A little less cynical these days, Steve Ohr offers new perspectives on programmable analog devices
Lattice Semiconductor won a Best-of-Show award at the Sensors Expo in Anaheim, Calif., this week. The award was given by the show's sponsors, Sensors magazine and Advanstar Communications, for Lattice's programmable analog module (the ispPAC80) that implements a programmable continuous-time fifth-order filter. The device was featured on the cover of Electronic Design earlier this year.
Lattice is among the latest and most ambitious of the IC makers that believe they can capitalize on circuits with analog functionality and in-system programmability. Among the others were Motorola, which marketed an "FPGA for analog" with a programmability technology acquired from Pilkington Microelectronics Ltd. (Northwich, UK); International Microelectronics Products (IMP, San Jose, Calif.) and Zetex plc (Oldham, UK).
Those manufacturers have provided IC packages that include op amps and a variety of passive components that can be "configured" - in the same way that field-programmable gate arrays are - to perform a variety of filtering and signal-conditioning functions on an analog signal. The programmability allows the user to not only set the gain and offsets for the amplifier, but to control the roll-off characteristics for any filter implementation.
Those guys have left a lot of blood on the floor.
With all the interest these days in analog signal conditioning, the market for these parts is actually a lot smaller than their designers imagined, and up to now programmable analog ICs have been pretty much market failures. Motorola and IMP are out of the business. Anadyne Microelectronics Ltd. (Crewe, UK) announced plans in March to revive the field-programmable analog device technology that Motorola took from Pilkington. (Motorola, in fact, has a minority stake in Anadyne.) But so far remaining players like Zetex have reported limited success.
I have my own opinions why, and have written pretty harshly about it in the pages of EE Times. Here I can afford to be a little more conciliatory than I can in the 425-word space allowed by that EE Times - but not much.
Programmable analog, it seems to me, fits into an applications space somewhere between the cheap-and-dirty op amp circuit (cobbled together with some resistors and capacitors) and the programmable DSP filter. An analog guru like National Semiconductor's Bob Pease or Linear Technology Corp.'s Jim Williams can craft a pretty good op amp filter blindfolded with about $1-worth of parts. As far as filters go, it might be pretty sloppy - say, 6 dB per octave - with some "raggedness" in the signal at particular frequencies. But it will do the job for a variety of front-end bandpass applications. The DSP filter will offer precision and adaptability, but it'll cost $5 or more (maybe $20 with the A/D and D/A converters you'll need to frame it with). The programmable analog devices now marketed by Lattice might go for $5 to $7, depending on what kind of deal you make.
I can see that the latest Lattice device makes a pretty good filter. It houses dozens of filter components, all user-selectable. In fact, Lattice's programmable analog parts were designed by Hans Klein, the same clever guy who built IMP's ill-fated Epac (electrically programmable analog component) product line. This time around he went for precision and speed. The Lattice ispPAC80 can tailor filter functions with up to 12-bit precision, and can operate at frequencies from 50 to 500 kHz. Incidentally, programmability is controlled by E2PROMs, a fact that has encouraged nonvolatile memory maker Xicor (Milpitas, Calif.) to approach the programmable analog market with a number of focused products, including a digital potentiometer later this year.
The argument I have with Lattice's programmable analog line is not over its cleverness and utility, but rather - in the big, live-and-die-on-the-Nasdaq business that IC production has become - there may not be enough users to make it worthwhile. To use Hans Klein's module, to pick the right resistors and capacitors in the feedback loop, you really do have to know something about analog filter design. This is not a product for a digital designer; you need some analog expertise to make it go. And if you have the expertise, what's to stop you from bypassing the ispPAC devices and simply picking your own op amps, resistors and capacitors from the parts bin in your laboratory?
True, if you don't have the analog design expertise you can sure learn a lot about filter design from the software Lattice provides for the programmable part. But you can learn as much from the filter design software sometimes distributed by op amp manufacturers in the efforts they make to get you to use their parts. In fact, a number of simulation software suppliers will allow you to demo their products online - and a couple of these would be useful for trying out an analog filter topology.
The programmable devices may have an advantage in some space-constrained applications, but considering that op amps and passives come in SOT23 even "chip-scale" packages it's hard to imagine a 28- or 48-pin ispPAC having any advantage in that place. "Where programmable analog definitely will have an advantage is in rapid prototyping," I wrote last fall. "It will sure beat a solder pencil or clip blocks for trying out amplifier and filter circuits with leaded components. Still, we're talking onesy-twosies."
I can only repeat Lattice's rationale here and wish them well. "The world thought we were crazy when we introduced a $35 PAL to replace $4 worth of SSI and MSI," marketing manager Andy Robin told me last year. "Now programmable logic is a $2.3-billion industry."
But, hey, I've been wrong before. In the late 1970s or early 1980s it still looked like a horse race between Intel and Motorola for dominance in PC hardware. When surveyed by a Portland, Ore.-based market research firm about which company I thought would dominate in 16-bit operating systems - the inventors of CP/M (the dominant 8-bit OS) or that little company in Redmond, Wash. (what's its name again? Starts with an M) - I said, with absolute conviction, Digital Research.