48V 5000W Oil cooled inverter build

After buying various inverter options I'm now opting for the monty python option of "and now for something completely different"

My plan now is to go non conventional, mixing commercial distribution transformer oil cooling with off the shelf parts and hopefully not create a glorified deep fat chip fryer.


Basic plan, 5kVA toroidal transformer, inverter board and about a gallon or so of whatever oil is knocking around at the time (less water).





Once tuned, hopefully with no load power use 40W or less and with the oil cooling I should be able to run the output at around 6-7kW for around 10 minutes or so providing the FET's are all ok. The following posts will be the progress so far....

---

First up and main issue the actual inverter board to use and the first selection criteria is it must be low frequency H-Bridge.




Not H-Bridge - might be good for 5000W at 48V but I can't charge my battery through it. Forgot to add the reason for H-Bridge is so that I can push power back from wind and solar grid-tie inverters back through the inverter to charge the battery..




H-Bridge and lovely Chinese writing to follow (must learn Chinese some day). BUT this unit only has 4 FET's per active leg, so the 6000W rating (special Chinese factory rating) of this unit at 48V would mean each FET is dealing with 31A (averaeg not peak switching amps) each.




Now this is starting to look better, 6 FET's per leg (12 x 2) and 5000W "rated" at 48V using RIFB4710 FET's, which are rated at 75A and have a gate resistance of 11mOhm. Not ideal, but a good start. Still, that 5000W at 48V spread across 6 FETs then drops to 17A per FET. Surge current per FET of 70% would then be 15.1kW. Hmmmm...




The layout of the PCB, however, may not be able to spread out the current effectively between the FET's due to the traces though.... without modification or additional "helper" wires...

---

Another board indicates that it can handle 7500W at 48V with 5 FET's and a surge capacity of 21000VA.... hmmm.. thats 21000/48 = 438A over 5 FET's = 88A per FET. Even the 7500W rating is 31A per FET.





The FET's indicated are either HY4008 or HY3810 and from the non specific indication of a board sold as 24V, 48V and 60V the HY4008 at 80V rating is not going to go into a 60V board.

The HY4008 is rated at 200A and would be curious to see the legs of the TO-247 cope with 200A for a long period of time.... the HY4008 does, however have a Rds of only 2.9mOhm so way lower losses.

The H3810, which I'm guessing/hoping would be in the 48V unit is rated 100V and 180A with Rds of only 5mOhm, brilliant..... but....




Using the same PCB layout as the board with 6FET's per leg AND the board in the previous post actually shows the under side PCB trace of a different board altogether with 5FET's per leg !!!!!!!

---

The powerstar W7 3000W units use 6 FET's per leg and I know that these units can take some abuse, my older unit can start my 3hp circular saw and cope with it stalling with my poor method of cuttingpallets up, which jams the blade on the odd occasion..





The 6000W version, which incoporates two transformers in the case rather than a single larger transformer, still uses 6 FET's and is a proven inverter for abuse but not so great on no load power consumption due to the E type transformers lossy design.





So, this makes me think that the 6 FET board is the way to go (can't find any larger boards yet....) and it should be able to cope with 5000W continuous without any problem.

As for larger boards....





This is rated at 10kW at 48V with the 24 FET's but they are not H-Bridge switched and will not allow power to flow back through the unit, which is one of may main criteria....




So, this appears to be my board of choice..... "rated" for 10000W at 48V (I'm designing for 5000W) and it has 6 FET's per leg so hopefully it will put up with my abuse.




The PCB trace also appears to be the same board this time round as well, showing the 6FET's per leg on one side and the 12 on the other. The traces look to have some better soldering additions as well.

Unfortunately the actual FET used is not listed so awaiting confirmation back as to the FET's to work out if this really is the 100% right board to order.

---

Next up is the transformer and this is the second critical item, ok there are really only two main big parts the board and the transformer....




From my time in the electricity industry many moons ago and playing around with a coil tyring to measure the magnetic field strength around buzzing transformers when I was even younger with no real idea what I was doing (not much changed) pushing the coil into the cooling loops on the metal case.




