Cheap TBE Inverter teardowns

In this page, we will have a look at:
TBE 3000W Modified sine inverter
TBE 2000W Pure sine wave inverter

       Up for teardown is a TBE 3000W modified sine wave inverter that was sent in by a supplier for me to check if it is viable for repair. These are notorious locally for easily failing even just by looking at it the wrong way. To verify these claims and confirm the poor performance, I took this chance to see how well (poorly?) these things are built.


This is the broken inverter.

It is a TBE 3000W modified sine wave model.

This is the inverter torn apart.

We can see there is a big problem here. All switching fets have burst into flames.

The bottom side looks pretty clean. There are two SOIC16s which are TL494s based on the pinout, the SOIC8 chip is probably an LM358 dual op amp. All of them have their markings erased.

Up front, the topology is your run of the mill step up DC-DC converter and H-bridge setup with minimal noise filtering.

PCB Version with a link to the manufacturer web page.

I guess they use what they can get their hands on. The caps are different, some are 16V, others 25V then they are different series too.

Other side.

The only thing they have in common are that they are all Rubycon branded and appear to be NOS.

The yellow wires go to the power switch. This got really burned when the switching devices burst into flames.

The white plug is for the fan. There is no fan control, it just runs when the switch is turned on.

Something is wrong here. The main switchers are marked as IGBT.

IGBTs aren't good in low voltage, high current switching. IF they are indeed IGBTs, I guess that is the reason this inverter blew up.

Thermistors unpopulated. Who needs thermal protection!?!

H bridge current shunt.

Which looks like two pieces of offcut component leads.

Output rectifiers also have the numbers erased. There is no thermal grease or silpads between the rectifiers and heatsink.

TO-92 devices are MJ13001 for the high side drivers.

The large devices are the HV H-bridge switches.

       Looking at the build quality, I can tell they use surplus/NOS parts or whichever is available for cheap and build it using those parts.

       Now, I can't do much testing as this is a burned up unit. It will most probably cost more to fix than get a new one so I will wait for the supplier's action on it.

Back to top of page

       Wait, I have another unit that is working which I have installed in our home RE system. Let's take a look at that instead! My unit is the same brand but is a 2000W Pure sine wave model. This will definitely be more complex circuitry wise but the build quality should be similar.

This is my 2000W pure sine wave unit

Removing four screws, we can pop the top cover off. The heatsink extrusion is a nice clamshell design.

The bottom looks a bit messy but they get a plus from me for putting copper wires to beef up the high current traces.

This wire is not soldered properly.

It, along with another, carries the high current B+ supply.

I resoldered it for reliability.

These fuses are simply stuck in faston crimp connectors and soldered into the PCB.

The OCD in me kicked in so I resoldered the fuses so they are neatly arranged.

An out of place SMD resistor.

Must have been nudged out of place. I resoldered it properly into the part lands.

The step up converter oscillator. This is branded TBE but I have seen the same board on a SUOER pure sine wave inverter (with different silk markings).

This uses an SG3525 oscillator and an LM324 quad op amp for over/under voltage and thermal protection(not used in this model)

The sine wave PWM control board. The main chip is an EG8010 driving two IR2110 mosfet drivers.

The other two SOIC8 parts are an LM555 and an LM393 for short circuit/overload detection.

This board is also the same as in the Suoer pure sine wave inverter. I'm guessing they are just copying one design and rebadging it or is from one designer after all.

Front panel sub board. This board holds the indicator LEDs, the USB jack and its 5V DC-DC converter.

The 5V converter is a basic MC34063 based switchmode buck converter.

The sub board also uses Rubycon caps.

Thermal switch. This turns the fan on when the heatsink gets hot.

If the other heatsink gets hot, then you're on your own.

The main switching devices. These have the numbers erased but I can see one of them marked IR. The problem is that these are all from different batches! No two have the same batch code! A manufacturer usually has all the same batch codes in one product as they buy in bulk. These appear to be NOS as the terminals are badly oxidised.

An important reason for having matching date codes is that these are operated in parallel. Matching date/batch codes means that MOSFETs in parallel will share current better. A mismatch in Vgs or RdsON will make one take more current than the rest and causing it to blow up, which then blows the others up and so on.

Another thing I found is that the devices are not even tightly bolted to the heatsink! Using silpad insulators make this worse as they need a fair bit of pressure to reduce the thermal resistance. They did not use silicone grease though which is correct for silpad insulators.

They erased the part numbers but wait, I spy... "1503"

The mosfets are indeed IRF1503's as confirmed by measuring gate capacitance and Vds breakdown voltage. These are rated at 30V 75A 3.3mOhm.

It has decent current capacity but are only 30V breakdown voltage.

With it driving a center tapped transformer primary, If the B+ voltage exceeds 15V (during battery equalization or bulk charging) then the 30V breakdown voltage of the MOSFETs are easily exceeded. We haven't even gotten into voltage spikes and ringing due to leakage inductance!

New devices vs the old.

I used IRF3205s which have a higher RdsON of 8mOhm but are rated 55V 110A. The higher voltage is needed for allowance from voltage spikes and high battery voltage. It also has a lower gate charge than the originals so it is easier to drive.

New MOSFETs screwed in place before soldering.

These are all properly tightened with the right torque on my HIOS high speed screwdriver.

