I probably see 50 or more announcements every day for new chips.
Most of these are digital in nature, although I am sure they contain some analog content. Most of them start with “world's smallest,” or “fastest,” or “lowest power.” Then they talk about the function that it performs. Many times the function is a commodity item, so they need to find some way to differentiate and show their chip as being special. They don't want to lead with other attributes such as “cheapest,” or “best trade-off” of all of the above. When it comes down to it, they are all trade-offs unless some engineer really does find some new way of doing something that has been missed by everyone else.
When I see the occasional release go through for something analog, and particularly when it is supplied as an IP block (which can mean integrated analog), it talks in much different terms. Size and power are discussed of course, but they are secondary to such things as noise immunity, sensitivity, and the process on which it operates. This last one would appear to be a substantial limiter to the success of an analog IP company due to the large number of processes available and the need to integrate it with other analog and digital content.
The companies specializing in analog IP tend to focus on PLLs, interface logic including SerDes, and of course DACs and ADCs. But compared to the number of suppliers of digital cores, their numbers are small. As a counterpoint, Allan Chin, CEO of Stellamar, in October 2012 said that he believes the analog IP market will grow more than 17 percent through 2015. It was not clear if that was 17 percent total or 17 percent per year. At 6 percent per annum, it would be nothing to write home about, but 17 percent per year for three years is a very substantial gain.
Analog blocks have to be shipped as hard cores — or as fairly hard cores with some amount of custom tweaking added as a service. While not always the case, you could think of the hard cores as the integrated analog parts. If tweaking can or must be done, you can think of the core as either programmable or as customizable. A discussion of programmable goes beyond the scope of this blog — just note there is a small amount of programmable analog being manufactured. For some thoughts on how customizable devices can be customized, see Design a Production-Ready Custom Mixed-Signal IC on Your Couch.
Most of the customizable devices will likely be considerably more expensive than their digital cousins. This makes the build versus buy decision a little more difficult. However, it appears that there are a limited number of analog designers available and with newer nodes being targeted these days, increasingly sophisticated solutions will be required.
Krishnan Ramabadran, senior president of marketing at Cosmic Circuits, recently wrote that analog IP has to be adopted in order to reduce project risk. He goes on to note that using the performance of the digital transistors also means having to accept the lower voltages that they use, but this is a necessary thing to do if the area occupied by the integrated analog functionality is not going to grow as a percentage of the total chip area.
The same is true for power consumption and interface cells. Standards are being defined with lower voltages as well necessitating the need to find better solutions. Incidentally, for some thoughts on how integrated analog devices can affect power consumption, and why that's a good thing, see Analog Integration Is Saving Power.
As another wildcard for the future, FinFETs may make for an interesting analog transistor structure. We can explore those devices in a future blog.
I will finish up with three questions. Do you use analog IP today? Do you expect to start using it in the near future? What are the biggest challenges that you think need to be overcome in order to make analog IP more reusable?