The basic options that I'm after are :
1. Better coolling which will provide the ability to run a 5000VA unit at 7000VA or more for longer periods of time.
2. Eliminaton of some buzzing noises due to wire vibrations.
3. Additional electrical insulation of the windings (not really needed or relevant just a nice to add in my mind)




Using an E type build trnsformer can work, however they tend to have higher magnetic losses and the grade of steel used in cheaper transformers adds to the losses, so forces a much more expensive transformer.




Toroidal transformers have lower losses but are not so easy to source for a specific voltage level, needed for the output of the inverter board.

They are normally are supplied wrapped and sometimes have the center hole filled with compound to help fix the toroid into a case, however I will be looking to unwrap the cover to allow the oil to circulate better, athough may have to compromise with some string binding in places if there is none underneath at the ends of the coils.




Which should then leave me with something like this....

I've often wondered about these boards. Then build the transformer or buy it locally. This would safe boat loads of the cost of high output inverter. But I have been leery as I don't know enough to go that route.
Glad someone is

I will be watching the progression closely.

This is one of the differencs in scale I'm hoping to get as a reduction in less wasted energy.




Lower no load power consumption




And higher efficiency when the cooker, TV, microwave, PC, Iron, washing machine, dehumidifier, fridge, lights, playstation are all overloading the inverter, bleeping away in the background getting ready to double up as a deep fat fryer..

Theaim is to have an inverter that is more efficient overall and completely non compliant with any grid code around the world, while being built in line with the standards that can be applied.... or not. lol.

Final goal is to take the house off grid 9 months of the year or more and only charge for 1 hour per day in winter (at 6kW) at most.

---

There is another thread on a forum where they take out the toroids from solar grid tie inverters, unwind the secondary and replace it with about 16 turns of very think wire like 2 or 3AWG...

Enough posting for now, have to actually get something done now, lol

So with the transformer type, E-core type are more forgiving to loads with some DC offset vs a torroidal one.
You can easily have loads like a typical hairdryer that half-wave rectify the AC when on half heat, etc. This means the AC wave gets current drawn on one half-cycle not equally on each + & - wave peak. Many other loads can have some DC offset too.
This will make the transformer saturate (becomes nearly a short circuit)

It's the losses that let the big power grid transformers soak up a few hair dryers without going into saturation.
Torroidal cores with less losses don't cope so well with the DC and get into saturation.
I built a few types many years back & saw all this in action.

To properly drive loads like this you really need a design with two transformers & FET systems one drives each half cycle separately.
My Victron inverter has two transformers.

Also when looking at the above FET drive boards, hake sure they're proper sine-wave drive not "modified square wave" ones

Oil cooling is not necessary for such a low power inverter - much debate on the use of ready made controller boards started elsewherea little over 10 years ago, eventually resultinginthe production of a viable open source inverter.

Most current thread -https://www.fieldlines.com/index.php?topic=148953.0







https://levivray.com/new-book--make-a-6kw-inverter.php

The control board that is used (EGS002) has transformer output feedback loop to alter the switching through the full cycle so that any offset does not occur and is true full sine, with carrier frequency around 23kHz.

The larger the transformer the larger the energy stored in the magnetic flux and this is what allows the larger grid transformers to cope better with high harmonic inducing loads switching part way through the waveform. This is also why for a larger inverter you can't just power the transformer part way through the waveform as the inrush (to charge the magnetic flux) would blow the FET's. The control board in this instance has a 3 second slow start approach. Zero crossing would also work in theory....

Are you sure the transformers in the victron are not in parallel or setup to handle split phase output ? Its more typical to install two smaller toroids/transformers so that they fit into a smaller casing with lower fixing load.....

---

For those interested in reading the original 139 pages from 2012 of the forum thread I found really interesting and what originated an open source inverter desing (using the same control boards in the pre-built board Iintend to use) you can read it here:

https://forums.energymatters.com.au/solar-wind-gear/topic3344.html

It's long and well worth a read to understand the behaviour of the inverter transformer and how to avoid saturation and decrease the no load power consumption.

---

Bit more research, learning chinese (or not) via google translate via the camera on a phone...




The 4 variable resistors on the board are for :

1. Output voltage adjustment - this has an impact on saturation if your output is being driven too high as the input voltage waveform will be pushing the primary into saturaction. Voltage sag on the secondary may also cause the primary into saturation if the primary to secondary winding ratio is not selected well for your battery voltage range. The inverter needs to be able to cope with your minimum and maximum expected battery voltage as an input and it is the minimum voltage that can create the real constraint on a true sine output at load.

