12meter xband cal values measured 07jun22
09jun22
Links to PLOTS:
Links to SECTIONS:
Why the cals were
remeasured.
The model used
to compute the cal values
The parameter values used
in the computation
Taking the data
Processing the data
Results
Total
power vs time by freq band.
calRatio (caldif/caloff) vs freq
Fitting
to calRatio vs freq
Computed
Tcal and tsys vs freq
Summary
Why the cals were remeasured.
The xband cal diode was stabilized with a peltier
cooler on 08Nov21. On 06jun22 a new diode was installed so that
the sband and xband diodes would fit in the peltier cooler.
Parameters used for Tcal computation
The values i ended up using for the
computations (see below for explanations) were:

value 
Notes

Tabsorber 
305K 
 89.5 F from temperature sensor in absorber at start.
 3rd abs scan, sun hit absorber
 99.5F at end of 3rd sb abs scan.

Tsky 
15K 
includes any scattered energy. see below

Treceiver 
74,68K

polA,B see below

Tomt

300k

omt temperature.. a guess.

g

.05

reflection coef for horn/omt see below

alpha

.07

omt absorption




Taking the data:
 The hardware winking cal was used:
 20 millisecond on
 20 millisecond off
 the cycle started on a hardware 1 second tick.
 The mock spectrometer was set up to measure the spectra.
 7 mock boxes were used each with the following setup:
 172.032 MHz bandwidth
 2048 spectral channels
 polA,B measured
 data sampled at 2 milliseconds (winking cal was 20
ms on , 20ms off).
 each scan (sky or absorber) took 2 minutes of data.
 The 7 band frequencies were:
 8219.00 8363.00 8505.00 8647.00 8789.00 8931.00 9073.00 MHz
(center freq 8646)
 The bands are spaced by 142 MHz
 the freq limits are (8219172/2) and (9073 +
172/2.)= [8133,9160] or 1027 MHz bw
 The freq range of the measured cals was limited to 8150 to
9150 since there is a 12Ghz IF filter.
 The measurement started at 11:07 and finished by 11:42 ast.
 We did 3 absorber scans and 3 sky scans.
 The scan sequence was:

Loc

ra , dec j2000

az,el

mockFile #

abs





200

abs





300

abs





400

sky

03:39:45/25:24:77

40,80

800

sky



900

sky



1000

Processing the data:
For each scan the following is done:
 Input the winking cal spectra and compute the ratio of
calDeflection(freq)/Caloff(freq)
 Average the 20 ms of calon, caloff, giving bCalOn[25*120],
and bcalOff[25*120]
 exclude the 2 ms spectra before and after each cal
transition. This gives 16ms of data after the average
 Compute the rms by chan for the bCalon and bCalOff data (for
debugging)
 Compute the avg calOn, calOff: bonAvg,boffAvg using
the median for averaging
 compute the caldeflection and then the calRatio (by Tsys)
 bcalDef= bcalOn  bcalOff
 bcalRatio=bcalDefl/ bcaloff
 this removes the IF bandpass.
 interpolate bcalRatio so that the data across the 7 bands is
a single spectra with fixed freq step (84Khz )
 This is then the cal value in units of Tsys (with cal off).
 Average calRatio for the 3 sky scans and the 3 absorber scans.
 Fit a curve to the calRatio for sky and abs.
 Compute Tcal using:
 Just the RatioAbs, RatioSky, and the combined
RatioSky,RatioAbs (to remove Trcv).
 play around adjusting gamma, alpha, and Trcv to try and get
the 3 different Tcal measurements to agree.
The Results
Total power vs time by freq band
Plotting the
cal Deflection total power vs time (.pdf)

