Why does the Sun2000/1000 keep having AC shorts when using it rated specs?

At first glance the Sun2000/1000 grid tied inverter appears to be a decent value inverter. But it seems the only way to make it work is running it at a derated spec with a lot of external protection making it fragile to use for most users. It has been around for a long time and even has a current 2nd generation version. You would think that over the many years and with a 2nd generation, improvements would have been made to increase its robustness and reliability. If you do the research, there are plenty of support from user communities such as Facebook and YouTube that have documented their experiences with using the Sun2000/1000 in various configurations. But overall, the reliability doesn't seemed to have improved. We've been playing around with one since April 2024 and we've had our share of failures. One of main failures is when the AC side is shorted which not only causes the inverter not to work, but also the LCD screen won't power up so it appears completely dead with no error messages or any ways to troubleshoot except to open up the unit and use the following guide.

So far, our only failures have been where the AC side is shorted causing the AC fuse to blow along with other components. This blog will attempt to document our journey and hopefully offer solutions to completely eliminate the AC shorts when used within the specifications. It will be updated over time since we are in the process of testing our solutions. We are mindful of people's time so this blog includes a quick summary section and a more details section.

A quick summary:

From a real simplistic explanation, the AC lines (L1 and L2) are each connected by an IGBT (fast transistor switch). Each IGBT turns on alternating at high speed to feed each AC line the current from inverting the DC input. If both IGBTs turn on at the same time, L1 and L2 are shorted sending too much current through the IGBTs causing them to be permanently on or shorted. It can take more than a second (at the maximum rated current for the IGBTs) before the 15 A fuse blows. During this second or so, the gate shorts causing the driver to blow, the SCRs can also blow along with some current sensing resistors and even the PCB traces can melt open. We think there are 2 causes to this condition and offer possible solutions.

Cause 1: The two IGBTs turn on while the inverter's microcontroller is in an unknown state.

Symptoms: AC short happens when unit is not generating or when there is momentary power loss on the AC side creating a brown out condition for the microcontroller.

Our Fix*: Add a battery powered UPS on the DC portion of the microcontroller circuit. This can be done by buying a mini UPS board from AliExpress (select 9V1A) and 3.7 V lithium polymer battery (at least 1100 mAh) and wiring it up yourself (wiring diagram) or buying a complete kit called SunScreen.

Cause 2: During normal operation, the IGBTs get too hot and fail as shorts. We believe they're getting too hot because the IGBTs are not switching fast enough.

Symptoms: AC short happens while the unit is generating at or near rated power.

Our Fix**: Change out the IGBTs to faster ones like these.

*Note, it looks like the latest version of the motherboard (2023-08-09) is attempting to fix this by adding 2 relays that engage the AC lines when the 12 V DC is powered on.

**Note, it looks like the latest version of the motherboard (2023-08-09) is attempting to fix this by using a higher rated full body TO-247 IGBT (1200 V, 25 A vs 600 V, 20 A), but it is slower by 2.5 times for the turn off delay.

Detail explanation:

For Cause 1

The signal to turn on the IGBTs are not being controlled properly. We believe under normal operating conditions, this cannot happen because the inverter's firmware controls the driver chip that is used to control the IGBTs. The only way is if the firmware is not controlling the driver chip as in the case of a brown out of the microcontroller. Basically, if the power supply voltage is below the microcontroller's recommended operating voltage, the microprocessor's outputs are usually in an undetermined state. Since the outputs are connected to the driver and then the IGBTs, the undetermined state could be turning on both IGBTs causing an AC short.

Our fix is to never allow the microcontroller to power down during a drop in AC voltage or brown out by having a battery powered UPS. If the firmware is always running during this time, there is no chance for the outputs to go into an undetermined state.

Our first failure was an AC short after less than a month when the unit wasn't even generating back in April 2024. We installed our brown out prevention fix in June 2024 and we have been running our Sun2000 (45-90 V) with 48 V lithium ion batteries without having any failures during brown outs or while it's not generating.

Since then, we have had AC shorts only when the unit was generating near or at its rated power which we believe is Cause 2.

For Cause 2

In attempt to finding the cause for this type of AC short failure, we installed an 8 A fuse instead of the 15 A fuse to see if we could isolate the failure to a single component since previously the AC short would destroy multiple components so it was hard to determine which actually failed first. The 8 A fuse was chosen due to the reaction time was about 0.07 secs when the max rated current of the IGBT (40 A @ 25 deg C) is reached. The 15 A fuse would take about 1.5 seconds to blow at 40 A. When we got the AC short failure this time with the 8 A fuse, only the IGBTs and fuse were blown. This tells us that the IGBTs were the first to fail and the other component failures were the result of the IGBT failures. So we are now focused on what caused the IGBTs to fail.

At first, we thought of trying IGBT with higher voltage and current ratings, but looking at the voltage and current waveforms on the AC side, there appears to be enough headroom with 600 V and 40 A for continuous operation. We know that ultimately too much heat is what kills it, but didn’t know how the heat was generated. The next thought was to look at the heat generated during switching. This would be looking at the turn on and off delays. The turn on delay is much less than the turn off delay and the turn off delay increases as temperature rises, but the turn on delay does not. So the more current it was switching off at high speed, the more the temperature would rise which would then increase the off delay which would then increase the temperature at a faster rate potentially causing a “thermal runaway” condition if not sufficiently cooled. We think that this would still slowly damage/degrade the IGBT every time it gets close to this runaway condition even though there wasn't a catastrophic failure. It seems to fit with the other symptoms that we have experienced and read about that others are having, more so than having under-rated components.

So at first, we put IGBTs with a longer turn off delay of 199 ns vs 125 ns (original IGBT) and got the AC short failure in about a month. We noticed that prior to a complete failure, the unit would intermitently turn off generating and then back on generating. It did this for a few days before the complete failure. Our theory is the longer delay in turning off allowed the IGBT to heat up more during the off transition thus increasing the chance for thermal runaway and chance to cause damage.

For our current test, we put in IGBTs with a 78 ns turn off delay vs the 125 ns and are waiting for the results. It is also a full body type TO-220 which has better heat dissipation. It is a 650 V vs 600 V rated part with the same current rating as the original one. However, the power dissipation is only 45 W compared to the original IGBT with 140 W. So far, we are already seeing overall cooler internal temperatures and it seemed to be more stable in tracking the power output with our limiter CTs. We started this test on 15th of Nov 2024 and we want to run this for at least 6 months in order to declare success. We don't see how this new IGBT can make things worse, so we want to go ahead and tell people what we are doing so that if you happen to need to replace your IGBTs, you might want to try the ones we are using to help confirm our theory that these IGBTs are better and may be the solution to fixing this cause.