GAIN CURVE)
21feb02:calibration runs on
scans) J0237+288 after pointing offsets.
09apr14: xmit at 8351 Mhz tone from the
polA jumps went away when bias box from cbh was used.
instabilities seen in pola and polb
installed new cal values that were measred ju12. (backdated to
position switching integrating up to 140 minutes. Showing bandpass
and total power stability.
08jan04: xb rcvr moved to new
position on turret floor.
in the xband receiver.
az,za pointing offsets using calibration data.
gain is modulated at 1.2 hz by the crosshead.
multiple drift scans. Tsys vs Time, DTsys/Tsys after integration.
21feb02:HC3N lines at
9097,9098,9100 possible detection??
- update az,za offset model 13.
az:+J0237+28827.4, za:+5.4 (from 13feb02 data).
13feb02: beammaps (turret scans)
13feb02: pointing error, squint, tsys, and sefd
02jan02: calibration run
o n J2253+161, J0137+331 (3C48)
27dec01: beammaps (turret
scans) of J2253+161.
27dec01: pointing error beam
widths from turret scans.
(out of focus) light.
The cal values and Tsys versus frequency.
20dec01: Power levels
in the upstairs if/lo.
19dec01: Rfi seen
using the xband feed.
19dec01: Measuring the turret
position of the feed.
On 09apr14 we transmitted a tone from the control
room towards the telescope. The setup was:
This will verify that the upstairs lo's are
locked to the station clock. It also shows the phase noise in the
- xb receiver. 2-12 ghz mixer.
- xmit at 8351 using hp 20Ghz synth (amplitude about 0 dbm). It
was locked to the station clock.
- xmit using xband helix
- use mock spectrometer to record data
- 160Mhz sampling freq, decimate by 1000 --> bw = 160Khz.
- 16 bit time domain sampling
- use 10 Mhz offset at a/d input, and -10Mhz shift in digital
mixer to get rid of spikes at dc
- offset input band by an extra 10khz so echo returns 10Khz
below center of downshifted band.
- Take data for 105 seconds (allows for a 16Megapnt xform).
The plot shows the spectrum of the
recorded time domain samples (.ps) (.pdf):
- a single 16 million point (2^24) length transform was done on
- The channel width is then .00954 Hz.
- PolA is black, polB is red (both are circular).
- The spectrum was normalized to the median value and then
converted to a db scale.
- Top: spectrum showing +/- 50 Hz about the expected receive
- midlle: +/- 5 hz about the expected signal
- bottom: +/- .5 hz about the expected signal.
- The upstairs if/lo is locked to the station clock.
- Most of the power in the tone is within 10 millihz
12jan06: check xband receiver after
The xband receiver was brought down to the lab
for some work in jan06. It was reinstalled on 11jan06. On 12jan06
and 13jan06 calibration scans were done on the source B1345+125
(2.17Jy at 9Ghz) to check the receiver. The same source was tracked
on both days to see if the values were repeatable. On 12jan06 there
were a few clouds. On 13jan06 it was clear. Data was taken 7am to
The plots show the
results of the xband calibration measurements (.ps) (.pdf):
The data shows that there is no large
pointing error caused by the re installation. The data from the two
days are repeatable which says that things are stable (the Tsys
difference may be from the weather). The gain variation with za as
well as the sidelobe variation is probably a focus/collimation
- Fig 1: gain, Tsys, Sefd, Beamwidth: The 12jan06 points plotted
with + while the 13jan06 values use an *. Colors are used to
show the 4 frequencies used (8500,8800,9000,9200 Mhz). Tsys
dropped by about 2 kelvins 12jan06 to 13jan06. The receiver was
cool by 12jan06 (at least the tsys measurments done by the
operators showed no decrease 12jan06 to 13jan06). The difference
may have been the atmosphere.
- Fig 2: Coma, first sidelobe height, beam efficiencies.
The first sidelobe increase 12 to 5 degrees za corresponds to
the gain drop. So were are probably going out of focus here.
- Fig 3: pointing errors. There is a 6 asec za offset (about .1
inches). With one source we can't tell whether this is a problem
with the model itself or a residual of the reinstallation.
- Fig 4: The pointing errors for the last year (black)
over plotted with the current errors for these two days
(red). The length of the vector is proportional to the
error (1 tick marck is 5 asecs). The direction of the error is
the direction of the total error (the radial direction is za
while the perpendicular direction is azimuth). These errors do
not look any larger than the errors from the previous year.
