I don't usually bother opening small power supplies - unless it's a quality-check kind of situation, there's almost never anything interesting or unique to see in a typical bench supply. However, this one has a high (600W) output in a surprisingly small (and strangely shaped) package, so it seemed worth a look.
This supply is really, really long and skinny.
The output rating of 100V @ 6A is actually pretty impressive. I don't know the model number because it isn't printed anywhere on it.
The first noticeable thing after sliding the shell off the back, is just how tightly packed everything on the inside is. The photos don't do it justice. There's lots of tetris-ing and pieces from both boards which almost touch each other.
Output is on a pair of busbars at the back end. Overkill for 6A, but hey, whatever. The physical structure is mostly just the two parallel PCBs themselves, attached to the front endcap and the back endcap as you can see here with through-hole-soldered angle brackets; these endcaps then attach to the outside shell.
There's also a single standoff in the middle of the two boards to attach them together and give a little extra rigidity.
In general, the mechies (rightfully) had a field day with this one. The dual boards are also insulated from the case with a pair of thin clear plastic insulators which fit around the whole thing; these insulators have some very nice tabs and slots to hold on to each other and various mounting bits. Wish I could get better photos, but photographing transparent things is, well, hard - it's bad enough already with the low light level in the lab, having to use flash and getting glare everywhere.
There's also an additional insulator piece sitting between the two boards.
In general, the two boards seem to be mostly separated into power vs. controls.
The power flow on the power board takes a kind of "U" shape: starting at the bottom-right of that photo, with the AC line input. It goes through a normal set of protection (fuse, MOVs, and an EMI filter) before reaching a pair of wires that go to the front-panel power switch. You can also see a nice soldered-down metal shield (with Kapton-tape insulator?) that separates the output section from the input section, presumably for EMI purposes.
It then goes through a bridge rectifier visible on the heatsink, a film input capacitor, and a boost-converter PFC stage: notice the toroidal boost inductor (with auxiliary winding)...
...with a single boost MOSFET and two smaller power devices, dumping the output into those large brown DC-link capacitors:
One of those power devices is a diode (2 leads), but the other has 3 leads. It turns out, looking at the bottom side of the power board (top-left, in this photo)...
...if you follow the traces you can see that it's a MOSFET, with its gate driven by the auxiliary winding on the boost inductor! We've got a synchronous rectifier here; unusual for off-line/high-voltage applications (because a volt or two across a diode, vs. 400V+ on the DC link, doesn't really affect the efficiency much) but does make a lot of sense for reducing the
absolute power dissipation in the boost diode, which apparently was important here with the compact layout. Tracing the (very simple) gate drive circuit gives this:
I can't guarantee the polarities/types of the diodes but this configuration makes the most sense by far. Using a zener in series with the gate gives a nice safe slow turn-on/fast turn-off, and the other diode shifts the output levels from negative-when-off/positive-when-on (AC-coupled, directly from the aux. winding) to 0V-when-off/more-positive-when-on (after a DC offset on that series capacitor) by charging up the series capacitor, to move its DC offset in the correct direction, whenever the gate signal goes negative. Without that level shift, the on-state gate voltage depends on the duty cycle, so especially over a wide duty cycle range as the rectified AC line voltages goes from 0V to ~170V, it can be hard to make sure that the on-state voltage always is high enough to turn on the MOSFET, but never high enough to damage the gate. Now since the series capacitor can be charged up during any cycle, but there's only that parallel gate resistor to
discharge it, you can run into "MOSFET never turns fully off" problems if the duty cycle increases more quickly than the series capacitor can discharge...but that's a different story.
You can also see, if you squint hard at that last photo, a bunch of black rectangles underneath the bridge rectifier; these are current sense resistors directly in the PFC input power's return path, used for current sensing by the PFC controller. There's also a diode across the sense resistors array, to clamp the voltage - because a boost converter has a direct DC path from input to output, when you throw the power switch, it will peak-charge the (large) DC link caps, and send a huge current spike through the boost inductor etc. This isn't a problem if you're sensing current from the boost MOSFET's source, but if you're sensing total input current, as in this case, that inrush current spike can easily blow up your current sense resistors when it puts an instantaneous almost-peak-line-voltage across them, current-limited only by the boost inductor (which is going to saturate very quickly and not do shit).
