L-band

Estimating the noise in the L-band measurements is slightly more complicated than at X-band. Because there are many detectable sources in the field at L-band, the noise cannot be accurately estimated directly from the fluctuations in the visibilities, but must rather be estimated from an image. Because the sources are sufficiently strong, the image must be deconvolved, and Dwaraka and I have both been using CLEAN (via MX, or, after the A configuration observation in 1995, IMAGR) to do the job. Table 5 shows values of the pixel-to-pixel rms variations in the resultant images for each channel and IF, for all observations prior to the D configuration in 1995. As mentioned in the introduction, these observations were done in 1 IF spectral line mode (switching between modes 1A and 1B), bandwidth code 0, with Hanning smoothing. This yielded 7 channels of 6.25 MHz each, in 1 IF at a time. The two IF's were centered at 1464.9 and 1385.1 MHz, respectively. I could find no noise numbers in Dwaraka's notes for the A configuration experiments of 1994, which is why they are not present in Table 5. Note that the absolute values of the noise should not be compared from IF to IF or from different experiments, since different numbers of visibilities go into each image. However, it is clear from the channel to channel variations that channels 4 and 7 of IF 1 and channels 6 and 7 of IF 2 are consistently high. This is interference, and will be discussed later.
The 1995 observations were all done with at least half of the L-band data taken in continuum mode. During the D configuration observation, some data were taken in 2 IF spectral line mode (mode 2AB), bandwidth code 0, with no Hanning smoothing. This again yielded 8 channels of width 6.25 MHz each, but in 2 IF's simultaneously. Table 5 shows the rms values from that portion of the observation. Also, the central frequencies of the IF's were changed to 1364.9 and 1435.1 MHz, to avoid the interference mentioned above, and to be compatible with the default observing frequencies. During the A configuration observation, some data were taken in 4 IF spectral line mode (mode 4), bandwidth code 0, with no Hanning smoothing. This yields 3 channels of width 12.5 MHz each, in 2 IF's, and in Stokes LL and RR simultaneously. Table 5 also shows the rms values for this data. These different spectral line mode observations were intended to be used as a comparison to the continuum data, to assess the performance of these relatively wide band spectral line modes vs. that of the continuum mode and hence decide in which mode the standard field observations should be done in the future. During the B configuration observation, all L-band data were taken in continuum mode. For this observation, the center frequency of IF 2 was moved up to 1485.1 MHz, which is the frequency used by the L-band survey. Table 5 shows the rms values for all of the continuum data, which are denoted by the asterisks. For the I, Q, U, and V Stokes, an image was made in which the 2 IF's were averaged, which was subsequently CLEANed (if necessary), and from which the rms variation was taken. The RL Stokes images were made with only IF 1.
Also shown in Table 5 is the inferred value of Rick's $K$, for all of the observations. For the observations prior to the D configuration of 1995, only those channels not affected by interference were used in this estimate. Since all of the observations are done at $\sim 35^\circ$ to $40^\circ$ elevation, and there is a well documented variation of $T_{\rm sys}$ with elevation at L-band (see Lilie 1994, and Bagri 1993), the value of $K$ needs to be corrected for that effect. The value of $T_{\rm sys}$ increases by a factor of $\sim 1.3$ from zenith to these low elevations, so the inferred values of $K$ need to be multiplied by $\sim 0.77$ to get the value of $K$ at zenith, which I have denoted as $K^*$ in Table 5. The value of $K$ supplied in the OSS is 7.7 mJy (note that this was the value in the 1994 OSS, and has been changed to 9.1 in the current OSS), which is very close to the values listed in Table 5, excepting the 1994 C configuration value, and the 1995 D configuration value (where confusion is starting to contribute to the ``noise''). So there is no problem similar to X-band in our published sensitivities at L-band. From the value of $K^*$, the value of $T_{\rm sys}/ \eta_a$ at zenith can then be obtained from: $T_{\rm sys}/ \eta_a = K / 0.1217$ (different from above, since the correlator efficiency in spectral line mode is $\eta_c \sim 0.77$). Using the nominal value of $\eta_a = 0.51$ at L-band gives values of $T_{\rm sys}$ near 30 K, which matches the engineering measurements at zenith. Again, no problem like that at X-band.

