In a perfect world, large telescopes would be developed and built in logical, sequential order. First, scientific
requirements would be agreed upon, vetted, and fully developed. From these, instrument designers would define their
own subsystem requirements and specifications, and then flesh out preliminary designs. This in turn would then allow
optic designers to specify lens and mirror requirements, which would permit telescope mounts and drives to be
designed. Finally, software and safety systems, enclosures and domes, buildings, foundations, and infrastructures would
be specified and developed. Unfortunately, the order of most large telescope projects is the opposite of this sequence.
We don’t live in a perfect world. Scientists usually don’t want to commit to operational requirements until late in the
design process, instrument designers frequently change and update their designs due to improving filter and camera
technologies, and mount and optics engineers seem to live by the words “more” and “better” throughout their own
design processes. Amplifying this is the fact that site construction of buildings and domes are usually the earliest critical
path items on the schedule, and are often subject to lengthy permitting and environmental processes. These facility and
support items therefore must quickly get underway, often before operational requirements are fully considered. Mirrors
and mounts also have very long lead times for fabrication, which in turn necessitates that they are specified and
purchased early. All of these factors can result in expensive and time-consuming change orders when requirements are
finalized and/or shift late in the process. This paper discusses some of these issues encountered on large, multi-year
construction projects. It also presents some techniques and ideas to minimize these effects on schedule and cost.
Included is a discussion on the role of Interface Control Documents (ICDs), the importance (and danger) of making big-picture
decisions early, and designing flexibility and adaptability into subsystems. In a perfect world, science would be
the big dog in the room, wagging the engineering tail. In our non-perfect world, however, it’s often the tail that ends up
wagging the dog instead.