I append a digest of discussions among ASAC members today. Clear skies, Al Mark Gurwell wrote: Date: Fri, 21 Sep 2001 16:57:04 -0400 (EDT) Hi Min, I was able to download the m51ha images from the iram site. I took all 10 images: model (ne) model (te) alma only (ne) alma only (te) alma short (ne) alma short (te) all (ne) all (te) hybrid (ne) hybrid (te) Where (ne) signifies "no errors" in the data and (te) signifies "typical errors". I've concentrated basically on the differences between all and alma short for each of the two cases, since what we really want to explore is the improvement brought by the ACA over alma-short alone. Now, for the no error case there is significant improvement; this is the same as found by the iram folks in their analysis (they find roughly a factor of three improvement in fidelity for the highest snr points). However, for the noise case, there is almost no improvement. In fact for the images provided I find (and it appears that the iram folks concur given their plots of image fidelity) that the inclusion of ACA data has very slightly degraded the image over the alma-short case. This is somewhat suprising an disheartening as well, though it may be a case where the imaging techniques are masking the true limit of the data. So, I don't know where to go with this...but it seems to me that the m51 image case will not lead to an example of a clear-cut improvement provided by the ACA. What do you suggest? Thanks, Mark Mark A. Gurwell -------------------------------------------------------------------------- Min replied: Date: Sun, 23 Sep 2001 17:30:06 -0400 (EDT) Hi Mark, I just updated my M51 test page with my new analysis. See if you would agree. I would like to quantify the differences in the ratio images better, but it is not obvious how to make it meaningful. I'm open for suggestions. Take a look at www.astro.umass.edu/~myun/aca/m51test.html. I believe the images I analyzed are based on "typical" errors the IRAM group considers. I should check that. These tests clearly demonstrate the danger of not having the ACA, but I agree with your general conclusion that M51 does not make the critical "make-or-break" case for ACA. -- Min -------------------------------------------------------------------------- Date: Mon, 24 Sep 2001 09:13:23 +0200 (MEST) Hi Mark, Min, and Al, Thanks for the e-mails. The effects that you are finding seem similar to those for the HCO+ images, namely that there is an effect on the line ratios but the improvement is small compared with ALMA + single dish. I am wondering whether we are too optimistic about the errors in the ALMA + SD case, and I recall that Stephane/IRAM was going to run some cases where the SD errors are larger. Al: have you heard anything from this yet? I think it would be useful to organize a brief telecon with the ACA tiger team by the end of the week, e.g. Thursday, to see where we stand. Cheers, Ewine -------------------------------------------------------------------------- Al replied: Date: Mon, 24 Sep 2001 09:50:10 -0400 (EDT) Hi Ewine I haven't heard from Stephane. I think a telecon would be very useful, as Mark has some surface error simulations we haven't discussed. I decided to cancel my attendance at the RAL meetings, partly to try to get this all in order. I think Stephane still will attend, at least some of them; I'll try to discover when he is available. Clear skies, Al -------------------------------------------------------------------------- Mark elaborated: Date: Mon, 24 Sep 2001 09:52:09 -0400 (EDT) Min, Ewine, Stephane, and Al, It seems to me that the real issue is at what frequency we are going to be performing the bulk of mosaics that can benefit from the ACA. From Min's analysis of 230 vs 460 GHz line ratio images (www.astro.umass.edu/~myun/aca/m51test.html), it seems to me that the low frequency benefits (in the case of "typical errors") are small to nearly unmeasurable, but that at the higher frequencies the benefits are more tangible. This suggests an important aspect to emphasize, namely high fidelity imaging at the higher frequencies (650 and 850 GHz bands). Clearly, the effects of atmospheric and antenna errors (pointing, anomalous refraction, phase screen, atmospheric attenuation and emission) are worse at these frequencies, and in these cases that we see significant improvement. This makes sense, as the errors are proportionally larger due to the much shorter wavelength. As Stephane pointed out at the meeting, the inclusion of the ACA provides a robustness to the imaging process that can overcome some of the atmospheric limitations, allowing precision imaging of larger fields at the higher frequencies. Note that "larger fields" is not really all that large, even 10" qualifies given the 7" primary beam at 900 GHz. Therefore, it is reasonable to assume that a large percentage of the science done at the higher frequencies will receive substantial improvement in image quality if the ACA is used. This is important, because of the limited time