Modified Sin Wave Inverter

[metrologist]

This inverter seems to work fine until you put too much load on it. It's rated 2400W and can barely muster 200W before the output voltage starts to drop and go into low voltage shutoff.

I was hoping to see some blown out fets but there is just typical electronics smell, maybe a slight hint of smoke flavor but I cannot detect where. I measured the fets in circuit and there are no shorts and give ~0.6 diode test.

Are there some other places to check before trying to get those fets out? I kind of doubt they are the problem because, don't they normally fail shorted, or blow up open? Is there someplace to probe when it starts to fail to see a problem? I'm not really excited about probing when it's lit but it'll be OK... 

I have several more detailed photos of the components. Fets are six P60NF06, then a HER1603GD rectifier(?), then a pair of 45N20B fets.

Thanks for any tips.

[shakalnokturn]

Yes, usually defective FET's will show one way or another.

Have you monitored what happens to the input voltage while the output voltage droops under load?

[TheMG]

First you need to determine which part of the inverter is causing the problem. There are two sections, the first is a SMPS to generate the high voltage DC rail (usually somewhere between 140-170V for a 120VAC inverter, double that for a 240VAC inverter). The second section of the inverter is an H-bridge to produce the "modified sine wave".

Check to see if the high voltage DC rail dips too much when you load it at 200W+ when the problem starts to occur. If it does, then the problem is in the SMPS section, if not, then the problem is in the H-bridge section.

Normally in this type of inverter the SMPS high voltage output is not tightly regulated, it goes up and down a bit. The output RMS voltage of the inverter is regulated by controlling the duty cycle: a longer duration of the positive and negative pulses results in higher RMS voltage.

A scope can easily tell what's happening all from the output of the inverter, as the peak voltage of the waveform will give you an idea of what the DC rail voltage is doing, while you can at the same time see what the duty cycle regulation is doing.

It's very unlikely to be failed FETs in the SMPS stage, there would definitely be some visual damage and/or blown fuses if that were the case. Those high current MOSFETs never go open-circuit without first releasing their magic smoke in a spectacular way! Bad capacitors or some type of fault in control/feedback circuitry is more likely. For example, I've repaired one where a faulty driver transistor caused one of the 4 H-bridge FETs not to switch on, resulting in low output voltage, reduced load capability, and a very unusual looking waveform at the output (more of a distorted square wave than a modified sine).

[MarkL]

Did this inverter ever put out 2400W and it's now broken?  What is the make/model?

It doesn't look anything near a 2400W inverter to me.

[james_s]

2400W is probably the very optimistic "surge" capacity. Even so, it should deliver a lot more than 200W without shutting down. At risk of being a silly question, is the 12V source feeding it capable of delivering sufficient current? Even at 200W it will be pulling at least 17 amps, and at a full 2.4kW it will draw more than 200A. You need some beefy wire to feed it too, something like welding cable.

[metrologist]

Thanks all for the inputs. I started thinking about caps on the control or output side too. I was just looking at some simple DIY transformer schematics where the 12V+ connects to the center of a transformer and the other two ends of the coils are alternately switched. The other side of the transformer is then stepped up to the AC voltage directly. This looks like a similar configuration on the input side with the red wires feeding the transformers, but then I'm not understanding the rest of the high current components. I also presume the 6 transformers are connected in parallel. (https://electronics.stackexchange.com/questions/546315/why-is-the-center-tap-of-the-transformer-connected-to-the-positive-terminal-of-t)

To answer some of the questions:

I don't recall the brand right now but I don't think it's currently on the market nor a known brand.

You are also probably right, the 2400W rating is probably the surge rating. It may have a 1200W continuous rating.

It was given to me so I don't know if it ever worked.

The input copper is about the diameter of a #2 wood pencil, probably #2 with more than 7 cores.

The supply voltage stays above 13V. I am using it on my 400W off-grid solar system and automotive wet cell. It does power a 700W "Travel" brand inverter to it's load rating for brief periods.

I'll try to get a scope shot later today when I get to the shop.

[metrologist]

Well, here we go, it looks terrible.

No/Low load look the same. 1st image

A few hundred watts and the alarm is on. 2nd image

I put one probe on line (hot) and another on neutral, ground leads on "earth" ground.

The brand is Vector - The Freedom of POWER! It is a 120 volt inverter, but why each leg has 150V?

There are split phase inverters, but then this is 300V p-p.

[metrologist]

Never mind that bit about 300V - lost my mind for a sec.

