"Improving" my Power Jack Inverter

[Jnjvan]

I recently purchased one of the ~$200 5000W Power Jack AMG inverters as a convenient means to do some battery load testing. This is the 12 Vdc in, 110/220 Vac out split phase version. SInce that task is complete, and the inverter is now for backup/play duty only, I am free to muck about with it. It should come as no surprise to most that there are a number of less than ideal aspects to the design of this inverter. Many of these items have been discussed on this forum, and elsewhere. I have decided to rectify some of these shortcomings.

Before I hear the "lipstick on a pig" type comments, realize that I am doing this for fun and education. I have no illusion that I will emerge with a Cadillac inverter, when all is said and done. With that said, here is a list of the mods to date:

1. Replaced the 15A glass fuses with pushbutton resettable circuit breakers. Fuses are fine, but I prefer breakers, and the fuse holders looked a bit sketchy. Plus, I know the time/current characteristics of the breakers.
2. Rewired the front panel so that the breakers were on the L1 and L2 terminals. As delivered, the fuses were on the L1 and N2 lines. If this inverter is ever connected to an electrical system, I prefer the overload protection be on the hot side, rather than on the neutral (possibly grounded) side. I also reconnected the 120V ac outlet on the panel so that the L1 and N1 connections went to the correct pins (they were reversed).
3. In addition to #2 above, I rewired the connections to the control board, so that the current sense transformer (overload monitor) was on the L1, rather than the N1 line. This probably doesn't make any practical difference, but it seems "proper" to me.
4. Added 4 uF of capacitance across the L2/N2 lines for noise filtering. I measured about 82V P-P ! of 24KHz switching noise across these lines prior to the addition of the cap. PJ has 9.5 uF on the L1/N1 side, and none on L2/N2. This change did increase idle current by 300 mA.
5. Added thermal paste between MOSFETs and heat sink. Maybe negligible difference, but I have no confidence in heat transfer without it. Note that the thermal compound must be electrically conductive, as the inverter relies on this interface to carry high current. I used a silver filled, electrically conductive thermal paste.
6. Removed RA3(?), the 3.3K current limiting resistor for LED4. Sid pointed this one out to me. This LED is used to indicate the presence of PWM drive to the FETs. Unfortunately, it significantly reduces the FET drive to one of the lower FET banks, so the FET drive is less symmetric. Removing the resistor restores the drive symmetry (but disables the LED). As a result, the inverter idle current was reduced by 40 mA, and the transformer buzz was reduced appreciably.
7. I did not like the ~18 volt gate drive power supplies to the upper FET banks, as the absolute maximum rating for the gate voltage is 20 V. To lower this voltage, I soldered 100K resistors across R7 and R30 on the driver board. This reduced the  drive voltage to 13.2 V at the FET gates. I would have preferred 15V, but 100K resistors are what I had on hand. 

This is what I have done so far. I do want to dig into the mysterious dip switch settings that determine current overload level. If I sort that out, I will report here.

I should also mention that I do not recommend any of these changes. They will certainly void your warranty, and may result in unexpected failure modes, decreased reliability, and/or hazardous conditions.

John Van

[Sid Genetry Solar]

7 minutes ago, Jnjvan said:

Added 4 uF of capacitance across the L2/N2 lines for noise filtering. I measured about 82V P-P ! of 24KHz switching noise across these lines prior to the addition of the cap. PJ has 9.5 uF on the L1/N1 side, and none on L2/N2. This change did increase idle current by 300 mA.

Considering how small the transformer is in the inverter, I'd suggest popping one of the 4.7uF caps off that PJ put on.  That way you'll have a balanced 4.7uF-per-phase filtration.

Or could just move one of their caps to the other phase.

 

8 minutes ago, Jnjvan said:

Added thermal paste between MOSFETs and heat sink. Maybe negligible difference, but I have no confidence in heat transfer without it. Note that the thermal compound must be electrically conductive, as the inverter relies on this interface to carry high current. I used a silver filled, electrically conductive thermal paste.

Ideally, the FETs wouldn't dissipate a lot of heat; however, on a PJ inverter, this isn't a bad idea at all 😉

Worth noting that the inverter in my house closet has thermal paste on it--until one of these days when I replace it with a proper GS inverter...

 

9 minutes ago, Jnjvan said:

Removed RA3(?), the 3.3K current limiting resistor for LED4. Sid pointed this one out to me. This LED is used to indicate the presence of PWM drive to the FETs. Unfortunately, it significantly reduces the FET drive to one of the lower FET banks, so the FET drive is less symmetric. Removing the resistor restores the drive symmetry (but disables the LED). As a result, the inverter idle current was reduced by 40 mA, and the transformer buzz was reduced appreciably.

