Got to open up a fairly beefy Chroma DC source at work (my other post has the context:
https://www.eevblog.com/forum/reviews/elgar-3000b-(ancient)-ac-line-conditioner-teardown/ Short version is, lots of extra time for doing not-too-invasive teardowns on
everything). This series of DC supplies has a variety of models, with a few different power levels, and different voltage/current ranges for each power level:
http://www.chromaate.com/product/62000P_series_Programmable_DC_Power_Supply.htmThe wide variety of similar models is probably responsible for some of the things that'll show up in here. This particular model can put out up to 600V or 8A @ 1.2 kW max.
(Yep, I'll be getting to the Chroma AC source underneath it later on)
Here's the overview with the cover off:
It's divided up into 3 boards: left, middle, and right. AC line power comes in at the back-left, so the left board is pretty clearly the input, and since the DC output leaves at the back-right, the right board is also pretty clearly the output.
Looks like overall, the left board ("PFC board") is a PFC stage: there's a boost converter which pumps up the AC line voltage to a 400V DC bus; it also has an auxiliary power supply for control power, gate drive power, etc. The output of the Chroma is isolated for obvious reasons, but this PFC board doesn't provide any isolation; everything is still referenced to rectified-AC-line.
The right board ("output board") is a DC-DC converter, most likely a full-bridge, which does provide the output isolation. This takes the 400V DC bus and steps it either down or up to the desired output voltage.
The middle board ("control board") has all the general control circuitry, minus all the stuff responsible for running each converter at a very detailed level.
I'll explain where this is all coming from as I go through the different sections.
Besides not making a single monster PCB which would be impossible to handle, the number of product variants is another reason for splitting up the boards this way. Control board can stay the same in every model, while there's probably one PFC board version per power level (600W vs. 1.2kW vs. 2.4kW), and different output boards (with different transformer ratios, etc.) for every voltage/current range option.
Control board first, since it's the most straightforward.
Date stickers cover a good number of chips, but there's a Lattice programmable logic device (no idea whether FPGA or CPLD) and what appears to be a processor and some memory, as well as a little auxiliary supply and an extra board on the back panel for some comms ports. The digital control here talks to the rest of the circuitry through a row of Analog Devices digital isolators, barely visible in the bottom right, near the ribbon cables which carry the control signals to the PFC and output boards.
The complexity of a processor AND an FPGA is justified here: this isn't just a simple "set a DC voltage output" supply, it can do controlled slew rates and programmed output voltage sequences/waveforms and talk to GPIB/USB/dolphins and all other kinds of fancy shit. I only ever used it to put out boring constant voltages with some occasional over-current foldback.
With an FPGA on the controller, the first thing that comes to mind for me is either fast number crunching or PWM generation for direct digital control: as in, putting the processor directly in the feedback loop (sensing output, calculating error, and generating PWM digitally) instead of having a traditional all-analog PWM controller. I don't think that's the case here though, as both the PFC and output stages have their own generic UC3xxx PWM controllers; instead, I'd guess that the digital stuff here keeps itself busy with the high-end features and communications, and just feeds a setpoint voltage to the output board's PWM controller, which quietly manages itself most of the time.
The AC line power enters through a hefty EMI filter:
There's then some relays, which probably are for bypassing inrush-limiting resistors. Those 3 black circles glued together are likely NTCs, which are the inrush-limiting elements.
A popular way to limit inrush currents into a large capacitance is to put an NTC (negative temperature coefficient) thermistor in series with it: the NTC starts off with a large resistance, which initially limits the inrush current. However, as current flows through the NTC, it slowly heats up and its resistance reduces to the point where the circuit is operating as if it wasn't there. It's a simple, cheap way to do inrush limiting, but there's two main problems with it:
1. You waste power: the NTC has to stay at a high temperature at steady-state to keep its resistance low, and even though its resistance is much lower than it started, it's still much more than a piece of wire, for example - especially important in a high-power supply like this.
2. The inrush protection takes a while to reset. The NTC has some thermal mass, and so when power is removed it doesn't instantly cool down, so its resistance is still low for a while. So an NTC-only inrush protection circuit doesn't limit the inrush current at
all if you flip the power off then turn it immediately back on.
If they've got space/money to spare here and can use inrush protection with a relay to bypass it when inrush is done, I'm not sure why they used NTCs as the limiting element instead of some power resistors. There may be a reason, or my guesses may be totally off here, but it's the only explanation I have given what I can see.
Anyways, back from that tangent, to the rest of the PFC stage:
There's a bridge-rectifier on the large heatsink at the far right, and a yellow film cap which I assume is the input filter for the PFC converter. There's also a giant toroidal inductor, two power transistors on the big heatsink, and a diode at the left of the same heatsink. There's also a metric fuckton of output capacitance at left: 7x 680uF 450V = 4,760uF total.
