Energy Matters Forums

[Andrew_electrix]

Hi All,
Been meaning to post some info on modifying powerstar W7 inverter for quite a while now...

To start at the begining i ordered one of thes units off EBAY for an off-grid system that i installed last year in West Wyalong - the model in question is the largest available 48V 6000W continuous 18000W surge

When it was delivered at my work and i took it home and unboxed it to find that the there had been some damage in transit - the unit had been dropped on it's end and the case had sheared off the
battery type selector switch and LED's.

transit damage

I took some photo's of the damage, then took the lid off and resoldered the LED's and BCD (binary coded decimal) battery selector switch.
As it turned out the chassis was actually bent but electrical the unit seemed OK so i decided to power it up and test it out!
Hooked it up to my 48V 600Ah battery, Initial results looked good powered up OK with 235VAC out, connected up a varity of load - 150W flood light, no problems. Although the 120mm fan at the end with the transformer was wired to run continueously which was a little annoying (the reason for this would become apparent later)

Next i though i should check the no load power of the inverter, this is where things went pear shaped in a big way, the next photo shows it all:

Inverter no load current

52V x 4.32A = 225W !!!!
Assuming the inverter is on contineously 24h a day this equates to 5.4kWh of energy wasted per day!!!
The 1.5 kW solar array that we had installed would not keep up with this blood sucker, on an average basis at least, the battery would slowly go flat.

A big Thanks to TIM from machinerydirect in Sydney (ebay seller) who was very helpful and accomodating when i explained the damage in transit and problems.

Becasue there was no rush to send it back i did a few more test/checks:

When connected to grid power the unit can work as battery charger, putting put approx 70A @ 48V, so i tried this out by plugging it into the wall via one of those cheapo $25 energy meters.
When switched on the reading on the LCD showed 3100W, which given it's max rating of 2400W was cause for some concern, however it managed to handle this without any drama which was a pleasent suprise

The battery was now fizzing along nicely and making it's way up to he absorb voltage of 56.7V, after a little while the charger had decided to move from the bulk stage to absorb and the input power levels started to fall away to more sensible levels as the charger held the voltage constant (indicated by the charge LED flashing)

All Good, except the 2x transformers were now very warm, after another 1/2 an hour of charging they were getting really very hot - so that's why the 120mm AC was set to run contineously?!
All Very interesting...

Original transformers

Going to spread this out over a couple of posts cause i have more pic's to add ect

[Andrew_electrix]

Next i got the CRO onto the output to see what was really going on...
This photo shows the output waveform of the inverter under load with a 1000W lamp

Original transformers waveform

Whilest the waveform looked fine with no load, with only moderate load the clipped sinewave clearly indicates that the transformer were being driven into saturation to a lesser (or greater) degree.
I did try it at higher loads and it gets even more ugly from there...
The ringing in the trace as the waveform clips also shows a fair amount of leakage inductance as it hits the wall.
This would explain why the unit was wasting so much power and getting so hot, both when working as an inverter and as a charger...

Beacuse this is a 48V inverter the primary of the transfomer is driven with 48/1.41 (sqrt2) or approx 34V RMS which is multiplied by the turns ratio of the transformer to 240V RMS.
Actually this value is less when the battery voltage is higher than 48V as the inverter does closed loop control to regulate the output voltage as the load varies.(more on that later)
'After a few more carefull measurements of the existing transformers and some calculations
So decided to subsitute a smaller 160VA 30V -> 240V toroidal transformer to see what would happen.
Powering up was a nervous moment, however there weren't any problems, switched on a 100W light, no worries.

Checked the no-load current:

160VA toroidal transformer no load

52V X 0.37A = 20W !!!
A TEN FOLD reduction in the no load current, now that's more like it maybe it might have some legs after all??

Here's the voltage on the primary of the transformer

160VA transformer primary voltage

At this point i hadn't made any non reversable changes so i still had the choice to:

1. Return the unit for a credit - put the $800 towards an expensive ($3k) brand name inverter (Latronics, selectronics ect)

2. Attempt some modifications to the unit to improve the performance and reduce the no load current.

After playing around with the subsituted toriodal transformer for a while, this gave me confidence that option 2 might actually be a good one.

The ebay seller Tim gave me $100 credit for the damage in transit so wasn't complaining about that at all.

[Tracker]

.
The very high "Stand-By" consumption is why they have the other mode of Power-Saver, where it pilses the load , looking for demand.
http://forums.energymatters.com.au/energy-efficiency/topic2950.html
You might also follow the above, whic is MY experience with the W7.

