The logic execution speed is a key differentiator. The PLC has been designed to meet the high-speed applications demands that require scan rates of 10 milliseconds or less, including operations involving high-speed interlocking, motion control, or control of motors and drives. Fast scan rates are necessary to be able to control these devices effectively.
The DCS does not have to be that quick most of the time. The regulatory control loops scan normally in the 100 to 500 millisecond range. It should be detrimental to have control logic execute any faster, possibly causing excessive wear on final control elements such as resulting in premature maintenance, valves, and process issues.
The extra cost for redundancy may be well worth it in the case of the typical DCS system, where high availability is mission critical. But, it is often not justified the cost to make a PLC system fully redundant.
Taking the PLC system offline to make engineering changes and configuration may have less impact, since the platform is not running continuously or because the process can be restarted easily. Configuration tweaks and changes to the DCS system are done online, while the process is running virtually non-stop. A lot of process applications may only shut down once or twice a year for scheduled maintenance, while others, such as a blast furnace, are planned to stay on-line continuously for 5-7 years.
The analog control issue is important, but confusing. DCS was designed originally for delivering analog control, but to say the DCS has a lock on the analog control market reiterates the problem with traditional thinking. Increasingly, the PLC is capable to deliver simple to complex PID control, but the DCS is clearly the choice or applications with a large amount of advanced analog control, including model predictive control, cascade loops, ratio, and feed forward loops.