2. Under voltage - battery protection shutdown level, basic and easy to adjust (in theory!)

3. Over voltage - not needed if your charge controller is dealing with things.... in my case ths pot will be set to a above max battery voltage level because the inverter can't switch off if the battery is being charged through the inverter..... automatic dump loads will be a last resort to prevent over voltage rather than shutting down the inverter as this would also switch the house off !

4. Overload setting - an interesting one to tune and set as I'm not entirely sure as to what the exact details are relative to any "level". Immediate shutdown, alarm level, 3x alarm level for x seconds ???? One for more research.... or more likely experimentation once built.

Very interesting threads there
Re my Victron unit I just squinted through the side, but it sure looks like there's two whole transformers in there.
Interested to see how the single transformer versions do handle a hairdryer - must be a trick to stopping the core saturating!

The cooling of a transformer as a result of the No-Load and Load losses is typically done with fans in your normal off the shelf inverter purchase or if your unlucky no fans and a shutdown sensor and if your really unlicky a shutdown when the magic smoke or fireworks appear.

Frome some peoples quantative experiments on the effectiveness of different cooling materials to add to my childhood experience of fortunately not burning the house down with a transformer, in a plastic tub, filled with used engine oil being short circuited for long periods of time.
https://hackaday.com/2018/11/15/measuring-the-cooling-effect-of-transformer-oil/

This gives some indication towards how much "overloading" above the air cooled rating of a transformer can be achieved for short durations of time without damage.




The main thing to to remember at this point is that the load losses and resulting heat are all related to the current and as long as the voltage stays low enough so that the core is not saturated. This means that your loading will increase the wire losses as a square of the current but have very very little impact on the actual core and you can wind the current up as high as you can "cope" with.

There is a good overview of oil cooling in a presentation "Properties behind effective Transformer Oil Cooling" given at a Transformer Life Management Conference in 2013....
http://www.brown.edu/Departments/Engineering/Courses/ENGN1931F/TransformerCoolingOil.pdf





The chart shows how much of a difference some oils can make, which are a lot thinner and therefore can flow and transfer the heat away more effectively.

Faced with the choice of cooling medium the ideal option is actually de-ionised water from a cooling effifiency perspective but due to the items being nice materials that would create some interesting electrolytic reactions the water would end up a highly toxic and conductive medium after a short period of time, besides making a big mess of everything. Which means for me it has to be some form of mineral oil, readily available and ideally clear.... which leads me to this...

That's what I was thinking, mineral oil. And there's several videos where ppl have made transformers, used a vacuum chamber to pull all the air out while submerged in the oil. Works pretty good.

Hahah, that oil cooled computer, 4 minutes of watching the system run in the fish tank
Didn't think about baby oil. That's actually cheaper than just straight mineral oil, for some odd reason. Altho, buying the MO in 5 gallon or larger quantities might drop the price drastically.

Yeah, he needed some floating plastic fish in the tank !

Still trying to work out the effectiveness of the seal / construction of electrolytic capacitors as the heatsink end of the inverter board may end up in the tank as well so that it eliminates the need for a fan there and helps with the cooling of the board. Capillary creep of the oil up the board to the variable resistors may make this a no go though...




Confirmation back that the FET's are HY3810 so all good, order going in shortly....




Next up, quick scan for Baby Oil and there is a deal on at Tesco (2 x 500ml for 3) so 3 per litre, which is cheaper than engine oil (but not as cheap as used engine oil from my very first experiment) ! So will be placing an order once I have a bit more of the volumes worked out which brings me to the actual toroidal trnasformer selection... The wife will probably give me a slap when the oil turns up for thinking that she is on massage duty for the next 5 years. lol. just waiting for the "WTF is all this ?"

---

Basic thought at the start of all this was why pay for a 10kW inverter when your only going to be using 10kW for a few minutes at a time or that some of that 10kW can be offloaded to a separate 2kW grie tied battery inverters switched on only when needed (to avoid no-load losses and extend the lifetime). This may sound a bit wrong at the moment, however a little diagram may help clear up what I'm trying to do....