You'll notice that we have no previous sources that went through
this same dec range.
13jun04: xband performance
above 10 Ghz. (top...)
The performance of the xband receiver above 10
Ghz was tested on 12jun04 and 13jun04. Calibration scans were
done with the source B1040+123 (.91 Jy at 11 Ghz) on 12 and 13
jun04. The source B1328+307 (3C286, 4.4 Jy at 11 Ghz) was used on
13jun04. The bandpass filter was removed from the receiver for this
test. The 2 to 12 Ghz mixer was used with a high side lo.
Since no filter was present, some of the other side band would
be mixed in. The cal values used were measured on the hilltop test
range. The plots
show the system performance (ps) (pdf).
different frequencies are color coded: black-10.5Ghz, red-10.7Ghz,
Green=10.9Ghz, blue-11.1 Ghz. The sources are plotted with different
The sefd is around 14 up to a za of 16
where it starts to rise. This compares with an sefd
of about 9 at 9200 Mhz. The setting track of B1328+307 has a
large increase of gain, sefd, beam width, and coma for this part of
the dish/rails. This increase is also seen at lower frequencies.
- Fig 1: The gain and Tsys for 10.5 Ghz are too high. This is
probably because the cal value used is too high. The Tsys for
11.1 Ghz looks a little low so its cal value is probably a
little low. The sefd for all of the frequencies looks pretty
consistant so the differences are probably coming from the cal
values used. The cal
measured on the hill show a peak at 10.5 Ghz. This peak is
probably not real.
- Fig 1: The SEFD for B1328+307 rises at za of 15 on the
setting track. The open loop for B1040+123 has a 120 degree
periodicity in azimuth so it is coming from the 3 azimuth
dependence of the optics. B1328+307 did not cover enough azimuth
to determine if its loop was coming from the 3 azimuth term.
- Fig 1: The beam width also increases at za of 15 degrees
- Fig 2: The coma is large. The first sidelobe gets up to about
-11db. The beam efficiencies need to be scaled by the error in
- Fig 3: The rms pointing errors are 3.65 (za) and 6.01 (az).
The 4.13 mean za error is similar to the mean error measured for
at lower frequencies. The azimuth pointing error
jumps by 10 arc seconds at an azimuth of 150 degrees. The
tension in tiedown 4 was decreasing at this point, but it did
not go loose until an azimuth of 135 degrees.
- Fig 4 shows the tracks across the dish. B1040+123 was done
08jan04: xb rcvr moved to
new position on turret floor. (top...)
The xband receiver was moved on 08jan04 to make
room for the alfa receiver. Data was taken on 05-07 jan04 before the
move with the turret at 161.4 degrees. On 08jan04 the receiver was
moved. Data was taken on 21-22jan04 with on the same sources with
the turret at 310.22. The average pointing error in azimuth
for the before and after runs were computed. The difference was made
zero by moving the turret from 310.22 to 309.94. The plots
pointing errors for the 3 measurements.
These are the mean (actually median) pointing errors (not the rms
error). These pointing errors are to be subtracted from the
computed position to have the telescope point on the source.
The turret was moved from 310.22 to 309.04 so that the mean pointing
error for these sources was the same as before the move. There was
not adjust possible in the za direction. To go from tur=310.22 to
309.94 using the above data:
- Fig 1: this is the azimuth za coverage for the 3 measurements.
Black is before, red has tur=310.22, and green is with turret at
- Fig 2: The pointing error for the 3 measurements. The median
pointing errors were:
The az error is within .5 asecs, the za error is
off by 4 asecs from what is was before. These numbers are
approximate since the final run at pos3 did not cover the same az/za
range az pos1,2. It's probably good enough until we get a chance to
survey the horn in position (sometime in feb04).
- Let Pos S have the telescope pointing directly at the
- Pos 1 is 1.42 asec greater than Pos S in az.
- Pos 2 is 11.64 asecs less than Pos S in az.
- To go from pos 2 to pos 1, you must move the azimuth 13 arc
- The turret coordinate system is counter clockwise with 45
asec/turretDeg. + 13asecs of azimuth is -13./45=.28. The new
turret position is then: 310.22-.28=309.94.
in the xband receiver. (top...)
On off position switching was done on 3C454.3
with the xband receiver to check for resonances in the receiver.
