History Indicates Analog Integration Is at an Inflection Point

A quick look over the past 50 years of analog design brings us the realization that we’re now at an inflection point where analog IC manufacturers are running out of volume applications for integrated analog, and now it’s time for We the People to take integrated analog to the masses. By taking this approach, we can identify ways to leverage the work of others for future accomplishments.

My background is in test instrumentation, so allow me to discuss integration from the perspective of a designer and user of precision circuitry. In this field, precision front-end amplifiers are quite important. These amplifiers should not drift with time or temperature. They’re often needed at various bandwidth and power levels and with various input characteristics such as leakage current or input capacitance (depending on the instrument). In an error budget, we find the specific sensitivities of these amplifier characteristics to changes in basically everything , such as open loop gain, common mode rejection ratio, power supply rejection ratio, feedback resistor tempco, thermal noise, self-heating, time drift, and on and on.

Now imagine the first test equipment designer to take advantage of the modern op-amp (compared to companies building operational amplifiers with transistors). They would benefit from 100x improvements in VOS drift (through better matching), higher and more stable loop gain to desensitize the amplifier to many ills (except for the beta network stability), as well as lower cost.

Nobody can build a discrete op-amp and achieve the VOS drift that the vendors can. The truth is that in some cases (as with the example of the op-amp), the integrated solution is a better solution. In some cases, the integrated solution is the same (considering only performance, not size and cost). And in some cases, performance is inferior to a discrete solution (think about many power applications).

So, here is the rub: For all these cases where the integrated solution is the same as, or better than, the non-integrated solution (assuming the analog vendor defined the part competitively, and it works as advertised), the integrated solution will be less expensive and will be easier to design in.

What a gift! Think about this for a moment. Of all the possible products you could decide to design, why not choose one previously screened by a giant like TI, Maxim, ADI, or LTC? In the case of analog integration, the vendors have chosen an application for you and completed most of your product definition and most of your analog design — and it all costs less than what your competitors are paying to design and build from scratch. Analog integration gives the first organization that uses the integrated part a distinct advantage over everyone else. The advantage depends upon how integrated the solution really is, as well as how well it was implemented.

From the perspective of the IC vendor, though, there is a catch: the more integrated the solution, the lower the volume for the analog vendor.

Consider that analog IC vendors have been in the business of integrating all along. The first IC vendor, just like the instrumentation company that used an op-amp (over discrete transistors), realized a great advantage over its competitors. Consider also that the new ICs they are marketing now are beginning to resemble completed products. In my opinion, our analog vendors are now running out of the high sales volume options for analog integration. As a result, from a marketing perspective, the advantage is shifting to us, the purchasers/users of the devices.

So, don’t you think that it would be wise for each of us to not only pay attention to all this new analog content, but also to contribute in our own small way?

5 comments on “History Indicates Analog Integration Is at an Inflection Point

  1. Scott Elder
    April 26, 2013


    You put together an excellent perspective of the issues at hand.

     I believe that the next turning point will be IC manufacturers providing soft IP and manufacturing support/services rather than internally defining a system solution with the hope that this product will have a large audience.  As you point out, the higher the level of integration the narrower the market.  I just don't see it working long term inside an IC company unless they are like Intel.

     Contrary to what many others will claim, prototyping a custom IC solution is very easy and cheap once a database is available.  It is the generation/qualification of the database where all the costs are incurred (i.e. product definition, design, simulation, layout, validation, qualification, etc.).  Physically making the IC die is cheap by anyone's definition. Even at sub 100nm when using MPW services to share costs.  If you don't believe this, read the annual reports of analog IC companies that explain where all the R&D dollars go – bodies and software tools, not mask costs.

     And one can already see this taking place.  First of all, the tool manufacturers like Cadence and Synopsis are slowly acquiring all types of IP including analog IP.  And then at the other end, you have companies like Apple that are hiring expert IC designers to architect unique proprietary IC systems.  Finally, and in the middle, you have IC companies that employ designers to take someone like Apple's architecture and putting it into silicon for just Apple.  Where is the value added for the IC company if the customer is inventing the system architecture?  The answer is nowhere because the patents on how to do opamps, data converters, etc. are all slowly expiring.

     Summarily, and in regards to analog IC companies, it will be those companies that have the transportable soft IP, and the skills to quickly turn that IP into a proprietary system for a system company, that will win.  They will still sell high volume proprietary silicon real estate; just in the form of placed soft IP not individual pieces of packaged hard IP (i.e. an 8 pin OpAmp). 

     I don't buy this idea that a handful of analog IC companies can employ experts in all disciplines of analog system engineering in the world.  Not reasonable.  Especially when expert designers can earn more money at a system company where the profits are measured in billions not millions.

  2. Slogan
    April 29, 2013

    James, nice article. The thing that seems most interesting to me is what exactly to integrate. Is it a PGA with an ADC? Should the ADC be a 16-bit SAR or 24-bit delta sigma? Does the digital isolation need to be integrated? And with all of these, there a number of different options for monolithic designs or mulit-chip modules for the big analog vendors. 


    The other catch is that the industrial and instrumentation markets often appear somewhat fragmented to me. So it can be tough to design a single integrated analog device that even hits 20-30% of the broader industrial market. But I suppose that's what makes this interesting, keeps the analog vendors designing new things, and coming to companies like Keithley to really value your opinion on what kind of integration solves your largest problems.



  3. Brad Albing
    April 29, 2013

    Steve – part of that (your questions in paragraph 1) goes to a marketing issue – or to the issues that the Marketing guys need to figure out. If there can be a way to do the multiple analog functions on a chip or in a module – and (better yet) make it easily and cheaply configurable during the manufacturing process – well then you've found the golden fleece.

  4. David Maciel Silva
    April 30, 2013

    If we think of integrating the application level have new integrated circuits still being developed.

    An old item is very common example is the TCA785 a stage controller which has been widely used before the uC expand considerably.

    The point x is integrated so that even a small proportion other associated components are used, not only passive components, but also on the nature and logical interfaceavel uC and digital circuits.

  5. Brad Albing
    May 2, 2013

    Scott – some excellent info in your reply – worthy of a blog in itself. So I'll likely borrow some of this info and use it in a blog in the near future. Thanks.

Leave a Reply

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