Interconnection resistance is proportional to signal delay, and is probably the most likely reason for the rarity of FPAA parts commercially available. All known FPAA architectures to date [1] have used RAM-based interconnect, in which two interconnection networks are connected by a MOS transistor acting as a switch. Switch resistance across such a device is typically in the range of 1000 to 5000 ohms.
Switched-capacitor circuits require MOS transistor switches, and several experimental architectures have taken advantage of that fact, getting double duty from the charge-transfer switches by using them as interconnect switches as well [4]. Consequently, there is no loss of bandwidth over and above what is normally associated with switched-capacitor circuits.
Nevertheless, modern metal-to-metal (M2M) antifuse technology [5] achieves resistances as low as 15 to 25 ohms per antifuse, which is an order of magnitude below the previous generation of antifuses and two orders of magnitude below MOS switches. Actel Corporation incorporates M2M antifuses into several chip series, most notably the RT-SX line of radiation-tolerant chips suitable for many spaceflight applications [6]. Use of M2M antifuse interconnects potentially allows fast (at least 10 to 20MHz) continuous-time circuits to be built with field-programmable modules. By contrast, a switched-capacitor circuit requires a clock frequency 50 to 100 times the maximum required bandwidth of the equivalent continuous-time circuit. To make use of continuous-time circuits, the analog array architecture must incorporate programmable resistor arrays in addition to the programmable capacitor arrays. Unfortunately, resistance tolerances are not well-controlled in digital processes, so the extent to which the FPAA architecture can be applied to continuous-time circuit design is questionable.
Last updated: October 13, 1999 at 3:05pm