meeting 23 Nov 04
Silva Laing Murowinski Shepherd Commissioning meeting
Week of 6 Dec. Computing face-to-face.
WVR radiometry won't really produce much good data. Mark: go with 90 GHz early. I could
do useful obs. I'd like multifrequency obs. It is important to do dual frequency obs
and to test these. DTE: Changing frequencies would be very difficult with the evaluation
receiver. To switch in seconds between bands may be doable but it isn't clear if the
equipment will be at the ATF to do multifrequency tests. DTE thinks this could be done with
relatively little effort in the preproduction receiver.
AI: MH tests on dual frequency operation.
AW gets data on cartridges and when.
Could preproduction rx come sooner if only B3 is needed?
CSV: Astronomical holography will be a bit limited by the one baseline available.
MH: 30s one sigma 5-10 mJy. 3C273
DTE: Need 100 000 points for a map, so it is non-trivial. For large scale structure
one could get by with 16 x 16 map, 1-2m resolution, which is doable. But for setting
panels this could become impractical. I think it probably isn't useful, esp if one
wants it on the scale of a panel. This needs further discussion. Another issue--is
there good visibility of the sky from the OSF? For some pads this is limited.
Radford: will visit tomorrow.
AI: where are the pads at the OSF? Is there fiber to them.
Baseline ripple and quantization noise.
Gain mod before digitization. The process of digitizing introduces additional noise.
Added noise is approximately flat across the passband. If there is modulation before the
digitizer, the added noise is more significant at the trough of a ripple than at the peak.
After it goes through the correlator and is normalized, S/N will not be constant. One can
work out ripple size and determine variation across passband. If we didn't mind the noise being
a factor of two worse in some places we could have a gain modulation of 10dB. If we wanted
less than 0.5 - 1 dB then we have constraints set. Current spec results in 2dB variation from
the current spec. A signal of 2sigma in one part of the passband is 4 sigma somewhere else.
If we said scientifically that 1dB was enough, that would imply 4dB of pre-digitization ripple.
DTE: We should say we need no more than 1dB change in S/N, we could prevent any further
degradation. This would give more credibility to the spec.
BG: I think eVLA has tighter spec achieved with extra filtering.
DTE: Added noise is cumulative. Most of the noise comes at the first quantization rather than
SM: 1.5 dB is a factor of root two, 1 dB is probably reasonable.
DTE: And this isn't too far from the spec we now have.
AI: AW - Propose some wording with some reasoning behind it.
- 23 Nov 2004