[evlatests] 3bit test summary

[Vivek Dhawan]

In terms of SNR, the new modules are not dramatically better than
the old. They do appear more stable, and the bandpass and cross-pol
look like the 8-bit. After chewing over the recent data, it looks
like the best 3-bit units are still 10% worse than the 8-bit, and
22 R&L is about 20% worse. The exact cause remains elusive.

The attached plot shows the breakdown by baseline (x-axis) with RCP
top, LCP bottom. red/black=8bit. blue/green=3bit. 2 datasets each.

Procedure: Nod between calibrators + blank field. Two 8-bit datasets
were acquired of 30minutes, and two 3-bit of 1 hour, all 5488-5616 MHz.
Calibrate amplitude scale on 3C147=0542+498 = 7.03 Jy at 5.424 GHz.
Apply calibration to J0217+7349 (4 Jy) and blank field. Extract
statistics of noise on blank field, for each baseline/pol.

Signal levels, gain slopes, and RQ gains were adjusted. Plots of
sampler histograms & the 2GHz spectra were made. Going over these,
it now appears 15R got mis-adjusted but does not show degraded noise.
22 R&L looks well-adjusted, yet it has lower quality.

The plan is to repeat this, with better adjusted gains
(a) with RSRO-type setup to cover 1 or 2GHz
(b) at K band.
(c) examine tsys and Sw.Power

However, I don't think the bottom line will change much: the 3 bit
data path is, at best, 10% worse than the 8-bit.

Vivek.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: uvhg1.pdf
Type: application/pdf
Size: 5133 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20110419/07acd2fe/attachment.pdf>