My install and review of the ConVerdant LFI30-220 3kVA Pure Sine Wave Inverter for Backup Power

This is intended as a companion thread to My install and review of the AIMS Prius 2kW Pure Sine Wave Inverter for Backup Power Generator started several years ago by AHetaFan.

It's starting to look like some kind of inverter farm around here, because it turns out I also have one of those. In fact, that's part of the story of how I ended up with this one.

See, there are some things the two have in common. First, they're both designed for a high DC input voltage, meant to run from a Prius's traction battery rather than from the 12 volt system. That's how they can be offered in power capacities that exceed what you can get from the 12 volt system of a Prius.

Second, they're both kind of rare. There just aren't as many inverters on the market made for 200+ VDC input as there are made for 12 volt input; those things grow on trees. This inverter was sourced (from somewhere) by ConVerdant, a business that was run by Randy Bryan for a while, then closed (and now Randy has opened a similar business called PlugOut Power, selling similar inverters but from a different supplier). The AIMS 2kW product was, as far as I can tell, first custom-ordered from AIMS Power by Jack Chen, who sold them on eBay; they were later also sold for a while by ConVerdant, and by The Inverter Store. Neither this one nor that one is in current production. I got both of them from other PriusChat members.

Then there are things they definitely do not have in common. Size, for one. The AIMS two-kilowatt unit is one that AHetaFan was able to tidily tuck along the back seat:



This one, not so much. Long-time member hobbit could make it barely fit under the hatch by taking the casters off the bottom first:



By the way, hobbit also wrote an extensive review of this thing, including waveforms and calibration details. That's all worth a look, for anyone who has or obtains one of these. What I'll try to cover here will be stuff hobbit didn't.

Another thing they don't have in common is weight. The little blue AIMS unit is about 6.8 kg. You can walk around holding it in one hand. Somehow going from that 2 kW unit to this 3 kVA unit, you go from 6.8 kg to 33 kg. That's right, this thing is 73 pounds.

Also, this unit has 120/240 VAC split-phase output. You can pull about half its rating, or around 1.5 kVA, at 120 VAC from either of its output legs, or both. The little blue AIMS can actually serve slightly more 120 VAC load, up to its 2 kW limit, but with that one there is no option to run any 240 VAC load.

I bought the AIMS unit first, when I saw another PriusChat member was selling one, because I liked the size and the weight and the capacity seemed enough for my purposes.

I did not buy this one because I liked the size or the weight, or because I needed the extra capacity. I bought it because the AIMS unit I was sold turned out to be DOA, and I wasn't sure I'd be able to fix it, and so it sat uselessly in a corner for a couple years and then I saw another member selling this one.

Then this one arrived. By whatever quirks of history and karma might have been involved, I saw that I have the unit with serial number N001. And, for reasons elaborated below, I promptly dismantled it.

Then one day I took another look at the little dead AIMS unit, and discovered it really wasn't that hard to fix. So that one is now fixed up and usable, only I am waiting while my friendly local auto electric shop looks into fabricating the high-voltage cabling for me, so while waiting on that, I've turned my attention back to the dismantled innards of this one.

Why is this thing so big and heavy?

That's pretty much covered in hobbit's review. There's a photo there that tells the story:



It's big because it's built in the same steel cabinet as its 5 kVA big brother, and contains a lot of empty space. It's heavy because there's a honkin' 3 kVA toroidal transformer in the bottom. True to this figure (credit to Bicron Electronics), that transformer all by itself is 16 kg ... 35 pounds. (Imagine how heavy this sucker would be if they had used a traditional laminated transformer instead of a toroid! Now just imagine the even larger 5 kVA version!)



The next top contributor to the weight is that comically oversized cabinet itself. Stripped of all innards and empty, that cabinet is over 11 kg. That's more than half again the weight of the whole AIMS inverter, just for an inert steel box.

(If you look at the picture, most of the innards are built on the steel 'shelf' midway up, all except the front-panel controls, back-panel connections, and the transformer in the bottom. When I dismantled the unit, I just removed the shelf and everything on it as an assembly, so what I give for the weight of the cabinet doesn't include the shelf.)

Maybe a better question is how the AIMS unit can be so light. It does not use any output transformer in the same sense. Sure, its circuit board does contain some small transformers and inductors, used for various internal purposes, but nothing like a single big 2 kW transformer that produces the output. The AIMS just directly modulates pulses into something that looks like a sine wave, filters the corners off with some capacitors and small chokes, and gives you the result.

So why does this unit use a big ol' transformer? I'm not sure of everything the engineers were thinking, but one reason could be to get the 120/240 split-phase output. The basic electronics produce just one AC output between two terminals, just like the AIMS, but that gets fed to the transformer primary winding. The secondary winding has a center tap that becomes a neutral, with opposite-phase 120 VAC legs either side of it.

That's a time-honored, very easy and, I guess?, economical way to get the split-phase output—if you don't care that it adds a giant 16 kg transformer to your design!

Could the same thing be done transformerlessly? Well, just imagine strapping together two of the AIMS units (that's the easy part), and synchronizing their oscillators to keep their outputs phased 180° apart (the harder part). Literally two AIMS units strapped together would still come in at a third of the weight of this thing, and would have 4 kVA total capacity rather than 3!

But this thing is what it is.

Why have I dismantled it?

