A. Station Board Configuration

1. It looks like the highest level configuration that we'll perform is at the station board level. The observer will be asked in the Observation Preparation Tool (OPT) to provide guidance on the number and/or identities of the antennas needed for the observation. In order to form a vciStbConfigReq, we need to identify the station board by giving its rack, crate, and slot IDs. This means that we to translate an antenna ID into a station board ID.

   Is there something (database, webservice, java class) that we can use to map an antenna ID into a station board ID?

   M.Rupen 2007-03-13: yes, there is something.

2. Is there a naming convention that all other software is using for antenna identification?

   M.Rupen 2007-03-13: I believe there is one. Bryan or Bill should know.

   B.Sahr 2007-03-15: EVLAconstants.java - antenna names

   Any idea what "atr" stands for?

   B.Sahr 2007-03-15: atr = antenna test rack, which lives in the basement of this building

B. Base Band Configuration

1. The Data Path property is mandatory and must have a value of either 0 or 1.

   In the way that the EVLA will be wired, is it true that any dataPath with the same value, regardless of baseband, will always refer to the same polarization? E.g., will "0" always mean "right [left] circular polarized"?

   M.Rupen 2007-03-13: TBD I think. Not sure why you care.

   DMH 2007-04-24: The dataPath is a mandatory attribute of the baseBand XML element in the VCI, so i must set it to one of its allowable values. While WIDAR does not require any particular convention (see Sonja's answer, below), i still wonder whether we need to at least remember what we sent down path "0" and what went down path "1".

   S.Vrcic 2007-03-13: Station Board input from the EVLA antennas can contain up to two basebands (for 3-bit samples). Basebands are time multiplexed, any of the two basebands can be forwarded to any of the two Station Board Data Paths. WIDAR hardware and software does not assume / require any particular arrangement.
   
   In the case when EVLA antennas generate 7-bit samples, there is only one baseband in the Station Board input.

2. Is it true that the baseband with dataPath="0" and the baseband with dataPath="1" will normally be configured identically for a given station board?

   M.Rupen 2007-03-13: Yes, but NOT always! Again not sure why you care.

3. The BaseBand ID is a mandatory property with valid values of 0 through 7. (There was a recent evla-sw-discuss thread on the issue of naming basebands started by B.Clark on 2007-02-06. I am not sure, though, whether discussion of T304 modules is relevant at the obs prep phase. (If it is, please tell me so.))

   Is it true, during observation preparation at least, that the assignment of a particular baseband ID value to a given frequency range is arbitrary?

   M.Rupen 2007-03-13: Yes.

4. The Number of bits property is mandatory and has valid values of 3 or 8.

   How should the software determine whether a 3 or an 8 is appropriate?

   M.Rupen 2007-03-13: Basically you use 8 bits if you have to or if you can get away with it. Fortunately these two are basically the same -- you really need 8 bits for high spectral dynamic range, which is important when there's strong interference, which is most likely at low frequencies where the total bandwidth is small and you can use more bits to handle the relatively narrow bands. It would be fairly easy to come up with reasonable defaults, but we might have to talk about how those should best be expressed (e.g., do you want to think in terms of the on-the-sky frequency or the baseband bandwidth?).

   S.Vrcic 2007-03-13: Number of bits in the input samples is probably part of the antenna configuration.

5. The Number of Bands property is mandatory and has valid values of 1,2,4,8,16.

   Does this property refer to the number of subbands in this base band?

   S.Vrcic 2007-03-13: For the EVLA antennas number of basebands in the Station Board input will be either 1 (7-bit input) or 2 ( 3-bit input). In the future WIDAR correlator may be used to handle data from different antennas. I believe that input from the VLBI antennas can have up to 16 basebands in the input stream. So in VLBI case EVLA subbands will in fact be basebands. WIDAR correlator must receive description of the input data.
   
   It is probably a good idea to keep in mind that eventually different antenna types will be supported.

6. The Offset property is mandatory and is described as: Local Oscillator (BaseBand offset) in Hz. Specified independently for each BaseBand of each antenna. Its valid values are "TBD". (See also questions II.1. and III.C.2.)

   a. Does this value represent the amount by which a received sky frequency range was shifted in order to match the correlator's input frequency ranges?

   S.Vrcic 2007-03-13: Yes.

   If the above answer is "yes", then the software can calculate this number by shifting the sky frequency range down to the range provided in the answer to question II.1.

   b. If the sky frequency bandwidth is less than 2GHz, does it matter where in the correlator's baseband this is placed? (I.e., if we're receiving a sky frequency range of 10-11GHz, and if the correlator handles a range of 0-2GHz, are we free to shift the sky frequency range down by any value in the range 9GHz to 10GHz?)

   S.Vrcic 2007-03-13: User may choose which part of the 2GHz will be processed. If the part of the 2GHz input is not needed, there can be less than 16 subbands.

C. SubBand Configuration

1. The SubBand ID is a mandatory property with valid values of 0 through 17.

   Same question as with BaseBand ID: Is it true, during observation preparation at least, that the assignment of a particular ID value to a given frequency range is arbitrary?

   (If so, the software will have a bias toward calling the subband with the lowest frequency "0", the next lowest "1", and so on.)