8000 to 10000 Ghz was covered. The OMT of the xband is cooled to
about 20 K. If a resonance occurs in the omt, then a fraction of
the input sky+source temp will be replaced by the 20K of the
omt. For off source measurements, 20K is close to the temperature of
the 3K background plus the scattered radiation so there will not be
a noticeable bump in the off position. The on source position
(probably about 12Jy or 40 Kelvins) will be replaced by a lower temp
so there should be a dip at the resonance.
The plots show
on/off - median(on/off) with the resonances flagged in green.
Black is polA and red is polB. The units are Tsys. The
discontinuities every 20 Mhz are from the different 25 Mhz
bandpasses (with had an overlap of 2.5 mhz).
05sep03: new horn
performance. SEFD old and new horn. (top...)
The xband receiver had a new horn installed on
04sep03. The x102 calibration data taken before and after the
horn change were used to compare the horns. The
before and after the horn change show how much
system performance improved. The improvement will come from a gain
increase and a decrease in Tsys (since Sefd is Jy/Tsys). Most of the
Old horn data was taken around april,may 2003. The new horn data was
taken in sep03.
The SEFD (Jy/Tsys)
improved 25 to 30 % with the larger increase at the lower
frequencies. At 9 Ghz the sefd is close to 8 Jy/Tsys. People should
be warned that this data was taken in the calibration mode which
searches for the peak. The 8 Jy/Tsys does not include any pointing
errors that normal observations would encounter.
- Figure 1: The SEFD vs za for the 2 datasets. Black is the new
horn, Red is the old horn. Each source has a separate symbol.
The 4 boxes are the frequencies for the calibration data:
8500,8800,9000, and 9200 Mhz.
- Figure 2 Top: For each common source, the SEFD of the old horn
was interpolated to the za's of the new horn measurements. The
ratio SEFDOld/SEFDNew was then computed. The top plot shows this
versus zenith angle. The colors are the 4 frequencies. The lines
are linear fits to the data vs frequency.The zenith angle
dependence of the ratio could be a pitch or roll change caused
by a weight change of the dome. The variation in the data
probably comes from Tsys changes caused by weather.
- Figure 2 Bottom: The linear fit for the ratio was evaluated at
10 deg za for each frequency. The lower plot shows the
change in SEFD (Old/New) for the 4 frequencies.
Compare (SEFDB/SEFDA) and (TsysB/TSYSA) using x102 calibration
The system temperature for polA and polB is
different. (see below). This could be caused by a real
difference in the system temperatures or it could be an error in the
cal values used to compute the system temperature. To check
this you can use the power from a continuum source rather than the
cals to measure the values for Tsys. This assumes that the
telescope gain for the two polarization's are the same and
that the sources are not polarized. Since xband is circular
polarization, source polarization is probably not a problem.
The x102 calibration data for sept
09,10,11 was used for this test. I re analyzed the
data since the normal processing combines the two polarization's
and computes Tsys for stokes I. I refit the data using a 2d
gaussian fitting for : Amp, offsets, major axis, coma, and coma
angle. I did not try to fit for the sidelobes so the data is
probably not as accurate as the default processing. The SEFD is
Tsys in Janskies. It only relies on the source flux to be correct
(the ratio sefdB/sefdA only depends on Tsys). You can compare the
ratio of TsysB/TsysA using kelvins from the cal or Janskies from
the SEFD and Tsys for the 3 days of xband data.
The tsys ratio using the cals does not match the Tsys ratio using
the source for frequencies 8500,8800,9000, and 9200 Mhz. It is worse
at the lower frequency. This agrees with the tsysRatio comparison
before and after the horn was installed. The Src ratio shows
TsysB to be higher by 2 to 7 % depending on frequency.
- Fig 1 : SEFD vs za. Top is polA, bottom is polB. Each source
has a separate symbol. Each color is a different frequency band.
- Fig 2: Tsys vs za. Top is polA, bottom is polB. Symbols are
different sources, colors are different frequency bands. This
data was taken after the new horn was installed (and after the
cal cables were cleaned). It is using the cal values from jun03.
- Fig 3: The top plot is the ratio of SEFD'S :
sefdB/sefdA. The center plot is TsysB/TsysA using the
cal values. The bottom plot computes the average ratio by
frequency (throwing out outliers). The black trace is
TsysB/TsysA measured using the cal values. The red line is
TsysB/TsysA using the source flux as the reference value.
05sep03: Tsys vs
frequency with the new xband horn. (top...)
