Lack of current sensing in the DC-DC converter:
Normally any switching converter uses current sensing. The better converters employ the current sensor to limit the pulse width on a cycle-by-cycle basis. This allows fast, precise and safe control of the current flowing through the system. Simpler systems do not use cycle-by-cycle limiting, but at least they sense the average current, and reduce the duty cycle when the current is reaching the acceptable limit, to avoid burning out components if something abnormal happens. The Power Jack inverter has none. If this is due to cheapness or incompetence, I don't know.

Unstable control loop:
The DC-DC converter in this inverter uses an SG3525 control IC. Its error amplifier was configured as a plain, simple integrator: A single capacitor as the feedback path. This results in the error amplifier having a constant -90 degree phase shift. But in the power section of the DC-DC converter, the dominating elements are the filter capacitance, and the input source resistance (batteries, cables, MOSFETs, transformers), giving the power system also a -90 degree phase shift throughout the low frequency range! This adds up to -180 degrees, which together with the inverting nature of any control loop produces in-phase (positive) feedback. In this case the positive feedback happens from less than 10Hz to over 1kHz. The limits vary depending on the load on the output, which is the third element influencing the power circuit's response! The filter inductance comes last in order of importance. It's much too small to have any significant effect.

With positive feedback, and positive loop gain, the stage is set for self-oscillations to occur. And this was the cause of the hiccupping I noticed. It was NOT by design. It was by poor design!

The control loop is unstable under almost all load conditions, resulting in the inverter drawing extremely high current pulses, followed by pauses, with the frequency and duty cycle given mostly by the load on the output. I measured pulses in the kiloampere range and millisecond duration, while the average current was just a few tens of amperes!

Flux walking:
Push-pull converters like these are prone to core saturation and consequent failure from flux walking. To make sure no damaging saturation will happen, the only way is to use pulse by pulse current limiting - which wasn't done in this inverter. Another method, not 100% safe but good enough in most cases, is making the error amplifier so slow that the lengths of subsequent pulses cannot change much, and then rely on the MOSFET's and circuit's resistances to balance out any remaining asymmetry. This was also not done in this inverter! The error amplifier has a gain bandwidth product close to 20kHz! As a result, this circuit is prone to core saturation and MOSFET destruction whenever there is significant coupling of the signal on the transformers into the feedback loop!

Inadequate filter chokes:
The high voltage filter uses two filter chokes, each consisting of 30 turns wound on a yellow and white toroid about 27mm in diameter. This seems to be an Amidon/Micrometals T-106-26, or an equivalent. Assuming it is the mentioned type of toroid, the inductance would be 81uH while the current is low. But the highest current these chokes have to take, at the waveform peak when a 3500W load is connected, is 21.5A, and at this current level the cores are already deep into saturation, with their effective permeability being about 1/4 of the original value, and the inductance having dropped to just 20uH! And at the rated 7000W surge power, the situation is even much worse.

This is just the beginning of the problem. At 81uH, the ripple current in these inductors is 16A p-p, which would be roughly 6A rms. This ripple current will theoretically increase to 24A rms, or well over 118A p-p, at full load! It follows that this DC-DC converter never enters continuous current mode operation at all! It's always in discontinuous mode, with the peak current in the diodes, MOSFETs, etc, over twice the average cycle current! In reality, at 21.5A output current, the inductor current will be swinging between zero and about 50A all the time, which means a peak current of over 160A in each single MOSFET, plus any magnetizing current for the transformers!

I leave it to the imagination of the reader to guess how long the filter capacitors will live, when they are subjected to an average of maybe 15A of ripple current, while their ripple current rating is about 4A maximum.

A quick calculation shows that these poor toroids have to work at an AC flux density of 0.35 Tesla. And this doesn't change very much over the power range. Even at a relatively low load power, the AC flux density is already almost that high, and it remains there, as the current continues to increase, while the cores increasingly saturate. The problem is that this high AC flux density, at the high frequency this converter operates on, produce a whopping 53 watts of power loss in each core! That's enough to make these coils unsolder themselves from the circuit, and probably to burn the insulation and/or meet the Curie temperature!