Reducing inverter idle current...that's significant.  Indicates that less power is getting converted to heat in the FETs.  Gotta love weak FET driving....

 

10 minutes ago, Jnjvan said:

I did not like the ~18 volt gate drive power supplies to the upper FET banks, as the absolute maximum rating for the gate voltage is 20 V. To lower this voltage, I soldered 100K resistors across R7 and R30 on the driver board. This reduced the  drive voltage to 13.2 V at the FET gates. I would have preferred 15V, but 100K resistors are what I had on hand. 

Did this quiet the inverter down any?  A GS inverter with a fully matched and balanced FET drive is almost dead silent at no load (there might be a tad of light "buzzing" from the choke[s], but not the main transformer.)

Funny that PJ ships some inverters with the transistor voltage regulator circuit SHORTED OUT.  I don't know how they don't blow FETs right then and there due to gate overvoltage, but it supposedly stops the inverter from blowing out under load...

 

12 minutes ago, Jnjvan said:

In addition to #2 above, I rewired the connections to the control board, so that the current sense transformer (overload monitor) was on the L1, rather than the N1 line. This probably doesn't make any practical difference, but it seems "proper" to me.

Gotta love PJ...I thought all along it was on L1, but PJ's found a new way to adjust things 😉.

 

12 minutes ago, Jnjvan said:

I do want to dig into the mysterious dip switch settings that determine current overload level. If I sort that out, I will report here.

This is nothing more than a bunch of resistors that the DIP switches connect or disconnect as a load resistor across the overload current sense transformer.  The CPU has a single hardcoded "limit" in it (regardless of the listed size of the inverter)--and this "adjustable load resistor" basically attenuates the signal from the current sense transformer.  The higher the load on the output of the transformer, the lower the signal at any given load--and thus the higher the overload limit.  (As well as charge limit, power save threshold, and anything else to do with output current!)

You should be able to measure the output of the transformer with a DMM set to AC Volts, and see what each setting does.  If you measure a "baseline voltage" with all the switches off and say a 10A load on the inverter, you should be able to mathematically calculate what the DIP switch combinations would do (based on the resistors next to it on the PCB).

The new 11.1 control boards add a little transistor circuit that's supposed to adjust the current limit for charge to a lower amount.  However, it seems to also significantly reduce the output current as well.

 

One other suggestion I'd make: add another full turn around any and all ferrite chokes on the transformer primary.  2 full turns seems to be about the best result--that alone can cut 250mA off your no-load current.  (They usually use just a single turn.)

[Jnjvan]

15 hours ago, Sid Genetry Solar said:

Did this quiet the inverter down any?  A GS inverter with a fully matched and balanced FET drive is almost dead silent at no load (there might be a tad of light "buzzing" from the choke[s], but not the main transformer.)

Funny that PJ ships some inverters with the transistor voltage regulator circuit SHORTED OUT.  I don't know how they don't blow FETs right then and there due to gate overvoltage, but it supposedly stops the inverter from blowing out under load...

I didn't notice any reduction in noise with this change. I believe that you may be correct in that the buzzing may be coming from the choke. Hard to tell where exactly it is coming from, I had just assumed it was from the transformer.

PJ (or whomever they copied the design from) made an odd choice of bias selection on the TL431. As delivered, the regulator was hardly doing any regulating. I.e. the setpoint was such that the regulated voltage was only lower than the input by a couple hundred millivolts. Hardly worth the effort. Interesting that they are shorting out the regulator now. Not sure how this helps avoid failures.

I don't think that I can add another turn in the ferrite choke. This transformer has four bundles of primary wires, probably 8 or ten in hand, wound through there. Not enough area to jam in another turn.

On another note, I have been thinking about converting this unit to 48V input. I see that there are 4 or 5 selector shunts on the control board that are marked 12, 24, 48. What else changes for control of a 48V system? Power supply? Is the software the same for the different input voltages? I would have to replace the fan, and the 40V fan FET with suitable parts. Of course, the transformer primary would have to be reconfigured. I think I could do this without too much trouble. What else am I missing? Does PJ use higher voltage bridge FETs for 48V? The current 80V FETs (NCE80H15) are probably marginal for a 48V nominal input.

[Sid Genetry Solar]

1 hour ago, Jnjvan said:

I didn't notice any reduction in noise with this change. I believe that you may be correct in that the buzzing may be coming from the choke. Hard to tell where exactly it is coming from, I had just assumed it was from the transformer.

With a PJ inverter, there's going to be a regulation oscillation providing a 15-30Hz hum from the transformer.  I'm used to hearing a pronounced AC hum from a PJ inverter; a GS at no load does not have this hum.