Well, the main reason was that it already got nearly dismantled on the way to my door. It comes with a manual that includes this helpful handling advice:



But that advice is inside the manual, which is inside the box, where it doesn't help when the freight handlers are wondering whether they ought to make the box upside down. When it was delivered to my porch I noticed that it had been made upside down, but couldn't tell whether that just happened when they put it on my porch, or it had been made upside down earlier in the journey. It must have happened earlier.

Imagine this 16 kg transformer hanging like a pendulum by its center mounting bolt, itself hanging from the center hole in that steel mounting plate, itself hanging from the bolts at its four corners, for 800 km of bouncing and jolting in a truck.



The center nut was coming loose, the square plate was bent down, the bolt holes were elongated, and I was thankful the whole transformer hadn't come crashing down into the shelf with all the electronics. The cabinet also had a kink in a back edge and one of the brass standoffs holding the main circuit board to the shelf had snapped.

So, first order of business was to get that transformer out of there, look it over very closely for any insulation damage (it's just lots of varnished wires barely covered with plastic wrap, bouncing around on that center mounting disc) and then carefully test it.

The transformer has this label:



First I checked for any obvious insulation failure between the windings, going about double its rated voltage with the 500 volt setting on a megger. The resistance between windings was offscale past 2 GΩ, both immediately and after one minute. So far, so good.

Next, I set a function generator for a 60 Hz sine wave and applied it between the two primary leads, at a very small amplitude (140 mV, exactly one thousandth of the rated voltage). Sure enough, there was a 260 mV sine wave between the two secondary legs (red and green), and 130 mV between each of those and the center tap (yellow).

In doing this, I also marked the two primary leads, which are both black. The secondary legs are red and green, so I mixed some red and some green Plasti Dip and put a blob of red on the primary lead that matched the red secondary in phase, and likewise green on the other one. I had been careful to disconnect only one of the primary leads from the circuit board (so I couldn't forget which went where), so after identifying them I also put matching blobs on the circuit board terminals. Then I could disconnect it completely.

Next, to find the actual resistance of the windings at DC, out came the rescued micro-ohmmeter:




The picture shows a test current of ten amps, but that was for something else. For transformers, about one percent of the rated current is recommended. I used the 100 mA setting.

That meter is not specifically made for measuring transformers. Instruments made for that can have special safety features to make sure you don't just disconnect the leads after measuring but while a magnetic field still exists in the transformer. Instead, they will create an internal short circuit when the measurement is complete, giving the field a safe way to dissipate, and then indicate when it is safe to unhook the leads.

That's obviously a big deal with substation-sized transformers. I wasn't sure whether it's a big deal with a 3 kVA transformer, but out of respect, I did use a separate wire to short across the tester connections at the end of each measurement, with an ammeter to watch the current decay to zero, and only then unhooked the test leads.

Sure enough, using a tester current of 100 mA, the transformer's primary winding could keep a current going through the short for 25 seconds, and the secondary winding for nearly a minute.

The winding resistances came out as 122.56 mΩ for the primary, 388.9 mΩ across the whole secondary, and 197.2 or 197.3 mΩ from each secondary leg (red, green, respectively) to the center tap (yellow).

Of course there's nothing on the label that gives specified resistances to compare these to, but this transformer seems to be ok, so these measurements may be useful in case somebody does end up with a damaged transformer and needs to find a suitable replacement. They should be an indication of the length and gauge of wire used. Naturally, the measurements from a damaged transformer couldn't be trusted!

It turns out there's a simple old-timer trick to check if a transformer has a concealed shorted winding internally. If you put some current through a coil and then stop, and you don't provide a path for the current to keep flowing while the magnetic field decays, it's going to find one anyway. (That's how ignition coils, for example, make sparks.) If it's a transformer—more than one coil—the path doesn't have to be in the same coil where you started the current. If another coil wound on the same core has a path available, the current will follow that.

So you hook a little neon light to one of the other coils:



When you interrupt the current in the first winding, the decay of the magnetic field will make current flow in the other winding, and flash the light. But if the transformer has a shorted turn of any of its windings, that will be an easier path for the current, and the light won't flash.

I never tried that exact test. But on the same principle, just the fact that this transformer was able to push current through my Fluke for nearly a minute after disconnecting the current source is probably a very good sign that there aren't any winding shorts in there to serve as alternate paths.

So it seemed like the transformer checked out, and I could move on to the other early orders of business, like using an air hammer to flatten out that mounting plate again, and ordering an assortment of M3 and M4 brass standoffs to replace the one that had broken.

Can this thing be repackaged with less ludicrous dimensions?

Another reason for buying the assortment of standoffs was that I kept looking at all the empty space in the cabinet and thinking "this could be made a lot smaller." Also, I think a smaller, less overbuilt cabinet could come in at a lot less than 11 kg. Given the transformer and the rest of the innards, this thing will never be light, but I think it could be improved a lot.

Looking at the space below the shelf:



There's barely anything down there but the transformer. (There is a very small current-sensor donut for the neutral wire, mounted on the underside of the shelf at the back corner; you can see the yellow neutral running through it, and the red/black twisted pair carrying the current signal back to the circuitry. It is on a small circuit board mounted by short standoffs at two corners: one just gets a screw from the top of the shelf, and the other shares the bottom of the standoff holding up that corner of the circuit board above, the standoff that broke. So that sensor was dangling by one screw.)

That shelf and everything on it could probably be lowered at least 12 cm with no problem at all.

What about the space above the electronics on the top of the shelf?

The tallest things on the big 'grunt' board are the big black electrolytic capacitors at the rear of the board (they can be seen there, standing about as tall as the top of the fan).

