These calculations are based on the assumption that spectral line data from ALFA are written as binary FITS tables using the SDFITS protocol in a manner similar to current AO line data.
A single dumped spectrum contains N spectral points which are written to column 1 as N elements of 4 byte real.
The remaining columns 2-52 are similar to those of a current AO SDFITS file, occupying at least 396 bytes (might be more if additional informational columns are added).
A single row has a width of at least (N times 4) bytes plus 396 bytes.
Spectra from all 14 ALFA channels (7 beams times 2 polarizations per beam) are dumped to disk every 1 second.
Each of the 14 spectra thus recorded is contained as a separate row within the table.
Thus a single 1 second dump would contain 14 rows of at least (4 N + 396) bytes.
And a single observation lasting T seconds would contain 14 times T rows of at least (4 N + 396) bytes.
Under the above assumptions, the minimum data rate for likely E-ALFA spectral line configurations is:
| Mbytes per sec | # spectral points | configuration | timeframe |
|---|---|---|---|
| 0.12 | 2048 | 50 MHz, 2 pols, 9 level | 2005 |
| 0.23 | 4096 | 100 MHz, 2 pols, 3 level | 2005 |
| 0.46 | 8192 | 200 MHz, 2 pols, 9 level | ? |
An E-ALFA program recording 4096 points/spectrum that was allocated 300 hours of telescope time in 2005 would thus generate > 250 GB of raw data.
These are not tremendously high data rates by many standards, but they do present challenges for typical university-based astronomers engaged in data processing tasks.
It would be advantageous to specify the format as soon as possible so that software development can work from a common, established basis.