Advertisement

Analog Angle Blog

Did You Say ‘Programmable Analog’?

Whenever I see a new product announcement saying an analog or mixed-signal component is “programmable,” my first thought is simple: “in what way?” Often, it's not called out clearly or until the end of the press release.

The reason I wonder is that “programmable” is a term that can take on many different realities. Back in the day, a programmable IC (such as an op amp or filter) was one where the user could set the gain, cutoff frequency, or other characteristics via external resistors. Soon, IC vendors incorporated these resistors inside the device, so the programmability was done by jumpers on the PC board.

There were pros and cons to this setup: On the pro side, lower total cost and a simpler BOM (no need to get your own resistors, which usually had to be tightly matched for best performance). On the con side, you were limited in your settings of gain or cutoff. For most users, the choice was easy: Using internal resistors with fixed settings was a very acceptable restriction versus the benefits.

Later on, programmable parts took on a different meaning. Users could set the operating parameters via a standard I/O port, usually I2C or SPI, rather than hard-wired with resistors or jumpers. In other implementations, the programmability was via a DC voltage or current on one of the IC pins. Either way, the programmability was dynamic and could be changed during run time, rather than static and fixed at the manufacturing stage.

Analog programmability can also be done in a way that is comparable to the digital world's FPGA. Instead of gates and logic functions, you have an array of analog functions and can configure their interconnect topology and some operating parameters in a one-time setup. (See Programmable analog ICs: designer’s delight or dilemma?)

Although the concept is attractive in many ways, it does not seem to have gained a large market share among designers, except in limited cases. I suspect there are just too many compromises that the user has to make, and the cost/benefit ratio isn't favorable enough. Once again, the analog world has a set of priorities and constraints very different from those in the digital world.

Our latest perspective on programmable analog comes with the availability of very highly integrated ICs, which combine analog I/O and peripherals with lots of digital functionality. Are these programmable? It depends on what you mean by programmable, of course. Many of these components are not programmable for general-purpose functions; instead, they target a very specific application. Consider these highly integrated mixed-signal IC examples:

  • Maxim's MAX14920/MAX14921, battery-measurement analog front-ends, which sample cell voltages in multicell packs, and provide level shifting to ground to simplify A/D conversion
  • Analog Devices' AD9671 octal receiver with 5 Gbit/s JESD204B serial interface, which reduces ultrasound system I/O data-routing requirements

Although they are programmable in terms of setup and operating modes, their internal topology is fixed by design. They can't be reconfigured or made to do something other than their intended, integrated function.

Is this fixed, non-programmable topology a limitation from the user's perspective? Probably not, since designers select them for a specific, well defined project, and not for broader, general-purpose use. In this sense, these highly integrated ICs (or we could perhaps call them SoCs, or system-on-a-chip, devices) are reducing a complex, multichip schematic to a single IC, rather than trying to simply pack more functions onto a given die.

Undoubtedly, there will be a few designers who wish these ICs did have more functional and topology flexibility. For the others, the gain in smaller footprint, simpler BOM, and higher likelihood of their overall product working both better and sooner is a worthwhile tradeoff.

Do you welcome highly integrated, functionally targeted ICs? Or do you chafe at the apparent restrictions and constraints in overall design flexibility they inherently must provide, to meet their functional and performance requirements?

Related posts:

19 comments on “Did You Say ‘Programmable Analog’?

  1. Brad Albing
    May 23, 2013

    >>you have an array of analog functions and can configure their interconnect topology and some operating parameters in a one-time setup…I suspect there are just too many compromises that the user has to make, and the cost/benefit ratio isn't favorable enough.

    That was my experience when I was looking at such devices for some possible applications in a CT scanner – not enough flexibility to fit the stringent requirements of my design. Apparently too many other engineers felt the same way.

  2. Steve Taranovich
    May 23, 2013

    I think that most highly integrated, functionally targeted analog ICs are born out of a vertical market need and sometimes even more specific than that such as in the medical market with an ultrasound system.

    Marketing engineers get feedback from suppliers to create these types of devices to save board space and cost, especially where there are multiple uses of a particular function (such as many ADCs) in a system. If the customer's idea will fit other similar systems designs and the revenue is high enough, then this becomes a product.

    When the IC is introduced and in production, that's when all the other worldwide electronics designers may come across the IC in time and see a fit in their particular system design. That's when the IC goes “viral”—when other users find applications for it that the original marketers and chip designers never even dreamed.

    This happened with the DSP and it has happened with so many other types of innovative, integrated designs.

    So, yes, I do welocme highly integrated, functionally targeted ICs–they can inspire the creative designing minds out there to use them in so many different ways, but they may not be right for everyone or every application, like a simple op amp might be.

  3. eafpres
    May 23, 2013

    Hi Bill–another direction of integration of analog that seems to just keep on going is embedded processors with ADC for sensing and related applications.  A great example is TI MSP430FR5724.

    This part has all kinds of digital features and memory, but includes analog-to-digital in the part.  From the datasheet:

    16-Channel Analog Comparator With Voltage Reference and Programmable Hysteresis
    – 14-Channel 10-Bit Analog-to-Digital Converter (ADC) With Internal Reference and Sample-and-Hold
    – 200 ksps at 100-μA Consumption

    Of interest is that according to their (really nice, IMO) configurator, you can get these for under $2.

  4. Brad Albing
    May 23, 2013

    OK, now that sounds pretty useful. I could have used that in several previous designs and might find a use for it in future designs.

