Hulse-Taylor binary data gap in the nineties I was wondering what is the reason there are no data points in the famous Hulse-Taylor plot of the period decay in the 1990s. 
Does anyone know why no one collected data during this period?
 A: For the sake of posterity, let me show the plot of interest:


Fig. 1, Weisberg & Taylor 2005.
The plot is from Weisberg & Taylor 2005, and I suppose the authors are the only people who can provide a definitive answer. I can at least hazard a guess, though. The data comes from (~1.4 GHz?) timing observations at the Arecibo Observatory. The gap appears to be from 1993-1997, which coincides nicely with a major upgrade to the telescope completed in 1997. The mid-90s upgrade added the now-iconic Gregorian dome containing two subreflectors, which reduced spherical aberration from the dish, and added a better S-band radar system and new receivers which expanded the frequency coverage to include the entire interval from 300 MHz-10 GHz.
My guess would be that observing downtime during the upgrade had an effect, particularly since the telescope had to be set at certain azimuthal positions for long periods of time. This was extremely helpful for a number of drift scan surveys, although it also imposed limitations on their ability to observe some sources (e.g. Foster et al. 1995, Camilo et al. 1996, Ray et al. 1996, Lorimer et al. 2005). The azimuthal restrictions, as well as some obvious downtime for installation, might have impacted the ability to observe the Hulse-Taylor binary.
After the upgrade was completed, things could go back to (a better) normal, particularly with the advent of the Wideband Arecibo Pulsar Processor (WAPP, see Dowd et al. 2000) in 2000, which, as noted by Weisberg & Taylor, was used as the backends for the last couple sets of observations shown in the plot. As expected, the observations picked up again in 1998.
(As an aside, this massive new Gregorian system required changes to the structure of the telescope, including the addition of 12 auxiliary cables. It was one of these mid-1990s cables that snapped on August 10, 2020, three months before one of the older main cables failed and ensured the dish's demise.)
Again, though, this is just a bit of a guess. It could be a coincidence.
A: Probably because the data were not terribly interesting. After the linear decay on the period was established, it was easy to predict future times of the recurring signal. After 20 years, a data point at 21 or 22 years only marginally improves the accuracy of the decay rate.
There is less information in this curve than what it looks like at first sight.  