I've found this out the hard way in one of my own designs. Putting a diode across the current sense resistors doesn't interfere with normal operation, if you size everything correctly, but in this inrush case it limits the voltage across the current sense resistors and forces the extra to flow through the diode.
Additionally, you can see the boost controller chip here: it's an ST L4891, a fixed-frequency average-current-mode PFC controller. Finally, notice the bits of glue underneath unpopulated components - looks like the power board was wave-soldered; an excellent choice here to get both the many through-hole and surface-mount components all in one process step.
Anyways, quick break from the power: now that we're up at the front end of the supply, you can see the fan that blows air from the front of the unit down the full length, over the boost converter heatsink, then the bridge rectifier heatsink, etc. until the exhaust comes out the back end. There's also a small board with some display drivers here too, for the voltage and current readouts on the front panel.
After the DC-link, the path now wraps around and starts to travel down the other half of the power board, towards the back end. There's a full-bridge converter, with 4 transistors, and a small daughterboard attached to the heatsink, which probably has a temperature sensor on it:
A zoo of magnetics follows the H-bridge:
There's a current-sense transformer to sense the primary-side current (right-hand small block), a power transformer which isolates the AC line side from the output side (left-hand large block), and the output buck inductor for the full-bridge (right-hand large block, half off-screen). There's also a gate drive transformer for driving all the H-bridge MOSFET gates; apparently 2 layers wasn't quite enough for routing all the gate drive signals, because they had to add on an extra little jumper board to get the final two where they needed to go:
Near the full bridge's output inductor are 4 diodes acting as a full-bridge rectifier for the transformer's output, and a few output capacitors, including some extras that were mounted on a right-angle daughterboard to squeeze out some extra room.
You can also see the output current-sense resistor here, which is cleverly using the positive output busbar as a heatsink. There's another mystery device here clipped to the busbar, which I'd have to guess is a discharge transistor for draining the output capacitors.
Now that the power board is clear, here's the control board:
Left end has an auxiliary supply (seems like a flyback converter driven by that TO-220 mounted to the small heatsink, which is a
Panasonic MIP2E4DMY all-in-one controller+power MOSFET, with the two side-by-side linear regulators above the transformer as post-regulation). Middle has some digital controls, in a separate "control ground" section isolated from the "output ground", followed by a bunch of analog sensing and a billion optoisolators.
The analog section...
...has an AD7798 16-bit sig-delta ADC, an ADR03A voltage reference, 2 DACs, and some other misc. parts (HC123 dual one-shot, 4x 358 op-amps, 2x OP2177 op-amps, HC4053 analog switches, 74HC4078 8-input OR/NOR maybe for fault logic).
The ADuM1400 and 1401 at the left end provide high-speed digital isolation, probably for the ADC and DAC, while the rest seem to be "normal" low-speed isolators for control or protection signals. The ADC is almost definitely used for sensing output voltage and current, while I'm not entirely sure what the DAC is for: maybe changing the setpoint on the DC link voltage (although that feels unlikely), or setting a hard-wired over-voltage/over-current threshold for a simple comparator somewhere in the analog mess (much more likely).
The processor driving the show is a reasonably powerful STM32 part:
This supply has USB and Ethernet interfaces, explaining the need for more than just a tiny little 8-bit controller to drive the front-panel displays. It shares a ground with the USB, which again is isolated from the other grounds, including the power output ground, which is separate.
Now, between the one-shot and the many op-amps in the analog section of the control board, I could believe that there's a switching controller for the full-bridge converter, with an error amplifier, PWM modulator, etc. built out of op-amps and digital logic. However, it seems more likely that the STM32 is directly creating the full-bridge's PWM signal with its PWM peripheral, and has a digital control loop - this could seem like overkill for a simple DC-output power supply, but if it needs a reasonably-powerful microcontroller anyways for the Ethernet and USB interfaces, then making it actually control the power supply, maybe going with a slightly higher clock speed, isn't a very large step (ST even specifically labels this part as being for "real-time control"). It's less parts overall, and having digital control could be very convenient for a supply with a wide range of outputs, to vary the loop gain with the output voltage or various things like that to give max performance over its full operating range.
Finally, the most important part: the control board has lots of pretty colorful resistors on the bottom side - KOA Speer parts I think:
Home
Help
Search
Login
Register
