[evla-sw-discuss] setting gains on the station board
Bruce Rowen
browen at nrao.edu
Mon Sep 13 12:25:57 EDT 2010
Based on priorities, I have only implemented the requantizer gain
setting for now. It seems that based on real data (though probably not
typical since it was from 'dead' antenna 26), the gain setting is
fairly robust all the way down to 32Khz sub band width. The adjustment
window takes about 1 second, and for perhaps an additional few system
ticks I could put an integrator on the gain to minimize error. As it
sits now, the test data from antenna 26 (8-bit) lets the gain be set
to within 1% of the target RMS without this extra step.
I'd like to test with some 4-bit data before building further on this
and perhaps get some examples of problematic data. I know that Dave's
delta function test data causes 'issues'.
The target RMS can be set per-sub band and each sub band adjustment
can be individually triggered so there should be no imposed
restrictions for granularity of control through the VCI. Allowing for
continuous ALC would produce a smoother and more accurate solution
(and could be gated on/off for 'set and forget'), but i'd like to see
how this simpler solution works in the mean time.
-Bruce
On Sep 10, 2010, at 11:56 AM, Sonja Vrcic wrote:
> Please disregard my previous message, I accidentaly prresed 'send'
> before the message was
> complete. See my comments below.
> Sonja
>
>
> Ken Sowinski wrote:
>> Getting the station board gains, especially the requantizer gain,
>> right will be important for the three bit samplers. I talked with
>> Bruce about this yesterday and summarize are conclusions and
>> questions
>> here.
>>
>> 2. We must decide how to communicate to the statino board when and
>> how
>> to do this. The obvious solution is for the executor to send a VCI
>> 'modify' document specifiying when to begin the adjustment, and what
>> to do. In addition to an activation time one might want to include
>> set points, integration time and, perhaps, desired ALC strategy.
>> this depends on implenting the VCI modify funtion. Is there any need
>> for an alternate path which might be available sooner?
>
>
> From the point of view of CM this should not be difficult to
> implement. Most work will be
> implemented on CMIB. CM would simply pass instructions to Station
> Boards. To speed up
> implementation it is important to define requirements.
>
>
> Here is my attempt to capture requirements (based on requirements
> for the Filter Gain
> panel in Station Board GUI):
>
> 1. Target filter gain (and other related parameters) to be specified
> (and set) for each
> subband independently, i.e. for each subband of each baseband pair
> (or baseband singleton).
> 2. Parameters are specified per subband and apply for all the
> antennas (stations) in a
> subarray (as VCI "modify subarray").
> 3. Alternatively, as for other parameters, user can specify a subset
> of antennas
> (stations) to be configured.
> 4. Activation time assigned per subarray (in the same fashion as
> when creating a new
> subarray).
>
> List of parameters that can be specified for each subband of each
> baseband:
> a. Desired RMS (mandatory)
> b. Desired RMS for re-qunatizer (mandatory)
> c. Number of integrations (for power measurements) to be used to
> calculate RMS (optional,
> if not explicitly specified = 100)
> d. Number of integrations to discard between the measurements for
> different stages
> (optional, if not specified it is ON, number of integrations to
> discard=100). GUI allows
> user to turn this on/off and to specify number of integrations. That
> was required in
> experimental phases. Is there a need to keep this as externally
> specified parameter(s) or
> should it be hard-coded in CMIB software ?
>
> Then there is a question of units: in GUI, one can select whether
> RMS is displayed in
> computer units or normalized units. Which one should be used in VCI.
> Is there a need to
> implement both options. If possible, it would be nice to simplify
> interface and
> implementation.
>
> This is just the first attempt, please modify the list as needed.
>
> Sonja
>
> -------------------------------------------------------------------------------------
> Here is a copy of the "user help window" for the Station Board GUI -
> Filter Gain.
