Correlator Capacity

There are two equations that are useful when planning observations that will be correlated on the EVN MkIV data processor. The first relates to total correlator capacity:

$$N^2_{\rm sta} \cdot N_{\rm sb} \cdot N_{\rm pol} \cdot N_{\rm frq} \le 131072 \cdot \mathcal{R}.$$ (1)

Here, from right to left, $\mathcal{R}$ is the recirculation factor (see next paragraph). $N_{\rm pol}$ is the number of polarization combinations wanted in the correlation (1, 2, or 4). $N_{\rm sb}$ represents the number of different frequency subbands, counting lower- and upper-sidebands from the same BBC as distinct subbands, but not multiple polarizations in the same sideband (these enter via $N_{\rm pol}$). The value to use for $N_{\rm sta}$ is ``granular" in multiples of 4: e.g., if you have 5-8 stations, use ``8". Independent of this equation, the maximum ( $N_{\rm sb}\cdot N_{{\rm pol}_\parallel})$ is 16 (a station-unit limitation), and the maximum $N_{\rm frq}$ is 4096 (a single baseline/SB/pol must fit onto a single correlator board). You should pick the various $N_x$ parameters in designing your observation such that the equation holds, otherwise you will have to compromise on at least one of them when it's time to correlate. Some examples of configurations that would use the full correlator capacity (without recirculation) include:

	8sta	1SB	1Pol	2048Frq	(EVN spectral-line)
	16sta	8SB	4Pol	16Frq	(global cross-polarization mapping)
	12sta	14SB	1Pol	64Frq	(geodesy)

When recirculation is operational, the constant term 131072 can be multiplied by recirculation factor, $\mathcal{R}$, for observations that don't use the full (Nyquist-sampled) 16MHz bandwidth per SB/polarization channel. In principle, $\mathcal{R}_{\rm max}$ would be set by $16 {\rm MHz}/BW_{\rm sb}$, but there may be a smaller absolute limit depending on the actual implementation. $N_{\rm frq}$ would still be subject to the maximum limit of 4096, as discussed above.