Thanks for your observations.. VERY interesting indeed..
I admit to NOT having analysed how it worked, by assumed that one transformer was for the charger and the second for the inverter.

Torroidals are always (I think) more efficient, but can you get the same power from the largest one that you can get into the box.. . As per my thread, I now have the W7 permanently setup supplying refrigerators and an RAC.
I have only just set it up and got my first days data. Not great, but encouraging..!

Something that seems very good with the W7, is that it can be powered up and draws very little power in the OFF position.
The function switch (Normal/PowerSaver) seems to happily control the operation, and via Time-Switch, ensure that minimal power is lost.. Remember that I am attempting to run on Batteries during PEAK and then CHARGE during the OFF-Peak.
..
.

[Tracker]

.
and... if anyone comes up with a circuit for the W7, please do share it ..

so that's why the 120mm AC was set to run contineously?!

Would not be very difficult to add some button-thermostats to critical parts, and only switch the fan, when heat-demands.. I think that I will look at that one myself.. the constant noise is a nuisance..
..
.

[Tracker]

.
BTW - - In case it's not noticed, there is an RJ25 phone type socket near the battery connections.
This is for a REMOTE control head..
The removable switch plate can plug in there , or, you can just hard wire a switch for function control..
I simply switch the Blue/Yellow (n/o) wires (6Wire Phone) via a relay. 3 wires for switch and three LED's...

If anyone needs the full connections, just ask.. (Save you the trouble of tracing the circuit)
..
.

[Andrew_electrix]

Thanks for the comments guys, i thought this might be of interest - i'm just getting to the good bit

Tracker: The original transformers are used in both inverter and charger mode. The low voltage primaries and wired in series and the high voltage secondaries are wired in parallel. I think because the current is doubled on the HV side for the same input current on the LV side, which helps reduce the losses (not that it does in this case)
These high power inverters are of the "full-bridge" converter topology where the H-bridge of MOSFET's switch the DC across the transformer synthsisng a sine-wave by pulse width modulation in inverter mode and work as a controlled rectifier in charger mode, controlling the output current by varing the pulse width.
In other words power flows both ways depending on what's required!!

This leads nicely on the next stage in the story...

After the success with the small 160VA toroid transformer, the ovbious thought was to scale things up with a really big toroidal transformer.
Also obvious was that it would cost something so i wanted to make sure it would work properly,
As mentioned previously the inverter does closed loop control to maintain the output voltage as the load varies or the DC input voltage varies with the battery state of charge.
It does this by varing the pulse width of the switching waveform to increase or decrease the RMS output voltage across the transfomer, however there is only so much headroom to this before the waveform would become clipped or grossly distorted.
To explore how this regulation mechanism worked i tapped into my 48v battery bank a couple of cells down to simulate a low DC input voltage (eg 44V) and do the opposite by adding a extra 12v battery to simulate a high input voltage senario.
Basically this showed that for the control loop to work properly the inverter needed a transformer with a high turns ratio than expected such that the inverter could maintain the 235V AC out with heavy load and low DC input voltage.
The expected turns ratio by calculation would be 240/34 = 7.05
However what i ended up going for was 240/30 = 8 which seemed to provide adequate margin for regulation given the worst case conditions.
After experimenting for a while i was confident that it was going to work OK with a transformer as specified, so i sent an email to HARBUCH transformers in Sydney explaining what i needed

http://www.harbuch.com.au/

The original 6kW rating for the inverter was highly questionable in any case, but i did want to replace the transformer with something fairly sizeable so i thought the 3kVA would be a good target to aim for which would result in approximately 100A flowing in the LV secondary for approx 12.5A output

Peter Terlich gave me a call back and after dicussing what i was trying to do, he agreed that the original transformer were probably garbage.
Then he told me the cost which was going to be about $900
It this point i did have thoughts about sending the whole thing back and questioning weather i was throwing good money after bad after spending about $700 on the inverter by it's self.
Confident that i'd done my homework i placed the order and tried not to think about expensive paperweights...

The HARBUCH order code was: HR2417FAP

[Tracker]

.
You had set my mind thinking (even if it don't work like it used to.. )
I too lifted the lid and noticed the // operation of the transformers.
I had some "Physically Right" toroidals with two 12V windings in //, and wondered about rewiring to 12+12 and then vary the turns ratio and then gang them in //..

Problem is I am unsure just what the toroids are capable of delivering pre-saturation.

I will be keen to read just how you go... You are clearly more up on UPS design than I am..

Don't forget that "Power Save" function.. It does help, but does cause issues if the appliance has computer control, as the drain is not sufficient to HOLD the inverter on, and thus incapable of ever "Starting"..
..
.