  5. DEREK.KOONCE
    May 23, 2013

    Based on the start of the blog, I would say one of the first programmable ICs were the INA series instrument opamps. It required 1 resistor to adjust the gain.

    Then came the digital potentiometer devices.

    I would say that many of these new devices as such, stem from a specific customer's requirement to shrink the electronics, that resulted in a custom IC. The semiconductor company would then tweak the chip a bit, to allow the original customer to have what they paid for be the best, and then offer the IC out to competitors. Though this would be at least 2 to 3 years after the original customer got into production.

  6. Brad Albing
    May 23, 2013

    Yep – a very simple form of programming the gain via 1 resistor. Then when you use a DCP and control it via an I2C bus, your programming starts to get a bit more sophisticated. 2 DCPs -control gain and offset. Add a mux for multiple inputs and control it with I2C too. Even  more programmability. Assuming this is all on one chip, it's also an example of integrated analog.

  7. Steve Taranovich
    May 24, 2013

    Roughly speaking if an electronics oriented problem can be solved digitally is should be because it is highly programmable using standard cells or the equivalent. Analog solutions by it's very nature is not programmable in the sense that digital cells are programmable because of the extreme uniqueness of each analog application. Analog solutions are required when some form of extreme performance is required. Furthermore the world is primarily analog therefore the I/O into the digital world has an analog component to it. The closest thing to programmable analog is IC design where standard models are used to describe circuit behavior. Programmable analog blocks would always be sub-optimal and wouldn't find much of a market. Programmable analog sounds like a plaything of some professor seeking grant money

    Mike Koen

  8. Steve Taranovich
    May 24, 2013

    I am in a way with Mike, but slightly disagree with his thinking of programmable analog blocks. They are as sub-optimal or ineffecient as an ordinary FPGA is. It all depends on the utilization of the blocks and the fact that the few solutions that have been seen in the market until some time ago where either lacking major blocks or where simply oversized. With the appearance of Silego's greenPak mixed signal arrays I personally think “programmable analog” became interesting again. With todays modern CMOS processes the shrinked die sizes make programmed analog becoming an interesting alternative if you can live with low voltage! And this is not just a professor looking for sponsorship, there is a reasonable profit oriented company behind the idea which I personally believe will fly.

  9. Brad Albing
    May 24, 2013

    I think I'll apply for a patent.

  10. Dirceu
    May 24, 2013

    “Do you welcome highly integrated, functionally targeted ICs? Or do you chafe at the apparent restrictions and constraints in overall design flexibility they inherently must provide, to meet their functional and performance requirements?”

        Maybe for products that do not have large production (special projects) or requiring a short time to market, IMO. In other cases, I think can be a way to abstract the project peculiarities to a team with no experience in that area. Also, when using a very specific product, we can get tied in a particular manufacturer or the cost may be very high (or not, since all hard work is already done for us). In some cases there is no alternative. As examples I can cite some motor controllers from IR, TI , Allegro or single chip electricity meters. These SoCs do almost everything for us.

  11. Brad Albing
    May 28, 2013

    Dirceu – So you're saying that the use of integrated analog ICs might be good for designs that will not have large production volumes? I can see where that is a valid use of the devices to get a product done without the need for much design engineering time, but it would seem valid for products with large manufacturing volumes, too.

  12. Dirceu
    May 28, 2013

    Brad,

        I was trying to answer that question at the end of the blog from my point of view (more worried about the cost), but I agree with you: There should be numerous reasons to use this type of integrated circuits, regardless of volume.

    Thank you for your remarks.

  13. jkvasan
    May 28, 2013

    @Bill,

    ADS1298 from TI is a highly integrated IC and has unique programmable interface. Though it aptly suits ECG and EEG measurements, in principle, it can be applied to multi-channel measurement applications, in general.

  14. Brad Albing
    May 28, 2013

    OK – I understand. Thanks for the additional comments.

  15. Brad Albing
    May 28, 2013

    JK – I'll take a closer look at that part to see what it does and to see if I can think of additional ways to apply it. Thanks.

  16. jkvasan
    May 30, 2013

    @Brad,

    'So you're saying that the use of integrated analog ICs might be good for designs that will not have large production volumes?'

    For example, the IC from Maxim MAX31855 Cold-Junction Compensated Thermocouple-to-Digital Converter is a teriffic one when used in a temperature control function with niche products. When it comes to the low cost temperature controllers (chinese ones cost around 12$ ) used in furnaces, the cost of this IC could make the manufacturer think twice.

    I would, however, use this IC in a deep freezer used in a clinical laboratory without a ny hesitation. The functionality and not the cost is relevant in this case. As designers and customers, we may have to find a way on bringing such nice chips to common applications.

  17. Brad Albing
    May 30, 2013

    @JK – that Maxim part is a nice one and takes a lot of the difficulty out of typical temperature controller designs. I note that you said you would certainly use this part in a refrigeration (freezer) application. Did I understand you correctly that would perhaps not use this in a furnace application?

  18. jkvasan
    May 31, 2013

    @Brad,

    I was just mentioning an application in clinical laboratory where tissues culture will be done at a controlled temperature of upto -40 degrees.

    I was not meaning that the chip may not suit higher temperature measurement.

     

  19. Brad Albing
    May 31, 2013

    OK – sounds right. I know that Maxim part is spec'd up to 125deg-C, so that makes it quite good for even the most stringent inductrial application – and again, I'm thinking of guys like Rockwell who always want that temp range – even if they don't really need it.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.