The xband horn was changed on 04sep03. On 05sep03
tsys vs frequency was measured by firing the high correlated cal and
stepping through the frequency range. The tsys vs frequency
with the old horn was taken from the cal measurements done on
06jun03. In both cases:
The differences in the measurements were:
- The frequency range 8 to 10 Ghz was covered 3 times.
- The high correlated cal was used (diode 1 to polA and polB)
- Blank sky was tracked during the measurement
- The same value for the cal was used in both cases (but see
Tsys vs frequency for the two measurements.
- sep03 used the new horn, jun03 used the old horn
- sep03 used the 2 to 12 Ghz converter, jun03 used the 10 Ghz
- sep03 data was taken at 15:20 (clear weather) while jun03 was
taken at 8:30 clear weather.
Something has happened differently for polA and polB between 8 and 9
Ghz. PolB is almost constant at 85% while polA goes from 50% to
85%. Above 9Ghz the ratios are the same.
- topPlot: This is Tsys vs frequency for both
measurements. Black, Red are polA,polB with the new horn.
Green,Blue are polA,polB with the old horn
- middlePlot: This is TsysB/TsysA for the newhorn (black) and
the old horn (red). Above 9 Ghz the two curves are identical and
close to 1 (tsysA the same as tsysB). Below 9 Ghz TsysB/TsysA
for the new horn increases. This is probably happening because
of tsysA being so low.
- Fig 1 Bottom: This is the ratio of
TsysHornNew/TsysHornOld. Black is pola, red is polB
I asked electronics if they had done anything with the xband
cals. The only thing they could think of was that they had
"cleaned" the cal cables into the dewar on 04sep03. The
performance difference caused by changing the horns should be
equal in polA and polB. The amount of cal signal getting into the
dewar has probably changed for one of the cal cables so we need to
az,za pointing offsets from calibration scans. (top...)
The calibration data taken apr03 thru 21jul03
showed pointing offsets of 8.04 asecs in az and -2.69 asecs in za.
These values were subtracted from the current model az,za offsets on
21jul03. The plot shows the pointing
for apr03 thru 21jul03 before the offsets were changed (the
src 3C48 was not included since the tiedown tension were close to
zero when the source was measured (10am).
jun03 a1704 recomb
lines, gain drifts. (top...)
Experiment a1704 was looking at recombination
lines in compact hII regions using xband. 5 minute position
switching with 1 second dumps were done. During this period the
correlator had problems with gain drifts (this problem turned out to
be the tubular filters in the correlator).
While monitoring the xband total power
(25Mhz , 1 second samples) some variations in polB showed up. The
plots show total power vs time for 8 OFF src scans (no
continuum). Each plot has the 4 sbc plotted bottom to top (with an
offset between sbc for plotting). polA is black while polB is red.
The total power has been converted to Kelvins (using the cals) and
then the median value for each scan was removed.
When all 4 sbc jump, it is probably not the correlator filters. It
is the IFLO before the downstairs mixers, upstairs iflo, or the
xband receiver polB.
- Fig 1 pattern 1,2 are pretty good
- Fig 2 pattern 3 has a jump in polB (red) for all sbc at sample
250 of about .1 K (Tsys was about 50K). Pattern 4 has a jump at
- Fig3 pat 5,6 both have drifts in polB
- Fig 4 pat 8 has a large jump in polB at sample 30.
The second set of plots has the total power
source deflection vs time for the 8 on/off patterns by. For each
pattern (on-off) was compute for each 1 second record and then the
total power was computed (in Kelvins). The plots
the srcDeflection vs time for the 8 patterns and the frequency
of the oscillations.
If the pointing error is large, the source will
be sitting on the steep slope of the beam. Any oscillations in the
structure will map directly into a gain oscillation at that
frequency. This is what is probably causing the high frequency
oscillation. At the start of each onsrc, the telescope has
just complete a move (from the previous off). There will normally be
oscillations for a while after this occurs. You can see this in the
red and black scans where the oscillations are present at the start
of the scans.
- Fig 1 Top. The src deflection vs time for
the 8 patterns. The data is sampled at 1 hz. The y axis is in
kelvins. Each color is a different on/off pattern. # of the
sources had continuum while the other 5 had little continuum.
- Fig 1 Bottom. Each 300 second source deflection was fourier
transformed and then the amplitude was plotted vs
frequency. The highest frequency is .5 hz. The power near
.5 hz is probably aliased down from above .5 hz. The platform
oscillations frequencies where measured to be .365 and .57 hz during
The slower drift may be weather or pointing.
