Sections:

SEE ALSO: monitoring alfa stability using TsysPolA/TsysPolB
 

---

Intro

• The median power level settting for each strip. The units are Measured/Optimum power. The values should be within 20% of unity for optimum sensitivity with 3 level sampling.
• The stability of the system is monitored using the median strip power for each strip taken during the day. The power by strip, the cal scale factor for each strip, and the cal corrected power for each strip are plotted. The first two show how fast the electronic gain is changing. When multiplying these two, the 3rd plot (cal corrected power) should be stable with time.
• The total power for each 1 second  sample is plotted for the first 10 strips of each day. You can see sources drifting through the beams. Occasionally you will see jumps in total power with long recovery times. These are caused in our system by the FAA radar. Some of the pixels also look "noisy" then the others.

---

How the data is processed: (top)

1. compute the rms/mean by channel for n secs (600) seconds worth of data. There must be a cal record in the file for the file to be processed.
2. Do a robust linear fit to the rms/mean by channel throwing out outliers. The remaining channels will be the mask used to compute the total power for each spectra of the 600 seconds (and the cal).
3. compute the  total power (mean) for each 1 second record using the above mask.
4. Over the entire strip compute the median of the total power using the data from 4.
5. Compute the cal scale factor : calK/(calOn-calOff). For the cal off use the last record of the strip (the one before the cal on). If the next strip starts within 6 seconds of the cal record, include  the first record of that strip for the calOff. If there are two cal offs,  interpolate their total power to the cal On time (if not, just use the single cal off.
6. Scale the total power data (from 4) to kelvins using the cal scale factor.
7. Save the following to disc (one save file per day). assume there are NstripsTot total strips:
• tpIAr[nstripsTot] (tpI struct), brmsAr[nstripsTot] (corget struct) ,bcalRAr[nstripsTot] (corget struct), and maskAr.

TpI struct:

   SCAN            LONG         503514951
   FNAME           STRING    '/proj/a2010/wapp.20050204.a2010.0007.fits'
   NPNTS           LONG               600
   RMSFITA         FLOAT     Array[2, 7]   (ao + a1*channum) fit torms polA
   RMSFITB         FLOAT     Array[2, 7]   (ao + a1*channum) fit to rms polB
   MASKFRACT       FLOAT     Array[2, 7]   (fraction of band each mask contained)
   TPCORA          FLOAT     Array[600, 7] (tot power deg K each point (polA)
   TPCORB          FLOAT     Array[600, 7] (tot power deg K each point (polB)
   TPMEDIAN        FLOAT     Array[2, 7]   (median tot pwr each strip cor Units (a,b)
   CALK            FLOAT     Array[2, 7]   (calval degK for [pols,pixels])
   CALSCL          FLOAT     Array[2, 7]   ( calK/(calOn-calOff) tp [pols,pixels]
   NCALOFF         FLOAT           2.00000 (number of cal offs used)
   AZ              FLOAT           359.638 ( azimuth feed)
   ZA              FLOAT           9.22350 ( za )
   JD              DOUBLE           2453405.3 (jd start of strip)
   ALFAANGLE       FLOAT           18.9999 (alfa rotation angle)

•  bcalRAr . This is a corget struct holding the calon/caoff spectra for each cal
•  maskAr . This holds  the masks used to compute the total power.

The save files can be input and merged using inpsav.pro. This routine creates one large array of tpI structs. It only takes strips that have 600 seconds worth of data.

