[Gb-ccb] CCB interface questions

[Martin Shepherd]

On Thu, 11 Mar 2004, Brian Mason wrote:
> -Do the timestamps give the start, stop, or middle of the integration (or
> something else)?

My plan is to use the time of the end of the integration, although
this depends on details of the hardware implementation that I haven't
attended to yet.

> -As far as we can see the ordering of the data in a single integration is
> not specified-- for instance, if you think of the 64 data for a single
> integration as a multi-dimensional array (N_dim=2: CCB input port and
> phase switching state), which is the most rapidly varying index, and how
> do you assign the data to specific input ports and physical phase switch
> states?  Does this specification exist?

The most radidly changing index will correspond to the phase-switch
states. Thus the first 4 integrations at the start of the 64-element
integration array will correspond to the 4 phase-switch states of the
first ADC, and the second 4 integrations in the array will correspond
to the same 4 phase-switch states for the second ADC etc..

The ordering of the 4 phase-switch states for a given ADC will be the
same as the time ordering of the corresponding phase-switch states
within a single phase-switch cycle, as dictated by the current phase
switch configuration. See section 3.1.1 of my ccb_network_interface
document to see how the phase-switch configuration parameters mandate
the time ordering of the states within a phase-switch cycle.

Although I specified an ordering for the CCB input ports wrt
detectors, frequency bands, and radiometer channels in my old hardware
interfaces document, clearly the actual assignment of ADCs to
radiometer channels depends on how the cables are hooked up between
the receiver and the CCB, and possibly on cabling or daughter-card
locations within the CCB. Since these assignments may occasionally
need to be changed to work around single channel failures or
crosstalk, I suggest that the manager read the assignments of ADC
numbers to radiometer channels from a configuration file, rather than
encoding them in parameters in a header file. This will also make the
differences between the 3mm and 1cm receivers easy to accomodate
without making any changes to parameters within the code.

Martin