[evlatests] Settling Times for Referenced Pointing

[Bryan Butler]

my argument was just a generic one for typical observing, since bob 
asked if anybody knew and i thought i could make a stab at it, and his 
argument was a more generic one.

if you want to argue that the special cases below drive a change to the 
servo systems (that will probably cost us more than a simple 
modernization [though, truth be told, i don't *know* that it will cost 
more - if it doesn't, then we all win!]), then you have to argue that 
they either occupy a significant fraction of the observing time (because 
if they don't, then we can just give them more overall time to observe 
to get the same result), *or* that they are special enough that they 
warrant an upgrade in any case, that is worth the cost.

	-bryan


Claire Chandler wrote, On 11/22/10 06:43 AM:
> I don't think the motivation for faster slew times comes from increasing
> the time on source for the "typical" observation described below. As
> Bryan notes, slew time is a small factor in the overall sensitivity for
> such observations. Rather, it comes from wanting to do (a) OTF mapping,
> (b) large surveys for which you only want to spend a minute per source
> (similar to the flux density run from which this discussion began; the
> larger bandwidths of the EVLA will make this a more common observing
> mode in the future), and (c) "fast switching" at high frequencies for
> which we know the observing efficiency can be 50% or worse.
>
> Claire
>
> On Mon, 22 Nov 2010, Bryan Butler wrote:
>
>>
>> a BOTEC (Back Of The Envelope Calculation).
>>
>> assume that for "typical" observing, you're slewing back and forth
>> between source and calibrator every 10 minutes (this is a rough average
>> between high frequency which needs it more often, and low frequency
>> which can be less often).
>>
>> these sources are typically fairly close together (5-ish degrees), so
>> say 20 seconds slew and settle for each, and you have two of them per
>> cal-source cycle, so 40 seconds of every 10 minutes.
>>
>> if you cut that in half you have 20 seconds of every 10 minutes, or
>> roughly 3% more time on source. but sensitivity goes like sqrt{t}, so
>> it's really like a 1.5% increase in sensitivity.
>>
>> that would be equivalent to about a .7 K improvement in Tsys for a 50 K
>> system (receiver + atmosphere + spillover + ...).
>>
>> it's all cost-benefit, of course - there is a clear gain, but is it
>> worth the cost? it's clear we need to modernize the system, but whether
>> it is worth making it faster (in slew and settle) depends on how much it
>> would cost...
>>
>> -bryan
>>
>>
>> Bob Hayward wrote, On 11/19/10 14:31 PM:
>>>
>>> Does anybody know how much of the time the telescopes actually spend
>>> slewing rather than tracking? My uneducated guess would say 10-20% of
>>> the time, depending on the observing program. If you could speed the
>>> drives up by a factor of two, you could get 5-10% more time on the sky
>>> looking at your favorite sources. That is the same gain in sensitivity
>>> you would get by reducing the receiver temperature performance of every
>>> one of the front-ends by 1 or 2 degrees Kelvin (which is an unlikely
>>> scenario as they're already as good as current technology allows). Put
>>> another way, I think it is equivalent to the sensitivity improvement you
>>> would get by adding an extra 25m antenna to the array (i.e., a full time
>>> 28th dish). So if you want to improve the sensitivity of the array,
>>> upgrading the drives might be a relatively cheap way to do it.
>>>
>> _______________________________________________
>> evlatests mailing list
>> evlatests at listmgr.cv.nrao.edu
>> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>>