There is nothing nearly that tall on the smaller 'smarts' board at the front of the shelf. But that whole board is lifted up on a tall steel platform-on-stilts, making it the tallest thing in the box! And what is taking up the space under those stilts?

You don't see it in hobbit's photo, but there's just one thing under there, a hefty toroidal inductor mounted on its own dedicated circuit board. It's the one at the top of the simplified schematic in the manual:


By the by, one thing you might notice about the schematic is that the voltage feedback from the output to the DSP is drawn confusingly. The lines don't line up, L2 appears twice, there's an L1 with no wire, looks like some botched edit of a different diagram. Which I think is exactly what it is ... more on that later.

I started imagining taking out that platform-on-stilts, turning it upside down, and mounting it beneath the shelf using the same holes. The smarts board could stay above the shelf, mounted on standoffs (from my standoff assortment!), again using the same holes. I just needed four longer screws (M3✕12) to go from below, up through the stilts, through the existing holes in the shelf, into the new standoffs for the smarts board. The big inductor could be kicked downstairs and mounted (still upright) on the stilt platform, placing it just below the shelf. The holes in its circuit board are drilled for M4 screws, so that just meant drilling four holes in the stilt platform for M4 screws, and putting four standoffs there. Only one new wire had to be made up, the one that runs from the grunt board to the inductor. I took the old one to the friendly local auto electric shoppe to have one made up just like it, but 8 cm longer than the original.



I'm pretty sure the whole thing could be rehoused in a box 27 cm by 48 cm by 30 cm high, with the control panel moved from the end to on top, where I think it would be more convenient for a box loaded into a car. It would still be a darned heavy box, but I'm picturing something with carry straps, passing underneath it, anchored by the four transformer mounting plate bolts, then coming up the sides.

The review by hobbit already established that the weird every-plug-shape Stupid Outlet originally on the back is an abomination that gets the hot and neutral mixed up for US plugs, and needs to go. (Amusingly, I found the AIMS inverter also managed to get hot and neutral reversed, and it wasn't even using a Stupid Outlet!)



The fate of the Stupid Outlet on hobbit's unit was to be "summarily diked out". I'm trying to keep the mods reversible, so the wires got unclamped from it and the ends insulated, but anyway, it's gone. The L14-30 twistlock connector is fine, as are the modular terminals on DIN rail. That's a construction technique I like and have already used in the car anyway.



I don't feel a need to expose the DIN rail terminals from outside the back panel like the original cabinet. That just makes my internal safety officer scowl. The DIN rail can live inside the box just fine, with proper knockouts and strain relief for bringing in wiring.

The Naked Inverter running a 25 watt light bulb—about the most it can run when supplied by my 0.2 amp bench power supply. If I set the supply for 250 volts (within the range of a Prius and within what the inverter can accept), then 0.2 A is about 50 watts. The inverter uses about 25 watts for its own entertainment and can light the bulb with what's left over. (If I set the supply for lower voltage, it will eventually go into current limit and lower the voltage. Then the inverter says "lower voltage? no problem, I'll just draw more current" and the supply says "no you won't, either" and that eventually ends with the inverter saying "Battey lack"—they didn't exactly overspend on proofreading—and shutting off.)



When working with the Naked Inverter, keep in mind that those black electrolytic caps are 680 µF each, 1360 total. There is a permanently installed 310 kΩ resistor on the board to make sure they eventually bleed to a safe voltage after you've unhooked the DC source. That gives a time constant around seven minutes. Figure on around five time constants, half an hour or so, to lose most of a 250 volt charge.

 

I'm not sure how expensive a 3kVA toroidal power transformer really is. I found this listing that isn't quite the same winding configuration but about the right size, power rating, and weight, and it's not as inexpensive as I thought. Granted, a Chinese one with none of the same approval logos on the label could come in a winner on price.

Meanwhile, back to the next dilemma: whether to buy or fabricate or have fabricated a new downsized enclosure, or just take a Sawzall to the original one.

There's an off-the-shelf one that is just about exact on length, and only a few cm over on both width and height. That would be pretty darned close. Still 5 kg, but at least that's less than half of the original.

I left a little mystery in the earlier post about the discombobulated mess on the right side of the schematic, looking like a hasty edit job from a schematic of something else.



What's really nice about this inverter is how clearly labeled and serviceable things are, thinking, for example, of the threaded standoffs installed on the circuit board where all the high-current wires attach with ring terminals and screws. That certainly makes it easy to work on and rearrange. (By comparison, the little blue AIMS pretty much has everything soldered right to the board. Even heavy wires, where that can really only work so well, and indeed that was behind the problem in the one I had to fix.)

I find the grunt board on this one especially nice; it is laid out so simply and clearly you could maybe use it in a class.



The action starts with the battery power coming in on the two ring terminals at the lower left, J1(BP) and J2(BN), for battery + and battery −, of course. The two legs of the final output are J5(OUTP) and J10(OUTN), which appear on the back panel as L1 and L2, respectively. (Also as AC1 and AC2 on the Stupid Outlet, before that was removed.) They have 240 Vrms between them. Each one should be 120 Vrms away from the transformer's secondary center tap, which doesn't even come back to this board from the transformer. That yellow wire from the transformer just goes straight to the back panel to serve as the neutral. It passes through a current-sensor donut on the way, which carries a measurement of neutral current back to the smarts board. But as far as this board is concerned, its only outputs are OUTP and OUTN.

But maybe it's better to go in order.