Remember this is a "and now for something completely different" monty python build....

Overall if I have two of the 2kW units (capable of 1900W) that can add 3.8kW to the 5kW of "base" I should then run my load upto about 12kW for a few minutes, if not longer, at a time while the inverter alarm bleeps away and the oil slowly warms up. The main limmiting factor is then only cooling if I keep the peak switching load below 15kW.

The two solar and one wind untis are the ones I have available at the moment and have so far tested a setup like the one above with just one solar inverter and the wind turbine grid tied inverter for a few weeks without any issue. Chargers are a work in progress after decoding the serial formatting for the Eltek units in a separate post.

This is all being posted now, but I have been doing research for several months so far and just getting to the actual nut's and bolts physical build stage.....

Now that 20 bottle of baby oil are on the way (10 litres)... lol.... I need a transformer...




This appears to be the "standard" off the shelf product from one suppliers, however they appear to be very willing to make transformers to request at a reasonable price.

Modifications needed (or lack of build steps for the manufacturer) are :
1. No core filling. This is nice in a normal install as the core then provides a nice bolt hole to fix the unit to the chasis, however my unit will be "free floating" in a sense... and will be bound to a fixing plate/mesh with cord.
2. No outer wrap. The final wrapping mormally provides a protective layer to the outer windings from scratches that remove insulation and provides an additional layer of insulation. In my install the oil will provide the extra insulation and as long as it is not damaged in transit or by me dropping it on the floor it should be all good.

Fixing the transofmer to something suitable will then be done with some paracord with thin layers/strips of padding to protect the windings from excessive pressure. This will leave theouter windings fully open to the oil and the inner windings with less distance for heat to transfer into the oil.

You could see if they'd be willing to wrap it in masking tape instead of the plastic, and then only on the outside of the 'donut'. Would make it easier to remove once you got it and was ready to drop it in the oil and would keep it protected from your Linus (LTT) moments

How much are the toriods?

completelycharged said:

Next up, quick scan for Baby Oil and there is a deal on at Tesco (2 x 500ml for 3) so 3 per litre, which is cheaper than engine oil (but not as cheap as used engine oil from my very first experiment) ! So will be placing an order once I have a bit more of the volumes worked out which brings me to the actual toroidal trnasformer selection... The wife will probably give me a slap when the oil turns up for thinking that she is on massage duty for the next 5 years. lol. just waiting for the "WTF is all this ?"

Click to expand...

LOL I'd love to see the video reaction of this moment
Then again, she might get a twinkle in her eye thinking you guys are going to start acting like teenagers again

LOL, that made me laugh !

Critical aspect for the transformer is turns ratio....

With the previous 139 page post the early conclusion was 8 to 1 to get a primary to secondary ratio (and later on slightly different). This is also where the FET selection Amp ratings come back in, as the calculations as to the current handling are not really based on just 48V if considering the average current delivered at the effective PWM sinewave voltage....




23kHz PWM switching into the transformer creates the sine wave and a feedback loop from the output also alters the timing so as to cope with the not so nice loads and with a reverse power flow of energy being pushed back in at the 230V side of the transformer back through the FET's to the battery.

The chart shows a simplified "slow" PWM just to give the idea (nice chart lifted from electric bike forum post by non other than.... spinningmagnets). The chart also shows where the FET's in a H-Bridge arrangement switch the phase aroud on the input to the transformer.

The "headroom" between the DC voltage and the peak of the sinewave is where your low battery voltage needs to be considered and all the other voltage drops along the way.

The sale page on aliexpress indicates a 220V to (24-28V) transformer selection... another inverter board quotes 26V and another quotes 24-32V, however a fixed toroidal transformer is not variable. So key criteria is then what voltage range do I want to handle at minimum charge ?

For my battery back my minimum charge will be 45V / 22 series cells = 2.05V per cell as they are LTO chemistry and below 2.05V you don't get much when you fall off a cliff and upper charge is 56V, which gives me something like this....




The black line represents the combination of volt drop caused by the current on each FET, the inline chokes and drop to the busbar on my system. This says that a 30V secondary is the highest voltage I can use, but this again does not always leave much room for high crest factor loads and a bit of de-rating still to cope with so 28V is looking like the choice of secondary.




