As reported in the previous note, the attenuators were set up on Cygnus A. (This was arranged by using an X-band dummy scan as the first observation). The resulting PSum values -- on Cyg A and on 3C295 (representing 'hot' and 'cold' sky) are given below in columns 2 and 3. Also listed is the ratio (a measure of the sensitivity) in column 4, the PDif compression , defined as: PDif(CygA)/PDif(3C295) in column 5, and the Psum value from last week's experiment (taken with the same correlator setup) when the power levels were based on the 3C295 observation, in column 6. I use 'R' and 'L' here, as I've yet to be told which one corresponds to the horizontal or vertical dipole channel... Ant. P(Cyg) P(3C295) Ratio PDifcomp OldP(3C295) -------------------------------------------------------- 2R 33 9.2 3.6 Var 18 2L 27 7.4 3.6 Var 14 3R 47 13 3.6 .99 20 3L 13 3.7 3.5 1.0 6 4R 29 8.9 3.3 .98 17 4L 61 19.7 3.1 .88 37 5R 18 5.2 3.5 .99 11 5L 19 5.6 3.7 .99 11 6R 36 10.6 3.4 .97 19.6 6L 17 4.7 3.6 .94 4.7 7R 40 13.6 2.9 1.0 21.6 7L 34.2 9.8 3.5 1.0 19 8R 40 13 3.1 .97 33 8L 42 12.9 3.3 .97 25 9R 32 9 3.6 .98 22 9L 39 31 *** 1.3 .97 24 ***Hugely insensitive 10R 37 11.1 3.4 Zero 22 10L 19.2 5.5 3.5 Zero 5 11R 17 4.9 3.5 Zero 7.4 11L 33 8.9 3.7 Zero 21 12R 10 4.2 2.4 Var 4.2 12L 14 5 2.8 Var 10 13R 30.7 9.7 3.2 Zero 19 13L 17.6 5.1 3.5 Zero 9.6 14R 41 13.3 3.1 .98 34 14L 27 8.5 3.2 .96 5.7 15R 34 11.6 2.9 .99 18.2 15L 4.1 12.9 3.2 .98 19.7 16R 23.8 7.7 3.1 1.0 12.4 16L 30 8.5 3.5 .96 5.8 17R 9.0 2.95 3.1 .93 3.8 17L 12.3 3.9 3.2 .83 5.4 18R 30 10 3.0 1.04 (!) 20.8 18L 35 14.7 2.4 .91 18.1 20R 21 6.9 3.0 .96 16.2 20L 25 7.8 3.2 .92 13.4 22R 24 8.1 3.0 .99 16.2 22L 35 11.2 3.1 .99 8.9 23R 30 10 3.0 .96 12.2 23L 7.3 2.3 3.2 .92 3.5 24R 24 6.7 3.6 1.01 13.6 24L 34 9.1 3.7 .98 18.0 25R 17 7.5 2.3 Zero 14.0 25L 35 10.9 3.2 Zero 16.8 26R 31 8.8 3.5 .96 27.5 26L 18.3 4.95 3.7 .98 8.1 27R 30/22 14/2 2.1 Zero 23 27L 28/20 8.6 3.4 Zero 22 28R 38 10.7 3.6 .97 21 28L 28 7.5 3.7 .98 14.6 ----------------------------------------------------- Some observations from this table are easy to extract: 1) The power level on Cygnus A (following the 'set-and-remember') is not the same as on 3C295 in the earlier experiment, where 3C295 was used for 'set and remember'. I had naively expected them to be similar. For virtually all antennas, the power level on 3C295, when it was used as the power reference, is between the power levels on and off Cygnus A, when Cyg A was used as the reference. The difference is presumably related to the increase in the primary beam at lower frequencies. The power level setting is based on the wide-band total, integrated over the entire receiver's bandwidth. Cygnus A is located in the galactic plane -- there is much bright emission from around it. The primary beam is 3 to 4 degrees in size at the lower end, whereas Cyg A is a measley 2 arcminutes. So the effect of Cyg A's power is increasingly diluted by the lower frequencies -- which is where the strongest emission is located. 2) The values of PSum, when on Cyg A, show much wider variance between antennas and polarization than I expect. Values between 60 counts (ea04L) and 7.3 (ea23L) are found -- nearly an order of magnitude... 3) The ratio of Cyg A power to 3C295 power (in the new experiment) is related to the antenna sensitivity. The ratios (column 4) are spread wide, but top out at 3.7. The sensitive antennas, having ratio of 3.5 or more, are: ea02, ea03, ea05, ea06, ea11, ea24, ea26, and ea28. (These are the 'poster children' for the system). Many others are close. The poor performers (ratio < 3.3 on both polarizations) are: ea04, ea08, ea12 (also has variable powers), ea14, ea15, ea17, ea18, ea20, ea22, ea23, ea25, and ea27. Most of these are near 3.3, so should not be considered 'bad'. 4) PDif compression (the apparent, but possibly false, reduction in system gain when observing a strong source) is less at this band than at the higher frequencies. The values are in the 5th column. Compression is generally less than 3%. Values listed as 'Zero' mean there are no non-zero PDif values (noise diode not switching). Finally -- what is the flux density of Cygnus A, when calibrated by 3C295/3C48? As reported last week, the values are quite far below the accepted, known values. At 300 MHz, the discrepancy is about 20%. At 400 MHz, it is definitely less, about 10%, and probably less than this at 450 MHz (harder to tell here -- there are very few RFI-free channels). This is far, far greater a discrepancy than can be believed due to errors in the supposed flux densities of the sources involved. An explanation arising within our system is more likely. To provide more information, I used an old and very good 327 MHz image of Cygnus A (taken in the 1990s), to enable a self-calibration gain table to be made. The goals were to see if there were variances over frequency or antenna (or polarization) in the shortfall. I used the Baars et al. expression for the 'correct' flux density of Cyg A. Solutions were made for each subband and antenna-polarization. The only visible trend is that mentioned above -- the discrepancy clearly increases with decreasing frequency. All antennas are affected about equally, although there is considerble variance. There *may* be a correlation with Psum. The highest power antennas (ea04L, ea07L, ea18L) have the greatest flux shortfall, while low-power antennas (ea17L, ea23L, ea26L) have clearly less shortfall. If so, perhaps we are saturating the 4-bit requantization. This setting was not part of the 'set and remember' regimen, but I had thought the appropriate levels were derived long ago which would prevent this effect. Perhaps we should set the levels lower? Walter offers the concept of self-noise. But the major effect of this is to increase the noise, not lower the cross-correlation. Furthermore, the Cygnus A antenna temperature, although significant, is not greatly dominant (such as it is for mega-masers). We wonder about the so-called 'van-vleck' correction. This was a big deal for the VLA's correlator, but with 4-bit correlation in WIDAR, the effect should be small -- far less than 20% (says here). Furthermore, this effect is strongly dependent upon correlation coefficient -- but the correlation coefficient at the low end of the band for Cyg A is actually less than the high end. (This is because the sky antenna temperature is rising much faster with decreasing frequency than is the Cyg A antenna temperature -- an effect caused by the increasing primary beam size, while Cyg A is the same size at all frequencies). Perhaps our next test should be to further lower (by artificial means) the input power to the samplers, and/or reduce the subband power to the requantizers, to judge whether the non-linearity is within the digital regime. Suggestions are welcome!!