Research purpose for group of students Technical University of Denmark

[RVale.4]

Hello, we are a group of students at the Technical University of Denmark

We are currently working on a project, regarding the company Lotus Microsystems, which would eventually be able to produce smaller sized power converter that would have better lasting quality.

We would like to ask if we could have a little bit of help to answer some quick questions that will help us write this project.

- In connection with todays semiconductors, we would like to ask if there would be interest for a company like STmicro to receive a license for the production of semiconductor (power converter) of a tiny size which would have the interest of prolonging the life of a product with the help of silicone implementation. The size would be smaller in today's power converters, which would result in implementing multiple converters for products of smaller size, such as earbuds, as well as giving the opportunity for the implementation of newer/more features

- If in the future a company like STmicro would have access to the production of a power converter of a tiny size, what would their potential market be, that they would supply

If anyone has any input regarding these questions, the help is very much appreciated !

Thank you in advance

[KiptonM]

I once worked for a large contract manufacturer, and we built a certain kind power supply for a very widely used commercial application and we already had a factory building these power supplies at between 10,000 and 80,000 per month. We with ST developed a custom power supply with 14 pins (if I remember right) the processor inside was based on an 8051 and there was a lot of hardware pulled into the chip. It was an amazing design. It was very efficient and cheap, because it had a very simple processor and was easy to make. And it worked well. It had Power Factor stage, then an LLC Resonant Power supply.

Unfortunately, the upper middle management changed and the engineers that were going to use the chip in Asia did not want to learn the new technology, and as a result the company not only did not use it but ended dropping that whole line of products (and many millions of dollars in revenue) because without the chip they could not compete with other power supply companies in that market.

The results are that you may be better off finding someone who actually builds power supplies in volume and show them what you have and see if they are interested. If you are not in the industry, you may have a very hard time selling your idea.

In our case we already had a power supply of a similar design in production. The new chip reduced the total parts count from about 215 parts to 78 parts. In addition to the power savings, especially when the power supply was plugged into the wall, but the device under power was not using any power. There was a tight specification to minimize power when it was plugged in, but the device we were powering was turned off, and the batteries it had were charged.

Maybe build the power supply with discrete parts, and then show how you could crunch it down. The fewer pins the package has the cheaper it will be to build. The simpler the processor the cheaper it is to build. We measured cost in the hundredths of a penny. $0.0001 because with the high volume it made a difference.

You have to look at the entire cost of the finished power supply. A lot of people go with high frequency switching to try to make it small and forget passing the EMI requirements are a big deal, and the filters needed to pass EMI are bigger, takes more parts, and result in more expense with the higher frequencies, which negates the cost and size savings of their "small" power supply. Do not forget it has to have an isolated output.

If you are using a 16-bit or 32-bit processor you will automatically be too expensive for high volume. I know that is not sexy for students, but it is what it has to be.

I did do one with a custom very simple RISC 16-bit processor with a 16-bit hardware multiplier peripheral that produced a result in 16 clocks. The good thing was I could get the peripheral loaded with the next multiply while the current multiply was working to maximize throughput. I needed this for the PID loop (actually only PI, no D) because it ran at 50 kHz. I ended up writing the PID routine in assembler because it was called so often and was the most computationally intense part. I discovered probably the ideal processor would have 24-bits for the PID, (for what it is worth). The rest was housekeeping and could be handled easily with an 8-bit processor.

Good luck. Prove the entire concept first with discrete parts (including input and output EMI filters, and do not forget about reducing input surge when you plug it in, and of course it has to have Power Factor if it is over 75W (ours were in the 90W to 240W range) and then have a plan to reduce it with custom silicon, with very few pins. If you can get below 16 pins you are in the ballpark, and you can get even better for every pin you can get rid of. The other key is to not run the processor clock any faster than needed to save power. 16 MHz is better than 32 MHz. 8 MHz is better than 16 MHz. 4 MHz is better than 8 MHz. etc.