Thread Rating:
  • 2 Vote(s) - 5 Average
  • 1
  • 2
  • 3
  • 4
  • 5
48V 5000W Oil cooled inverter build
Proof of the pudding as they say is in an actual test (with all wires visible and no convenient off camera loops).

Sometimes I do wonder if these types of tests are great to show that it can really work in the middle of your kitchen or a near complet lack of accurate instrument readings to actually prove anything other than it appears to work. but they do seem to show something common.

Some of the boards used are lacking the jumper wire when advertised as they will work with EI transformers that are not as brutal and as stiff as toroidal units.

When shown under test with a large toroidal unit they all (so far) have the jumper modification wire attached.
Korishan likes this post
Bit more of an update on the transformer as I managed to work some figures out in relation to what the overall efficiency may end up looking like.

Because the inverter is going to be used to charge the battery pack it will spend quite a few hours ar load levels upto 6kW and over (including cloud edge events) so the sizing was also to take this into account. If it was just to be an inverter for supplying loads in the house then a smaller transformer would work quite well and increase the efficiency a little at the lower average load level.

The sizing is also to allow for future expansion as with oil cooling it should run quite well at around 8kW with only a further 1.9% drop in efficiency, so EV charging at 6kW while the kettle is on can work.

For the 6kVA unit (now indicated by the manufacturer at just under 31kg) the efficiency will hopefully look something like this

The blue curve is just the wire losses (winding resistance) and the red curve is the winding wire losses plus the idle power loss

Higher loads efficiency drops to around 88% at 12kW so at this point 1.4kW of heat is being generated by the transformer...
Korishan and mike like this post
If you can't quantify how much they cost, it's a deal, I'll buy 5 of them for 3 lumps of rocking horse ......
One thing that I should point out this satge is the cost of the transformer and also to put it into context as to what it really means..

The cost of the transformer on it's own is around £350. For around 26kg of steel, 4kg of copper and just over 1kg of insulation...

The reason for the two lines on the chart in the previous post is to highlight the impact that idle lossses can have on the precieved overall transformer efficiency. This is however a perception and the red line is not necessarily the way to view your efficiency when your switching loads on and off.

WIth an inverter you either run it 24x7 or swtch it on when you need it. The times that your inverter is switched on (24x7 or not) you are paying a fixed fee (idle losses) per hour of that availability and you should view it that way. It's a flat opportunity cost to use power from the batteries. The core losses are fixed, "almost" regardless of load.

The way to then view your efficiency is more towards the blue curve, because losses are mainly current derived due to I x I x R and you should not really look at always trying to run over a given wattage to "gain" efficiency. Separate the opportunity cost (inverter switched on) from the load cost passed through it.

The idle (no-load) losses are threrfore a lot more of a critical aspect to an inverter and a lot of people frequently overlook the significance of selecting an inverter with a low no-load idle power use. The high idle losses can unfortunately create a situation that running a system 24x7 has a very high energy cost and then ends up being switched on and off as a result. Your inverter selection should have a significant value (for a solar battery setup) placed on lower idle loss inverters if you want to run them 24x7, it is quite likely worth the investment... value it..

The transformer that I will hopefully pickup soon was altered a little bit from the standard specification with the aim of reducing the losses further by adding in an incremental cost now that I know will be saved in reduced energy losses over time.

For the UK the electricity price is around 16p/kWh and 1W over a year (8.76kWh) is worth £1.40. Over 15 years your then looking at £21 of energy "value" for each an every 1W that you are loosing in efficiency. The modification made reduced losses by around 4W or over £80...

For an inverter with just 100W losses can add up to £2,100 over 15 years of 24x7 running, which is close to, if not more than, the price of the inverter in the first place, something not alway sconsidered when buying...

Should inverters be valued on the cost per kW over the lifetime of the inverter. Say a 10yr life 3000W inverter with 90W no-load losses (0.090kW x 8760h per year x 10 years x £0.16 per kWh) the lost energy value is £1,261. The on a per kW basis this is then (£1,261 / 3) a lost value equivalent of £420 per kW of capacity.