available for high frequency work we want to provide every reasonable measure to allow for good imaging at these frequencies...its not like we can just go back to each project until we get the perfect weather needed. To demonstrate this, it might be useful to have someone compile a 650 or 850 GHz image of the same source to see the improvement that the ACA adds at the high frequencies, including a "line ratio" test ala Min. The downside of this is that we seem to have a limited use for the ACA at the lower frequencies, at least given the simulations that I am aware of so far. Given that the lower frequencies, particularly 230 and 345 bands, will be the 'workhorse bands' that will likely have the highest overlap of proposals and available weather conditions, this means that a lot of science may not benefit from the ACA. We also need to think about this question: if a high percentage of high frequency projects need the ACA (e.g. more than 25%), will the ACA be fast enough to provide the uv data? I recall that Stephane calculated (Al, I think you came up with a similar number) that 25% of all projects would need the ACA, and that the ACA would be spending essentially 100% of its time doing the uv spacing observations for those projects (the factor of four is to match the sensitivity of the base array with the ACA). What if it seems that 50% (by time) of the high frequency projects need ACA data? I assume that for the base array, there will almost always be a high frequency project done as long as the weather permits its, e.g. a 100% use of high frequency weather to do high frequency observing. This suggests that we will not have enough good weather to obtain the ACA data needed to satisfy the 50% of the projects that need it. I don't know if this is a problem, in that I don't have a feel for how much of the high frequency projects will require the ACA. However, the simulations that show significant improvement mostly at the high frequencies, and the small field of view of the array in a single pointing, suggest that a fraction larger than 1/4 may not be unrealistic. If so, what do we do? I mean besides doubling the size of the ACA? Sorry for the long email. Mark -------------------------------------------------------------------------- Min replied: Date: Mon, 24 Sep 2001 11:00:18 -0400 (EDT) I think all estimates on the projects requiring the ACA are just guesses, and it will be hard to justify any number. Regardless, I concur with Mark's assessment that the most significant contribution by ACA will be made at high frequencies. It was implicit in my understanding of the simulations, but I was focussing more on the mosaic technique than the effects of the error arising from the atmosphere and hardware earlier. Frederick and Jerome earlier mentioned that producing larger mosaics would take a very long time. On the other hand, the 230-vs-460 test suggests that the high frequency image dominates the effect, and it would be worth doing 19 field mosaic simulation at 650 or 800 GHz, using a scaled down version of the model image. Ewine's question on the estimate of total power error is also quite valid. It is dangerous to approach the technical people with this question, but we might query a sensible person and ask what might be a conservative estimate of TP error we can safely assume (as opposed to what we really like to see). Needless to say, the ACA will provide a real security blanket for the imaging performance if we end up nowhere near the target TP performance. -- Min -------------------------------------------------------------------------- Al wrote: Date: Mon, 24 Sep 2001 10:52:36 -0400 (EDT) Hi Mark I agree. After all, the specifications were set to make imaging very good at 300 GHz and below; I think the simulations suggest that that goal was achieved. Mark has the surface error simulations calculated and is in the process of evaluating them. Bryan is working on the debris disk science, including those images. I think Mark is still working on the M51 images, which require more computation owing to the number of point sources which must be cleaned. I think that if we had examined a CI 809/492 GHz line ratio image of M51, the benefits of the ACA would have been more marked. Mark addressed the issue of 'what frequencies the ACA observes' somewhat in his protomemo on surface errors of 20 July. That wasn't circulated outside the tiger team I think but it is at www.tuc.nrao.edu/~mholdawa and it basically reports the philosophy behind Mark's simulations. I suspect that the 25% estimate is low for the ACA but I expect that for some of the time when the 12m array is in an extended configuration, the ACA will not be used as often with data collected in that configuration. During good weather, it may be playing catchup. There will be some periods when the ACA can make good use of high frequency weather but during which the large array will be better employed at lower frequencies. This needs a more careful examination, to be sure. Clear skies, Al