Consider using the actualDurations table to store one weight per polarization product per spectral window per baseband per baseline. Assume, further, that the number of phase bins and APC bins is fixed to be one in all spectral windows. Note that the number of spectral channels may vary with spectral window, but there is only one weight per polarization product that applies to all spectral channels within each spectral window. The key phrase in the foregoing sentence, which is emphasized, is a sign that the spectral channels axis may be omitted from the table in the data stream. The following XML snippet would be part of the sdmHeaderData element that describes this table.

Example A.1. Omitted axes table header snippet

This example needs some updating.

<actualDurations size="18144" ref="actualDurations"
                 axes="a1 a3 a4 a5 a6 a7"/>
<numAntenna>27</numAntenna>
<numAPC>1</numAPC>
<baseband>
  <spectralWindow numBin="1" numSpectralPoint="512" numPolProduct="2"/>
  <spectralWindow numBin="1" numSpectralPoint="1024" numPolProduct="4"/>
</baseband>
<correlatedData type="float" size="15482880" ref="crossData"/>

	8 ⋅ 378 ⋅ (4 + 2) [using Equation A.1, “Binary table size” and the section called “actualDurations table”]
	“a2”, the spectral channel axis, is missing.
	The number of spectral channels and polarization products vary among spectral windows within the baseband.
	2 ⋅ 4 ⋅ 378 ⋅ (512 ⋅ 2 + 1024 ⋅ 4) [using Equation A.1, “Binary table size” and the section called “crossData table”]

The actualDurations table will be indexed as A_{bl,bb,sw,pb,ab,sc,pp} . However, for optimization, the table in the data stream has omitted the sc axis since the data have no dependency on spectral channel; that is, the data in the table are indexed as T_{bl,bb,sw,pb,ab,pp} . The mapping of an index for A to the equivalent index for T is simply (bl, bb, sw, pb, ab, sc, pp) ↦ (bl, bb, sw, pb, ab, pp) . For example, the following statements hold.