[Andrew_electrix]

Next installment:

You correct Tracker - the AUTOSTART feature on this inverter is not very useful at all, becauase it requires at least a 60W load to cause the inverter to actually start and stay on.
However - by the end of this little blog you'll see it's not really neccessary to use it anyway!!

So the 3kVA transformer arrived actually while i was away from home doing the rest of the off grid install at west wyalong.
What a beauty it was!! I can't say i've ever seen such a large toriodal transformer !!
The primary windings are actually 2x separate 30V 50A windings which can be connected in parallel to double the current rating, the secondary is a single 240V output.

3KVA transformer

So anyway i set about carefully connecting it up to the MOSFET banks (heatsinks).
To increase that chances of the NOT blowing anything up i had 4x 7Ah small SLA batteries connected
by multimeter set on the 10A current range and a couple of very thin strand of wire to serve as a fuse

These precautions were justified as i flicked the switch to "ON" and the inverter soft started, the transformer growled, the multimeter beeped as it showed me over 10A flowing from the battery and i saw the strands of wire move as they heated up to a dull red colour

Quickly switching off to preventing disaster i sat back in the chair as the dryness in the back of my throat subsided.
Playing it safe i reconnected the small 160VA toroidal transformer and powered up to make sure very thing was still OK - and it was.

Then i reconected the 3kVA toroidal transformer but with only one of the 30V windings in use.
This time the result was a little less severe with only 8A flowing, this was clue as to the nature of the problem...

After a fair bit of head scratching, meaurement with the CRO and experimentation the picture started to become clearer as to what was going on:
Transformers can be modeled by the PI equivalent circuit:

http://en.wikipedia.org/wiki/File:Transformer_equivalent_circuit.svg

For a toroidal transformer the magnetising inductance (Xm) and core loss resistance (Rc) are very large because the magnetic circuit is so tightly made and couples the windings very well, which means the no load current and losses are small.
The tight magnetic coupling also means that the leakage inductance on both the primary (Xp) and secondary side (Xs) are very small - at full load where less of the applied voltage is stolen by these parastic components the result is less losses and higher efficiency.
Transformers that are less than ideal compromise these characteristics.

And of this should help improve the inverter's efficiency and reduce the wasted power, but it didn't, so why not??

Directly across the output of the transformer secondary exsists a 4.7uF polyester capacitor.

4.7uF polyester capacitor

As it turns out the original transformers are made deliberately lossy with a disproportionately large amount of leakage inductance, as this forms a filter with the 4.7uF capacitor on the output of the transformer to help filter, block and smooth the switching waveform generated by the inverter into a nice clean sinewave.

Refered to the primary side by multiplying by the (turns ratio)^2 this capacitor appears as approximately 300uF. And due to the ideal nature of the toroidal transfomer with very little leakage inductance, this quite large apparent capacitance was virtually short circuiting the MOSFETS as they switched at each transition from on to off.
Hence the reason for the high current draw.

Removing this capacitor and powering up saw the no load DC input current drop to about 0.25A, which was great except now the PWM switching transitions and hash from the MOSFET's appeared at the output

Problem indentified - but not not solved

Obviously some low pass fitering is required some distill the fundemental sine wave from the square switching waveform
http://en.wikipedia.org/wiki/Fourier_series
Absolutely beautiful mathematics discovered over 200 years ago, with totally practical applications that make all inverters - stand alone and grid tie possible.

Some considerable trial and error and measurements resulted in the following changes:

1. The 4.7uF capacitor on the secondary side was reduced to 0.47uF - This Capacitor MUST be an X2 class self healing MPP (metalised polyproplene) capacitor ie one rated for contineous safe opperation at mains voltages

0.47uF capacitor

2. Some inductance must be added to the primary side of the transformer to help block and filter the switching waveform.

3. A final LC filter was added on the secondary high voltage side to wipe off any traces of the switching noise to make the sinewave as pure as possible.

[Tracker]

.
Your observations conjure up vague memories of long-long-ago (40 years ago), that have not ever been needed and hence forgotten...

Very interesting indeed, but the classic $64K question remains for the next installment...

What real difference did it make..?

Looking forward to the findings.. and - boy, they got that Tfmr made quickly..

PS - I assume that the inductors are the first attempt with filtering.. You got them made up quickly too..
Sounds like your in that industry.. Lucky You..
..
.

[Tracker]

Removing this capacitor and powering up saw the no load DC input current drop to about 0.25A, .

One wonders what difference would be made, to the original device current, if that Cap was reduced, and perhaps an external L/C filter used to clean up the act..
..
.