BP passes through a 40 amp fuse, a familiar automotive ATO blade fuse. Ok, in an inverter where pretty much everything impresses me, that right there does strike me as goofy, considering another very recent PriusChat thread established that those fuses are rated for no more than 32 volts. The AIMS inverter also has that same style of fuse inside. At least this one is replaceable, inserted into terminals soldered to the board. AIMS soldered the fuse right to the board.

BN and BP then encounter some capacitors to clean up transverse-mode noise, and they pass together through that toroidal common-mode choke to suppress noise that rides in on both together. In a Prius, that can include the 5 kHz signal that the battery ECU couples onto the battery outputs in order to detect isolation faults. (However, hobbit noticed that there is another path for that noise to pass through the inverter and reappear on the output, unless the inverter's neutral output and chassis ground are bonded together.)

From P13, the plug with the red/black twisted pair, the incoming battery voltage is carried to the smarts board to be measured. (I don't know whether it is also used for any other purpose, such as bucked down to produce operating voltages there; see also J12 later on.)

Next comes the capacitor precharging circuit. which was mentioned in hobbit's review. It is like what's done in the Prius itself, where the battery is brought online first through a resistor that slows the current rush into the car's inverter capacitors, and only then does the relay close that bypasses the resistor.

Here, there is the black relay K1, which can be seen in the schematic on the 220 VDC input, with the diode D2 and resistor R3 arranged around it. At first, K1 is open, and the big caps can only be charged through the diode and resistor. The resistor is 200 Ω, so limits the charging current to not much more than an amp.

Here, the yellow trace is watching the charge in the capacitors, and the blue trace is the voltage across the precharge diode and resistor, so it roughly shows the precharge current. Starting with the caps completely discharged, I connected the input at point A.

Because my bench power supply limits its output to 0.2 amps, the precharge voltage only jumped up to around 40 volts (point B) and stayed there a while, producing that slower steady ramp-up of the capacitor voltage. At point C when the caps came within 40 volts of full charge, the charge current then decayed along the exponential path it would normally follow. In the actual car, that charge current would have jumped about five times higher at first and then come down an exponential slide the whole way.



At point D where the caps are about fully charged to the incoming battery voltage, the little glitch coincides with some clicks from the smarts board as it realizes it is awake and has power. The LCD display and backlight turn on. About three seconds later at E, relay K1 closes and squashes the blue trace to zero as it shorts around the precharge components. The caps are now charged and directly connected to the battery input, and the inverter can be turned on.

When I turn off the output of my bench power supply, the capacitor voltage declines pretty rapidly at first. I think it is actually my power supply doing that, as the capacitor charge flows back out through it. That stops at the point where K1 opens and no more reverse flow is possible, because of the diode in the precharge circuit. From that point (shown here when about 40 volts remain in the capacitors), the capacitor charge can bleed off only very slowly through the 310kΩ resistor R5. After several minutes it will be down to 20 volts, after several more down to 10, and so on.



I suspect that if I were to just disconnect my power supply, rather than turning its output off leaving it connected, the caps would continue to hold 200+ volts of charge, with nothing but the 310kΩ resistor to bleed through. So after the first several minutes they'd be down to 100 V, several minutes more down to 50, and so on, and that's also what would happen in the car if simply unplugged from the battery cable, or the car was turned off.

Turning the car off and leaving this inverter connected would allow its caps to still present voltage on the hybrid system wiring, at least until K1 opens and puts the diode in the way. But that shouldn't be a problem, as the car's inverter caps already do that anyway.

I have not checked whether K1 always opens when the caps have dropped to 40 volts as shown here (which could take a really long time if the caps are bleeding down slowly), or if it opens just some fixed number of seconds after the inverter has decided its power is lost.

There's one more neat feature of this precharge circuit. The manual says this unit is protected against connecting the battery input backwards. And sure enough it is, because as long as K1 is open, that diode is in the circuit, and connecting the battery backwards won't make anything happen. And K1 doesn't close until the caps are charged, which will be never, if the battery is backward and nothing is happening. And there's no cost to this protection in normal operation, as K1 is closed and the diode is out of the circuit.

Anyway, so you've connected the inverter and made the Prius ready, battery power is coming in, the capacitors are charged, the control panel is active and you've pressed ENTER to get past the "Welcome to use the 220VDC invertor" display. Now you're ready to hold the ON/OFF button for a few seconds to start making sine waves.



Hidden on the underside of this board is a Mitsubishi PM75CL1A060 Intelligent Power Module.



The DC power comes into it on the − and + DC rail connections on the left, and it contains three pairs of transistors, one each to drive the U, V, and W outputs on the bottom. (The leftmost connection on the bottom has no function in this IPM and nothing is connected to it. It can be seen faintly through the board here. The W connection on the right has a contact screw and some copper on the board so it could be used, but this inverter doesn't use it for anything, and so also omits the U6, U7, and U12 driver ICs. Through the big holes in the board are the screws that fix the IPM to the heatsink.)

While this IPM with its three pairs of IGBTs could be used in a three-phase power project, this inverter only uses the U and V outputs ... the two pairs of IGBTs shown in the inverter schematic, each one with its freewheeling diode pointed backwards around it. Those are all built into the IPM too, along with overtemperature and short-circuit detection for each IGBT.

