Most of my time is spent writing about digital design and verification and I am about to start running a series on intellectual property (IP). Sadly, I can’t say that I am terribly surprised that I have had no submission on the subject of analog IP.
Why is that? We know analog IP does exist and probably should be in wider use, but for some reason it is another area where analog design and integration has stuck around in the old world. It's time that changed.
Clearly, without abstractions in the analog world, the only IP that exists is hard IP. This is IP that has already been through the layout process, has been characterized and verified on a specific foundry process, and is ready to be used. This means that the business is unlikely to be as profitable as the digital IP business where soft IP can be shipped and the layout and routing left to the chip designer. But then again, hard IP, which clearly contains more of the back-end effort, is probably not as cheap as soft digital IP. Some of these analog IP blocks contain very deep technical knowledge about the interface or subject that is unlikely to be shared by too many people in the industry, meaning that it is possibly a better implementation than most companies would be able to create on their own.
These days, almost every complex chip targeting cellphones, tablets, set-top boxes, automotive, and many others will contain one or more IP blocks. There may be blocks such as SerDes, PLLs, PHYs, DACs, and ADCs. An increasing number of sensors are being included these days, and this phenomenon will likely increase as additional chips are made for the Internet of Things (IoT). These chips require very high levels of integration and ultra-low power consumption especially since many of them have to scavenge power. Assuming that design times do not increase, there will be a growing pressure to start using pre-existing analog blocks.
The first question to come up probably is can we leverage a previous design and thus do internal reuse. This is probably going on a lot already and resembles soft IP in the digital world. You may have the design, but you will have to redo the layout, and it is likely that several other things will change between applications as well. This could include:
- different clock frequency
- different voltage
- new technology node that will affect the sizing of everything and thus all of the parasitics
- changes in chip layout that may affect signal integrity.
But there could also be changes in the requirements for the device that may require small changes in the design. As soon as this happens, all of the verification that had been performed goes out the window and you really are back to square one.
Many foundries supply a library of analog blocks, and there is a clear reason why they would like you to use those — it makes it more difficult for you to go to a different fab. Using these could also complicate the ability to get a second source if that is important to you. But there are third-party analog IP suppliers. Some of these are small companies that may give pause for thought, but large established IP companies also offer analog blocks. There are also companies that will customize blocks for your specific needs.
Given the many sources, what are the biggest issues or obstacles that you see preventing more widespread adoption of analog IP?