The gravitational lens CLASS B1608+656 is the only four-image lens system for which all three independent time delays have been measured .
This makes the system an excellent candidate for a high-quality determination of H _ { 0 } at cosmological distances .
However , the original measurements of the time delays had large ( 12–20 % ) uncertainties , due to the low level of variability of the background source during the monitoring campaign .
In this paper , we present results from two additional VLA monitoring campaigns .
In contrast to the \sim 5 % variations seen during the first season of monitoring , the source flux density changed by 25–30 % in each of the subsequent two seasons .
We analyzed the combined data set from all three seasons of monitoring to improve significantly the precision of the time delay measurements ; the delays are consistent with those found in the original measurements , but the uncertainties have decreased by factors of two to three .
We combined the delays with revised isothermal mass models to derive a measurement of H _ { 0 } .
Depending on the positions of the galaxy centroids , which vary by up to 0 \farcs 1 in HST images obtained with different filters , we obtain H _ { 0 } = 61–65 km s ^ { -1 } Mpc ^ { -1 } , for ( \Omega _ { M } , \Omega _ { \Lambda } ) = ( 0.3 , 0.7 ) .
The value of H _ { 0 } decreases by 6 % if ( \Omega _ { M } , \Omega _ { \Lambda } ) = ( 1.0 , 0.0 ) .
The formal uncertainties on H _ { 0 } due to the time delay measurements are \pm 1 ( \pm 2 ) km s ^ { -1 } Mpc ^ { -1 } for the 1 \sigma ( 2 \sigma ) confidence limits .
Thus , the systematic uncertainties due to the lens model , which are on the order of \pm 15 km s ^ { -1 } Mpc ^ { -1 } , now dominate the error budget for this system .
In order to improve the measurement of H _ { 0 } with this lens , new models that incorporate the constraints provided by stellar dynamics and the optical/infrared Einstein ring seen in HST images must be developed .