Look at the specs on this thing! If you've got a 600 volt, 75 amp, three-phase power project you've been planning, this is the IPM for you right here. For this inverter, using under 300 volts, under 30 amps, and only 2 of the 3 pairs of IGBTs ... clearly the engineers just said "look, there's an off-the-shelf IPM and it will definitely work." There's where the money they didn't spend on proofreading went.

So the IPM's U and V outputs are used to drive the primary winding of the giant toroidal transformer downstairs, via the cables attached to J3 and J6. Well, that is one of the transformer's primary leads on J3, and the cable on J6 goes to that largish inductor that's shown at the top of the schematic, and then to the other lead of the transformer primary.

So what do we see between J3 and J6?



It's like some geeky op-art painter wanted to remind you of a sine wave by shading, drawing different-width pulses all the same height. Well, this is pulse-width modulation, after all.

Zooming way in on those individual pulses ...



they are about 35 microseconds apart. These guys are using a pulse frequency around 29 kHz. You probably can't hear it. Your dog or cat may.

Those pulses are very very skinny. There is no load on the inverter right now. It takes only very tiny kicks to keep the sine wave rolling along.

There's already something interesting in that first scope trace. You can see which kicks are aimed up and which ones are aimed down. It could have been otherwise.

When you look at an H bridge as shown in the schematic, and you think how to make a pulse of either direction, one easy idea is to pick the right two diagonally-opposite transistors and turn them both on, and when you're done with your pulse, turn them both off. By which transistor pair you turn on, you will either make U look positive and V look negative, or vice versa.



At the end of your pulse, if you turn both transistors off, what happens? The current is not going to stop flowing; that inductor will see to that. Just to keep the current going in the same direction, it will end up using the freewheeling diodes of the opposite pair, sucking current through one of them and pumping it back up to the + rail through the other. That would make the voltage across U to V appear to switch polarity completely. The current pumped back (representing energy that went briefly into the inductor's and transformer's magnetic field but wasn't consumed by the load) goes back into the capacitor.



On the scope, it would be hard to tell the up kicks from the down kicks, because every pulse would start with either U at the + rail and V at the − rail, or vice versa, and would end with both U and V flipping to the voltage of the opposite rail.

But if we're going to make 29,000 kicks every second, and we want the result to be an AC waveform that only changes direction 120 times a second, that means we'll always be making a couple hundred kicks in the same direction, before turning around and making a couple hundred more in the other.

Why not pick the right pair of transistors to kick in the direction we want, turn one of them on for the whole half cycle, and just make pulses with the other one? Then repeat with the other pair. The picture during a pulse (both transistors on) doesn't change, but something different happens when the pulse ends, because only one transistor shuts off.

The inductor is still going to make sure the current keeps flowing, so it will have to suck through one freewheeling diode of the opposite pair, but the current path can include the still-on transistor of the currently selected pair.



Now that's some freewheeling. It doesn't even require bouncing some current back to the capacitor between kicks. The inductor gets to keep the electrons zipping around a closed course between kicks, through one diode and one turned-on transistor, at the same supply rail. The voltage difference U to V will appear positive or negative during a kick, and close to zero in between, and on the scope you can tell the up kicks from the down kicks.

So that appears to be the way this inverter is driving the transistors.

So far we've been looking at the voltage from U to V, that is, just what the IPM is supplying to the inductor+transformer combination. To see what that looks like across just the output transformer primary itself, the scope reference can be left at U (J3), and the probe tip moved from V (J6) to the far terminal of the inductor, where the other primary transformer lead goes. There we'll see what the transformer is seeing, including the inductor's effect of keeping current moving between kicks.



It is totally starting to look like a sine wave being rolled along by hundreds of little kicks.

So that's what the transformer's primary winding sees. What happens to the output of its secondary winding?



The two ends of the transformer secondary winding return to this board at the terminals in the right corner there, J7 and J9. They get filtered by capacitors and common-mode chokes, twice. There is another 310 kΩ resistor between them, perhaps again to make sure capacitors eventually discharge when off.

Then the leg from J9 makes a right turn and meets U11, an interesting blue IC. From that IC it proceeds to J10 (OUTN) and that wire takes it to the back panel as L2 (and AC2 on the Stupid Outlet). Meanwhile, the leg from J7 goes to pin 3 (INV) of J14, that green plug at the bottom edge, and also meets the vacant outline of K2, a relay that isn't there.

Here's where a mystery starts to unravel. There are several signs that this gadget was modified from something that was originally a single-phase UPS. The front control panel, which still has a "BYPASS" LED that lights up when the inverter is off, is one giveaway. There is no "bypass" function in this inverter, no place where you would bring in power from your normal grid supply and let the inverter switch from that to inverter power when the grid failed.

But if you did have such an incoming line connection, you would have brought it to the screw terminal in the lower left here, J4 ("LP", the hot side of the incoming Line), which connects to the normally-closed contact of the missing relay K2, and K2 is what would do the switching. Whatever it selects becomes OUTP at J5, with the red wire taking it to the back panel as L1 (AC1 on the Stupid Outlet).

You can tell the original UPS was single-phase because K2 just has one pole, only switching OUTP to be either the incoming LP or the locally-generated INV. OUTN would have been simply connected to the incoming neutral LN, for all time. The blue U11 IC is a QBC50LX current sensor. You would only need one for a single-phase unit, and it wouldn't matter whether it was on OUTP or OUTN since both would see the same current. It happens to be on OUTN.

Repurposing this design to do split-phase means the single-pole bypass feature can't be used; it would be like putting a transfer switch in a split-phase electrical system that only switches one phase. So K2 is just not even installed on the board, and on the underside there is a big permanent jumper soldered in place taking the INV output to J5/OUTP.