While testing the inverter at very low power (just 100 watts, less than 3% of the rated output), where the inductors are not yet close to the full AC flux density they reach at higher powers, in a matter of one minute they got too hot to touch, and after five minutes they started discoloring, and smelling burned!

To operate correctly, this inverter needs toroidal cores many times larger than these, so that they can be wound to provide a few hundred microhenries, and not saturate until at least 50A, to enable them to handle the rated 7kW surge load too. This would also make them work at much lower AC flux density, producing acceptable power loss and heating. Alas, Power Jack in Taiwan was too cheap or inept to do that.

Optoisolator drift:
The feedback signal of the DC-DC converter comes through an optoisolator, to afford the required galvanic insulation between the DC link, which is at output line potential, and the control circuit, which is at battery potential. In principle this is fine. The problem is that the designer of this inverter apparently didn't know that optoisolators have a strong temperature coefficient! When they warm up, their current transfer rate goes down markedly. And in this Power Jack inverter, the optoisolator is used in such a way that its drift directly affects the "regulated" DC link voltage! As the optoisolator warms up, the DC high voltage goes up, and up, and up, and.... BANG! This is one very likely trigger for the chain reaction that burned out my inverter before two hours were over! To compound the problem, this optoisolator sits in the top of the case, in a pretty warm spot.

Unsafe feedback:
Good design practices dictate that circuits should be made so that in the event of a malfunction they will shut down without further damage, whenever possible. Unfortunately the Power Jack inverter is made in the opposite way. The control loop uses an auxiliary power supply, which is derived from another auxiliary power supply, both of which use SMT components stressed beyond their absolute maximum ratings. Specially the ripple current rating of the small electrolytic caps is far exceeded. And if any of these two auxiliary power supplies fails, the feedback loop is broken, in such a way that the DC high voltage will soar far above the capacitor and transient protector ratings. BANG!

Also, the high voltage is sensed through three tiny SMT resistors in series. Each of these resistors is rated at 100mW. One is 360k, another 200k, and the third is just 20k. The DC voltage is about 340V. This results in a total power dissipation of 200mW. If the three resistors were of the same value, this would mean 67mW on each, which is a bit tight for comfort, but OK. But with the values used, the 360k resistor works at 124mW, well above its absolute maximum rating! If it fails, the feedback loop is opened, the voltage soars, and BANG!

Selecting such different values for the three resistors in series is not a matter of cheapness. It's plain, simple, gross incompetence.

Optoisolators are less prone to failing, and this one is used in the low area of its rating, so it should be pretty safe. But anyway, if it fails open, which is the most common failure mode, BANG, again!

Lack of overvoltage protection:
The unsafe feedback loop, using an optoisolator so that it causes severe voltage drift, would in itself warrant the inclusion of an independent overvoltage protection circuit: Anything that in the event of the voltage rising to a value that is dangerous to the inverter or to the devices powered by it, would activate a save shutdown mechanism. But in this inverter, there is even more reason to use it: The filter capacitance on the DC link is so small, that even a quite moderately reactive load on the output would drive up the voltage on the DC link to a dangerous level! Sure, a cheap inverter doesn't HAVE to handle highly reactive loads. But if the load is too reactive to handle safely, it should shut down, instead of exploding! Well, this one explodes. It would be simple to include an overvoltage detection circuit that shuts down the inverter in the event of overvoltage. This would provide protection both in the event of a control circuit problem, or an overly reactive load. And on top of all that, don't forget that this inverter's DC-DC converter has two sections connected in series, and each of these sections consists of three converter units, and each of these six converters is individually fused! If the fuse for just TWO converters open from overload, which is a perfectly feasible scenario, the DC-DC converter would end up badly unbalanced, with one group providing 3 times as much voltage as the other. With the total voltage regulated to 340V, and each filter capacitor rated at 200V, this would cause 255V to show up on one of the 200V capacitors: BANG! But no, despite all this, Power Jack people were too cheap to include any overvoltage shutdown. The necessary components to afford full protection cost about two dollars. Not including them is either being outstandingly cheap, or being incompetent.

