Principles of Top-Down Mixed-Signal Design
4 of 31
The end result is that analog designers have no equivalent to RTL, a relatively high-level language in which they can describe their design and from which they can synthesize an implementation that is guaranteed to be functionally correct and have near optimal per- formance. As such they must transform their designs from concept to implementation by hand, and so the design process is naturally much slower and more error prone that the design process for digital circuits.
The outlook for providing the equivalent to logic synthesis for analog designers is bleak. However, things cannot continue as they are; the current situation is becoming untena- ble. While a complex digital chip can be designed correctly on the first try in a few months, designing a complex analog chip can require 3-4 respins and up to a year and a half to get right. This is problematic for many reasons:
The tremendous mismatch in schedule and risk between the analog and digital por- tions of a mixed-signal makes it difficult to justify combining analog and digital on the same chip.
The high risk makes planning difficult. It is hard to predict when product will be available, and when valuable analog designers will free up.
A long time-to-market makes it tough to react to changes in market trends and com- petitive pressures.
Analog and mixed-signal product development demands large investments of time and money. This makes it difficult to justify developing new analog products, espe- cially in tough economic times.
Analog and mixed-signal designers are scarce and hard to recruit. Compounding this problem is the inherently low-level of productivity of the current mixed-signal design process, which makes it difficult for small design houses that are not focused on analog to field an analog design capability.
Some mixed-signal designs are becoming so large that, with the low productivity of the analog design process, a team of analog designers that is large enough to take on the project and complete it in a timely manner simply cannot be assembled.
Clearly a change is needed. It is interesting to note that when digital designers were try- ing to design systems of a size comparable to today’s mixed-signal designs, their design process was not that different from what analog designers are using today. But it was at that point that they began to transition to a more structured and more automated design methodology. For analog designers, substantial automation may not be in the cards in the near future, but the need to transition to a more structured design methodology that is both more efficient and that allows designers to handle the growing size of analog and mixed-signal circuits is clearly needed.
The availability of logic synthesis tools was not the only enabling factor for digital designers to move to more efficient design methodologies. By moving to FSM and RTL, digital designers also gave up considerable performance in terms of speed and power. They made this sacrifice to be able to design the larger and more complex systems quickly. This sacrifice was a critically important enabling factor. Analog and mixed-sig- nal designers have not demonstrated the willingness to make a similar sacrifice. In those cases where performance is not critical, the tendency is to instead convert the circuit to a digital implementation in order to gain flexibility. In the remaining cases sacrificing per- formance is not an option, however it is also not clear that such a sacrifice is needed. Analog designers do not have the equivalent of logic synthesis, so they will continue to use custom design methodologies. While moving to IP (intellectual property) reuse may
The Designer’s Guide Community