If they were pointing errors, then the oscillations that are
sitting on top of the slow drift should decrease as the slow drift
reaches a maximum (no longer on the edge of the beam). The data
doesn't seem to do that. but..It could also be that the
oscillations are dying out so that when there are no oscillations
it doesn't necessarily mean that we are on the center of the
jul02 Tsys vs time.
Dtsys/Tsys after integration. (top...)
During jul02 drift scans were done over multiple
days. The correlator setup was 50Mhz by 512 Channels, 3 level
sampling, 2 polarization's, and 4 frequency channels. The data was
sampled at .5 seconds with each drift scan lasting 10 minutes giving
1200 samples per drift. 14 drifts were done per day on the same
strip. The data was hanning smoothed. A source was positioned 28
samples in from the start of each drift (it is the spike you see
every 1200 samples). Its strength is roughly 1.7 Jy at 8800 (this
number not very accurate).
- The first plot shows Tsys
sample for the 1200 samples by 14 drifts. The x axis is a
complete set of samples for 1 day. The data has been smoothed by
3 (1.5 seconds) to reduce the size of the plotfile. Each 1200
points, the telescope would move to a new za (rise at the left ,
transit in the middle , and set on the right). Each color is a
separate day (black1, red2, green3..). Each plot of the 4 plots
is a separate frequency band. Within a plot, the upper set of
lines is pol B (it has a higher Tsys) and the lower set is polA.
The numbers 1-14 across the first plot are the 14 drifts
done. Some things to notice are:
The second plot is a blowup
the first showing only the 8750Mhz data with no smoothing.
Each day has been offset by 2 K for plotting. You can see that
the weather is increasing the tsys by up to 2.5 K or about 5%.
The third set of plots shows spectral density (in units of
Tsys) versus strip position and frequency. It has been
averaged over all of the strips off a day, all of the days, and
both polarization's. The data has been weighted by 1/sigmaTsys^2
(no weighting for telescope gain was done). For each drift scan:
- day 1 (black) strip 3,4 the gain in polA is varying.
- day 1 strip 3, day 2 strip 8 there is rfi at 8800 Mhz (2nd
plot).The bumps in the total power (eg day2,strip 4)
are constant across frequency (all 4 sbc).
They occur more at the edges (hi za). It is not rfi since it
is wide band. It is probably not a source in the sidelobes
since it did not repeat day to day. It is probably not a gain
variation since it was identical in both polarization's (and
they have different amplifiers etc ..(except the first lo is
common).). It may be atmospheric variation of Tsys that is
occurring on the order of minutes.
After accumulating all of the data, it is multiplied by 1/weights.
An average bandpass is divided into this data and channels 50:450
are used to remove any time variation.
- The 1200 spectra are normalized in frequency by the average
bandpass (excluding the continuum source).
- Each spectra is then divided by the total power (computed
over channels 50:450). This is to remove any gain variation
(and to some extent weather).
- sigma^2 is computed over channels 50:450 and records
100:1200. The spectra are then multiplied by 1/sigma^2.
The images have been scaled to 8 Sigma (min to max in the
greyscale). Each image is about 800Kbytes.
The rms was computed using strips 100 to 1200 and frequency
channels 50 through 450. The expected rms is:
Where driftsPerDay=14, Ndays=7, npol=2, Han=2, and 3level=1.23 for
3 level sampling.
Expected rms (7 days): .00028 dTsys/Tsys.
The measured values were a little less than this (see the plots).
The 2nd image (at 8800 Mhz) has interference that was there for
2 days. The large drift in frequency may point to a satellite in
4 5 minute on/offs were done tracking the
position ra: 043841,dec:253445 (1950) in TMC1. A1.56 Mhz bandwidth
with 1024 channels was used for a velocity resolution of .05
km/sec. The rest frequencies tracked were: 9097.0346, 9098.3321,
9100.2727, and 9024.004 all with a lsr velocity of 5.5 km/sec.Each
on,off had a linear baseline removed and then the 4 onoffs were
averaged (with no weighting for g/t). The plot
the line strength of the 4 lines versus the rest frequency and
versus lsr velocity. The last plot of HC7N has been smoothed
by 11 channels in frequency and velocity. The temperature scale was
arrived at by using 45K for Tsys.
on B0518+165(3C138),B1040+123,and B1328+307(3C286). (top...)