Because there's only one diode, and the inductor doesn't seem to have any kind of isolation (some of the windings even overlap), I think this is a boost converter with two paralleled MOSFETs. This would also explain why there's a need for inrush limiting: with almost any other kind of switch-mode converter, you have no direct path from input to output when the converter isn't running, and so inrush isn't a big deal because at turn-on the AC line only has to charge up comparatively-small filter capacitors. With a boost converter, however (which incidentally is popular with a 400V output for PFC applications!), notice how even when the transistors are off, any input voltage is forced to charge the output capacitance through the inductor and diode. With this much capacitance on the output, you'd definitely want some inrush limiting for the sake of the poor AC line which has to charge 4,760uF, and the poor diodes in the way!
Check out the current sense resistors:
Not sure what the giant white power resistor is for (10W 500 ohms): maybe it's part of an RCD snubber with the round diode behind it and one of the many (!) nearby caps for the MOSFETs and/or output diode. You can also see the controller chip here, which even though the markings are faint, I confirmed is a UC3xxx-series part. The other through-hole transistors are probably gate drive for the MOSFETs, and the large blue film cap just on the left side of the heatsink is probably the (high-freq.) output capacitor for the boost.
After the massive capacitor bank of doom, the 400V DC bus goes through a common-mode choke and maybe a snubber (?) with another white power resistor before taking the scenic route through a long cable over to the output board:
It's also worth mentioning that on the PFC board is what looks like the auxiliary supply; this may take its input from the 400V bus:
Really though, even the auxiliary supply for this DC source is bigger than a lot of the power supplies I've dealt with! Especially the transistor:
(Although it's probably high-voltage but low-power; if the auxiliary supply is a flyback with a high step-down ratio, the transistor will see a drain voltage of 400V bus + (high turns ratio)*(output voltage) = easily 600V-900V. You can get high-voltage parts in much smaller packages though; maybe this part was on the BOM anyways for the output stage, or it really just needs a whole lot of aux power for the gate drive!)
Next up is the output stage:
The power from the PFC stage comes in over that cable and connects to a daughterboard with an unnecessary-feeling number of capacitors (except they probably were necessary):
As a sidenote, even though I've been designing switch-mode power supplies for a while now, I was still surprised by just how many capacitors this entire supply has inside it. Not the big guys for the various converter inputs and outputs and the EMI filters, but just tiny little capacitors sprinkled around
everywhere. Maybe they were either playing it really safe with EMI or had some serious issues with it at first.
The power transistors are hidden underneath a plastic sheet which directs the fan's air over the heatsink, so the best I can do is to get a shot with the sheet pulled aside a bit, showing the large packages, gate drive transformers, and a little saturable magnetic bead on each drain lead for suppressing switching spikes (good for reducing both EMI and peak drain voltage).
There's also a couple large magnetics:
The 4 power transistors and lack of large series DC-blocking capacitors (as far as I can tell) makes me think this is a full-bridge converter, with the output diodes further back along the heatsink where I can't see them. That would make the larger front piece of magnetics a transformer, and the slightly-smaller rear piece of magnetics the output inductor:
There's also a current transformer near the power transformer: this is probably for measuring switching currents to do peak-current control. The white power resistor near it might be its burden resistors, as I put in the schematic, or it could be part of a snubber across the transformer's primary winding, to absorb spikes created by its leakage inductance.
There's also a mystery clump of power resistors just past the output inductor.
The round diodes behind them suggests they're part of snubbers, which could be for the output diodes I guess? (as shown in the schematic above, with the center-tapped secondary configuration) They're kind of far away from where the output diodes would be though, on the heatsink (between the transformer and output inductor) so that seems unlikely. No idea on this.
It's also possible that the diodes are a red herring, and that these resistors are actually under the control of some unseen nearby power transistor to discharge the output caps on output-voltage-setpoint reduction and power-down.
Either way, the output goes through a couple stages of filtering, with a couple pairs of aluminum bulk capacitors (which must be in series, as they're rated for 400V and the output goes up to 600V) and a large common-mode choke. EMI must've been a massive pain in the ass with something like this in the kW range.
A highlight of the final output is the current-sense resistors (presumably used to sense output current for display, limiting, foldback, etc.) which are essentially just two hand-sized pieces of sheet metal. A ribbon cable is wound through them, interestingly, which either suggests bad design or that the resistors aren't expected to get actually hot.
As to my earlier point about capacitors: look at the output sense connector board, which is just overflowing with them!
The cable from the output sense board is the one that winds through the current sense resistors; it goes to a little vertical daughterboard (only visible in the main overview as I didn't get a separate shot of it) which likely does the scaling/buffering/etc. for the output voltage sense.
While the control for the PFC stage was pretty simple (all it has to do is slowly regulate a DC voltage), the control for the output stage understandably has much more going on:
Lower section has op-amps/comparators/digital logic plus a generic UC3xxx PWM controller in the DIP package (although the flash obliterated the part number here):
The daughterboard has the same, with a good old 555 timer and an AD633 multiplier (
) in the DIP-8 making an appearance:
Can't think of what you'd need a multiplier for, unless it's using some kind of interesting custom control scheme (if anything, I'd expect to see that in the PFC, not in a DC-DC). Maybe it's for either a weird non-linear control loop, variable gain scaling (although that feels error-prone and overly complex) or extrapolating measurements from other measurements: for example, predicting an individual MOSFET current (MOSFET current = D*xfmr current). So yeah, this is another piece I'm not sure about. Any ideas on this bit of weirdness?