Digital controllers have been promised for a long time, and we are starting to see some inroads in real products. How many of you are using digital? And what kind of projects are you using it on?

Do they live up to the promise?
What are the surprises encountered?

I look forward to hearing your experiences.

We use digital controllers almost exclusively. We have done for perhaps four years. The main reasons for this is that the flexibility in the control lets converters run at very low losses at all operating conditions and modern low loss converters show significant plant variation with operating point.

Typical surprises encountered are problems with digital precision, timing precision, processing delay, signal phase shift, digital filter group delay, non linearity of digital system and vendors who sell solutions that are almost OK but not quite.

In terms of what you can do with digital control yes they live up to the promise.

In terms of the new mistakes you can make they exceed most peoples, and most people's managers, expectations.

Interestingly the use of digital control and software control of power electronics is long established in the DC and AC variable speed drives industry (early 90s). Lots of inspiration and solutions to the problems in digital control have been sorted and can be found there. It is worth a look for inspiration.

With new complexity of the phase shift and zvs topologies , digital control seems to be provide much needed adjustment for given application. I have not used myself digital controller yet but I have involved in some project. What I have seen , future will be dominated by digital controller. We need to align our self and embrace it.

I have worked on and used digital for a long time now, since 1992. First products shipped about a year later. It has it's place; especially in complex power systems where fault logic, sequencing and telemetry are required.
If you just need 3.3V form 12 or 5 use a simple analog controller.

A couple of years ago I led a project to develop a wireless inductive charger for electric vehicles. We were pushing 4 kW across 3-4 inch air gaps. We used a digital controller because we had to transmit the error signal over a wireless link from the vehicle to the charging station controller. The transmission delay was (relatively) very long so we ended up with a very low loop bandwidth. The bandwidth was not an issue as we were charging a battery and had a fairly large capacitor on the output of the vehicle mounted secondary coil and rectifier. Also, because the bandwidth was low, we used a simple PI controller algorithm. The team successfully demonstrated the prototype to the client. We haven't heard anything from them since but they are apparently still in business.

So here is a question for you digital-experienced designers.

If I have a 300 W or so offline converter with all the control connections ready to go, and its running with a conventional PWM control chip today, where would would you go to get it "digitized".

Who makes the most accessible part, and has solved noise problems, interconnect problems, programmability problems the best.

I choose offline as an example since it has a lot of noise, and needs good protection.

I have a patent in digital control from the 80's. I developed it before there was a micro that was usable for practical product and relied on discrete logic at first. My intent was NOT to mimic an analog solution.
Ray, You asked the question, "Who makes the most accessible....?"
This may not be a good time, yet. The digital controllers today are still just mimicking analog schemes, such as PID.
The number one limitation on digital control is the analog peripherals providing the sensing.
Number two is the available MIPS.
A superior control IC may be best constructed from an FPGA that allows "soft cores" and makes advantage of parallel logic paths. The analog portions depend on the technique desired... my preference is not to use ADCs or fixed frequency.
What is cost-effective is another conversation.

In any converter I would still have hardware protection. So if I add some logic and sensing circuitry would it compete with the custom IC?
On the other hand, if the control law is complicated and it requires, for instance, table lookup, would this justify DSP or PGA or micro per se?
Thirdly, one can not usually choose different micros for similar task - you chose what
seems to be most suitable/universal to some degree and stick to it, unless you want to utilize numerous incompatible development suits. Sure it can be small lead count micro for axillary purposes.
I did developed custom sigma-delta ADC to fight the noise but nowadays components are so much better - no need for it.

PWM timer resolution is one of those issues where the controller device manufacturers have responded to the market but the market really does not need it.

Digital timer based PWM modulation is a precision reduction to a lower number of bits The solution to this precision loss is to use a precision retention system. These systems typically trade short term (switching frequency) precision loss for increase long term or DC and low frequency precision. The high frequency switching spectral components are slightly spread and the low frequency precision is now significantly higher. Simple to implement in any micro or DSP and one of the enabling little pieces of the early digital motor drives industry.

The precision increase from the device manufacturers to this PWM precision loss is an unnecessary blunt hammer.