90W is the no-load loss of my 3kW PowerStar unit, which I can't leave running 24x7...... not enough solar in winter to start !

The majority of the losses are in the transformer and directly attached components. Plus, the typical inverter will not last 15 years ! This is another benefit in this build method because you will only replace some parts and not the critical "costly" transformer, the transformer is more of an investment for life and should still be running even when you are not.

The losses in the transformer to be used in this build (will be tested as proof) are indicated under 20W and closer to 16W. With the rest of the parts going into the inverter I am hoping to have the overall idle loss below 30W (0.72kWh per day or 11.5p/day @ 16p/kWh) and have it running 24x7 without thinking about switching it off.
rottenkiwi and Korishan like this post
If you can't quantify how much they cost, it's a deal, I'll buy 5 of them for 3 lumps of rocking horse ......
Board arrived....... lo, behold, no instruction paper, lol.

Thought I would do some large images so that you can see all the details.... and problems to fix....

Notice the additional connecting wire, added after the original design of the PCB had been done, hot fix......
The blue capacitor sat on top is a 5uF 400V rated cap X2 for 275V operation..... but it's the wrong value to use for my transformer.
This is also one of the common issues and reasons for failure of these boards in DIY builds, the requirements of the parts change depending on the transformer used.

Nice row of FET's, 2 x 6 switched H-Bridge.... but......

The soldering does need some small additions... the gates are all fine as they all have good solder on the reverse side of the board, the drain and source however need to be checked and added to on some FET's.... These are the small details, which can result in one FET overheating and then causing a subsequent failure cascade. Simple quality control issue.

Heatsink and capacitors, of which I may need to add to or replace them if they do not provide enough smoothing at higher load levels (above 6kW).

Not entirely sure that the connections across the top are ideal due to the very limited thread available on the screws. Might consider bolting a thicker bar to the front of the heatsinks (10mm thick) and then connecting to that as it will provide a firm connection point and add to the heatsink mass and cooling surface area. Again minimal modification, potentially a large impact on the long term degredation of the FET's due to insufficient cooling or additional heating from poor connection points.

This board is the 2nd major cost item in the build and something that will need part replacements over time (or full replacement) if you want it to last 20 years. The capacitors will fail first under normal use, so being able to swap them out easily is a win. I'm looking to replace the board in around 5 years with a 14-16 FET (or higher rated FET) board to allow for a higher load level and potentially the addition of a second transformer in parallel.... 12kW, 30kW surge. That would then be a true 100% house supply replacement kit, on-grid or off-grid.

Last thing I forgot to point out is the 12V transformer as it's rated 220V and I think it is used also for the voltage regulation feedback, whcih if it is running 235V the transformer is already into a fair bit of saturation and then if it is backfed and the supply voltage raises further to say 250V then the transformer may create some other issues. Another part, whcih may require a replacement in order to eliminate another potential cause for problems.... unless I run the output at a lower 225V (which would also reduce transformer losses a little more) and I2R losses overall.

Thought an additional photo would be usefull... with the flash for the camera at the back of the PCB so that the traces show up a lot more Big Grin

Looking at the incomming voltage resultation it is a separate path and not via the transformer as this is just for the onboard electronics and fan supply....

And reverse side.

These will be useful for any part replacement (caps) and understanding the board.
Korishan and Redpacket like this post
If you can't quantify how much they cost, it's a deal, I'll buy 5 of them for 3 lumps of rocking horse ......
With any inverter battery voltage is key, however I bet that not many people with powerwalls actually know how many Wh they have available for a given voltage level or how many Wh do they get per volt drop on the pack, from full to near empty...

Left axis is Wh per Volt drop, bottom axis is battery pack voltage.
This was a test I did last year on part of the pack where I switched off the solar input and then had some electric heaters on for the day, periodically noting down a few variables (yes lack of automated logging....), otherwise the curve would have been a lot lot smoother...