---

Median power for each strip. The system stability: (top)

    For each month the median power, calscale factor, and Tsys are plotted for each strip taken. The four sets of plots are:

• Col 1: The median power for each strip versus date.  The plots span 2 pages. Data is in correlator units (measured/optimum power level). It has not been scaled by the cals. This data shows when the data was taken and whether the power levels setting of the wapps were ok.  The  dashed line is at +/- 20% from optimum (1 db off). The power level can be adjusted in 1db steps at the wapp input. The optimum value  should be 1. .
• Col 2: The median power level of each strip is plotted versus the strip number for the observing run. Each plot is divided by the median value for the observing run (and 1 subtracted) to show the fractional change in power. This data has not been scaled by the cals so changes in level could be from sky, gain variation, or a change of the attenuator setting during a run. Each pixel is plotted on a separate page. Consecutive days are offset for plotting with the yymmdd label on the right. The dotted lines are spaced by 5% of Tsys.
• Col 3: The cal scale factor for each strip is plotted versus the strip number for the observing run. This value converts from correlator counts to degrees K for each strip. The data is computed as :  CalK/(calOn-calOff) .  Each plot has been divided by the median value (and 1 subtracted) to show fractional changes. Variations in this value will occur because of gain variations or sky changes between the calOn and the calOff (sources drifting through). Each day is offset for plotting and labeled with yymmdd on the right.
• Col4: The median strip power corrected using the cal value (from col 3) is plotted versus stripnumber of day. The multiplication by the cals should get rid of the electronic gain variation of the system. The median power is an estimate of the average value of the power in a strip. The cal scale factor is measured at the end of the 600 seconds. If there is a linear gain change then the plot will be flat (but offset). If there is a non-linear change in the gain then there will continue to be curvature in these plots.

• jan08 alfa reinstalled after painting..

• apr07 alfa brought down to lab for maintenance.
• mar07 stripPwr vs stripNum Beam 5 shows clear temperature dependence. Probably because data taken in early morning.

---

1 second total power samples vs position in strip  (top)

• 080919: finished modifying amp 3 bias voltages for long recovery times. Looks like this has made beam 4b less stable. It now looks like one of the noisier beams.
  
• 080317: jumps > 4% with long recovery times many beams,pols.
  
• 080328: beam 3 dropouts both pols (wapp problem??).
  
• 080331: beam 5a jumps up 1% stays for a few 100 secs then jumps back down.
• 080816: beam 6b dies.
  

• 05mar06 scan 606477112 beam 6 jumps polB and then polA
• aug06: 18,19,20,22,23 beam 6b power is drifting/jumping 2 to 5%.
• sep06 lots of jumping/drifts.
• 03sep: 5b
• 05sep: 6b
• 08sep: 4b
• 09-14sep: 5b (large jumps, drifts) on 09 and 12 sep. On 11sep06 the 12 second radar seem to be magnified for 5b.
• 16-17sep: beams 3a,4a,5a start drifting together.
• 30sep: beams 3a,4a,5a drifting together.
• 08nov: amp on beam 1b died again
• 22nov: warmed up/ cooled down alfa dewar. beam 1b back

notes 2005:

• 06apr05: pixel 2a,b had 10 db amp installed before 1st mixer on 05apr05.  06apr first day of data.
• 08apr05: 10db amps were placed in front of 1st mixer for all pixels on 08apr05 during the day

---

The electronic gain change versus turret room temperature:  (top)

    The median power for each strip versus Room temp(.ps)   (.pdf):

•  The temperature in the gregorian room was interpolated to the time when the cals were measured. The median total power of the strip (in cor units) was then plotted versus the room temperature. An offset of .05 for each day was added for plotting.  * is polA while + is polB. This data has not been scaled to the cal deflections.
• You can see that the temperature for the first 11 days had very little temperature change. The outside air temperature was less than the room air conditioner setting,  so the AC never had to turn on. The last 14 days have  larger temperature change with a corresponding change in the power.  The linear ramp shows the temperature dependence of the electronics.
• The gain of the alfa receiver is changing with the gregorian receiver room temperature by  up to 12% over 5 degreees F. This is probably coming from the 1st mixer/fiber optics electronics.
• Applying the cal deflections will remove this change. Since the cals are only measured once per strip you might want to interpolate the cal correction between the adjacent strips (assuming you don't change any of the attenuator values.

 home_~phil