Recent Improvements

Since the Barcelona meeting, we have continued to improve the throughput, reliability, and astronomical capabilities of the correlator. The following points represent some of the accomplishments within this period:

$\bullet$

A TRM byte slip is an event where one track of data spontaneously shifts by 8 lags with respect to all the other tracks. For our typical 2-bit fan-out data, such an occurrence results in a multi-peaked fringe, and hence ringing in the frequency-space spectrum. Recorrelation of a pass containing a byte slip was the only recourse; ensuring all of them were found significantly lengthened the post-correlation review (cf §3.3). We implemented a mechanism to detect and repair TRM byte slips and to flag associated data in the logic of the responsible station-unit boards. We have thus cut down on the number of recorrelations required and streamlined the review process, while ensuring data integrity.

$\bullet$

Observations at $BW_{\rm sb}=16$MHz could cause read/write pointers of a circular buffer in a board in the station unit to fall on top of each other, resulting in a shift of the data- and model-times of $\simeq$131ms and an associated jump in the residual delays and rates. Use of a servo feature for the buffer pointers overcame this problem.

$\bullet$

We increased correlator output rate by a factor of 8, so that the whole correlator can be read out in 1s (cf §2.3). This improvement came about primarily by incorporating software to read out the four correlator crates in parallel, and also by reducing the size of the auxiliary information output along with the correlated data.

$\bullet$

The van Vleck correction for 2-bit data depends on the fraction of high- and low-magnitude bits within an integration, but this dependence was not being taken into account. If a BBC at a station had such ``sampler statistics" different from the expected values, a non-unity autocorrelation peak would result. The amplitudes on baselines containing it would be affected to a lesser degree. If the sampler statistics were especially bad on both stations in a baseline, even the closure amplitudes could be affected at a $\sim$1% level. We now compensate for the observed sampler statistics per station/BBC in a post-correlation program.

$\bullet$

Positions of EVN stations that don't usually participate in geodetic campaigns have been significantly improved, in some cases with adjustments on the order of 5m [6], [7]. In addition, a better tie between the Wb single-dish and array positions was determined, and an unrelated effect that introduced discrete phase jumps in Wb$_{\rm arr}$ (typically once per several hours) was detected and repaired during this process. We contacted PIs of previously correlated experiments to help them incorporate these station-position improvements directly into their AIPS data [8]. The EVN performance in phase reference experiments has significantly improved following these improvements (i.e., more reliable detection/imaging of fainter sources).

$\bullet$

JIVE has begun the process of consolidating experiment information of interest to PIs into a web-accessible archive. This will contain station feedback, a summary of the correlation and post-correlation review, various standard plots, preliminary calibration results, and the FITS data (cf §3.3).