   M.Rupen 2007-03-13: For the OPT this is arbitrary. We've discussed occasionally whether 0 will always be the lowest frequency; I think we decided NOT.

2. The Central Frequency property is mandatory and has a valid range of 0..2048MHz. The valid range given here seems to imply that the answer to question II.1. is "0-2GHz".

   No questions here. Refer to answers to questions II.1. and II.2.

3. The Use Mixer property is optional and has a default value of No. When left in its default state, subbands must be placed in specific slots in a baseband. When set to Yes, however, subbands may be placed anywhere within a baseband.

   M.Rupen 2007-03-13: I believe this refers to the mixer in the Stage 2 filter. If so this is only available when that filter is in use, i.e., for sub-bands of less than the maximum bandwidth.

   a. Is there a penalty for using the mixer?

   [put answer here]

   b. If there is a penalty to using the mixer, under what conditions should the software decide to use it?

   [put answer here]

4. The Number of spectral channels property is mandatory. The description for this property states that: The maximum number of spectral channels that can be obtained per SubBand is calculated based on the number of BaseBands, number of SubBands, number of bits to correlate and recirculation factor. There is also an absolute maximum of 262,144, a minimum of 64, and a quantized list of valid values in between these limits.

   a. Should the software calculate the maximum value and ensure that it submits a value no greater than this, or may it submit its desired value, even if that exceeds the calculated maximum? (Subject, though, to the absolute min, max, and allowable intermediate values.)

   M.Rupen 2007-03-13: To which "software" do these questions refer? The VCI has to have the actual number of channels, and that number has to be possible. The astronomer would like to be able to ask for a certain number of channels (or frequency resolution) and then be told of the possible correlator setups and what they'd give. At that stage (s)he should be able to enter any value.

   DMH 2007-04-24: By "software" i mean the object (currently called "WidarConfiguration") that creates VCI messages. (One input to the configuration is a "science view" -- a specification of spectral chunks.) This object would probably be used by the OPT in order to give the observer feedback as to whether the correlator could fulfill her/his request. It might also be used by other applications.

   S.Vrcic 2007-03-13: If the number of spectral channels specified by the user can not be obtained (due to lack of resources) WIDAR correlator will allocate the maximum available amount of resources. The correlator will start allocation from the first subband. If there is no enough resources to obtain the specified number of channels for all the subbands, the correlator will configure the specified number of spectral channels for the first couple of subbands (depending on the number of spectral channels required), until all the resources are used. Resulting output data may have different number of spectral channels for different subbands, and some subbands can be completely dropped. It seems to me that better results will be achieved if the user calculates the number of spectral channels.
   User is aware of the number of basebands and subbands in advance and can assign priorities: only user can decide whether to get more spectral channels for less subbands (only part of the bandwidth) or less spectral channels for the max. number of subbands (the full bandwith).
   
   Note that correlation can be performed for 16 subbands. Remaining two Station Board filters will be used for on-going testing and as spares.

   b. If the software needs to calculate this maximum, what is the formula?

   (One of the inputs to the formula is the recirculation factor, which is an optional field. If the recirculation factor is left in its default state, the correlator will calculate the factor needed to achieve the desired number of spectral channels. Would this not indicate that we should put the desired number into this field (subject to the abolute min, max, and allowable intermediate values) and leave the recirculation field in its default state, thereby allowing the correlator to adjust the recirculation factor to accommodate the desired number of spectral channels?)

   S.Vrcic 2007-03-13: I will address the "formula" in separate e-mail. Several factors influence the allocation of the correlator.

   c. If we are permitted to submit a desired value that turns out to be greater than the calculated maximum allowed value, how will the Configuration Mapper react?

   [put answer here]

5. The Number of bits to correlate property is mandatory and has valid values of 4 or 7.

   How should the software determine whether a 4 or a 7 is appropriate?

   M.Rupen 2007-03-13: See above. Generally we'll use 4 when we're fed by the 3-bit sampler, and 7 when we're fed by the 8-bit sampler.

   S.Vrcic 2007-03-13: The "user" should specify whether 4 or 7 bit correlation is required.

6. The Scaling Factor property is mandatory and has valid values that are "TBD". Its decription is: May be specified either as a set of five scaling factors (one for each filter stage plus quantizer scaling factor) or as signal to noise ration[sic] (percent of signal in the input data). If specified as signal to noise ratio, the correlator will derive scaling factors.

   Is signal-to-noise ratio something we should ask from the observer when they are working with the science view?

   M.Rupen 2007-03-13: I believe this is waaaay too low-level for the astronomer, and should be derived from the software. It sounds like this has to do with the proper settings of the quantizers to maximize the sensitivity for a given input signal, which is determined basically by the strength of that signal relative to the noise. Usually the signal will be much weaker than the noise. I don't see that this is of any interest or importance at the ObsPrep stage.

   DMH 2007-04-24: Again, because this is a mandatory property, the creator of the VCI messages is obligated to place a value here. It may well be that the OPT does not solicit anything here from the observer, but the object that converts the observers' wishes into VCI messages must know how to value this quantity.

   S.Vrcic 2007-03-13: Scaling factor is still TBD. Signal-to-noise ratio will be specified by the user or operator. I imagine that there will be a default value.