What it shows is that the voltage range on an inverter for my pack needed to support down to 46V as a starting point in my case or I would loose access to a reasonable chunk of energy. The range also needs to take into account any additional incremental voltage loss that your pack will show under higher surge loads.

Why bother covering surge loads properly ? Under surge conditions if your voltage falls away the sine wave on your inverter could go from a nice smooth sine wave to more of a square wave as the input side has insufficient voltage and the PWM duty cycle ends up 100% for the central part of the sinewave, resulting in a flat top and bottom appearing to the "sinewave" output like this.....

Plus, potentially under a distored sine situation some devices may have issue, which were otherwise humming along nicely......
Lookup sinewave clipping for more information..
If you can't quantify how much they cost, it's a deal, I'll buy 5 of them for 3 lumps of rocking horse ......
Very brief update.....

Redpacket and LEDSchlucker like this post
If you can't quantify how much they cost, it's a deal, I'll buy 5 of them for 3 lumps of rocking horse ......
Massive! Iron Donuts! Tongue
Proceed with caution. Knowledge is Power! Literally! Cool 
Knowledge is Power; Absolute Knowledge is Absolutely Shocking!
Certified 18650 Cell Reclamation Technician

Please come join in general chit-chat and randomness at Discord Chat (channels: general, 3d-printing, linux&coding, 18650, humor, ...)
(this chat is not directly affiliated with SecondLifeStorage; VALID email req'd)
After a drive over to pick the units up and returning home I decided to weigh the larger unit as something did not seem right.

I had been given a quote for a heavier unit and was expexting it to be closer to 31kg, which with anything you buy is a case of being given something less than you were expecting... therein followed some emails, spreadsheets, more reading, giving up for a time and then curiosity. What went wrong, or not ?

The next step was a basic power it up and see what the idle losses were and also the secondary voltage, to check the winding ratio is as expected. In order to avoid the surge blowing the fuse I ended up wiring a 2kW heater in series in a separate loop to switch in to charge it and then that was switched out so that the unit was fed with grid power directly.

After the breif sub second faint hum, almost silence, switched grid in and fan heater out.....

So, I now knew it needs just 14.5W at idle (withou the rest of the idel losses yet to be added from the inverter board and filter parts)  this makes me think the hole inverter will hopefully idle below 20W around 235V.

After a few more calculations and the actual core size, wires, etc. it started to become a lot more clear that even though it is smaller than the original quote it seems like a really nice 6kVA unit and with oil cooling can then be pushed for EV charging at 7kW while still feeding the house on top..

More on calculations...

When the unit is loaded to 12kW this is then going to be pulling over 260A and is the sort of current level I'm needing to consider tripping the circuit, but timing determines the breaker rating. Type-C breakers should be able to run for nearly a minute at around a 30% overload.

From looking at the chart profile and a rough finger in the air estimate, my conservative starting point for the live breaker will be 40A (4 pole crossbar connected, so 160A). This then gives me a 7.6kW baseline with a 60 second overload trip level of around 10kW. I have ordered 50A and 63A for later... and a 6A for stage 3 testing....

Breakers are slow. They are mechnical, unless you put enough energy into them and then the mechanical makes way for plasma, gasses, flashes and lots of smell.

Fuses, however are faster.. but, they are quicker in overloads (less to heat up).

Using both a breaker in series with a fuse should result in a nice balance...

With 6 x MOSFET in the board they are package rated for 90A continuous, so that is 540 Amps. That is 25kW.. surge rated...
Idm is quoted 800A but way too big to consider as the FET will melt before even any fast blow fuse.
Continuous drain is around 170A at 60C so that is then 1020 Amps. This is the sustained point above which damage will occur.

Short circuit currents will be high, so a fuse in addition to the breaker will be used when surges reach 16kW, so that''s around 350A for the fuse.

Basically a 160A circuit breaker in series with a 350A fast blow fuse. Hopefully the house wiring will have enough resistance to bring faults well below 8000A and give the fuse a chance to then save the FET's.. overload issues, the breaker handles. That's the plan anyhow...