The total power is computed for each 20ms calOn, caolOff
spectra.
 This total power sequence was then normalized to its median
value
 Each scan was normalized to its median value.
 This shows the gain variation during a scan (assuming the
noise level remains constant).
 There are 4 frames per page: polA,B 1 freq band, followed by
polA,B of the next freq band.
 In a single frame there are 6 scans (horizontally):
 ab s,abs,abs,sky,sky,sky
 blue lines separate the scans. (there is a short time lag
between each set of scans).
 What the plots show:
 most freq bands show similar variation in level.
 The absorber varies by about .5% (of Tabs)
 the sky varies by about 1% (of Tsky)
 the 8789 and 8931 bands on sky show rfi (this is later
removed with a robust average).
Calratio (calDif/calOff) vs frequency
The 2nd set of plots shows the
calRatio vs freq plotted for each scan (.pdf)

compute the median spectra for the calon, caloffs
 compute (calOncalOff)/calOff spectra
 The division by calOff removes the IF bandpass.
 Page 1: calRatio vs frequency (CalonCalOff)/CalOff spectra
 Each color is a different scan:
 black,red,green : on abs
 blue,pink, light red : on sky
 The dashed vertical lines (purple) show where the rfi lies
in the 8363 MHz total power band.
 The cal is .1 to .05 * Tsys on sky
 the cal is .025 to .01 * Tsys on absorber
 The attenuators that were installed in the cal line where
(7,7) db . They should have been (3,2) db .
 You can see ripples in the cal values vs freq in the sky
measurements. They may also be present in the abs measurement.
 Page 2: Normalize each calRatio scan to their median value
 top frame: polA
 bottom frame: polB
 the ripples are now the same size for both the sky and
absorber.
 Since we divide by the calOff, these ripples are probably
not coming from the IF bandpass.
 If the ripples were in the cal cables, then you would
expect the ripples in the abs to be much smaller than
the ripples in the sky (since the Tcal value is the same,
but the Tsys is very different).
 Page 3: ACF of calRatio spectra. Ripple spacing.
 the acf of the calRatio spectrum shows the spacing between
the reflection points (if it is a standing wave).
 Top: polA, bottom: polB
 The lag peak is about .015 usecs.
 With n=1 this is a distance of 2.26 meters.
 with n=.7 the distance is about 1.6 meters.
 Page 4: rms/mean by channel for the calon, calOff
spectra
 black polA, red: polB
 top frame: on sky
 bottom frame on absorber
 The rms by channel is computed for the 3000 calOn and calOff
cycles for a 2 minute scan.
 The 3 sky scan and 2 abs scans are over plotted.
 You can see the larger rms in the 83198350 mhz spectra for
the on sky calratios.
 the lines going off scale are the edges of the band passes.
 The mean value is a bit higher than the expected value,
probably because of the gain variation during the 2 minutes.
Fitting a curve to the calRatios.
A harmonic fit was done to the calRatio vs
freq and then the fit was used to compute TCal.
The plots show a fit to the CalRatio measurements
(.pdf)
 The 3 sky measurements and 3 absorber measurements were
averaged to give a sky and absorber calratio vs freq.
 A robust harmonic fit was then fit to the calRatios vs freq
 the fit included:
 a linear term
 A 17th order harmonic fit (fitting ripples of 172MHz down
to 172/17 = 10MHz spacing).
 I looked at harmonic fits of 10 to 22 order and settled on
17 since the fit residuals did not improve much after this.
 The fit was iterated throwing out outliers and then redoing
the fit.
 Page 1: calratio vs freq and fit
 black is polA , red is polB, green is the fit
 the blue * points were excluded from the fit.
 Top: On sky calratio vs freq and fit
 2nd frame: on sky fit residuals
 The residual rms was around .0008*TsysSky.
 3rd frame: fit to calRatio on absorber.
 bottom: absorber fit residuals
 the residual rms was around .0005*"TsysAbs
Computed Tcal and Tsys using the cal values.
The plots show the computed Tcal
and tsys in deg K using the new cal values (.pdf)
 Page 1 Tcal vs frequency
 Tcal was computed using calRatio from the y factor as
