[evla-sw-discuss] EVLA WIDAR: Delay Model - format and content

[Bruce Rowen]

Barry Clark wrote:
> There's a lot of content here, and I suspect I'll have much more to say
> when I've had time to do some work on these (I'm a bit tied up at the
> moment in my other hat as scheduling officer).  However, a couple of 
> remarks that occurred to me on a rather cursory reading.
>
> 1.  Number of derivatives.  We are working with two derivatives
> at the moment.  This seems to work pretty well for VLA antennas tracking
> translunar objects.  (And the code can be easily changed to handle NMA 
> antennas.)  The cases that probably require more (VLBI and satelite 
> tracking) have not had any serious design work done on them, so it's a 
> little unclear how we want to handle them.  I strongly suspect that they 
> also will turn out to have a number of derivatives set by the application, 
> not something decided on the fly.  My preference would be that the number 
> of derivatives be moved to a setup message, not sent with every delay model.
> (Presuming that it's needed at all, that it's inconvenient to just use
> whatever arrives.)
>
> 2.  Double.  Why the preference for a hex encoded double, instead of the
> native 'double' form?  (Just in case somebody decided to read one of these
> godawful things.)
>   
>> Each coefficient is 64 bit (8 byte) floating-point number encoded as hexBinary.
>>     
>
>   

As a partial answer to #1 & #2, the coefficient encoding is nicely 
compact and XML-able when all coefficients are encoded in
hexBinary. As part of the XML, an attribute specifying the length 
(number of coefficients) is handy.

> 3.  Timestamps.  We work internaly in a double precision MJD and fraction
> system (LSB is about one microsecond).  Converting that to ISO 8601 form 
> is easy of course, but I rather suspect that you also will be working in 
> some similar system internally, and it might save a few a little bother at 
> both ends just to send that.
>   

Internally our time is in seconds since the "epoch" with a time 
resolution of 10ms. These are internally carried in
two unsigned integers (32 bits) as that works very nicely with the OS.
> 4.  Delays.  Who organizes the delay center?  Do I send numbers which might
> be negative, or should I add something to them to make them always positive?
> If so, what?
>   

TBD, but my feeling is the correlator is "dumb" in the sense it doesn't 
care about signs or offsets from its delay buffer
"center" (If the delay buffer center is somehow tied to the array delay 
center). My only concern (the correlator speaking here) is that the 
delay never exceeds the buffer range.

-Bruce

> _______________________________________________
> evla-sw-discuss mailing list
> evla-sw-discuss at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/evla-sw-discuss
>