Housekeeping regulators for the sinewave controller. The step up converter runs open loop, these optocouplers transfer protection signals from the Sine modulator to the main SMPS oscillator.

Output rectifiers (in grey rectangular silicone bags), output HV bulk caps and output switching transistors.

The output switching IGBTs. These also have mismatched date codes and are heavily oxidised.

Cleaned up, these are Fairchild SGH40N60 IGBTs.

Devices are rated at 20A 600V.

Matching is not too important in the output stage as they are not in parallel.

Main HV output capacitors. These also appear to be NOS due to the scratches and the markings on the sleeve starting to rub off in places.

I promise, it came that way!

It is already dented but rubber seal still intact and appear to work fine.

Yep, they also erased the numbers on the output rectifiers.

And yes, they all have different batch/date codes too!

I thought that the rectifiers are fully insulated types so I just used new silpads and kept the fully enclosed ones for future projects.

Output X and Y class capacitors for noise suppression.

Output common mode choke is omitted to save a few cents.

Found a bunch of common mode filters to improve EMC performance. The left one is too big(but the pins line up to the PCB holes), the middle one fits perfectly but is damaged and the last one just might work with a little persuation.

This fits like a glove but it is a burned up unit. Besides, even if it still worked, the windings are too thin to carry 2000W.

I decided to use the toroidal common mode choke instead but it needed a bit of lead bending to fit.

I also added a wire to connect the chassis and outlet earth pin to the filter capacitors for better shielding.

       Now, at first I was skeptical for this to be a pure sine inverter as those cost a fortune a few years ago so I take out the scope and took measurements.

Since the output is galvanically isolated from the battery input, the probe ground is connected to the HV caps negative terminal. This is what we get at the NEUTRAL pin on the outlet.

Wait, this is supposed to be a pure sine wave inverter, right? Don't worry, it will make sense later.

This is the one measured at the LIVE pin on the outlet.

The inverter modulates only one line so that the output filter is greatly simplified. Instead of using a lot of LC filtering and common mode chokes, doing it this way reduces the output filter to just one inductor and one capacitor and still get a sine wave across the load.

During the fast transitions, the voltage across the load is zero. There is only voltage during the curved portions which results to a differential sinewave voltage between LIVE and NEUTRAL.

Check out the EG8010 datasheet for the schematic of the output switches and LC filter to see how it was done.

The two phases overlapped on the scope screen.

Output waveform at 10.5V battery voltage.

Don't mind the trigger frequency counter. It could not trigger cleanly on the signal so it was showing an incorrect frequency.

Output waveform at 12.8V battery voltage.

Output waveform at 15V battery voltage.

You can clearly see that the main DC-DC step up converter is running open loop and unregulated which greatly simplifies its design and improves efficiency. The output voltage is regulated by the PWM modulation of the LIVE output terminal.

Incase you can't figure out how it becomes a sine wave, I edited the image so that one side is referenced to the NEUTRAL pin and this is how it would look like.

To derive a common point at the output, I used a center tapped primary of a transformer as grounding point for the oscilloscope probes.

And this is the waveform that we get. See, it is indeed a pure sine wave inverter!

       I have run a 1500W microwave oven on this before the mod for 3minutes and it survived just fine drawing about 116A from the battery and it did get pretty warm. I have not done some power testing after the modifications as it is cloudy and raining so there isn't enough juice in the battery bank. I will update when I get a chance and hopefully add some thermal images at high power operation as time and weather permits.

       20 Dec 2014:

       Today is a sunny day. Now we proceed with some high power testing! Battery voltage is at 14.5V before the test, ambient temp at around 29degC. The load is a 1500W rated microwave oven. Measured to be around 1200W / 1400VA.

Test setup.

The oscilloscope measures output waveform via a small 220 to 12V transformer. The PV is also charging the battery during the test.

The clamshell case makes it convenient to remove the cover for access on the internals during testing.

Waveform at no load condition.

I would trust the trigger counter for running frequency as it is sampled for longer periods. The measure function just estimates it from the displayed waveform.

Waveform at full load. There is slight noisiness in the output but is still an unclipped sine wave.

There is also very little noticeable voltage drop showing that output voltage is well regulated.

MOSFET temperature at over 1min into the test.

This is not a fair comparison as the fans installed suck air out of the case. With the case open, the fans are actually not helping in cooling parts of the inverter. I would assume then, that with the inverter fully assembled, the internal temps should be a little lower.

The hot spot in the middle is the housekeeping regulator for the SPWM modulator board.

The output filters are warm but didn't get really hot.

The common mode filter that I added handled the load well and did not get warm.

Output IGBT temperature.

The heatsink for the output IGBTs are a tad cooler to the touch than the main switching MOSFETs.

The hot part here appears to be the one set of the fuses in the middle.

Seems the other set of fuses are getting hot too.

       The test appears to be a success. There are a bunch of hot spots but my MOSFET upgrade seems to be holding up. I don't intend to power anything heavier than this microwave oven so the inverter will have an easy time on my usual loads. I have also found another Chinese company selling the same inverter under a different brand. Same size, same topology, same heatsink extrusion! same endpanel layouts and connectors but different case color. The specs are the same as what I measured - Idle current, over voltage and under voltage protection points, etc but is instead rated at 1000W cont/2000W peak.

Page created and copyright R.Quan © 16 Dec 2014.