And here, between OUTP and OUTN, we see the final result: the op-art pulse train, fed through the inductor, in the transformer primary, back from the secondary winding, through the two stages of filtering, and finally arriving here:



OUTN in this unit is not a neutral, it is an opposite leg to OUTP, with 240 Vrms between them. The real neutral, the center-tap yellow wire from the transformer, doesn't even return to this board where the red and green do; it just goes straight to the back panel neutral terminals.

The control panel can give you three measurements of current: Iac L1, Iac L2, and Iac N. Naturally, if you measure any two of those, you can math to get the third. But you have to measure two. There is only the one current sensor, U11, on this board, measuring the current on OUTN, which is actually L2 at the back panel. (Note to self: check whether what the LCD calls Iac L1 and Iac L2 agrees with what's labeled L1 and L2 on the back.)

That problem is solved by the additional current-sensor donut underneath the shelf; the yellow neutral from the transformer goes through it on the way to the back panel, and the sensor output is wired back to the smarts board. So the board has real measurements of current on L2 and neutral, and can compute current on L1.

There's a similar issue for voltage measurement. The display can show you Vac L1, Vac L2 (both being the voltage of the respective leg to neutral), or just "Vac" which is the full voltage between the legs.

Again, it only needs to be able to measure any two and it can calculate the third. In this case, no new sensor had to be added, because the original design could already measure two voltages: LP, the utility voltage coming in, and INV, the voltage being generated by the inverter itself. By comparing LP to INV, it could decide when to switch K2 to select one or the other.

The green plug at J14 carries three wires to the smarts board. The labels are only at the corresponding connector on the smarts board: LP, LN, and INV, respectively. In the original design, LN would have been the utility neutral and also the inverter OUTN (that assumption is etched in copper as a trace on this board). The two voltage ADCs on the smarts board would measure both LP and INV, referenced to LN.

In this repurposed design, the middle terminal ("LN") of course is still connected to OUTN, which is no longer a neutral but is the leg opposite INV. So the ADC between LN and INV can be used to measure the full leg-to-leg voltage Vac.

The other ADC, between LN and LP, can still be used for something. In fact, it has to be used for something. So something has to be connected to that J4/LP screw terminal. And something is. Have you guessed what's connected there?

Yes, there is a suspiciously skinny wire there (which is fine, as it has no purpose but voltage measurement), and it is yellow, and it goes to the back panel and meets up with the neutral terminals.

So now the ADC between LN and INV can be used to measure the full leg-to-leg voltage Vac, and the other ADC between LN and LP ends up measuring the single-leg voltage of L2 with respect to neutral—backwards, because now LP is really the terminal considered neutral, and LN isn't. And the Vac L1 display can be computed by subtracting that from Vac. (Note to self: check whether what the LCD calls Vac L1 and Vac L2 agrees with what's labeled L1 and L2 on the back.)

Those details of which sensors are really measuring which switcheroo'd values, and which values are computed, may be useful to fill some gaps in the calibration details that hobbit reported earlier.

I'm sure they also pretty much explain why the voltage feedback details shown in the schematic in the manual look like they were hastily drawn in by someone who was thinking "huh?!".

 

Interlude: calibration, more calibration, and then some calibration

While doing the early smoke tests of the Naked Inverter, I did notice that what it was showing for "Vdc" disagreed by a few volts from what my bench supply said it was supplying. Didn't really get in the way of testing, so I filed that away to think about later. I didn't have an immediate way to decide who was right: this is obviously an old, beat-up unit, the bench power supply was also a vintage eBay rescue, and my handheld Fluke was brand new and certified when I bought it, but that was 30 years ago, and no, I have not been sending it for annual calibration.

Well, I managed to get access to a recently NIST-traceably-calibrated instrument and compare my 30 year old Fluke to it, and on DC V it checked out spot-on. Remarkable. AC volts might be a fraction of a volt off, certainly still close enough for anything I do, but outside the original accuracy limits. Maybe I should get it a calibration for its 30th birthday.

Having confirmed that the Fluke has a good idea what a DC volt is, I then used it to go through the calibration procedure of the old eBay power supply, and tweaked it by barely a fraction of a volt, over a 350 volt range. Again remarkable. So it was definitely this inverter that was displaying Vdc incorrectly. (And not just displaying incorrectly; its DC input under- and over-voltage shutdowns are triggered, of course, by what it thinks the DC voltage is.)

A nice thing about this unit is that calibration is done digitally. I don't see a single adjustable trimmer on any of its circuit boards anywhere. Whatever is tweakable can be tweaked through the "Running Set" menu on the control panel. (By comparison, the little blue AIMS unit has the usual smattering of factory-adjusted trimmers adorning the circuit boards, painted with glop at wherever the factory set them, and without any identification of what each one adjusts.)

So here, one of the "Running Set" parameters is "Vdc Adjust P3014" and it's a factor multiplying the raw reading to get the displayed one. If the displayed reading is off, you pick a new factor new = (old ✕ correct reading ÷ displayed reading) and enter it as the new value of P3014.

After a couple of iterations, I got the inverter displaying the exact same Vdc shown on the power supply and the Fluke, and that happened with a "Vdc Correct" value of exactly 1000. That is, the sensor hardware in the inverter was also spot-on, but at some point, someone put in an odd value for "Vdc Correct". Maybe they were trying to make it match some less-accurate meter.