Next up, filters.
Redpacket, Wolf, rottenkiwi And 2 others like this post
If you can't quantify how much they cost, it's a deal, I'll buy 5 of them for 3 lumps of rocking horse ......
The inverter consists of two main parts, namely the transformer and the inverter board... between them sits a small component, or few...  and then some other parts, whcih will be covered in a later post.

With the default listings on various seller sites they have a generic component value given of anywhere between 47uH and 60uH but none of them define why that value is applicable to any and all transformers, if it actually is even stated or not in this example...

"Magnetic loop inductance" - brilliant !

So, what is needed here and why ? It's a combination of two component characteristics.....

For an inductance value basically the "ideal" value changes depending on the inverter board and transformer, using 60uH labeled inductor "should" be fine but may well result in higher idle power losses and earlier failure (years, not imminent doom of magic smoke in hours....) of the FET's. The reasons are a little more involved and have to do with the way a MOSFET switches on and the resulting small period of time the gate presents a decreasing resistance value. The inductors main role is to prevent any excessive current during these first few 10's of nano seconds as MOSFET's are not designed to cope with any significant current during this phase and can fail easily with very few amps.

For my inverter board and transformer combination the characteristics of a single E70 type of size ferrite core of 60uH is insufficient.  It works, but my idle losses are somewhere over 10% higher than they need to be (87uH losses were 6% higher and close to the knee point in loss increment and significant FET current during transition).....

Using a single E70 ferrite core 350uH is closer to an ideal balance for reducing idle power. But, if you add in two ferrite cores in parallel things change and the new ideal values are closer to 120uH each (returning to 60uH total...) and this matters because a single 350uH inductor needs 6 turns very thick of wire and two 120uH inductors only need 3.5 turns of wire half as thick. 

The difference is 6 turns of very thick wire may not physically fit through the core depending on your wire....

Add in 4 cores and things change again, add in air gaps into the ferrite cores and things change yet again...

This does not really address the reason why a larger single inductor "works" the same as two smaller inductors in parallel presenting around 25% of the inductance of a single ferrite. Here it get's a bit more complicated trying to explain because at the moment I have a working theory (lots of math and some test data) but it's not definative at this stage so I will post the details for this later when I am 100% certain (or as certain as I will know at the time !).

The aim here is to find off the shelf parts that will work repeatedly without home wound gapped ferrite cores and that's where it gets a little more complicated because I'm still looking for a particular source that meets the criteria, which I need to prove.

There are a lot of people putting these inverters together and then getting "random" failures and not knowing why... this testing will hopefully resolve and detail a lot of the reasons.

So far in my testing the idle load of the whole inverter is around 23W while it is sat on my desk..... 50V in and 235V out... not bad for a 6000W 100% duty inverter, even without oill cooling !

Next, output filters.....
Wolf, Chiptosser, Korishan And 1 others like this post
If you can't quantify how much they cost, it's a deal, I'll buy 5 of them for 3 lumps of rocking horse ......
A few more tests and idle power is now around 18W or just under, so now around 20% drop from last testing... or an extra 34kWh per year not wasted... or around 100 cups of tea  Big Grin


Scope traces are typical switch on and switch off event... the 10nS rapid rise is key to curtail and slow down.

The two adjusted cores also visible, scaled back from using 4 cores....

The power supply is showing the output 30.6V @ 630mA. This feeds a boost converter, which delivers 50V to the inverter for testing purposes (easy way to limit the current in case of issues and it saved the day once so far.....) If we say the boost stage is 95% efficient then idle power works out to be 18.3W... think it's closer to 19W though.

The smaller transformer is 240/29V rated at 1kVA in comparison and will be used in a second inverter build, which may well have idle losses around 8W or lower, hoping for 6W. This may end up the winter standby equipment / computer supply.
Korishan, smaresca, rev0 like this post
If you can't quantify how much they cost, it's a deal, I'll buy 5 of them for 3 lumps of rocking horse ......

Forum Jump:

Users browsing this thread: 1 Guest(s)