Calibration runs were done on 3 sources using the standard heiles
scans (see stokes
calibration). The pointing
and telescope performance are shown in the plots. It can be
compared with the 02jan02
data before the horn was moved in focus by .44 inches and the
turret position by .35 degrees.
- Fig 1. This has the pointing error in arc seconds for the
azimuth and zenith angle directions. For the 3 sources there
still appears to be an offset. Part of this is do to the
calibration scans fitting for the coma while the turret scans
don't. This moves the position a little. We also may need more
calibration sources to get a better average.
- Fig 2. plots the Gain (K/Jy) and Tsys(K), SEFD (Jy/Tsys), and
average beam width (asecs) for the 3 sources (different symbols)
and the 4 frequencies (different colors). The SEFD at low
za is well behaved (the 02jan02 data used 3C454.3 whose flux is
not known but the shape seems to have improved).
- Fig 3 shows the coma, 1st sidelobe height, main beam, and then
main beam + 1st sidelobe efficiency. The first sidelobe heights
has improved by maybe a db from the 02jan02 data.
21feb02:HC3N lines at
9097,9098,9100 Mhz. possible detection? (top...)
2 five minute on/offs were done on B0638+095
looking for the HC3N line transitions at 9097,9098, and 9100 Mhz. 12
Khz resolution was used. The onoffs and the polarization's were
plot shows the spectral density versus the topocentric
frequency. The expected topocentric frequencies are flagged. The
spectra has a .011 Kelvins rms. So the 9097 and 9098 lines are a 3
sigma detection (assuming of course that it is supposed to be in
emission and not absorption!).
(turret scan) J0237+288 (top...)
were made near transit of J0237+288 using turret scans. This
was after the pointing offsets were updated. The pointing errors for
these maps was 1-2 asecs in za and 2->18 asecs in az. The
sidelobes are better than the corresponding maps before the offsets
were corrected (see the beam maps from 13feb02). The sidelobes now
range from 11 to 13 db down. The pointing error for each beam map
was: (2,.9,7,5,7,11,10,18,10,12) asecs.
13feb02 beammaps J2253+161,
Beammaps were made of the turret scans on
J2253+161 and J0137+331 (3C48). Each source is split up into rise
and then set (each ps file is about 3mb..). The color contours are
spaced by 3db. The J2253+161 is about 3 times stronger than
3C48 so the beamaps show more dynamic range. For both sources there
is a significant sidelobe on the lower right side. The az pointing
errors for these source were large (about a beam) so the turret was
off the paraxial rate for the maximum. This may have contributed to
this sidelobe. The setting az error for 3C48 got up to 55
asecs so sidelobe should have been worse here. We can check this by
redoing the strips after the pointing offset is updated and see if
the sidelobe decreases.
13feb02 pointing, beam squint, sefd, and
beammaps. 3C48, J2253+161 (top...)
The turret position was moved by .35 degrees and
the horn was moved up by .44 degrees from the horn survey. The
pointing model was switched to model13 with a "best guess" of the
az,za offsets. J2253+161 (3C454), J0137+331 (3C48), and
J2123+055 were tracked using turret scans. The pointing
squint,Tsys, and sefd are shown in the plots:
- Fig 1. The top two plots are the pointing error in az,za
versus az and za. This is using model13. The average offset is
27.4 asecs in az and 5.3 asecs in za. The model13 offset for
xband will be updated by this amount. The bottom plot is the
beam squint (pointErr PolA-PolB). It less than an arc second.
- Fig 2. top is the normalized Tsys for polA (black) and polB
(red). I normalized the curves by dividing by the average Tsys 0
to 12 degrees. Jumps in the plot may be from the fits or any
gain changes (I assumed the gain remained constant for the
entire set of observations). Apart from the jumps, tsys is
relatively flat out to 15 degrees (unlike some receivers that
show a linear ramp at low za).
- Fig 2 center is the source deflection (as a % of Tsys for the
3 sources). The large hook for J2253+161 below 5 deg za is from
the prf errors (see the next plot).
- Fig 2 bottom is the SEFD versus za. 3C48 is the only source of
the 3 with a well known flux (the other two are flat spectrum
and probably variable). The fluxes used in K/Jy are next to the
source names. The best value for 3C48 is 15 Jy/Tsys. With a 45 K
Tsys this gives a best of 3 K/Jy gain. The pointing error of
almost a beam will cause the gain to be down by .5 to 1
db. We will get this back when the pointing offsets are
3C48 and J2253+161 were
tracked using the standard calibration routine (heiles scans). The plots
xband calibration run on J2253+161,3C48 (top...)