Table: L-band Standard Field noise measurements (map based)

date	config	IF or	1	2	3	4	5	6	7	$K$	$K^*$	$T_{\rm sys}/ \eta_a$
		Stokes								(mJy)	(mJy)	(K)
		1	135	134	138	142	142	143	155
1/1/93	A									10.0	7.8	64.1
		2	137	138	142	142	146	149	162
		1	120	122	124	134	129	127	146
3/29/93	B									9.5	7.6	62.4
		2	131	133	137	135	137	156	163
		1	130.3	132.8	134.4	153.9	136.5	131.9	152.7
8/21/93	C									9.2	7.2	59.2
		2	115.7	120.1	126.5	120.8	123.7	155.7	158.8
		1	172.7	168.2	168.2	276.3	174.7	173.7	255.9
11/24/93	D									11.4	8.8	72.3
		2	184	177	178	187	187	341	316
		1	127.7	128.5	131.5	146.2	138.4	139.8	159.3
8/19/94	B									10.4	8.1	66.2
		2	140.5	141.1	144.3	143.7	145.6	192.4	198.2
		1	186.6	188.4	192.9	240.9	194.2	200.4	246.1
11/12/94	C									14.7	11.4	94.1
		2	237.1	243.2	242.1	242.0	244.7	296.9	287.7
		1	303.8	324.8	266.8	296.3	295.0	305.8	326.3
3/18/95	D									19.6	15.1	123.9
		2	346.8	352.0	320.4	316.1	303.6	323.2	359.3
3/18/95	D$^*$	I	172.9							41.9	32.2	265.0
3/18/95	D$^*$	Q	68.57							16.6	12.8	105.2
3/18/95	D$^*$	U	81.38							19.7	15.2	124.9
3/18/95	D$^*$	V	62.2							15.1	11.6	95.3
8/9/95	A	1+2 (V)	62.4	75.7	81.5					11.2	8.6	70.5
8/9/95	A$^*$	I	146.3							50.2	38.6	317.0
8/9/95	A$^*$	Q	32.2							11.0	8.5	69.8
8/9/95	A$^*$	U	31.9							10.9	8.4	69.1
8/9/95	A$^*$	V	33.8							11.6	8.9	73.2
8/9/95	A$^*$	RL	63.8							10.9	8.4	69.1
10/27/95	B$^*$	I	100.7							34.5	26.5	217.6
10/27/95	B$^*$	Q	36.9							12.6	9.7	79.8
10/27/95	B$^*$	U	38.1							13.0	10.0	82.4
10/27/95	B$^*$	V	45.0							15.4	11.8	96.9
10/27/95	B$^*$	RL	71.0							12.2	9.4	77.2

$^*$ continuum observations

As far as the interference in the early line observations is concerned, there is no particular mystery surrounding it. The interference in IF 2 was caused by the well-known and documented internal birdie at 1400 MHz (see Crane 1982, Perley et al. 1983, and Janes 1995). Since the frequency responses of both channel 6 and 7 were significant at 1400 MHz, the interference was picked up in both channels (see Figure 5). The interference in channels 4 and 7 of IF 1 were probably caused by U.S.F.S. microwave transmissions (see Janes 1995). The ``channel edges'' of channel 4 were 1461.775 and 1468.025 MHz, and of channel 7 were 1480.525 and 1486.775 MHz, which picked up two of the U.S.F.S. microwave transmission frequencies. The interference in IF 2 was much stronger than that in IF 1, evidenced by inspection of the images. The interference showed up in the images as striping, but at a much lower level in IF 1. As a matter of fact, if you averaged the 7 channel maps into one map, in IF 1 the interference stripes were at the level of the noise (you couldn't see them by visual inspection). In IF 2, this was not the case, and in the averaged map, the stripes were clearly present. The significant thing about the interference in IF 1, in my opinion, was in the repeatability of the effect. This was not intermittent interference, but seemed to be present in every observation.

Figure: Frequency response for channels 6 and 7 of IF 2 in the L-band standard field observations. The 1400 MHz birdie is also shown.