Analog and mixed signal design goes programmable

Electronic News, August 12, 1996 by David Gillooly

There are few systems today that do not take advantage of digital circuit programmability. Microcontrollers, field programmable gate arrays (FPGAs), programmable logic devices (PLDs) and microprocessors have become the raw material of designs in the 1990s. A digital circuit designer from the 1980s transported to the present would discover a vastly different and more rich design environment. A similarly transported analog and mixed-signal designer would feel at home and would not see a vastly different design landscape. However, a glimpse of the future would be seen in handful of programmable analog circuits based on EEPROM technology.

So far, in the dance between programmable technologies, digital has led. Microprocessors, digital signal processors (DSPs) and PLDs have delivered radical gains in price/performance and ease of use. By contrast, most analog designs today still rely on a mixed bag of low-integration parts such as amplifiers, voltage references, filters, switches, comparators, multiplexers and data converters. Besides the unit cost and manufacturing implications, the design and debug effort and effect on time-to-market of traditional analog design is increasingly problematic for managers. The lack of engineers trained in analog design and with real world experience makes even simple tasks or system changes arduous.

Furthermore, a single device can be programmed in different ways to support multiple applications. It was these factors that sparked the creation and fueled the ascendance of programmable technology as embodied in the microprocessor or programmable logic.

The programmability concept really took off with the invention of the EPROM, which moved final configuration from the read-only-memory (ROM) chip supplier to the customer. The benefits of programmability from this manufacturing perspective are noteworthy as well.

Up-front NRE costs and leadtimes are eliminated to the obvious advantage of the bottom line and schedule. Inventory risk is limited. Further efficiency is gained by utilizing a single part number across multiple projects, minimizing paperwork and the penalty for forecast errors. By becoming expert users of a few parts, the efficiency and capability of individuals and groups increases. Notably, bug fixes--the bane of factory-programmed parts--are quickly and easily slipstreamed into production.

Though subtle, another difference between factory and field programmability shouldn't be overlooked. A factory-programmed part is essentially "custom," while a field programmable part is a "standard product." From a procurement point of view, customers can take advantage of the credit, inventory and competitive service offered by a robust "standard product" distribution network, but not give up the basic fact the device provides the benefits of a custom device.

Having moved from the IC suppliers to customers, programmability is now moving into the products themselves, a concept sometimes referred to as in-system programmability. Instead of ROM or EPROM, in-system programming utilizes EEPROM or SRAM to allow repeated reconfiguration.

In-system programmability further refines the manufacturing and risk reduction advantages of field programmability. For instance, production flow is simplified by complete elimination of the separate programming step associated with EPROMs (not to mention banks of EPROM programmers). Instead, in-system programmable devices can be programmed post-assembly, eliminating a lot of handling of the increasingly small and fragile chips. Whereas EPROM technology allowed fixes and updates as far as the shipping dock, in-system programmability enables on-site service.

Equally compelling, in-system programmability allows designers to create products that are intrinsically flexible and adaptive by utilizing dynamic reprogramming as part of normal operation. Examples include compensating for wear, aging or other environmental changes, controllers that tune themselves for optimum performance and transformation of formerly specialized products into ones that are more "universal."

Once again, when it comes to in-system programmability, the digital side has led the way with products like the Xilinx SRAM-based PLDs and recently emerging EEPROM-based micros from Motorola, Microchip Technology and Hitachi.

What's needed is mixed-signal chips that catch up with the integration, ease of use and in-system programmability of MPUs and PLDs. Now let's take a look at just such devices, the aptly named "Electrically Programmable Analog Circuit" (EPAC) family from IMP, Inc.

An analog PLD?

In one sense, an EPAC can be thought of as an "analog FPGA." However, unlike digital FPGAs, there's no analog equivalent of the two-input AND gate. Thus, the "fine-grained" approach espoused by some programmable logic suppliers is not suitable. Supplying a "sea" of resistors, capacitors and op-amps would not meet the desired analog performance specifications, or ease of use goals.

Instead, the EPAC functional modules are actually complete subsystems that each encapsulate the equivalent of dozens of traditional analog discretes (in high-level functions such as multiplexers, programmable gain amps, filters, comparators and ADCs and DACs). The blocks are designed by analog experts so the user doesn't have to be. A digital designer can tackle an analog design since years of analog expertise is encapsulated in the programmable function blocks. With IMP's EPAC offering "correct by construction" capability, traditional testing and "tweaking" chores are reduced. The function block approach allows a greater proportion of any firm's design staff to execute an analog or mixed-signal design task successfully.