Nonsensical output filter arrangement:
This inverter uses an H-bridge of four IGBTs, with pulse width modulation at 20kHz, to produce the sine wave output. The output filter consists of two really large toroidal inductors, and two film capacitors. All this seems good and fine to me. But when looking at the board, which you can see in the first photo on this page, I would never have thought that The Power Jack engineers chose to place BOTH inductors on the SAME side of the bridge's output! Of course, on a piece of paper, it doesn't matter whether both are on the same side, or each on one side. Electrically, they end up in series anyway. In practice, however, it makes a whole world of a difference, in terms of interference to other devices! As it is in this inverter, one side of the chopper appears directly connected to the output, and the negative side of the DC bus is capacitively coupled to the negative of the battery. This results in 340V of square wave 20kHz riding on the output during one semicycle, relative to the battery circuit, which is typically grounded! Talk about a big noise bomb!

And yes, there is an EMI filter in the output line. Unfortunately, it is configured the wrong way, and so it's completely useless: The small ground bypass capacitors were installed on the side of the filter that goes to the chopper, rather than on the output side! They surely blew up when at the lab they actually connected them to ground, because that results in a powerful 340V 20kHz signal connected straight to two little ceramic caps: The one from the "hot" bridge side to ground, and the one from the negative DC link side to ground!

Do you want to know which was Power Jack's solution to this problem? Simple: They left the ground connection of the EMI filter disconnected!!! The result: Absolutely impressive, all-overwhelming radio interference! This is particularly bad because many people buy sine wave inverters to feed sensitive studio audio equipment, or laboratory instrumentation, with a clean AC supply! Ha ha!

Risk of death by electrocution!
I wouldn't have written this web page bashing on Power Jack, if their product was simply unuseable. But when I noticed the gross, extremely dangerous blunter they did by joining the AC input to the output, I decided I had to write this page! I hope that nobody has been electrocuted yet by one of these inverters!

As explained in the beginning, this inverter includes a battery charger, and switchover circuitry for UPS (Uniterruptible Power Supply) service. So there is an AC input and an AC output. The AC output is switched between the AC input, and the inverter output, and in addition the AC input goes to the charger. So far, so good. The problem is that the geniuses at Power Jack chose to use a single pole relay to switch just ONE side of the 220V line, while the other side of the 220V remains connected at all times, to the input, output, charger and inverter! As a result, when using this thing as a standalone inverter, one of the 220V output poles is connected to the male AC input plug!!!

The concept of this was probably to use this permanently connected side as the neutral, but that doesn't work out, because the input and output connections don't use polarized connectors. The output universal jacks can be connected in any polarity, while the input connection comes from Taiwan just as a stripped wire, to which the user has to install a plug of the kind used in his country. And the manual has not even one word of warning about phase, neutral and the like!

So it's perfectly possible that someone connects this inverter to his house, leaving the input disconnected, and uses a standard plug to mate with the jack on the inverter. The house wiring will normally have the neutral grounded. Depending on which way this plug is put in, one pin of the 220V input plug will have either the neutral, or the phase on one of its pins, exposed for anyone to touch!

Folks, this is more than being cheap, or being incompetent: It wouldn't be hard for a lawyer to bring up charges for attempted manslaughter, or, if someone actually gets killed, erase the word "attempted"! I can only hope that each of these Power Jack inverters burns out quickly enough to become "safe", before getting a chance to kill anybody. And I'm glad I didn't touch the exposed male plug on mine, in the short while it worked!

After this major scare, let's go back to what may look like petty issues.

Lack of EMI filtering in charger:
The charger is a simple half-bridge circuit. It follows the trend of crappy Chinese PC power supplies, and has absolutely no EMI filtering whatsoever! The AC input goes via an inrush current limiter directly to the rectifier, filter caps, and from there to the IGBTs. As a result, the charger causes severe radio interference. Not as brutal as the one the sine wave chopper causes, but still severe enough to disrupt HF and MF radio reception.