I should have engaged a math function - a crude mental plotting tells me the no load operation is actually putting out 20V with brief -100V pulses. Under load the 20V period fall off at both ends, still with -100V pulses.

I found the manual and this description

The first stage raises the 12VDC input to 145VDC. The advanced design
of the DC-to-DC first stage uses high-frequency conversion techniques
that replace bulky and heavy transformers found in older inverter technology. The
output stage changes the 145VDC to 110/120VAC. This solid-state circuitry
ensures excellent overload protection and the ability to operate reactive loads such
as those found in inductive motors. The inverter’s output stage uses multiple banks
of metal-oxide semiconductor field-effect transistors (MOSFETS). This stage functions
as a high-power bipolar switch, alternately applying opposite polarity to the AC outlet
HOT and NEUTRAL terminals.

[TheMG]

To get a proper waveform on such inverters you can't use the ground connection as a reference, since it is usually floating.

What I usually do with these on the test bench is power the inverter from an isolated bench power supply or a 12V battery that is not connected to anything else. Then can safely connect a single scope probe between neutral and line.

[metrologist]

I figured I could, but also wanted to see each leg. Now I was not expecting to subtract.? Here are more...

no load

load

''

[james_s]

The output waveform looks about right, I think that part is working more or less correctly. I would find the DC bus and see what that's doing, check it with no load and while applying a few different loads.

[NiHaoMike]

Is the DC/DC part actually going to full output? Vector has a V/Hz ramp up on some models (which I recall turned out to infringe some patent) to ease starting of motor loads. I have the 400W version but I worked out that the V/Hz part worked using a RC delay to change the frequency as well as the output duty cycle. The bigger one you have has the inductor on the DC/DC output that allows it to work well scaling down the voltage, so I think yours is doing the voltage scaling at the DC/DC stage instead of reducing the output duty cycle.

[TheMG]

That waveform actually says a lot about what's going on. Firstly, the H-bridge inverter is doing what it's supposed to. Note how under load the waveform becomes almost a square wave, with very little dead time between the positive and negative pulses, that's the inverter circuitry trying to compensate for a low DC bus voltage by increasing the duty cycle, which in turn increases the RMS output voltage.

The peak voltage, which is representative of the high voltage DC bus of the inverter, looks quite low, and seems to sag quite a lot under load. With no load it's right where it should be at about 150Vp, but it looks like it drops to around 120Vp under load which is much too low! I would definitely turn your attention to the DC-DC converter part of the inverter.

But first, make sure the input voltage going into the inverter is within operating tolerances, measured at the PCB. Quite a lot of current there, so a bad connection somewhere in that path will cause a lot of voltage drop under load (and a lot of heat).

[metrologist]

Well, the only thing I've determined is that with a 100W load, the inverter draws a whopping 40A, or 500W input. As I increment the load, the supply voltage continues to drop until the inverter's threshold voltage is reached.

So something must be getting hot, but the only means I have to tell which components is by touch, and I'm not touching anything in there soon enough after cutting the juice before the caps discharge and whatever cools down.

[james_s]

If it's dissipating that much power you should be able to tell what's getting hot without even touching the part itself. You could try the Louis Rossman trick of dripping some alcohol on it then powering it up and seeing where it evaporates first.

[MarkL]

How are you measuring the 40A current and input voltage to come up with 500W?

Looking at the photo you posted, there aren't even any heatsinks on any of the transistors.  If it's really dissipating 400W from input to output, it's going to cook something on that board.

One way to get an accurate power measurement is to set up the voltage and current waveforms on your scope simultaneously and multiply the two waveforms together.  I'm going to bet the average is a lot less than 500W.  And looking at these waveforms at the inverter input may provide further clues as to what's happening.

[metrologist]

To determine input power, I had my DMM connected to the interter terminals measuring DC volts, and a 100A clamp ammeter around one of the supply cables at the inverter terminal. So, after seeing 40A I looked at the DMM and saw 12.5V. To measure the load, which is a heat gun, I plugged it into my AC wall out through something like a kill-o-watt and it said 104W on its lowest setting.

I have not run it more than a few seconds under any load without the heatsink, which is a single [-channel aluminum piece that spans both sets of fets with the fins going up.

I'm wondering if one or some of the transformers may have melted a bit internally? I expect the fets to be getting hot.

