Equalizing gain to group0,group 1 spectrometers
16feb09
19jan09:Before attenuators installed.
16feb09:After attenuators installed.
Intro:
The pdev spectrometer is split
into two groups of 7 boxes. Group 0 is the primary observer and group 1
is the commensal observer. The signal path is:
- 250 IF -> variable attenuator -> mixerToBaseband -> amps
-> split signals to go to grp 0 and grp 1.
There is only 1 variable attenuator (per freq band) that is used for
both group 0 and group 1. When adjusting the power levels, the
attenuators are set using the
values from the group 0 read back. Any gain differences after the
baseband output split will cause the power levels for group 1 to be a
little off.
Grp0,1 gain dif before attens added.
Data was taken on 19jan09 to measure the gain
difference after the baseband splitting of the signal in the bbm. This
was prior to inserting 1db pads to equalize the gains: The
setup was:
- Use the if2 noise source as the input
- adjust the power levels for group 0 to have and rms count of 30
- measure the rms and mean for group 0 and group 1 about 50 times.
The plots show the results of
the rms, mean measurements (.ps) (.pdf):
- Page 1: Rms measurements
- there are 14 bands with group 0 band (b0s0g0) followed by group
1 band (b0s0g1).
- s0 is the low band, s1 is the hi band (looking at the front
of the pdev boxes)
- For each band there are 4 values for the 4 digitizers. The
order is:
- i0,q0,i1,q1 (0 is polA, 1 is polB)1 Db pad
locations that are needed
- TOP: rms measurements in counts (target value was 30).
- Black is group 0, red is group 1.
- 2ND: db's that the measurement was from the target value (30)^2
- Positive means that the value was higher than expected.
- Group 0 (black) was adjusted via the attenuators and the
group 0 read back. The attenuators have a 1 db resolution.
- Group 1 (red) values result from the attenuator setting for
group0 and any gain difference in the group 1 path.
- 3rd: Power ratio group0/group1.
- This is the ratio of the group0/group1 measured values.
- The dashed green line if +/0 -5 db.
- Values outside of .5 db should have a 1db pad placed in
either group 0 or group 1.
- Locations where 1db pads are
needed:
- b1s0g0: dig q1 (beam 1 (grp 0) lowband Dig Q1)
- b1s1g1: dig q0 (beam 1 (grp 1) highband dig Q0)
- b2s0g0: dig i1 (beam 2 (grp 0) lowband dig I1)
- b3s0g0: dig q0 (beam 3 (grp 0) lowband dig Q0)
- b6s1g1: dig q0 (beam 6 (grp 1) highband dig q0)
- Page 2: digitizer offsets
- TOP: measured offsets in a/d counts (the dashed vertical line
is the start of the new band).
- 2ND: a/d offsets in millivolts
- 3RD: the a/d offsets after the cal.conf correction.
- The measured cal.conf offsets (top plot) are subtracted from
the output of the a/d. t The plot shws the measured a/d offsets after
this subtraction. These are the offsets that the spectral computations
in the fpga see.
Grp0,1 gain dif after 1db pads added.
The 1db pads were inserted in the digitzer path
specified above on 20jan09. Data was taken on 16feb09 (after
b6s1g1 dig i0<->q0 were unswapped inside the spectrometer box).
The if2 noise source was used. The cal.conf file was generated
before taking the data. Data was taken as described in 19jan09.
The plots show the results of
the rms, mean measurements (.ps) (.pdf):
- Page 1: the rms values in adc counts. The bottom plot shows that
the ratio of grp0/grp1 power is now less that abs(.5db).
- Page 2: the digitizer offsets: after the cal.conf file correction
the dc offsets are less that .5 digitizer levels.
processing: x101/090119/getrms.pro
<-
page
up
home_~phil