I propose a simpler variant of original idea of Gibson & Welch, which could lead to a much simpler implementation of an accurate absolute calibration. Gibson & Welch performed absolute flux measurement using a complex instrumental setup which uses a standard gain horn on the side of one of the BIMA dishes, connected through a waveguide switch to the BIMA receiver. The success of this method, as implemented, makes use of all or several of the following properties, depending exactly on the implementation details: 1) The accurate a-priori knowledge of the gain of the horn, which is well characterized by Schelkunhoff's gain formula 2) An accurate measurement of the waveguide losses using a network analyzer 3) The common receiver gain shared by the large antenna and the standard gain horn after the waveguide switch 4) The proximity of the gain horn and the large dish, which ensures that similar atmospheric conditions are applicable to both signals, provided the switching between the gain horn and the large dish is fast enough 5) An accurate calibration of the antenna temperature scale of the large antenna 6) An accurate calibration of the antenna temperature scale of the standard gain horn 7) An accurate correction for atmospheric extinction 8) An accurate correction of any gain-elevation dependence of the large antenna. 9) A negligible decorrelation on at least 3 baselines on the timescale required to detect with sufficient signal to noise the astronomical calibrator The experiment was done at 28.5 GHz, where the BIMA antennas have no gain-elevation dependence, thus satisfying point (8), and where the opacity is small and can be measured with sufficient precision using a tipping curve, as for (7). Decorrelation (either instrumental or electronic) is also negligible at that frequency, as required in (9). Welch proposes a similar scheme for the ALMA. However, because of the widely different frequencies and atmospheric opacities, this scheme cannot easily be transferred at all ALMA bands. While it is conceivable that the scheme could be simply applied at Band 3 and below, waveguide switches may not be available at all for Band 7 and above. It is worth reviewing why all properties 1 to 8 are required in the original method. The absolute calibration actually produces two different results: a) The absolute gain of the large antenna, and b) The absolute flux of the astronomical calibrator. Result (a) requires a comparison of the (absolute) gain of standard gain horn (Req 1), corrected for the waveguide losses (Req 2), to that of the large antenna. Since the same receiver (Req 3) is used alternatively, at a relatively high switching rate, the ratio of cross-correlation measurements gives the ratio of the voltage gains of the large antenna and standard gain horn, provided the atmospheric effects remain the same (Req 4). Atmospheric decorrelation is assumed to be statistically identical on both measurements in this case (electronic decorrelation being identical, if any). Once Result (a) is obtained, Result (b) can be established in several different ways. - In Method I, the large antenna is subsequently used in Single-Dish mode to measure the flux of one calibrator. This makes use of Req 5, 7 and 8. At first view, it has the advantage of avoiding any potential problem with decorrelation, but this is not true: in obtaining Result (a), the gain of the large antenna is affected by decorrelation linked to some of the antenna mechanics... In addition, it requires good antenna temperature calibration... - In Method II, the standard gain horn is used directly to measure the flux of one calibrator. This cannot be done by Single-Dish mode, and thus requires to isolate the interferometric horn gain (using closure relations on 3 baselines) and make sure decorrelation effects are negligible (Req 9). The gain horn antenna temperature may be easier to measure than that of the large antenna, since a load covering the horn is quite feasible (note that in their original experiment, Gibson and Welch actually used the waveguide switch also to calibrate the antenna temperature scale of the large antenna). Correction for atmospheric extinction remains required. Thus, in addition of Req 1,2, this method makes use of Req 6, 7 and 9. Method II can be used to establish the absolute flux of a strong astronomical calibrator independently of whether Result (a) has been established or not. From the description of Method I, one can sees that the process can be reverted if needed, i.e. one can derive Result (a) using Result (b) and the properties 5,7 and 8. Hence, providing decorrelation can be adequately controlled, it is possible to avoid the delicate switching mechanism between the gain horn and the antenna