Maxim has a go at programmable analogue, but will it succeed where others have failed?
4 mins read
Whilst programmable analogue devices have made it from the lab into the 'real world' over the last decade or so, it's fair to say the approach hasn't been a raging success. Products have come and gone, but few have been successful commercially.
But that doesn't stop a company taking another 'kick at the can'. In this instance, it's Maxim Integrated, which has developed the PIXI programmable mixed signal technology and applied it in the MAX11300, said to be the first in a new range of products.
Martin Mason, executive director of Maxim's Industrial and Medical Solutions Group, explained. "We've come up with something which addresses the problems of making analogue programmable."
In Mason's view, there are two problems; apart from making analogue programmable, there is also an integration issue (see fig 1). "We have been talking consistently, as an industry, about analogue integration. That's fine for a product running at 100million units a year, but it's not so good when you're making 10,000 a year for an industrial application. The economics become challenging, just as they were in the digital domain before programmable logic helped to take the cost out."
On programmability, Mason said: "Making analogue programmable is tough. While programmability might be good, people in the analogue world are looking for robustness and precision and it's hard to do this well."
His solution is a 'halfway house'. "PIXI devices aren't programmable," he admitted, "they are configurable. The difference is important and goes to the heart of why I think these products will be successful."
Mason explained the approach. "The MAX11300 contains predefined building blocks which can be configured in a deterministic fashion. Determinism is important," he contended, "and that's why I think these parts will succeed where other approaches – such as field programmable analogue arrays – have failed."
It would be stretching the point to claim the road to programmable analogue is littered with failures, but there has yet to be a convincing success. Nevertheless, Mason acknowledges products from companies like Cypress.
Some years ago, Pilkington explored the field, while Zetex developed a form of field programmable analogue array with limited success. Lattice Semiconductor – inventor of the 20V10 – developed the ispPAC range, but has withdrawn them. Perhaps the most successful company was Anadigm, whose products are now end of life.
Mason explained why he thought the approaches failed. "In many cases, they went to too fine a level of granularity. Every time, a switch in the path changed the fundamental performance of the circuit you were trying to build. It was too difficult to determine what you would end up with in terms of bandwidth and throughput – and those are parameters about which people in the analogue world care. With the MAX11300, all specs hold, no matter what parts you use."
Mason not only uses the word 'determinism' regularly when talking about the new product range, but he also refers back to the early days of programmable logic to explain the concepts behind the MAX11300 range.
"In the early days of programmable logic, it was all about determinism; when you bought a PLD, you bought a speed grade and knew what the pin to pin delay would be.
"With the MAX11300, we're going to a higher level of abstraction. It feels like we are marching down the same path in the analogue world today as the digital world did 20 years ago. We want to do the same thing for the analogue world."
The inspiration for the MAX11300 range is said by Mason to have come during a visit to a Maxim customer in the wireless comms sector. "One of its biggest problems is power amp (PA) biasing. Not only does it need to have many different PAs in its systems, all with different biasing requirements, it also needs hundreds of different radio configurations around the world. One European customer has to support more than 130 combinations of PA and radio configurations. Previously, it would have had to design biasing circuits for each combination.
"That problem is a servo loop and can be solved using a configurable architecture in which a bias point can be set while power and temperature vary."
Mason said it then occurred to him that there were many other 12bit servo loop applications. "Millions of 12bit I/O cards are sold for use with PLC systems," he continued, "and there are 12bit servo loops for timing and thermal management in many power systems. And there are common applications in the industrial, medical and test & measurement sectors."
The MAX11300 PIXI is described as the industry's first configurable 20 channel, -10 to 10V mixed signal data converter. It is said to be suited to use in applications requiring multiple mixed signal operations, including basestations and industrial control and automation.
The device integrates a 12bit, multichannel A/D converter, a 12bit, multichannel buffered D/A converter, high voltage analogue switches and digital I/Os (see fig 2). Each of the 20 mixed signal I/Os can be configured for the particular application, allowing designers to optimise the layout of their PCB.
"With 20 configurable ports," Mason noted, "it's similar to a 20V10." The 22V10 had 12 input pins and 10 pins that could be configured as either inputs or outputs. "MAX11300 can be configured to do any combination of analogue I/O," he continued, "and has temperature measurement functionality. It's a real 72dB product with good 12bit performance."
Any pin can have any functionality, so a single ended 12bit A/D converter running at 400ksample/s can be shared between any number of channels. "You could have 20 channels at 20ksample/s," Mason offered, "but it's smarter than that. Because some channels can be oversampled, you can use the sequencer to have higher sample rates on these channels. And each pin can have a different voltage."
An application for the part which Maxim hadn't considered is using the D/A converter to provide a small power rail (fig 3). "You can use the scaling block to drive the D/A at a range of voltages," Mason said. "It's an interesting use we didn't see. For example, this part could supply an MCU with 10mA at 1.8V through the D/A port. It's an attractive additional capability."
The MAX11300 is programmed using a 'drag and drop' software package. "You can build a circuit in five minutes," Mason claimed. "Once you've done so, you save it and generate a config file, which you load into the device or into an evaluation kit."
As to who will be interested in using the part, Mason said customers would fall into two camps. "There will be grey haired analogue designers, who don't want to use an MCU because they do things using discrete components, and there will be MCU firmware people who don't understand analogue. We're playing to both camps, because both want an easy to use, deterministic product," he concluded.