An enigmatic architecture, interesting parts, an incredibly basic revelation, and overall, a nice looking module.
A whole series of interesting and seemingly well-designed and -made power supply modules have come out of Rui Deng (Hangzhou Ruideng Technologies) of China over the last year or two. If you read Help me build an electronic load & bench supply, you know I plan to use two of their DPH3205 buck-boost modules to round out my electronic load (EDN CDEL) chassis. If you plan to build the CDEL along with me, you are of course free to add whatever you like, or nothing, to your box.
But before committing to use these modules, I wanted to take a closer look. Join me as I explore the performance and design.
A primary concern of mine is input ripple rejection. I’ll likely use a linear supply to drive the modules, and there’ll be ripple. I was especially interested to see what the modules did when their output was set to a voltage between the input ripple peaks!
Figure 1 With VIN swinging from 13-21 V, VOUT was a flat line in all modes: buck, boost, and alternating.
Of course, I’ll design for a worst case ripple of perhaps 1V, but it’s nice to see that 8V doesn’t faze the PSUs.
Figure 2 Output ripple/noise in buck mode, 2A load.
Figure 3 Output ripple/noise in boost mode, 1.4 A load.
The outputs aren’t super-quiet, but fine for general use. I may add output filters.
So, what makes this baby tick?
Figure 4 The control panel: What you see is what you get: LCD module, buttons, and encoder.
Figure 5 Most of the PSU’s key parts.
As I researched the various parts, I became extremely curious about the design’s architecture. Things just didn’t seem to fit. But now I think I’ve cracked the code, at least partially. Play along at home as I describe the components.
The input MOSFETs, Toshiba TPH2R306NH, are 2mΩ N-channel parts, 100% in parallel. They aren’t directly involved in the voltage converter, as there’s no switching activity on them. The drains are connected to -IN. Perhaps they are reverse-polarity protection. They can’t be over-voltage protection, as the body diodes can conduct when the module is properly connected. The maximum input current is 10A, which means each FET will dissipate…umm, will dissipate…umm… No way!
Have I been in electronics for 45 years without this bit of magic being ingrained in me? One FET would dissipate 10A2·2mΩ=200mW. So each of two paralleled FETs will dissipate 100mW. But wait. Each FET sees 5A, so: 5A2·2mΩ=50mW. That’s crazy, but I see what’s happening. Dealing with fixed resistances, this wouldn’t happen. Combining resistors to obtain a given resistance and power rating, the individual powers do add up to what a single equivalent resistor would burn. But with the FETs, current is the constant. Halve the total resistance, halve the total power (also consider that adding a second FET halves the voltage drop across them). Hence, each FET dissipates one quarter of what one FET would. Well, well. I shall consign that to my subconscious intuitive stash. Obvious no doubt to anyone doing lots of power work! Okay, to almost everyone (he said, sheepishly).
The LTC1871 on the little daughtercard is a boost converter running at a bit over 100 kHz. The left toroid and Schottky also have 100 kHz signals on them, so it appears they all form one subsection.
The TL594 is a general PWM control IC, similar to the classic TL494, and runs at 68 kHz, as do the right Schottky and toroid.
The SGM8582 is a precision op-amp with 100 µV VOS(max). That would be pretty darn good for a standard chip, but this appears to be chopper-stabilized, so that’s actually pretty poor. It amplifies signals from input and output current shunts.
The Rohm RB085T-60 5A Schottky rectifiers.
The STM32F100C8 is an ARM Cortex-M3 microcontroller with all the usual trimmings.
What’s conspicuous by its absence? Yeah…switching MOSFETs. There are two SMT parts clamped under the bottom edge of the heatsink. I think we’ve found our transistors. No…I’m not removing the sink.
As I said, this crazy arrangement of three switching converters had me perplexed at first. I originally assumed there’d be one switcher – a controller that was able to switch from buck to boost, or in SEPIC mode perhaps. But after more poking and prodding, I think I know what’s going on. Any guesses?
Here’s a clue: The left pair of lytics are the input caps. The middle pair always have 35.7 V on them. The maximum output voltage of the module is 32V.
Yes… It seems the input (6-40 V) is converted to an intermediate 35.7-40 V, which is then stepped down to the desired output. That’s a bit surprising, given that losses tend to be exacerbated with high buck step-down ratios. I guess the designer felt this to be the best trade-off. If it works, who am I to complain?
That third switcher, the XL7005A? It appears to generate 5V, and there are two (I think) 3.3V linear regulators as well. The µC, LCD, and op-amp need juice too.
The fan is controlled, but as best as I can tell, not by temperature. It comes on when the output current hits 3.5 A, and turns off at 3.4 A. Perhaps not the best control scheme. The sink can get fairly warm when supplying highish voltage at lowish current, so I suggest rev.2 of the firmware employ a bit more intelligence here. But, with a case fan running too, all should be fine.
One thought this analysis brings home: The wealth of Chinese-manufactured semis that must be out there, yet that very few of us know about. Have you manufactured in China? Did you use Western parts, or search out Chinese ones where possible?
—Michael Dunn is Editor in Chief at EDN with several decades of electronic design experience in various areas.