Converting position switched data to Kelvins
Posit ion switched data can be converted from
spectrometer counts to kelvins using the cal values (see cnts
to kelvins). The processing includes:
- let gIF(f) be the
normalized IF bandpass function. I'll pretty much ignore the Tsys and
Tsrc frequency dependence.
(Tsrc_cnts(f) + Tsys_cnts(f) )*gIF
- Bsrc_tsys(f) = (Bposon_cnts(f)-
BposOff_cnts(f))/BposOff_cnts(f) = Tsrc_cnts(f)/Tsys_cnts(f)
- where Bsrc_tsys is now in units of Tsys.
- Bsrc_k(f) = Bsrc_tsys(f) * mean(BposOff_cnts(f)) *
- The 2nd term (mean..) converts the Bsrc_tsys in tsys units back
to spectrometer counts.
- CalInK(avg) is the measured cal value averaged over the entire
172 Mhz band of the spectra.
- CalDifCnts(avg) is calOn-calOff averaged over the 172 Mhz
Averaging over spectra.
The processing takes averages over the posOn, posOff
spectra as well as the calOn, calOff spectra.
The solution is to use the same rfi mask for the on,off position
data as the cal on,off data.
- The data for these averages are taken at different times.
- If there is rfi in any of the spectra then the averages will be
biased by the rfi.
Example spectra: 3C48.
On off position switching was done on 3C48 using the
lband wide receiver. The setup was:
The plots show 3C48 in Kelvins
across lband (.ps) (.pdf):
- Track 3 minutes on and 3 minutes off position
- fire the calon,caloff for 10 seconds in the position off position.
- Use lband wide in linear polarization mode (3C48 .4% polarized at
lband .. from nvss).
- The data was taken near: az=201, za=16. This is close to transit
for the 33 deg dec source.
- Use the mock spectrometer.
- 1 second integrations
- 4*172 Mhz bands covering 1115 to 1735 Mhz (this is the entire
lbw band). band center frequencies were:
- 1200, 1350, 1500, and 1650 Mhz.
- a/d blanking enabled.
- The idl routine masposonoff() was used to reduce the data
- Black is polA, red is polB.
- Vertical dashes show the edges of the 4 bands (there is only
about 20Mhz overlap).
- The first page used the mean when averaging the 1 second records.
The 2nd page used the median.
- Page 1: mean to average the 1 second records:
- Top: When computing the averages, use the entire spectra (no
- the 1200 Mhz polB band is off scale (it is negative).
- Middle: use calOn/calOff to compute frequency channels with rfi
to be excluded. The mask was then applied to all averages.
- the 1200 Mhz polB is not on scale, but it is offset from the
1350 Mhz band
- Bottom: Compute the rfi mask from the calon,off and well as the
position on,off data.
- the 1200 Mhz band now has a smaller offset from the 1350 Band.
- Page 2: same data as page 1 but the median rather than the mean
was used when averaging 1 second spectra.
- Large offsets can occur when the rfi is strong.
- Using no mask in the 1200 Mhz band, the source strength for PolB
- the aerostat radar was (1242,1256Mhz) was stronger in the
calOff than the calOn.
- <calOn -calOff> then became negative so the cntsToKelvins
conversion factor became negative.
- Using calOn/calOff and the data records to compute the rfi mask
did the best job.
- There is still an offset between the 1200 Mhz and 1350
- The jump between the 1350 and 1500 Mhz band around 1400 Mhz may
related to the resonance in the feed around 1398.
- If the cal values used are wrong, then there can also be a jump
between bands. We average the cal value over the entire 172 Mhz band.
- It wouldn't surprise me if the lbw cal values for the 1125-1270
Mhz band were off since the radars were present in the band when the
cals were measured (although 20K/170K= 10% is a bit much..)