After that, I moved on to "Vac L Correct P3017". This one's important, because it determines what actually comes out of the inverter. You adjust it the same way, multiplying the current parameter by the actual measured output voltage and dividing by what is displayed. But the effect is a little counterintuitive, because you don't end up seeing a different displayed value.

The inverter is always trying to control the output to make this displayed value be 239-240. You aren't even offered a parameter to adjust that target. So if you change the "Vac L Correct" parameter, the result is that the display still shows 239-240 but the actual output voltage changes. If you increase "Vac L Correct", the output voltage goes down, and vice versa.

This one had been set so it was actually producing somewhat over 240, meaning the "Vac L Correct" was too low. So I raised it a bit, to get exactly 240 Vrms out.

Another adjustment is "Vac L1 Correct P3016". Recall that the unit only has two hardware sensors, one for the full 240 leg-to-leg voltage Vac L, and one for Vac L1 to neutral. It will display Vac L2 by subtraction.

One way to set this parameter would be to make the displayed Vac L1 as close as possible to a meter reading of L1, and not care about any slight error between the computed Vac L2 display and a real reading of L2. Another way would be to compare both displayed Vac L1 to real L1 and Vac L2 to real L2, then set this parameter to split any remaining slight error equally between them. I went for the second approach. It ends up looking better, and both readings end up agreeing with a meter within its limits of accuracy anyway.

At first there was a perplexing issue where the inverter, the Fluke, and the scope all agreed on what Vac L was, but had wildly different ideas for the RMS value of Vac L1 and Vac L2. The inverter was thinking they were right around the expected 120 Vrms, the Fluke (which uses a dedicated RMS converter chip) was coming up with something like 150, and the RMS display feature on the scope, which I'm sure is in software, was showing more like 180.

Sounds like these things were not looking at a clean sine wave! But how could that be? Didn't I already scope the 240 volts leg to leg and show how nice it looked?



Nothing to see there. Hmm, did I ever actually look at either leg to neutral?



Yuck!

Well, maybe I shouldn't be too astonished. Didn't I mention already that both ends of the transformer secondary are brought back to the grunt board and go through multiple stages of filtering by capacitors and chokes, but the yellow center tap just runs straight to the back panel terminals? Make two nice clean legs across 240 Vrms and then reintroduce the switching noise on the neutral. Hmm.

Well, there's another thing. This is all happening on the Naked Inverter. In hobbit's review, it was already shown that bonding the neutral and chassis ground helped a lot to keep the noise down (though in that case, it looked like the noise of interest was coming from the car).

As a Naked Inverter, even though I still have the ground jumper installed on the DIN rail terminals ... without being in any kind of metal box, that's doing nothing to bond the metal shelf with all these innards on it. My oops.

A couple alligator clips between the DIN rail and the shelf, and it all gets a lot cleaner.



Still some noticeable hash, especially around the zero crossings, that isn't there on the 240 volt leg-to-leg, so it has to be reappearing from the neutral. It clearly wouldn't hurt to add some kind of filtering on that straight-to-the-back-panel center tap lead.

But this at least is clean enough for the three different instruments to agree what RMS voltage they're seeing.

On to ... calibrating the current sensors (with limited power!)

There are two parameters for calibrating the current displays: "Iac L1 Correct P3019" and "Iac N Correct P3018". Recall here that the unit only has two current sensors: the QBC50LX chip on the grunt board, which measures the current in "OUTN" which becomes "L2" at the back panel, and the extra donut sensor added for the neutral. Also, I had this note-to-self:

But while testing the inverter on my dinky 0.2 amp bench power supply, where about the biggest load I could get the inverter to power was a 25 watt light bulb, the trouble was, that draws so little current it's about useless for calibration. It wasn't even enough to answer the note-to-self question: both the displayed Iac L1 and Iac L2 sort of wandered around low tenths of an amp at no load, and even plugging in the light bulb several times on each side, it never made enough difference to say for sure "ah, that showed up on L1". Especially when I wasn't sure how well calibrated the displayed numbers were to start with.

I needed something that would draw an unmistakable amount of current. Full-scale would be best for calibration, of course, but that would be around 12.5 amps; out of the question. But heck, I'd settle for something around a couple amps; that would be enough to see for sure which of L1 and L2 was which, and at least get the calibration right to two figures.

But there I was, stuck with a power supply offering 50 watts tops, and the inverter using about 25 of those for itself, leaving a firm budget of 25 watts for any load I wanted to hook up.

Then it hit me: a capacitor!

A 40 µF capacitor, if connected directly across 120 Vrms 60 Hz, would draw about 1.8 amps. 120 V at 1.8 A would be 216 VA, but exactly zero watts (or as close to zero as the capacitor is to being purely capacitive).

Meanwhile, if the inverter's about 90% efficient as it claims, that 216 VA out would have to be about 240 VA in and the extra 24 would be real watts used up in the inverter, but that should be (just barely) in the power budget available.

The capacitor, of course, would have to be something happy on AC (no polarized electrolytics!), and with a sufficiently high voltage rating. An HVAC motor-run capacitor fits the bill nicely. (I would not use a motor-start capacitor; those are only intended for a split-second of use when a motor starts. A motor run capacitor is built for the long haul. It also usually has some safeties, like something to break the circuit if something goes wrong and its pressure rises.)



