Advertisement

Analog Angle Blog

If Integration Is the Solution, What Was the Problem? Part 1

The shift towards higher levels of component integration is not a new trend, as I noted in a previous blog, but it has taken on increased momentum. (See: Getting Over My New Integration Fears.) There are many reasons why designers look to an integrated solution.

Before we look at those reasons, what do we mean by “integrated,” anyway? In some cases, it's a chip set comprising a set of standard and application-specific components, selected and combined by a vendor, and which work together to provide a good fit for a given application. As a result, on the hardware side:

  • You no longer have to pore through websites and selection guides trying to find the “right” or “best” component for a function;
  • You don't have to make sure the various ICs you've selected are compatible in signal levels, I/O, power rails, and timing;
  • You get a reference design, PC board layout, and basic evaluation/exercise code, yielding hardware-related confidence;
  • You don’t have generate the bill of materials (BOM) from scratch, and with purchasing asking if it is better to go with ICs from one vendor or to instead use a mix of suppliers (whichever way you go, they'll argue at the design review that it should have been the other way, I assure you).

The other kind of integration is when the vendor offers a single, larger IC, which directly incorporates many of the functions of the application (in some cases, of course, more than a single IC is needed, but you get the idea). This reduces cost and PC board footprint, and it may limit your implementation flexibility or options — but you may not care about this constraint.

Either way, whether you choose integration door #1 or #2, you hope to reduce design headaches and time to market (TTM).

Why so? Reality is that today's applications are so ambitious, with user expectations so high, and with such aggressive cost and TTM goals, that such integration is probably the only way to succeed for many designs.

Don't think I'm denigrating the efforts or outcomes of earlier generation of scientists and engineers, not at all — in fact, I stand in admiration of them. For example, in the 1950s, they were able to determine the spin moment of the electron to eight significant figures using laughingly crude instrumentation by our standards (and their first six digits are still considered correct, by the way).

But there is a major difference: these were specialty projects, not mass-market, low-cost products popped out by the tens of millions. The world of electronics has gone so far beyond the few mass-market products of that period — it was primarily five-tube AM radios and a some early monochrome TVs — that the product-design tools, techniques, and engineering approach that worked then would simply not be viable now.

Do you feel the availability of higher levels of component integration, along with the necessity of using it, makes your design task easier? Or do you think it constrains your flexibility and creativity?

Next, we will continue exploring more details of analog integration.

Related posts:

14 comments on “If Integration Is the Solution, What Was the Problem? Part 1

  1. Davidled
    May 3, 2013

    Integration of high density functionality chip provides a simply and less BOM and less effort of PCB layout. But, I am concerning about the reliability of the chips including all functionality. Secondly, there are more possible defects in inside of customized chips, even though internally chip validation is executed in the vendor.  Troubleshooting is not easy for chip itself. Lastly, I am rethinking for process integration. Process integration looks a good approach to get TTM. It makes sure that all process of product should be thoroughly reviewed in the top to bottom to get a quality product in the timely manner.   

  2. Scott Elder
    May 3, 2013

    With all the analog PCB engineers on line, I wonder if they can chime in and tell us what can be done better on a PCB with parts than in an IC.  Better can be cheaper or higher performance or both.  I'm simply curious.

  3. Netcrawl
    May 4, 2013

    Interesting topic, thanks for sharing us. My concerned is on chip reliability, I'm sure we got some plenty of defect here, and also the possible drawback on functionality, that's the nighmare of customized chips. 

  4. goafrit2
    May 4, 2013

    Integration is the solution because cost is the problem. The more you integrate, the more you save cost. That is the way I see it in the industry and that is what is driving the market. The challenge though is that too much integration cripple a system as when failure hits, nothing can be saved.

  5. Netcrawl
    May 4, 2013

    @goafrit2 you're right, it something to do with cost, integrating more we could solve one big problem here-the cost. But it has a serious drawback on product's reliability(I do think). 

  6. amrutah
    May 4, 2013

    DaeJ,

      “But, I am concerning about the reliability of the chips including all functionality. Secondly, there are more possible defects in inside of customized chips, even though internally chip validation is executed in the vendor.”

       First the process should have good Yield, it that is right then the defects that you are talking will be a small factor.  every component has a FIT rate and having most of the devices integrated on a chip and with a good FIT rate should make the system solid and more reliable.

  7. amrutah
    May 4, 2013

    ” Troubleshooting is not easy for chip itself.”

      Any integration is complex and integrating many functionality on a chip is further complex.  Testing and validating the chip becomes very complex, but I think that will be the chip vendor's headache.  The org. or person building a system can concentrate more on making the product a success.  This will help reduce the cost and TTM.

      But when it comes to integration of many functionalities there needs to be a through understanding of the system so that all the possible scenarios is considered building a chip.

  8. Davidled
    May 5, 2013

    Chip performance might be slightly different depending on the CMOS process used to fabricate all circuit such as cascade amplifier. They try to squeeze all circuit block as minimizing package size of IC. Secondly, I am concern about signal noise inside of chip. If they could provide output port as noise measurement and information and offer tuning parameter, it will be beneficial for customer in all aspect. By tuning parameter, all circuit function is changed to meet system goal.  Application engineer would access parameter and changing it via PC through USB, with a nice GUI. GUI will display all behavior of circuit inside chip.

  9. Brad Albing
    May 14, 2013

    DaeJ – it's always good to be concerned about IC reliability and to have your Quality Control people monitoring it. And you're right that it's tough to trouble-shoot defects inside the IC. But I'm confident that you will see better reliability with an integrated device that you would with the equivalent circuitry built with conventional ICs.

  10. Brad Albing
    May 14, 2013

    Goafrit2 – I agree with your point about more integration will save/lower cost. To your point about “too much integration [could] cripple a system… when failure hits [and] nothing can be saved,” a quick comment. While I guess it's true strictly interpreting what you stated, I think you'll find that since your failure rate goes down overall, even with the occasional board that goes bad, you will just scrap it out and still be money ahead.

  11. goafrit2
    May 15, 2013

    >> But it has a serious drawback on product's reliability(I do think). 

    When you wholly-integrate, it is work or nothing since you cannot salvage anything. The risk model increases as the chip must perform as there is no other way to salvage it.

  12. goafrit2
    May 15, 2013

    >> But it has a serious drawback on product's reliability(I do think). 

    When you wholly-integrate, it is work or nothing since you cannot salvage anything. The risk model increases as the chip must perform as there is no other way to salvage it.

  13. Brad Albing
    May 15, 2013

    See my other reply time-stamped 5/14/2013 2:38:12 PM. I think it's less of a concern than you make it out to be. You have to evaluate the cost*probability for failure vs. success.

  14. goafrit2
    June 3, 2013

    >> But it has a serious drawback on product's reliability(I do think)

    Hopefully as yield improves and processes mature, integration becomes very compelling. I hate it when yield is low and one is packaging everything in one die. I like the MEMS guys that are not crazy over the quest of moving to new processes pursuing feature size as old processes with long channel devices just work fine.

Leave a Reply

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