 

1 hour ago, Jnjvan said:

PJ (or whomever they copied the design from) made an odd choice of bias selection on the TL431. As delivered, the regulator was hardly doing any regulating. I.e. the setpoint was such that the regulated voltage was only lower than the input by a couple hundred millivolts. Hardly worth the effort. Interesting that they are shorting out the regulator now. Not sure how this helps avoid failures.

That was what they told me, shorting it out somehow stopped MOS failure at high load.  Goes to show how weak the driving is if that was a solution!

The original design used a zener shunt across the output of the transformer.  Which if you're only seeing a small change, might be why the original design did that.

Rumor has it the original design (which is ubiquitous in ALL Chinese LF inverters, showing a very low inventiveness there!) was copied from an older Xantrex inverter.  I have no idea which one...that's just what I've heard.

 

1 hour ago, Jnjvan said:

On another note, I have been thinking about converting this unit to 48V input. I see that there are 4 or 5 selector shunts on the control board that are marked 12, 24, 48. What else changes for control of a 48V system? Power supply? Is the software the same for the different input voltages?

PJ uses ONE firmware for ALL their inverters, regardless of size, spec or voltage.  They just adjust the signals going to the CPU at a hardware level for different setups--which makes the inverter extremely frustrating to write firmware for.

Power supply circuitry does not change, but you may need to replace the filter cap right next to the heatsinked FET (by the switching power supply transformer).  A 48v inverter will need at least a 63v cap here; better with 80 or 100v.  (A PJ 11.1 mainboard here for 48v has a 220uF 100v cap.)

Might also check the heatsinked FET there to make sure it's got a sufficient voltage rating...

 

1 hour ago, Jnjvan said:

Of course, the transformer primary would have to be reconfigured.

Rewound is more like it.  You'll likely have to completely remove the transformer primary, and string it together in twice the length, half the thickness...then rewind it.  Let's just say I did that on one transformer (actually rewound the whole thing, both primary and secondary), and that pretty well cured me from wanting to rewind toroidal transformers.

I know what you're probably thinking, "it's got 4 leads on the primary, so I can just put them in series."  Count yourself extremely lucky if that is possible--because PJ winds the transformers...then just stuffs 8-10 wires into each terminal, and crimps 'em down.  There is no separation of "bundles" between the terminals: if you pick 10 strands in one terminal, they'll likely be divided randomly between 2-3 different terminals on the other end of the coil.

If it's a 24v inverter, you MIGHT be able to get away with trying to separate the primary into 2 parts (requires cutting the terminals off one side of the coil, removing the enamel with sandpaper, and then recrimping new terminals on the proper wires.  You'll want a hydraulic crimper for this; you can't solder it, as the transformers are wound with aluminum.)

 

1 hour ago, Jnjvan said:

Does PJ use higher voltage bridge FETs for 48V? The current 80V FETs (NCE80H15) are probably marginal for a 48V nominal input.

80v will be fine; there isn't much "splash" on the FETs.  Firstly, the transformer is driven with a push-pull topology--and secondly, the body diodes in the FETs will "dump" any overvoltage right back into the main filter caps / battery. 

I do have my concerns with the 80H15 FETs, as they are relatively fast switching (50nS turn-on time)--but have a HUGE reverse transfer capacitance ("Miller capacitance") of 460pF.  (Compare to the GS FETs at 31.5nS and 26pF, respectively.)  The higher the battery voltage, the worse this problem becomes.  But as a PJ inverter drives the FETs quite weakly to begin with, it might be OK...fingers crossed tight...

[Steve]

I also have an inverter I'd like to improve

Mine is a rebranded power Jack (upower 10000W) 24V input

I blew the fets and they sent me more so it works now

I'm wondering what FETS would be suggested as an upgrade 

Original are RU7088R

[Sid Genetry Solar]

5 hours ago, Steve said:

I also have an inverter I'd like to improve

Mine is a rebranded power Jack (upower 10000W) 24V input

I blew the fets and they sent me more so it works now

I'm wondering what FETS would be suggested as an upgrade 

Original are RU7088R

For a PJ, the FETs are often the most obvious collateral damage resulting from an entire upstream set of design issues.  Putting more expensive FETs in the inverter will pretty well just make the blowups more expensive.

We at GS tried starting out by improving a PJ inverter--but each time we found a problem to solve, we'd have to redesign another part of the inverter to fix it.  Which is part of why we no longer sell "upgrade kits"...because you might as well start completely over if you want a truly better inverter.

 

Depends how you blew the inverter, if you care to elaborate?

[Steve]

Well I had it running the booster pump in my well setup.

Motor pulls about 9A running but has a heavy inrush (56A)

Runs under a minute at a time.