I'm not prejudiced - the digital guys are going to take over many, many areas. Their strengths are:

* flexibility to changing power stage characteristics

* communication with whatever the mother ship might be. this is a BIG deal.

3) tweaking of design performance at the various operating points though this will require a LOT of prep work and empirical testing. Don't underestimate this.

 Sometimes I think doing analog is like supporting the LP versus the CD - a lost cause if you read the digital proponents!

But it is not like that at all.

In order for the digital to have anything to control, it must have a well-designed analog power stage. You can't avoid this, and if anything, it must be built better than it ever was for analog control in order to prevent the noise from getting in the way.

It must also be characterized and understood much better than before if you want to fully apply the digital adaptability to the system.

I suspect we will see a lot of digital power control guys coming to our workshop in the future. To rediscover the "lost" art of getting the power stage right first before they can do their magic (and some of it is truly magic!)

Already we see an interesting shift of attendees away from classic power supply companies. It started with Intel coming because motherboards were failing with inductor problems.

Then Microsoft showed up, they wanted to get a handle on their hardware design. And, most recently, Google and Amazon are coming to power supply courses - amazing!

I agree with Ray, the power section must be in order before the method of control and its subtleties can be tweeked.. The power circuitry defines the physical "plant" (from control theory). How one does this is to the imagination of the Si definers/designers and the end design engineer. The housekeeping functions need to be factored into the process. Many of you digiofiles get upset with differing data resolutions, processing time for digital noise problems, etc. As My Italian Grandmother-in-law (from New York City) used to say "Its life, get used to it!"
As a humorous aside, my Grandmother-in-law, when I was dating my future wife, asked me to fix her vacuum tube TV (a typical anode oxidation problem) and gave me a thumbs-up when I did it. Hey, if you can't mix a little personal stuff with the stuffy academic stuff, what the heck. Most of you will face this sooner or later.

Ray, I think you need to distinguish between different "digital controller" implementations. Broadly speaking I can think of 3 different categories:
1) Complete hardwired digital implementation just with the ability to change control parameters but not algorithm.
2) Complete DSP/Microcontroiller implementation with control all done in software for ultimate in flexibility and system level integration.
3) A hybrid between 1 & 2

Personally I only have experience with (2) which has been a very natural progression for me and my colleagues as we came from a variable speed motor drive background originally. For the kind of things we work on digital implementation makes sense ususally for one or more of the following reasons:
1) System level integration - demands and feedback are often via serial coms
2) Complex control (resonant and / or multiphase conversion) and the ability to add in all the bells and whistles such as optimisation of low load efficiency, frequency dithering for EMC etc
3) Flexibility so when (not if!) the customer changes the spec it can be more easily accomodated.

Should we get stick to our well known converter for developing new products?

In fact, it all depends of the phase of the moon, your boss, the design’s budget, the schedule and the product’s cost. But, for those who have the opportunity to evaluate new topologies, the DSP controller approach is worth to study.

Personally, this evaluation is part of my objectives and I’ll jump on it in a near future.

As usual this site always has interesting discussions. My design career started in the middle 60's until the present so I've seen Power Design go through a lot of changes, and the use of Digital is just another. I think Ray is correct when he says the major topologies haven't changed (or much else). My opinion is this adds another wrinkle to Power Design in general, but won't impact our day to day jobs. You still have to meet EMI, Efficiency and the myriad of other specs that are almost pure analog problems. You can't solve these from a keyboard:)

William, it is worth the time and money when you are between a rock and a hard spot like I was recently. When you can't use electrolytics and you have to "ladle out" energy in fairly accurate chunks into a variable load without messing up the PF. I don't know why one needs standards for everything, I am actually glad for every area in engineering where there isn't one. More freedom :-)

Anthony, fully agree, digital is not only easier in some areas, it is sometime the only practical way. I wouldn't want to design a grid-tie inverter that has to provide reactive power on occasion without some sort of controller. There will be more and more requirements for grid-tie and doing all that analog would result in a huge hairball of circuitry that cannot be upgraded when the rules change again. Hard to say this as an analog guy but t'is what it is and we all have to be willing to change.