well as just calRatioAbs and calRatioSky.
 the values of Trcv, alpha, and gamma were modified to get
the closest agreement with the 3 methods.
 PolA
 black: used y factor calRatio
 green: used calRatioAbs
 blue: used calRatioSky
 PolB:
 red: used yfactor calRatio
 lightBlue:used calRatioAbs
 purple: used calRatioSky
 We ended up using:
 Tsky + Tscattered: 15K
 Trcv(a,b): 74,68K
 gamma=.05,alpha=.07
 Page 2: Tcal vs freq and Tsys vs freq
 Top: Tcal vs freq
 this is from the y factor measurement after adjusting
gamma and alpha.
 TcalPolA is about 10k to 7K
 TcalPolB is 8K to 4.5K
 The + are the cal values sampled every 10MHz.
 they are entered in the cal table that users will use
for the cal value
 routines will then linearly interpolate the cal value
between these 10MHz points.
 Bottom Tsys vs freq
 the cal off data was averaged, divided by the cal
deflection and then multiplied by Tcal to get the system
temperature on Sky.
 Tsys ranges from 95K to 120K with polA being a little
higher than polB
 The elevation was close to 80 deg.
 elevations close to 10deg will see a Tsys increase of
about 20%.
Summary:
 The single cal diode feeding polA ad polB of the 12meter was
measured on 07jun22 in the morning.
 the tcl routine
/share/megs/phil/svn/aosoft/p12m/x101/220607_meascalsbxb.proc
was used to take the data.
 The hardware winking cal data was taken for 2 minutes on each
scan.
 There were 3 scans on sky
 there were few clouds and the elevation was around 80deg
 There were 3 scans on absorber (the absorber was 89.5F during
the measurement)
 Looking at the total power vs time showed up to a +/ .5%
variation in the calratio during a 2 minute measurement
 the noise diode is temperature stabilized. This change
is problem temperature variations of the electronics.
 The calRatio (calDeflection/calOff) vs freq showed ripples vs
freq.
 this was seen on sky and absorber.
 this rules out a reflection from the structure.
 dividing by CalOff would cancel most ripples in the if/lo
after the cal injection point.
 the period corresponded to a length of 2.26 meters
with n=1, or 1.6 meters if n=.7
 I've previously thought that the ripples could have been
coming from the cable from the cal diode to the coupler where it
is injected into the signal path.
 It no longer looks like the ripple is in the cal cable.
 we inserted a 7db bad in the cal cable. this should reduce
the ripple by 14db if the reflection passes through the
attenuator twice. We did not see the ripple decrease by
a large amount
 If the reflection was in the cal cable, then it's amplitude
would be a function of the calValue.
 Ripple amplitude (after normalizing to Tsky orTabs) had
similar amplitudes. It should have differed by the Tabs/Tsky
ratio.. It didn't
 If The reflection is the horn input to somewhere before the
cal injection then the abs,sky reflection amplitudes (after
normalization to Tsky or Tabs should be the same.
 We see this ripple in Tcal since we divide by CalOff.
 This division cancels ripples after the cal insertion
 It will create ripples in the cal if the ripples come
before the cal insertion.
 Rfi was seen from 8319to 8350 MHz while on sky
 a 17th order harmonic function was fit to the calRatios vs
freq and then used to compute Tcal
 the fit residuals were: .002*TsysOnsky, and .0008*TsysOnAbs
 The Tcal was then computed using:
 the Y factor (abs hot load, sky cold load)
 this removes Trcv but needs gamma(reflection coef
horn,omt) and alpha (absorption in omt).
 Using just calRatioAbs
 using just calRatioSky
 Trcv, alpha, gamma were adjusted to get the 3 measurements
as close as possible.
 Tcal values
 Tcal polA about 10K to 7K
 Tcal polB about 8K to 4.5K
 a table of values spaced every 10MHz is stored in a file
 /share/megs/phil/svn/aosoft/idl/data/cal12m.datR11
 the idl routines:
 calinpdata.pro will input the entire table
 calget1.pro will input a value at a single freq.
 Tsys vs freq.
 using the calOff measurements on sky Tsys was measured.
 It varied 100 to 125k over the entire freq band with polA
being higher than polB.
processing: x101/220607/xbcal/measure_cal.pro
12
meter page
home_~phil