>
> Commands:
> Refresh reads from hardware and displays:
> Filter ID and status
> Station/Baseband/Subband IDs inserted in the output data stream
> (subband)
> RFI Detect Level
> Scales for used filter stages and for re-quantizer
> Number of bits for re-quantizer
>
> Set RFI detect level sets RFI detection level in the filter FPGAs
> (as specified in the table)
>
> Measure RMS calculates RMS based on the power counts
>
> Adjust filter scales to get desired RMS optimizes RMS as close as
> possible to the desired
> RMS (specified by user) RMS is calculated as follows:
> RMS = sqrt(powerNdOn + NdOff) / numberOfIntegrations)
> Use normalization units applies to RMS. If selected, RMS is
> normalized:
> normRMS = (RMS / (2^(numberOfBits - 1) - 1))
> For stages 1 to 4, numberOfBits = 16. For re-quantizer, number of
> bits is obtained from
> the Filter FPGAs (and displayed in the table).
>
> Save/Load Configuration to/from File on the "Main" menubar allows
> you to optionally save
> or load "RFI Detect Level" (column RFI) and "Desired RMS" (columns
> D_RMS and D_RQ_RMS).
> Only the filters with their "Select" box checkmarked will be saved.
> When loading, the "Select" checkbox will be used to indicate which
> filters were updated.
>
> RMS Units:
> RMS displayed in computer units can be converted to normalized
> units. Conversion from
> normalized units to computer units is not performed. If "Use
> normalization units" not
> selected, columns for measured RMS are filled with blanks, and
> desired RMS is set to the
> default values.
>
> RMS is calculated for the stages that are used (see the Filter GUI's
> "Format" box, Primary
> Signal) and for the re-quantizer.
>
> When the Start button is clicked, the following is performed:
> Get filter status, scales and number of bits for re-quantizer;
> For each stage {
> Configure all the (green) filters for stageX;
> Collect power counts (for all the filters);
> For each filter {
> Calculate RMS;
> Display RMS;
> If (adjustScales is selected) {
> Calculate new scale;
> Configure filter for the new scale;
> }
> }
> If (adjustScales is selected) {
> Collect power counts (for all the filters);
> For each filter {
> Calculate RMS;
> Display RMS;
> }
> }
> }
>
> Get RFI and scales from hardware;
> Number of power counts integrations used to calculate RMS:
> User can specify the number of power count integrations used to
> calculate RMS.
> By default, this number is the same as the power count integration
> factor obtained from
> hardware; which means that, by default, one report (table) is used
> to calculate RMS.
>
> Delay between stages:
> RMS is calculated for one stage at a time. Before calculating RMS
> for each stage, filter
> configuration is slightly modified. To ensure that the power counts
> used to calculate RMS
> are only for the data processed after the configuration change, the
> power counts received
> immediately after the filter configuration change should be
> discarded. By default, the
> number of integrations to be discarded is the same as the power
> counts integration factor
> set in hardware; which means that the first report (table) received
> after the
> configuration change will be discarded. This feature can be disabled.
>
> -------------------------------------------------------------------------------------------
>
>
>
>
>
>
>
>>
>> 3. We must decide how the executor knows when to tell the correlator
>> to set gains. Is it explicitly in the scripts? Should the executor
>> understand enough about the meaning of the script to discern when to
>> send the cause a gain setting? Getting this right will interact with
>> setting up the antenna, setting downconverter gains, and antenna
>> motion.
>>
>> _______________________________________________
>> evla-sw-discuss mailing list
>> evla-sw-discuss at listmgr.cv.nrao.edu
>> http://listmgr.cv.nrao.edu/mailman/listinfo/evla-sw-discuss
>
> --
> Sonja Vrcic
> Software Engineer
> National Research Council
> Herzberg Institute of Astrophysics
> Penticton, BC, Canada
> Tel: 250-497-2309 / 250-497-2300
> Sonja.Vrcic at nrc.gc.ca
> http://www.nrc-cnrc.gc.ca/eng/ibp/hia.html
> _______________________________________________
> evla-sw-discuss mailing list
> evla-sw-discuss at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/evla-sw-discuss
More information about the evla-sw-discuss
mailing list