[Gb-ccb] recent ccb lab test results

Brian Mason bmason at gb.nrao.edu
Wed Dec 7 13:58:07 EST 2005 

We have been hard at work in the lab for the past week or so but a little 
remiss about sharing our results. Below is my $0.02. Randy et al can fill 
in a bit more on the hardware details this afternoon.

  cheers,
   Brian

I. Levels

There were some concerns about the stability of power levels seen at the 
CCB (partly due to the erroneous misidentification of a scale inversion 
for variable offsets, but also due to real differences in the detector 
amplifier output levels seen when these were spot checked in our 
on-telescope investigations). We have been monitoring this in the lab. 
Two main tests:

A) we set up the receiver to run for ~5 hours over Thursday night with a 
hot load. The mean levels and the cal diode signal as well were extracted 
from the FITS files. See attached file tpLevels02decOvernight.ps and 
calLevs02decOvernight.ps.  Levels were very stable.  J14 cal shows some 
fluctuation; as it's uncorrelated with TP fluctuations it appears unlikely 
to be gain. It could be a circuit offset, or a fluctuation in the power or 
phase balance of that radiometer at that frequency. These could be 
distinguished by examining J10, which I have yet to do (ports with *any* 
phase saturated are completely excluded from my analysis... there are some 
useful phases in the J10 data that I can extract)

B) Every day (Fri/Sat/Sun/late Sun PM/Mon) we collected raw sample 
statistics with Martin's ccb_test_sample_stats program, against a hot load 
again. See attached plot sampsHotAll-swNONE.ps.  Reasonabl stability over 
the weekend is seen, with fluctuations well-correlated between channels 
suggesting RF gain variations or common-mode stuff in the detector 
circuits (ambient temp variations say). The fluctuations are roughly 
consistent with being gain fluctuations not variable offsets-- the peak to 
peak variation in counts of the two low channels (J4 and J8) are roughly 
half that in the other channels.  The dropout of port 1 is due to the 
cable being disconnected during one measurement.

At present the continuum section has been disassembled for spectral line 
work. It should be reassembled shortly, and the real test will be to see 
the levels after that.

II. Phase switch blanking

A number of improvements have been made in the phase switch drive circuit 
which Randy will describe (faster op-amps, adjusted monitor circuit 
levels). These result in needed phase switch blanking times of 8 or 9 usec 
(pending quantitative analysis). A "smoking gun" has been found by 
comparing switch on->off to switch off->on transients in a channel (J1) 
where the switch on and switch off power levels are comparable. It is 
found that the ringing residual anticorrelates in these two cases.  Since 
nothing downstream of the phase switch knows about the polarity of the 
phase switch drive signal this points to the phase switch or its drive 
circuitry (not say the detector intrinsic time response, or bessel filter 
imperfections-- both say driven by the negative power spike during the 
switch transition). Randy believes the observed ringing is due to an 
impedance mismatch between the phase switch itself and its drive circuit, 
and has a design to fix this. It will require a new board to be 
manufactured and we will proceed without it, and 9 usec blanking, at 
present.

See attached plot phiSwDt02dec05.cps.  Red and Blue are two separate 
acquisitions of On->Off transition dump to assess repeatability and noise. 
Green is an Off->On transition. Power levels are fortuitously balanced. I 
would say beyond 9 usec there are no significant anticorrelations 
(blue/green) within the noise level indicated by red-blue-- but 
quantification is needed.

III. Noise vs thermal noise and dependence on beam switch frequency

I acquired 3 scans with the manager against a cold load, with phase switch 
frequencies of 4, 10, and 20 kHz. Differenced data RMS vs the theoretical 
minimum RMS ranged from 1.04 to 1.40. There is a notable improvement in 
noise performance in J1-J8 as the switch frequency increases; however we 
believe that this radiometer has probably been thrown out of (phase or 
ampltidue) balance by recent modifications and are will be rebalancing it. 
Improvement in J9-16 is pretty small.

See plot realTheoryNoise05dec05.ps

Two clear contributors to better results here: better cables. And better 
IDL analysis routine (I was using a single phase's mean level to compute 
the "expected RMS" before, instead of the apropriate combination of phases 
given my differencing-- a few to 10% overestimate of the noise ratio had 
was generated by this error).

IV. Power spectra

I computed power spectra of cold load data obtained late last week, see 
http://wwwlocal.gb.nrao.edu/~bmason/ccb/pspec05dec05/ By way of 
comparison, the earlier measurements (with the receiver) are 
http://wiki.gb.nrao.edu/bin/view/Projects/CcbNineteenOctSpectra

The old data show a good deal more residual RFI (presumably) in the 
differenced data. That was with the old bad cables.  The new cables appear 
to have improved the rejection of this stuff quite a bit-- the new data 
look quite good in differenced power.  There are some fairly strong low 
frequency spikes in the total power data, interestingly, mostly in 
radiometer 1 (worst are: J8, J3, J4, J5; but also J12 and J16).  This is 
worth noting as well, but again given the level, and the cleanness of the 
differenced data, I would not classify as a Real Problem. For instance J3 
is the worst, with one spur of 600 counts variance (25 counts rms). By way 
of comparison the *rms* of the differenced signal overall in that channel 
is 315 counts.

I have but have not circulated comparable spectra obtained with the CCB 
itself, taken in the RFI chamber with no inputs. Those are clean at the 
level we are discussing here, and agree with our "canonical" 0.5 count RMS 
from the CCB itself.



--------------------------------------------------------------------
Brian Mason                           | office: +1(304)456-2338
Associate Scientist                   | fax:    +1(304)456-2229
National Radio Astronomy Observatory  | mail:   PO Box 2 
bmason at gb.nrao.edu                    |         Green Bank, WV 24944
http://www.gb.nrao.edu/~bmason/       |
--------------------------------------------------------------------
-------------- next part --------------
A non-text attachment was scrubbed...
Name: realTheoryNoise05dec05.ps
Type: application/postscript
Size: 15170 bytes
Desc: 
URL: <http://listmgr.nrao.edu/pipermail/gb-ccb/attachments/20051207/6b74e1f2/attachment.ps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: phiSwDt02dec05.cps
Type: application/postscript
Size: 21817 bytes
Desc: 
URL: <http://listmgr.nrao.edu/pipermail/gb-ccb/attachments/20051207/6b74e1f2/attachment.ai>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: sampsHotAll-swNONE.ps
Type: application/postscript
Size: 17059 bytes
Desc: 
URL: <http://listmgr.nrao.edu/pipermail/gb-ccb/attachments/20051207/6b74e1f2/attachment-0001.ps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: calLevs02decOvernight.ps
Type: application/postscript
Size: 18897 bytes
Desc: 
URL: <http://listmgr.nrao.edu/pipermail/gb-ccb/attachments/20051207/6b74e1f2/attachment-0002.ps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: tpLevs02decOvernight.ps
Type: application/postscript
Size: 19131 bytes
Desc: 
URL: <http://listmgr.nrao.edu/pipermail/gb-ccb/attachments/20051207/6b74e1f2/attachment-0003.ps>

More information about the gb-ccb mailing list