The 2012 Applied Power Electronics Conference and Exposition (APEC 2012) is being held in Orlando, Florida this year, and Chipworks has sent two of our folks to survey the show and keep our ear to the ground. Expect a few blogs on the show and conference this week – here’s the first!
Contributed by Rob Williamson.
Digital Power Controller – “it is all about design”
At APEC 2012, Intersil’s Chris Young, Senior Manager of Digital Power Technology, presented “Digital Power Mythology”. It walked us through a number of myths related to digital power management and control (henceforth called “digital power”) and explained how those system engineers who have traditionally steered clear of digital power should trust it to be at least as effective as analog power.
Chris Young, Senior Manager of Digital Power, Intersil
In presenting the argument that digital power should be the preferred solution he cited some examples, “many UPS solutions today use digital power and that it is an option on the PM (12 and 12.5) bus for Intel processors. By 2015 it is my opinion that all of Intel’s processors will be specifying digital power.”
With regards to performance, “many experiments have been done with the same board set-up but swapping out digital and analog control and they are often shown to be highly similar”. In the examples presented digital had a slight performance benefit. He was clear to point out that although a slight improvement may only save 1 watt, when you multiple the benefits across multiple power supplies on board, and then multiply that in an environment like a server farm then you have the potential to save the customer a lot of money. For systems designers, where a lot more of the potential can be realized is that they do not need as many components on a board and therefore the overall systems efficiency is better.
The presentation aimed to dispel the myths associated to digital power, and present them as no longer true. Subdivided into three categories- performance, economic, and usage. We’ll focus on the top two because their descriptions were a bit more technical.
Performance Myths
1) Is digital power slower than analog power because the additional A-D / D-A conversion? Author, says “NO, since one can just use an intelligent circuit design and make the whole thing as fast as you want”
2) Does digital power need a high clock frequency for high speed response? Author, says, “No”
3) Does digital power require higher quiescent current than analog control? Author, says, “No, again, some digital controllers have lower quiescent currents than analog controllers”
4) Is digital power is more efficient than analog? Confirmed as digital power uses the central clock system
5) Is Digital power is not as accurate as analog ? Author, says NOT, since most digital controllers uses 14-bit DPWM controllers. In fact, digital controllers can be much better calibrated than analog”
6) Does digital power generate more noise than analog? Author, says , first analog controllers are not immune to noise, and in fact, there are several digital controllers are now being marketed on that basis of low noise compared to other digital controllers.
This was a 3 hour session, so we’re going to focus on a few “myth dispelling” highlights. For the full presentation, you may want to find Mr. Young on LinkedIn – he was a good speaker and certainly passionate about digital power.
Analog vs Digital Power
Looking at the typical analog controller, you have a simple well understood design with a short path. However, there is not a lot of bandwidth available so it is not infinitely fast. A digital power controller uses a clocked system with an ADC (with some latency) a digital filter (with some kind of group delay) then you have the PWM with latency. However, it is worth noting that analog can have latency and bandwidth as well. A typical analog power supply control loop has loop bandwidth <200 kHz and digital power is more than up to the task.
In digital power, with 1 MHz sampling, 1 sample delay then at fsw/5 has 75 degrees of phase lag. So quite a significant lag? But is this lag an issue? Not if intelligent design is used. Most likely method for design is to do over sampling. “Demonstrated a close loop regulation of >12 MHz fsw. Using a 0.18 um CMOS digital logic cell is between 20 and 50 ps.”
And it is not a simple matter of turning up the clock. Putting a clock of 1+ GHz in a $1 or $2 pwm chip doesn’t make economic sense, but you don’t need > 1GHz clocks. By using a tapped PLL you can make a full oscillator to increase resolution. You can also dither the lowest significant bit of the PWM pulse over 8 to 32 cycles providing another 3-5 bits of resolution. The result is a 9 or 10x reduction in clock frequency required. Additionally, you can reduce the clock speed when the frequency is at steady state. The bottom line, states Chris, “For power management, you can do digital circuitry as fast as you want and as fast as you need.”
Myth #734 (there were quite a few and we lost track)
With respect to resolution, increasing resolution to greater than 10 bits (there are already plenty of 14 bit controllers on the market) results in more than sufficient resolution over the voltage ranges. In addition, digital controllers can be much better calibrated than analog and can consider a very wide range of temperature variability and thermal compensation. And most particularly, there is no drift. You never have to worry about “0 drifting to .01 in the digital world.”
With proper design, issues around resolution, limit cycling, duty cycle dithering, noise, and compensation are not relevant and digital can be equivalent or better. With digital solutions you can build smaller power supplies and have significantly more features.
The basic and oft repeated statement was - It is all about design.
Economic Myths – Digital controllers are more expensive than analog controllers
It is true that you can get some analog controllers for under $.10, but you can also get full featured analog controllers that cost $5. At the lowest end, digital controllers are more costly, but digital controllers have a wide range of pricing. Additionally, isn’t an apples-to-apples comparison, because digital controllers are typically full featured with capabilities for data analysis, autoconfiguration / calibration, monitoring, etc. Chris said “comparing a digital controller to an analog controller is like comparing a car to a bicycle.”
Comparing cost and value for digital vs analog
When considering the full system, digital power offers features for margining, power management, fault management, etc. and reduces the total BoM of devices, inventory costs, failure rates, manufacturing costs etc. since many digital boards can have less than ½ the total number of devices.
From a chip design point of view, the more you can put on the MCU the more you can scale the solution from process generation to generation. And that makes good sense since we are talking about a huge number of sockets.
Design density of analog and digital power
From a usage standpoint, digital power was presented as easier to use, as having less inherent risk (compared to analog), and being quicker to design. On the risk side, the main reason was that the overall BoM could be lowered with fewer components to buy, inventory, manufacture, and test. From the ease of use and design perspective, a good deal of time was dedicated to demonstrating Intersil’s simple GUI interface that designers could use to monitor and tweak the power supply.
In summary (if you haven’t heard it enough by now), it is all about design.