and yet provide an accurate absolute calibration. All absolute calibration methods require proper correction for the atmospheric extinction. They also all rely on an intermediate antenna temperature scale which is determined using calibrated loads. Applicability to ALMA: ---------------------- ALMA antenna / Gain horn ratio : 3 to 10^4 (assumption) ACA antenna / Gain horn ratio : 10^4 (assumption) Sensitivity issues (Single Baseline, one minute integration, typical weather) Frequency \ ALMA-ALMA \ ALMA-Horn \ ACA-Horn \ URANUS \ ACA SNR \ Required time 90 GHz \ 2.5 mJy \ 0.43 Jy \ 0.74 \ 6.61 \ 9 \ 125 min 230 GHz \ 6.3 mJy \ 1.10 Jy \ 1.91 Jy \ 29.9 \ 15.6 \ 41 min 270 GHz \ 8.3 mJy \ 1.45 Jy \ 2.50 Jy \ 38.3 \ 15.3 \ 43 min 345 GHz \ 12.4 mJy \ 2.14 Jy \ 3.72 Jy \ 55.4 \ 14.9 \ 45 min 670 GHz \ 64.7 mJy \ 11.2 Jy \ 19.4 Jy \ 136 \ 7.0 \ 204 min 860 GHz \ 87.5 mJy \ 15.2 Jy \ 26.2 Jy \ 175 \ 6.7 \ 223 min The required integration time is inversely proportional to the number of baselines which can be used to derive the horn gain. In a proper geometric setup (horn surrounded by a number of antennas), it could easily be divided by 3 to 6 without resorting to baselines affected by significant decorrelation. At the longest wavelengths, using brighter sources like Mars would increase the S/N by a factor 4 or so. However this is obtained by using larger sources. The angular resolution of the ACA is around 9"/100 GHz (that of the ALMA compact configuration is 3"/100 GHz). URANUS is about 3", so at the highest frequencies, some source model may be required, although a very limited one (elliptical uniform disk) is probably sufficient at this level. The strongest quasar are only 20-30 Jy, so using planets is required for sufficient S/N. Proposed Technique: ------------------- It is proposed to use a comparison method in which the gain ratio of the horn and a nearby antenna is measured interferometrically. The proximity of the two receiving systems will ensure that the same atmospheric correction is applied to both, and that the atmospheric decorrelation effects are similar. This setup does not guarantee that electronic decorrelation is identical, but this should be easier to control. The technical setup would essentially be an antenna mount and a special receiver (and receiver cabin) but devoid of any reflector. Each receiver band would be connected to a standard gain horn, rather than to the common optics in the other antennas. Large loads (hot and ambiant, perhaps rather than hot and cold) would be inserted in front of the horns to define the antenna temperature. The receiver cabin should have a special enclosure and receiver fixation system so that the horns can look at the sky without blockage. This "calibration antenna" could be correlated with any other antenna, or either ACA or ALMA, as part of the 16 (ACA) or 64 (ALMA) antennas that the correlator could process. Because of the size of the strongest sources, as well as of the adverse effects of decorrelation, it would be advantageous to do so with ACA antennas which always sit in a compact configuration. Using this "calibration antenna" would only occur occasionally. However, for practical purpose, it would be advantageous that this antenna could be remotely connected to the correlator at any time, because it would guarantee use of the best weather circumstances to perform the absolute calibration. This an be done either as always considering it part of the array (thereby sacrifying one antenna), or by having an electronic switch build in the system. Both LOs and sampled data should be controlled by this switch. An hybrid solution in which the LO is always available, but the sampled data can be analyzed or not is acceptable. Such a dedicated antenna would probably cost less than a fully equipped ALMA antenna. The antenna cost would be minimal (200-300 k$ perhaps): since the mount of one of the prototypes could be re-used for this purpose, only refurbishing of the cabin being required. But the cost of the receiver could be substantially higher than those of the series. A modified cryostat would be required, plus dedicated horns, significant studies and well designed loads. To that, the cost of the automatic switch should be incorporated if feasible. Although this may look costly, it is perhaps not much more expensive than trying to implement switches in one of the antennas to connect horns located at the periphery of the main dish, since the antenna mount is essentially available at no cost. Moreover, it avoids potential problems with the modification of the collision diameter of one antenna. And a cost of 5 M$ is less than 1 % of the cost of the project. I suggest we evaluate the feasibility of such this approach.