- Fig 1. top is the Gain in K/Jy. The flux for B2253+158
(3C454.3) is probably off by up to 30 %. The next plots show the
Tsys, SEFD, and average beam width.
Fig 2. top is the coma parameter that measures the asymmetry of
the main beam. The next plot is the peak of the 1st sidelobe,
followed by the main beam, and then the main beam plus 1st side
lobe beam efficiencies.
27dec01 beammaps of
After the heating resistor was added to the
waveguide, J2252+161 was tracked doing turret scans from za=9.8 to
19.2 degrees. Contour maps where made of each 2 minute strips of
maps are 3 arc minute square (the x axis is labeled with the
turret motion rather than the azimuth offsets). Each colored contour
is separated by 3db. I used the simple contour filling algorithm
which makes smaller files, but gets a little confused with the
colors. The edge of the inner most color is 3db down from the peak.
The best sidelobe is about 9 db down and is always to the right side
in azimuth. The coma lobe is opposite this and gets worse as we go
up in za (since the pitch,roll are increasing).
Pointing error and beam
widths from turret scans: (top...)
Pointing runs using
turret scans were run on 25dec01 and 27dec01 using the xband
receiver at 8.8 Ghz. 500 MHz was detected about this frequency. The
first 5 sources (on 25dec01) were taken before the heater was
installed on the wave guide so there was condensation in the horn
(and a high Tsys). The signal to noise was not great. Fits with beam
widths below 30 arc seconds are above 55 arc seconds were excluded
from the data. The pnterr,beamwidth
figures show the results:
The zaError versus za increased linearly above za
of 10 degrees for all sources. This is also the region where the
pitch error increases linearly. The beam maps above show that the
coma is in the uphill direction or positive za. There is no coma
term in the fit, so it is biasing the 2-d gaussian uphill.
- Figure 1 plots the za error versus za and azimuth.
- Figure 2 is the az error versus za and azimuth.
- Figure 3 is the total error (az and za) added in quadrature
and plotted versus za and azimuth.
- Figure 4 is the beam widths in the za (top) and azimuth
There is not enough data, but it looks like there
might be a large 1 azimuth term in the za and az errors.
The beam width for the za direction is close to
the value expected. Scaling from the measured za beam widths at
sband, the xband beam width should be 35.6 arc seconds. The az beam
width of 40 arc seconds is too large by 25 %. This is larger than
az beam width discrepancy measured at sband.
First (out of
focus) light B1611+343: (top...)
B1611+343 was our first shot at tracking a source
with the xband receiver. Turret scans were used to measure the
pointing error, gain, and SEFD. The conditions were not ideal:
The tiedowns were not working so the platform was 2.5 inches
The source didn't come below 15 degrees za so the pitch, roll,
and focus errors were large.
The source has a flat spectrum and is a low frequency variable
so the flux is not real accurate.
The horn location still needs to be surveyed into position
with the theodolite.
In spite of all of these "shortcomings" there was still
something to see at 8800 Mhz.
and Pointing error. The flux used was 4 Jy at 8800 Mhz.
The gain was computed using the source deflection/TsysDeflection
and Tsys of 43K and 48K for polA and polB (taken from the tsys
plots above). The pitch and roll error added in quadrature is
probably close to .2 degrees. At 10 Ghz this would cause a 3db
gain loss. The focus error is around 1" which probably
contributes a db. A 2 mm rms reflector surface would reduce the
gain at 8800 MHz to 58% of theoretical. The pointing error on
page 2 is about 38 asecs in az and 18 asecs in za. These offsets
can be added to the xband pointing model offsets.
of B1611+343. Each contour map is a 2 minute turret scan
integration. The dome moves along the y axis +/- 1.5 arc
minutes. The x axis is +/- 2 turret degrees (+/- 1.5 arc minutes
on the sky). The contour colors are spaced by 3 db. The azimuth
and zenith angle is plotted at the top of each map
cal values and Tsys versus frequency: (top...)
Tsys versus frequency for the xband receiver was
measured by tracking blank sky and taking data with the
correlator while the cals were fired. Tsys was computed as
CalInKelvins. The frequency was stepped from 7800 to 9800 Mhz.
