Anablog

My IC designer buddy Don Sauer copied me on an email he sent to a fellow asking about the art and practice of IC design. Like The Unwritten Rules of Engineering, an ASME book from the 1940s, Don points out that most problems are organizational, not technical. I am less pessimistic than Don regarding the ability of people to come up with interesting applications. There are a lot of clever people willing to apply or mis-apply technology, I am compelled to think of engineer Suzanne Nell, who has provided Design Ideas that used a linear regulator as an audio amplifier as well as other novel circuits. I also encourage the fellow that wrote Don to check out THAT Corporation, a company that bought up the silicon part of the old DBX Company. They then bought the old Sipex fab in Milpitas, a nice counterpoint, a fabless company that buys a fab to provide analog technology not available at the CMOS digital fabs. THAT has some neat parts that may be useful for analog synthesizers. First, here is the letter that Don received:

I was referred to your website by a discussion on a mail list I get called synth-DIY. The topic was your section on your design of the LM13600/700. As you have this background, I wonder if you could answer a few questions.... Years back us synth geeks were in love with custom IC's made by SSM and Curtis that modularized common subtractive music synthesis functions; like the VCO, VCF, envelope generators, and VCA's. Almost all of these IC's have gone out of production - mainly because of the switch in interest from analog synthesis to digital synthesis. But now the analog synthesis has experienced a resurgence. What has happened to IC design in the meantime? - Can such IC's be now re-introduced in a more cost-effective process - or perhaps done by some small company in Taiwan or wherever? I am slightly familiar with the PSOC chip that Cypress has - that can be configured in programmable fashion to do some interesting things. Could something be manufactured along these same lines that would include the frequency stability, v/oct. accuracy, and functionality of these older IC's, but at a price a small company could pay?

Even if a company were to re-introduce a LM13600 type IC now - have costs gone up or down to do so? Is the reason many of these IC's - like the CA3080 - have been EOL'd that the processes are no longer compatible with current ones (equipment obsolete and too expensive to maintain etc.)?             Thanks, Barry

Now Don replies:

The problem is working on anything like an LM13600/LM13700 has always remained the same. Almost nobody can see anything from a technical potential viewpoint. For people to be able to see a large enough market such that they won't stop you, you have to have the product out creating its own market. I lucked out with the LM13600/LM13700.

The silicon costs have really gone down and the quality has really gone up. If you look at the layout of the LM13700 in my web site, a minimum NPN is now about the same size as the little dots that are emitter contacts in the LM13600 layout. The speed of the LM13600 used to be defined by the ft of the lateral PNPs, which was around 2Mhz. Today that ft is greater than 250MHz. The dynamic range for current used to be from milli amps to nano amps. Now it goes from milli amps to pico amps. And the BiCMOS process is common. Now you can build buffers for operational transconductance amplifiers which have the input impedance of glass.

I know analog and digital, bipolar and CMOS, hardware and software. I too am sensing there may be some unexplored opportunity in the area of continuous-time signal processing. But even after you discover and develop a great opportunity that is out there just waiting to be exploited, most people in the electronic industry don't have enough technical experience to really understand the potential. I may have known, in my 30 years of IC design, less than a handful of managers that could understand. So the real challenge here has always been human nature and not the technology.

However, a significant numbers of barriers have been removed. During the time of the LM13600/LM13700, the cost of producing silicon was very expensive. Now a days, a design engineer like myself can afford to personally design, layout and tapeout a GDSII file to be added to a "test wafer" for some local or offshore IC fab. Today I can afford to personally produce my own working silicon. The price for testing hardware has greatly gone down as well. With the web, the marketing of datasheet and application notes has removed the cost of dealing with publishing companies. I would guess that the total cost of designing and manufacture an analog IC is maybe less than 5% of what it was during the LM13600 days. This is providing of course you have people doing the work who know what they are doing.

You are asking for an IC for a small company with price they can afford. This puts you out of the market for companies that are in the high volume/low price business. You would have to prove that there will be a large number of small companies which will buy an IC to just get their attention. One the other had, there is a huge reduction in resources needed for developing an analog IC today. And you can do incredible things today that you could only dream about in the LM13600 days. Can one make a profit today in new ways by just thinking small?

Don had told me how SPICE, which when run with VBIC models can even do a half-decent job of predicting temperature behavior of an IC, can make the design process much easier. What needs to be mentioned is that in order for you IC designers to have an easy time of it a company does have to spend tens, if not hundreds of millions on CAD, Process, models and tools. Still, Don and I plan on showing how to design an analog IC on the cheap as well, which would be perfect for the fellow that wrote in.