My design consideration will also take into acount a load derating when the battery pack is below 48V as I will want to avoid continuously pulling 8kW when the battery is not above 48V. This still allows for a very comfortable draw of 5000W down to around 45.5V or about 10% full without distortion.

---

Efficiency means loss, so if I want to see 8000W on the output terminals I will have to put more than 8000W in so I re-run the sheet with 9000W as an inpout and now get a rough indication as to what the losses will be as a starting point. The actual losses should be more than this because the switching currents will be higher and the core losses of the transformer are also not taken into acount in the chart.




Still, as a starter the low load efficiency should be ok and with 8kW out 1kW of heat will be generated in various places with 800W at the heatsinks of the inverter board when the battery is at 49V before de-rating, so the inverter board may yet end up in the oil with the transformer after replacing the post with fixed resistors once the levels are set.

There is still yet the option of two smaller toroids as per many manufacturers (like Victron) using two smaller units rather than a single larger unit.

For reference a standard 2000VA Toroidal in the UK with 230V Primary and 30V secondary is listed as 132 each from Airlink Transformers. They each weigh in at 12Kg.... the 5000VA units are around 23Kg and a big difference in fixing load in a case. One transformer gives less no-load losses, which is the plan..

Looking further into the ratings and heat dissipation of the FET's in the various data sheets the critical chart can be the drain current de-rating due to "package limitations" or put another way watching the legs melt off.

The HY3810P is "rated" for 180A (5mOhm Rds) but the "package limitations" mean only 75A in the real world, there is another near identical board made with HY4008W FET's that are "rated" for 200A (2.9mOhm Rds) and again the "package limitation" is 90A, but the main downside is the voltage rating with the HY4008W which is 20V lower at 80V compared to 100V on the HY3810. Running the 80V FET's at 56V on a full charge does not leave much room for back EMF spikes on shutdown, so would possibly look to add surge suppression between the board and transformer.

This is the HY4008W FET based inverter board of choice so far
The Rds of 2.9mOhm vs 5mOhm also drops the losses by 42% to 470W on the FET's at 9000W input (30A per FET) and improves overall efficiency by 3%.

At 3000W I'm then hoping for over 95% discharge efficiency to grid (LTO full cycle efficiency at 5A is around 98.5%) on 3.5A per cell.




Modifications to make it "oil" proof.... so it can then join the transformer.

I'm curious about the losses in the FETs you're calculating?
The transformer has a single low voltage winding not a centre tapped winding right?
The board has an H bridge design with 6x FETs on each leg of the H right?
So at conduction time, you'd have 6 FETs in parallel to ground on the "low side H bridge leg" & on the other "high side H bridge leg" you'd have another 6 in parallel.
So a total of 12 FETs would see the average current for each half cycle.
All values approx.
Working back from a nominal 8000W output on the AC side @230VAC, I get:
230VAC side current = 34.78A RMS @230VAC.
Assuming turns ratio of 8:1, a peak DC side voltage of 40.7 & RMS current of 278.3 A
So 278.3 divided across 6 FETs per leg = 46.4A/FET
At 5mOhms Rds, that's a power dissipation P = I(squared)*Rds = 46.4 x 46.4 x 0.005 = 10.75W per FET
At any one time, 12 FETs carry this current & so dissipation is 12 x 10.75 = 129W
This is for conduction losses & doesn't include switching transition losses.
Efficiency losses also not included for simplicity.
129W is a lot less?
Did I get something significantly wrong?

You are indeed correct, the lack of sleep + using watts as voltags in a column = way more than it should be... will re-work the sheet after a large cup of coffee. lol.


Found out the transformers in the Victron are in parallel but one is apparently switched in as the load increases, also part of the toroid is left exposed to increase leakage to provide some additional inductance that then eliminates the need for an second in line inductor. learnt something new on that one...

Delivery no 1 complete.... but only 8 litres as I guess they ran out of stock for delivery. Think I may need some more given the case size and a circuation allowance for cooling when EV charging for 3-4hrs at 6kW.




I was going to carry out some more detailed flamability tests, although I guess this is about as far as I need to know... lol.

Summing up so far, this is the effective circuit, components and wiring




The 50uH inductor core will then need 3-6 turns or with two smaller cores, one per secondary leg, which may be more practical.