I still have not completely visualized the circuit of this either. I haven't found a diagram online that seems to be similar enough. This one is close, so I assume that the output waveform is created by 4 more fets after the rectifier: https://www.homemade-circuits.com/wp-content/uploads/2021/05/MOSFET-inverter-circuit-compressed.jpg

Many that I've seen just use the output of the transformer for 120V or 240V.

[james_s]

Yes as someone else described, the inverter will consist of a DC-DC converter followed by a H-bridge that generates the AC waveform using 4 mosfets, or more likely 4 groups of 2 or more mosfets in parallel. I doubt there is anything wrong with the transformer, I would suspect a damaged mosfet or other semiconductor. Did you use a true RMS meter to measure the current? The draw from the battery is not going to be smooth DC.

[metrologist]

The meters say true RMS - the AN8008 and the UT210E clamp meter.

When there is no load, there was only 1A supply current on the clamp meter (100A range). Could that mean that the 12 P60NF06 fets (6 each side) that switch the current through the transformers would be OK?

Then perhaps the pair of HER1603GD rectifiers would be fine too?

Now, the two pairs of 45N20B fets would drive the output waveform, so maybe one or two are shorted?

I let it run on a light load (2A input, 21W fan) for a minute and used my IR temp gun and it showed the areas around a few of the input fets got about 2 degrees warmer than the rest, but hard to tell. I felt everything with the back of my fingers (after disconnect) and only the six transformers all felt slightly warm.

Am I just going to have to bite it and start taking out fets? Where would you start?

[shakalnokturn]

I think you're running in circles trying to blame the FET's here.
If they short you know about it. If they fuse after shorting you usually know about it too because it can be seen. Having part of the MOSFETs just open or badly soldered is unlikely because that usually means excessive heat at the weak solder or the remaining MOSFETs popping because they handle too much current.

TheMG had good advice I don't know if you saw it.
To me the output waveform clearly shows some capacitor discharge.
Re-check the HV DC voltage under load before and after the toroidal choke. Consider checking the large electrolytics on the HV side.

[metrologist]

[edit]

I took one 220uF out to measure:
C=198 with D=0.147, Q=6.01, ESR=0.11(@1kHz)
C=205, D=0.037, Q=27.1, ESR=0.2 (@100Hz)

Measured in circuit, the capacitance of the larger electrolytic caps are in tolerance. The two larger one's on the output ESR was ~0.25. I could not get a reading on the input side.

[metrologist]

I lost another inverter. This one was working well. I was running it near max load when my solar system was not keeping up and the inverter spent about 10 minutes hovering around low voltage shutdown where it was toggling in/out of shutdown. I heard slight cracking sounds and saw a wisp of smoke just before I shut it down.

The reveal



The burned part looked like it was once a tantalum capacitor, but I'm not 100% sure. I whacked in an SMD 10-35 N2 and the inverter powered up fine, about 10W standby power, draws ~20% more power than the applied load, tested up to 100W load. So it looks like it's fixed, but it's not.

The SMD cap I put in is getting hot and starting to bubble.  polarity is correct... seems like more load = more hot.

I noticed the 4 caps around the transformers are a little bulged and the plastic labels have shrunk and split - all 4 are 1000uF 35V and I ordered new Panasonic caps. I don't think they are the problem because the inverter had been working before though.

Any ideas what to check?

[SeanB]

Is the capacitors, that little tant is being asked to supply 20A of ripple current, and is of course objecting to it. Of the others, 3 are pretty much guaranteed to be over 1R ESR wise, and the top left one, along with top right, will be dribbling electrolyte out of the bottom.

[TheMG]

I noticed the 4 caps around the transformers are a little bulged and the plastic labels have shrunk and split - all 4 are 1000uF 35V and I ordered new Panasonic caps. I don't think they are the problem because the inverter had been working before though.

Oh I can guarantee the capacitors are part of the problem. Without those capacitors to decouple the DC-DC switching converter from the DC input source, it will essentially be trying to pull significant current at tens of kHz from the DC power source, and most likely the wire inductance won't let that happen and/or the power source will be unhappy about it. Not to mention the fact that it may cause unwanted high frequency oscillations to occur, as well as significant RFI (as if these cheap inverters aren't already electrically noisy enough!).

Looks like a very cost-cut inverter. It may simply not be rated (or it is overrated) to deliver the load you had connected to it on a continuous basis (more than a couple minutes).

[metrologist]

I am getting suspicious my capacitors may not come. 10pc Spamasonics from an in-region seller with 99.1% + and a few thousand sales with decent history.

Today is 1 week past later delivery date, and no tracking was supplied.