Anablog

In addition to the Darnell Digital Power Forum near San Francisco airport this week, there is the Embedded Power conference in San Jose. The shows overlapped on Wednesday but since most of my friends and associates were going to attend the Darnell conference Wednesday, I stayed up there and filed my story about those great presentations. I was able to make the Thursday Embedded Power conference. Unfortunately there were out of the conference proceedings and I cannot find them online so I will only be able to give you a rough outline from memory. This is the first year for this show and it was rather lightly attended, which actually turned out to be a benefit since both the speakers and fellow attendees were more available for questions and networking.

The Thursday morning session I went to was a National Semiconductor presentation about 2-switch forward and flyback converters given by Youhao Xi from National’s great Phoenix design center. The only downside to the presentations was that he spent too much time on very basic things like the definition of a single-switch buck converter, and that meant there was no time for the detailed analysis at the end of the presentation where he skipped 10 or 15 slides with the detailed performance and operating specs. From years as an analog consultant I can tell you what come to my mind when I hear flyback converter: Cheap, noisy and good cross-regulation between multiple outputs.  It turns out that the form of a two switch forward and flyback is very similar. By putting a high-side switch and two diodes in the circuit you can eliminate the snubber winding and get a better fill factor in your transformer. Better yet, the FETs do not have to withstand twice the bus voltage like a single FET converter, they can be rated to withstand the bus voltage alone. Helping this is the fact that the leakage energy does not appear as a spike across the FET, there are no turn-off overshoots. Another advantage is less EMI. I apologize for not having a schematic, as I said, there was no CD or web link I can find that has the presentations. Oh wait, here is the same presentation from almost a month ago from the same people that put on the seminar, and here is one from EDN from 8 years ago by a person that also works at National’s Phoenix design center. Good to see we scoop the competition by 8 years.

Since the 2-switch presentation was starting out so basic, I peaked into another track and saw Mike Speed presenting “Boosting Energy Efficiency through advanced Power System Design”. The title was pretty generic but the details are more clear when you realize that Mike is showing off Fairchilds’ power FETs. I did not stay too long but I am sure he explained one of the neat things I saw at another Fairchild media event—that you don’t use the same FET for high-side and synchronous FETs in a buck converter. On-resistance is the big parameter in the high-side FET, but low gate charge and fast speed is the beneficial parameters in the synchronous FET.

The next presentation was by Harry Zhang from Linear Tech. It was about buck-boost converters so I really wanted to see it since I just did an article about this topology. He was showing off the LT module part that could run a discrete inductor. By hanging a big inductor on it you can get efficiency around 97% over a large range of input voltages. Indeed a Linear Tech application note covers (pdf) some of the same material and shows a remarkable 99% efficiency at certain loads and input voltages. And here is a neat paper from Allegro on their buck-boost part (pdf).

Lunch was great, as we had a chance to gossip about analog companies and many of the people and bosses that have circulated though the industry. Rumor has it that Maxim is coming of the Pink Sheets and Intersil is hiring. I also got an insight into that outfit Powervation that I excoriated yesterday. It seems like their real mojo is not really diode emulation or duty-cycle knowledge. What they are trying to do is cycle-by cycle compensation for the loop with not just one, but two DSPs. I guess if cost is no object the switcher can run at a low frequency you can see the benefit. First off it would accommodate the aging and loss in value of the output electrolytic capacitors. Apparently a real value would be for module companies that could sell one module to all kinds of different applications and the module would tune itself for what every load and output capacitance the system has. TI and Zilker do auto-tuning where you can tune the system on factory-build or power-up, but this cycle-by cycle tuning is what Powervation is trying to patent and control,. I guess they saw what a success National Semiconductor had with their Simple Switcher line 15 years ago- an analog switcher that needed no compensation. One day I will hunt down Larry Sample who did the buck and John Bittner that designed the boost and ask them what tricks they used so the part was unconditionally stable.

I missed the sponsored session since I was having a great conversation with Dave Freeman from Texas Instruments. Dave goes back to the Benchmarq days and was able to reinforce all the good feeling I got when I interviewed Gregg Lowe last month. Dave had nothing but good things to say about Gregg including a couple of stories that shows Gregg really as a genuine person that is keeping Ti humble and customer focused, no small task when you are one of the biggest analog players in the world not to mention some silly, what do they call them Dee Ess Pees or some such thing. I guess us analog dogs are going to have to brush up on our DSP firmware if we what to stay up with digital power. TI offers both state-machine digital power that has an ARM core to do the monitoring and control, as well as having their C2000 series DSPs that you can configure to do things that may differentiate you product. The thing that impressed me about Dave Freeman was that he seemed to understand that digital; power faces a social and organizational problem, not a technical one. This is one reason to stay abreast of what TI is doing, since they have always had world-class development systems for their DSPs you can expect them to set the bar for software GUIs and such for digital power. Indeed, I had such a good time with Dave that we gabbed right thought the first afternoon sessions, so I only caught the end of Mike Thompson from Actel talking about using FPGAs for monitoring and control. Here is a link to an Actel presentation with the same title by Wendy Lockhart. These FPGAs are like the fast logic cousin of the Cypress PSoC family—you just define the analog that you want and now you can have a 5-dollar part that will not get obsoleted next year like Atmel did to me (twice). 

After catching the end of Actel’s presentation I went to see Dave Freeman and it pretty much the same presentation that I saw up at the Darnell conference earlier in the week. There was a lot of explanation of compensation and discussion of the kinds of systems where digital power can help. Once again, the server market where you might have a multi-phase VID supply for a CPU seems to be a good application, Dave showed one that had 16 phases. Another thing that came up was how digital control has been used in motor inverters for years and is nothing new there. As I have said before, it will be interesting to see if digital needs to keep up with faster switching speeds. Even the International Rectifier presentation about gallium nitride FETs admits that they only expect to capture a few percent of the market, so there may well be many many supplies in the kilohertz switching range that can benefit from digital power. I fell pretty good about what TI is doing since they have plenty of analog and both flavors of digital so they seem pretty agnostic about digital power. Dave obviously thinks it is a good solution or he wouldn’t be tramping around the country educating us about its benefits. But Dave is experienced and grounded enough to know that there will always be a place for stupid old analog power as well.

The only sad note is that Dave Freeman told me his is not going to be writing any more columns for Portable Design Magazine. Darn, I will miss him. Type “Freeman” into the search box in the link to see some of Dave’s stuff. Which reminds me about a nice compliment I got about my bad website editorial—we all agreed that TI does a great job; I hope your company does as well. One person at the lunch table told a story where his company said that they would institute training on how to use the web site. He asked who would give that site to the customers. And we all agreed that semiconductor company websites should be geared to the engineers looking for parts, not investors or media or even job postings.