Instability of the charger's control loop:
The control circuit for this charger has both voltage and current regulation, which is fine. The problem is that the control loop is misdesigned, resulting in instability when operating in current-limited mode. The charger pulses heavily, and if left doing that for long enough, it would fail. It requires re-design.

Here you can see an oscillogram of the charger's output voltage. The main square wave happens at about 160Hz, and the triangle wave on the upper side of the square wave is at roughly 3kHz. The "fog" around the waveform is noise fed through from the main 100kHz switching process. As you can see, the charger's control loop oscillates at two frequencies at the same time, with some modulation on each, and on top of that there is about a half volt of switching frequency noise! How's that for crappy design?

It looks like the designer copied the control loop from some application note or other source, without understanding it, and not realizing that it was meant for a power supply, not for a battery charger with its immense de-stabilizing "capacitive" load attached!



Poor design of the thermal attachment of the power components.

Near the beginning of the page I showed you photos of how the TO220 cases were mounted such that their tops separate from the thermal pad, and how these people smeared thermal grease over it, in a vane attempt to fix the problem. Well, here is a top-down view of one of the larger parts. These have their mounting bars at the proper height, but use just one tiny screw in the middle to press down two big components. Before having any chance to apply enough pressure, the bars bend, and apply pressure only to one edge of the part! The result: Very poor thermal contact, because a big portion of the part's seating surface ends up with a layer of air between it and the heatsink!

You can see the the board through the wedge gap between the pressure bar and the part!

And one more problem: Even electronic hobbyists, let alone engineers, know that power devices first must be mounted to the heat sink, then soldered to the circuit board! This inverter is designed the other way around. There is no way to solder the parts once installed on the heat sink. It becomes necessary to solder them first, and then install the whole board with all the parts in the case, and bolt down the parts as the last step! This results in poor seating, and severe mechanical stress on the connection pins.

By the way: The DC-DC converter board, with its heavy transformers, has no mounting screws whatsoever. It's attached ONLY by the pins of the TO220 MOSFETs and diodes!

Oh, and while we are looking at the cooling, let me tell you that the case of this inverter, which acts as the only heatsink for the big power semiconductors, has the fins on the outside, in horizontal orientation, while the fan draws air through the inside, where there are no fins! Not that this would matter much, because anyway the fan is essentially decorative: The openings through which air can enter the inverter have a total combined area of less than 5% of the fan's area

You might say I'm almost there... But not so. Stability of the control loops has been a nightmare. In part it's because I still haven't replaced the much-too-small filter inductors in the DC-DC converter. As they saturate to different degrees, the power loop of the converter changes its behavior. And for the charger, I only managed to obtain conditional stability under most operating conditions, but at that point where it moves from controlling the current to controlling the voltage, during the charge cycle of a battery, it's still unstable, and it seems that nothing I do can stabilize it!

Also the thermal coupling of the parts remains to be improved, the filter inductors need to be replaced, and many other such "minor" things, but the really big obstacle is this: I don't see any way in the world to stabilize the control loop of the DC-DC converter, while at the same time assuring that it cannot develop flux walking problems. The issue is that not having primary current sensing, flux walking cannot be prevented through cycle-by-cycle current limiting. That requires using a slow error amplifier, but such a thing has a 90 degree phase lag, which adds up with the basically RC behavior of the power circuit, to create instability conditions. With a fast error amplifier, that has phase lead instead of lag, it would be easy to stabilize this circuit, but then there would be severe flux walking problems! No solution seems possible other than implementing primary side current sensing, and that's difficult, given that it would have to be done AT LEAST in each of the six converter groups separately, and that the currents involved are several hundred amperes, and that the room available is tiny! Any "solution" without primary side current sensing will be unreliable.

For the moment, I have given up, and intend to bury the shiny almost new Power Jack inverter in a deep, damp hole, then plant a tree on it, the most thorny sort of tree I can find.

Mai intai si mai intai verifica sigurantele fuzibile .

Mare atentie la pozitia in care se afla in caz ca o sa gasesti una arsa.