A/D rms values vs mixer gain & compression

24apr08

The A/D rms values versus mixer gain were measured on the sky with alfa to see how much range we had around the optimal operating level (an A/D rms of 30 counts). The A/D rms counts were also measured using a noise source to check the dynamic range of the mixer/AtoD combination.

Terminology:
Gain vs attenuation in the mixer chassis:

The mixer chassis has a programmable attenuator that can be set from 0 to 31 db. The chassis also has a fixed gain of about 14 db.
When programming the mixer chassis you can set the gain from 0 to 31 db. I've used this variable for the plots. Setting the gain variable 0 to 31 db is actually setting the attenuators from 31 to 0 db of attenuation. I've italicized gain when referring to the setting of this variable.
The mixer chassis gain vs the gain variable is:

gain variable=0 : bbmchassis gain: 14db - 31 db Attn = -17db of gain
gain variable=31: bbmchassis gain: 14db - 0 db attn = + 14 db of gain.

Measuring A/D rms counts vs mixer gain on the sky:

The setup and processing was:

Alfa wide band mode on the sky. Cfr 1370 MHz. LowB rfFreq was1445 MHz and HiB rfFreq was 1295 MHz.
Telescope sitting at az=270,za=10.
Adjust the A/D rms 5 thru 50 counts (in steps of 5 counts) recording the mixer gain at each step.
Adjust the A/D rms 10 thru 500 counts (in steps of 10 counts) recording the mixer gain at each step.

The plots show the A/D rms counts vs bbm gain. Each color is a pair of A/D converters (I,Q). Red in lowband polA, green is lo band polB, Blue is hi band polA, purple is hi band polB.

On the sky A/D rms vs mixer gain (5 thru 50 counts) (.ps) (.pdf). Each panel is a separate beam of alfa. The dashed blue line is the rms that we will typically use for a measurement (around 30).
On the sky A/D rms vs mixer gain (10 thru 500 counts) (.ps) (.pdf). The maximum gain is 31 db. Not many beams arrived at rms=500. The jumps in the measurements come from the radars.

Measuring A/D rms counts vs mixer gain using a noise source

The downstairs if/lo noise source was used to measure the linearity of the bbmixer , A/D system. The setup was:

The if2 noise source was selected with +30 db gain. The power into the mixer chassis was about -17 dbm.

When the gain variable was set to 0db, the chassis output power was: (-17dbm -17db chassisGain)= -37 dbm.
When the gain variable was set to 31db, the bbm output power was (-17dbm + 14db chassisGain) = -3 dbm. (assuming no compression).

The A/D rms was set 25 to thru 1000 counts in steps of 25 counts.
Rms counts vs bbm gain using the noise source (.ps) (.pdf):

Each plot is a separate spectrometer box.
The maximum bbm gain is 31 db.
None of the boards got to an rms of 1000 counts with the given inputs.

Compression vs bbm gain levels (.ps) (.pdf):

A Linear fit was made to 20*alog10(a/d rms) vs bbmGainDbs and then subtracted from the data.
The plots shows where the measured rms deviates from linearity. This occurred at a gain value of about 22 db.

Compression vs A/D rms counts (.ps) (.pdf):

The same linear fit of 20*alog10(a/d rms) vs bbmGainDb was plotted versus a/d rms values.
This shows how large an rms we can use before compression sets in.

Mixer levels at 1db compression:

The table below shows the computed power levels at stages in the mixer chassis. I've used the lowband with the 250 Low pass filter and the 500 Mhz noise input bandwidth. I've used attn=3 (gain=28) w(close to the 1db compression).
Computation uses lowband(250 lowpass filter) and (gain=28 or attn=3)

Input  Output  Component
-17.0   -21.0  H1 splitter
-21.0   -25.0  LPF1
-25.0   -29.5  H2 ProgAttn set to:  3. db (gain=28)
-29.5    -7.5  H3-1 Amp
 -7.5   -12.5  U5 Attn
-12.5     7.5  H3-5 Amp
  7.5     3.5  H1-2 Splitter
  3.5    -0.5  U4-Attn
 -0.5   -11.5  H4 Mixer
-11.5   -19.5  U3 Attn
-19.5   -22.7  LPF2
-22.7   -28.7  R2 attn
-28.7    -8.2  H5-1 Amp
 -8.2   -17.2  U1 Attn
-17.2    -2.2  H6-1 Amp
 -2.2    -8.2  U2 Attn

The measured output level from the mixer chassis (beam5a low band) was -4dbm so there is 4db more gain in the mixer chassis than expected.

Conclusions:

On the sky there is a large variation (7db) in beam0 polA and polB which is caused by the failure of beam0 polb.
Gain levels needed for a standard A/D rms level of 30 counts ranges from 7 to 15 db for the various beams. This is an ok range to be in (unless we end up sitting on a large continuum source).
The rms (blank sky) at maximum gain ranges from:

500: polA low band beam 1
200: polB hi band beam 6.
This is spread of about 8 db in the beams (I've excluded bm0b).

The gain levels for beam 5 are surprisingly equal for the high and low bands, polA and polB.
The bbm, A/D system linearity:

All chassis start to go non linear with a gain setting of 22 db. The 1db compression is around gain=28.
The 1db compression has an output power of -17dbm +9db chassisgain=-7 dbm. (low band)
This corresponds to an A/D rms of 400 counts.
The maximum rms was in the 800 to 1000 a/d count range.
The high band is compressing before the low band. The power Level in the Hiband is 2db higher than the power in the Low band. This occurs because the low band has a 250 Mhz low pass filter while the hi band has a 400Mhz Lowpass filter (and we're inputting 500 Mhz of noise).

processing: x101/080424/chkadjp.pro,chkadjpNoise.pro

<- page up
home_~phil