This cycle was repeated 4 times. The figures show the cal
and Tsys versus frequency:
There was interference around 9450 Mhz. Each
frequency has four * plotted for the 4 separate measurements
(separated by about 5 minutes). There is no scattered in the
measurements so the rfi does not look like it made any difference.
The rise starting at 9500 Mhz may be partly do to the fiber optic
transmitter adding to Tsys because of the IF signals low spectral
- Figure 1 plots the cal values versus versus frequency for all
of the cal types. The values were measured on the receiver range
using liquid nitrogen as the load. The top plot is the low cals
while the bottom plot shows the high cal values.
- Figure 2 is Tsys versus frequency. Black is polA while red is
polB. The telescope zenith angle varied from 12.5 to 9 degrees.
At each frequency there are 4 separate measurements (*) plotted
on top of each other.
20dec01: Power levels in
the upstairs if/lo: (top...)
The IF signal is sent
from the platform to the control room via fiber optics. For the
transmitter to not add to the system temperature the spectral
density of the signal should be greater than -103 dbm/hz. The
FO xmter will start to compress around 0 to 5 dbm total power. Blank
sky was tracked around 10 degrees zenith angle and the power at the
input to the fiber optic transmitter was measured versus frequency.
levels are shown in the figures:
The top plot is the total power versus frequency at the input
to the fiber optic transmitter with all of the adjustable
attenuation removed (11 db before the mixer, 11 db after the
mixer). The horizontal line is the total power level to give
-103dbm/hz over 1 ghz bandwidth.
The middle plot is the gain needed in the postamps to bring
the power levels up to -103dbm/hz at the FO xmter and still have
6 db attenuation in the rf attenuator (to allow for some
adjustment). The spectral density was computed from the total
power by assuming 1 Ghz of constant spectral density about each
total power measurement. This will underestimate the spectral
density on the edges (since there is not 1 Ghz of total power).
The bottom plot shows the power levels at the input to the IF1
mixer chassis if 12 db gain is added to the postamps. It assumes
a 35 db gain for the mixer chassis.
12 db seems like a reasonable gain to insert. Larger gain could be
used by increasing the programmable attenuation (as long as you
don't lose to much head room in dynamic range on the amps/mixer
chassis). With the current setup the the system temperature
for the higher frequencies (9.5->10Ghz) is probably being
increased because of the fiber optic transmitter noise level. At the
top side there will probably be 15 do 20 db of headroom before the
fiber optic xmter begins to compress.
Rfi seen using the
xband feed: (top...)
After the feed was
installed an rfi monitoring session was run using the xband receiver
on the telescope. This lasted for about 2 hours (11 am to 1 pm). 60
1 second integrations were done at 100 Mhz intervals with a
resolution of 25Mhz/1024 channels. There were three 60 second
runs (spaced by about 40 minutes) at each frequency.
Interference was seen in two of the 100 Mhz bands. A peak hold
spectra was made for these two bands (taking the maximum value in
each frequency bin over the 180 samples). This is the worst the rfi
got in a 1 second integration. Images of time versus frequency of
the spectral density were also made. Each image was normalized to
the median spectrum over the 180 samples. The figures are:
density versus frequency for the two bands with rfi. The
vertical scale is linear in power. Red is polA and green is
image at 8120 Mhz (time versus frequency).
image at 9350 Mhz (time versus frequency).
The images have been normalized to the median spectra over the 180
samples. They show the time variability of the rfi.
19dec01: Measuring the
turret position of the feed: (top...)
The turret position of the feed is measured by
moving the turret in a sine wave above the tertiary while
tracking blank sky with the telescope. It turns out that the peak of
the power corresponds to the turret focus position (this has been
verified by surveying the resultant positions of other feeds). You
would expect the turret position for focus to correspond to a
minimum in power since the sky is the coldest thing around. This
peak may result from the match of the horn into the tertiary
maximizing the transmitted power into the receiver at the focus
(thermodynamic arguments aside..). Figure shows
2 sets of turret swings using a sine wave with amplitude 8
turret degrees and period of 15 seconds. Each measurement lasted for
2 minutes. The top two plots are polA and the bottom two are polB. A
fifth order polynomial was fit to the data and then the peak was
flagged. The two polA swings and the two polB swings agreed to .001
degrees. The polA swings differed from the polB swings by .17 turret
degrees (45 sky arc seconds per turret degree). This disagreement
could be a delay difference in the A, B channels (The time constants
in the square law detectors were set to .02 seconds). I used the
average of: 161.99 as the turret position for xband.