2 FET boards shorted on the same leg of the transformer

Clipped out the shorted fets and unit ran on what was left so driver board survived

 

 

 

 

 

[Sid Genetry Solar]

1 hour ago, Steve said:

2 FET boards shorted on the same leg of the transformer

Clipped out the shorted fets and unit ran on what was left so driver board survived

Have to ask...by any chance could you measure the resistors on the gate lead of the blown FETs, and see if they are open circuit?

[Steve]

10 ohms all good.

I identified the bad fets by checking for gate shorts

 

[RobertM]

9 hours ago, Steve said:

 

Motor pulls about 9A running but has a heavy inrush (56A)

 

 

 

 

What size is the motor in hp?

[Steve]

That I don't know

 

[Steve]

Good tip on 2 turns in the choke dropped my idle from 24W to 19W

Makes a bit more noise though 

[Steve]

[dickson]

What size is the motor in hp?

1/2  horsepower  if  120vac motor  .         1 hp if  240vac  motor .  

[RobertM]

9 hours ago, dickson said:

What size is the motor in hp?

1/2  horsepower  if  120vac motor  .         1 hp if  240vac  motor .  

 If a 230v pump motor draws 9 amps when it's running, probably it's 3/4 to 1 horsepower? Starting amps are 56 amps. I have essentially the same inverter, and want to start a 230V 3-wire deep well pump. Mine is 1/2 hp.

Edited July 17, 2022 by RobertM

[dickson]

want to power a 1/2hp 240V 3-Wire pump. My initial attempts were failures. The inverter shut down within a few seconds.          a 230v pump motor draws 9 amps when it's running, probably it's 3/4 to 1 horsepower .

The  1/2 hp  pump   at 5 amps  is  1200 watts  and should start .      Try to start a  microwave oven  which is 3000 watts .        Try 4/0 cable  copper wires  .         The battery connection inside may be loose .       Try  turn on a hair dryer first  and  then  turn on the pump  second .       Your inverter look the same  but mine 8 kw  PJ is different .   

[RobertM]

The well pump is less than 15A to start, and it also has a start capacitor in the pump control box. My guess the starting surge is about 3000W. I will try again when it's cooler.  I recently learned how to start my 800W 115V refrigerator, and it was your hair dryer trick.

[dickson]

 I recently learned how to start my 800W 115V refrigerator, and it was your hair dryer trick.

My  8kw  PJ  with the rev 10.3 control board  can run a 5 hp air compressor  by  running it on the grid  for  2 minutes  to charge the capacitor         then  connect  directly th the 8 kw PJ   to run the 5 HP shop compressor  .    Running  2 minutes  on grid  is a trick to charge the  capacitor on the  compressor   and it also work to start  the 4 ton  heat pump  with my  15kw PJ  with the rev 10.3 control board  .       I  run the heat heat pump 3 minutes on grid and then  switch  to directly connect to the  15 kw  PJ  .   I found  that my   15kw  PJ  advertise on ebay auction  was actually  a  20 kw PJ  prototype  from Walnut  with the ASL9.0  transformer .       Your 8kw PJ  should easily  start any  1/2  hp  motor  buy never know what is actually  inside any  PJ case .      

[RobertM]

2 hours ago, dickson said:

 Your 8kw PJ  should easily  start any  1/2  hp  motor  buy never know what is actually  inside any  PJ case .      

It has 2 big caps and the ASL-2 transformer. I am slowly learning the tricks, so my next attempt might work.

[Steve]

My 10k upower is also 2 caps and asl2.

Running RU7088 FETs which failed fairly quickly

 

 

[RobertM]

30 minutes ago, Steve said:

My 10k upower is also 2 caps and asl2.

Running RU7088 FETs which failed fairly quickly

Did the FETs fail while you were starting the pump? The inverter didn't just safely shut down? Do you think a 50uH 20amp choke on the AC side would have helped?

 

 

Edited July 18, 2022 by RobertM

[Steve]

I suspect it was during pump start as that was my only load and running current doesn't get it hot.

I wasn't watching it

I wouldn't expect the choke to help.

My clamp meter showed a 56A inrush current when running it from utility power so the initial pulse is very heavy

[RobertM]

1 minute ago, Steve said:

I suspect it was during pump start as that was my only load and running current doesn't get it hot.

I wasn't watching it

I wouldn't expect the choke to help.

My clamp meter showed a 56A inrush current when running it from utility power so the initial pulse is very heavy

I have a 1/2hp pump meant to be protected by a 15A fuse. Wouldn't that mean the inrush current would be less than 15A?

[Steve]

Inrush can be well above fuse as it takes time to heat the fuse

[RobertM]

5 minutes ago, Steve said:

Inrush can be well above fuse as it takes time to heat the fuse

 I'll  continue to try to start and run the pump. I'll add a choke, and just replace the mosfet boards, if they blow. Live and learn. My initial problem was probably voltage sag, but I think I've solved that.