I was kind of alert for surprises when I pushed the button the first time, but it ended up doing exactly what the math said it would, drew 1.8 amps and stayed almost within the power budget. My power supply went slightly into current-limit mode and that would eventually get the "Vbattey lack" inverter message, but it held up easily long enough to read the displayed L and N current values and compare to the value on the Fluke, and worked just as well again when repeated on the other leg, and repeated again after adjusting the calibration settings to confirm all the readings agreed on 1.8.

And by the way, the answer to the note-to-self question is no: the currents displayed as Iac L1 and Iac L2 do not agree with the L1 / L2 labeling on the back panel.

The reading shown as Iac L1 (and adjusted by Iac L1 Correct) is indeed the hardware reading from the QBC50LX chip on the board. As seen earlier, that's in the path of "OUTN", which appears as "L2" on the back panel. Iac N is N, of course, and it and "Iac L1" are used to calculate "Iac L2", which should match the current called "L1" on the back panel.

I still have one other unanswered note-to-self question, namely, do the Vac L1 and Vac L2 readings match the panel labeling, or are they also swapped like the Iac values? If they too are swapped, it would be pretty simple to just relabel the back panel and call it good.

I'm still trying to think of a way to test which displayed Vac represents which leg.

 

Thought of a way: just temporarily put a 10Ω 10W resistor (had one handy, really 1W should be plenty) between the yellow transformer center tap and the neutral terminal block on the DIN rail. The neutral voltage sense wire (the skinny yellow that runs back to the LP terminal on the board) is taken from that terminal block, therefore downstream of the resistor, and plugging the 25W light bulb into one leg or the other will then pull "neutral" toward that leg by a couple volts, without causing the board to compensate the voltage, as it would if the resistor were inserted in either non-neutral leg.

Then the front-panel Vac L1 or Vac L2 display that corresponds to the leg the light bulb is on will read a couple volts lower, while the other reads a couple volts higher.

And sure enough, what's labeled L1 on the back panel is what's shown as Vac L2 on the display, and vice versa.

This agrees with the note "the line "L1" in the inverter menu is the wire terminal "AC2" printed on the back of the inverter. And vice versa" in the calibration email posted by hobbit, but that email didn't say how that was determined, so I wanted to double check.

It also agrees with what I found earlier for the current displays: Iac L1 goes with AC2/L2 as labeled on the back, and vice versa.

It's convenient that both the V and I displays are switcheroo'd the same way, as it means leaving everything as is and just fixing the back panel labeling would straighten it all out.

By the way, hobbit also mentioned in passing the green three- and six-pin connector blocks on the back panel, which are described in the manual as three pairs of pins that can signal (by dry contact closure) "invert fault", "bypass fault", or "DC fault", and three more pins that constitute a TIA-485 (or, by arrangement, TIA-232) serial interface using Modbus protocol, for querying and controlling the thing.

There's no wiring from this back panel bulkhead connector to anywhere, which led hobbit to call it "a vestige of some different configuration". On the other hand, the smarts board does have an identical six-pin connector right next to three relays that can be heard clicking to indicate faults, and an identical three-pin connector in the corner near where the board communicates with the front panel.

There are two LEDs there that sure behave like transmit-data and receive-data indicators, regularly flickering when the unit is on, in time with a similar pair of LEDs on the front panel board. I would bet that the front panel controls are just built around a simple microcontroller that uses the same Modbus protocol to query and control the smarts board.

So it would seem that all of the functionality represented by that back-panel pin block is really available at the board, and only some cabling from there to the back-panel connector is missing.

 

While I felt it was important to mention that, I realized that for not everyone (including me!) is it obvious how to translate "capacitance of 1360 µF on the input of this inverter" into "so how does that compare to a) a mousetrap, b) a firecracker, c) a falling cartoon anvil, etc.?"

So it turns out the NFPA 70E "electrical safety in the workplace" standard contains an "informative annex" section on working with capacitors, with some helpful advice on gauging the hazards involved.

So that would start by looking at the input voltage. The nominal voltage of a (regular liftback) Prius battery is 201.6, but if being charged near full capacity (such as during balancing) it could be nearly 270.

1360 µF charged to 270 V represents ½ * 0.001360 * 270² or just under 50 joules of energy. Way down at the battery's nominal 201.6 volts, that would be about 28 J. At an intermediate voltage like 220, the inverter's nominal input voltage, the stored energy is about 33 J. So, there's at least 28 J, possibly nearly 50 J, of energy stored in these caps, at a voltage of 200 or more VDC.

The hazard of shock is divided into categories, and even the lowest figures above easily fall in the top one (voltage over 50 V and more than 10 J stored energy), presenting "a significant risk of shock and fibrillation". These caps can definitely give you your last heartbeat.

The thermal hazard categories start at around 100 J or more (where you could get a serious burn if it shorted through a ring you were wearing or a tool you were holding), with higher categories for thousands or tens of thousands of joules. So these caps don't present that category of hazard.

At 50 J, you'd have to be within a cm of the arc to be at risk of arc flash injury (within a couple cm, using the confined-in-a-box formula). So that's not a big hazard here.

The risk of injury from arc blast pressure is bupkis at 50 J, though you could have a one percent chance of eardrum rupture if your ear were within 13 cm of the arc.

In a compliant workplace, given the stored voltage above 50 V and stored energy above 10 J, a written lockout/tagout procedure would be needed for this animal, and the warning label would have to give the wait time for the capacitor charge to bleed below 50 V. Because of the 310 kΩ resistor on the board, from 270 maximum volts, that wait time is 12 minutes. (It only takes six minutes for the stored energy to drop below 10 J.)