Thread Rating:
  • 2 Vote(s) - 5 Average
  • 1
  • 2
  • 3
  • 4
  • 5
48V 5000W Oil cooled inverter build
#51
The filters are to stop higher frequency (well above 23kHz) harmonics and noise leaving the wires and transmitting to the world as noise or causing issues with other items plugged in.

With the switch on rise and too much inductance you end up with oscillations as shown (as a more extreme example - and a slow switch on event due to no load as the inductance is slowing the voltage rise more)


This is the voltage switch on event for the FET's with too much inductance, under load and shows an additional distortion in the ramp up of the voltage.



The filters will then these higher frequencies from going down the wire to the house and also block any noisy applicances back feeding into the inverter and potentially causing issues.

The E70 cores (ferrite) take out a lot of the high frequency noise but you can still end up with residual noise/harmonics after the E70's.

Filters are only dampening or reducing the extent of the signal as they can't take out 100% of any unwanted component, unless you are dealing with a digital system.... so with analog you end up with a compromise.

The 5uF is there to do two jobs, one to smooth out the 23kHz and also to create an out of phase resonance with the toroid (75Hz) to help stop the 50Hz wave from being distorted and creating harmonics. If you use one 5uF or 5 x 1uF caps they do the same thing. Caps are the same as batteries in that regard, parallel or series they give the same way. I still need to measure and test for the exact value for 75Hz resonance with the transformer and have a bunch of smaller caps to add or combine in different ways to get any value I need.

So far after I set it up to test live I have not powered it down (now at 773kWh through) but need to do this soon, or in about a month under my deffinition of soon, lol.
Korishan likes 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 ......
Reply
#52
Yeah I'm aware of how caps parallel :-)
I was thinking of how some caps differ internally (some better @ HF than others, internal inductance, etc, etc) & given that 5uF cap is getting the full brunt of the transformer's output! Must be under some serious stress!

Re the filters, was wondering about this position in the cct:
FETs > E70 inductors > main transformer > 5uF cap > {insert new filters here} > CW4L2-20A-T.

Re the waveforms, the second one with the wrinkle half way up the leading edge, that wrinkle is FET switching point oscillation, gate drive resistance also affects this.
Agree the second trace has too much ring at the top.
Some of that ring might be from inter-winding capacitance.
Great to see the traces there :-)
completelycharged likes this post
Running off solar, DIY & electronics fan :-)
Reply
#53
Teaching to suck eggs... :-)

The capacitor is polymer X2 self healing. Has to be self healing so there is no sudden cap failure and the inverter suddenly exposing everything to a 24kHz PWM. Under load the for a ripple of 15V (235V to 220V) stores 15.9 milli Joules and at 24kHz, 50% duty the load would then be 762W (reduce the duty for lower loads...). Suppose this is a bit simplistic as it's not really taking into account the larger phase offset at lower loading (at idle the apparent load is near 100% capacitance, the 5uF).

Increasing to 90% duty, maintaining a maximum 15V ripple the load tops out at 3.8kW, that is if all of the PWM smoothing were to be handled by the cap on it's own..... 4V ripple and 98% on duty the cap can handle 5568W.... This is partly why i am suging the board at maximum 6kW, even though it is 'rated' (chinese deffinition) for 10kW as I don't want a high ripple on the output under high loads as I think this would lead to failure quite quickly.


Adding a separate filter in I had looked into that, specifically something around a 8kHz low pass rather than the higher frequency units I have ended up with. For a 8kHz low pass you then need something like an inductance of 850uH and a 0.47uF cap.
The 850uH would then need to remain unsaturated for currents upto say 25A (minimum) for a 6kW load at 235V, which can then make for a rather large choke. Using a single T-400-26D core (AL262) you then need 57 turns on a single T400 and that saturates at 21A so....

Stack two T400's and 41 turns then saturation is raised to 29A. I only have one T400 and did try it on the output with 56 turns but the wire I used for the test was 0.75mm2 (5A) and did not go back to re-winding it.

I may end up ordering another T400 and wind them, ideally I was initially hoping to source all the parts for the build off the shelf at a sensible price so that they could then be used as a nice easy standard kit.. a plug and play alternative to the way over $1k inverters for 5kW.

I'm not an electronics engineer by proffession so this has all been an interesting learning curve... (this is the disclaimer that I'm not sure I am 100% right in all or any of what I have said !) Still a lot to learn, test and implement...
Korishan likes 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 ......
Reply
#54
(06-08-2019, 08:29 AM)completelycharged Wrote: Teaching to suck eggs... :-)
Definitely learning from reading your progress path here! :-)

(06-08-2019, 08:29 AM)completelycharged Wrote: The capacitor is polymer X2 self healing. Has to be self healing so there is no sudden cap failure and the inverter suddenly exposing everything to a 24kHz PWM.
Good point.
(06-08-2019, 08:29 AM)completelycharged Wrote: Under load the for a ripple of 15V (235V to 220V) stores 15.9 milli Joules and at 24kHz, 50% duty the load would then be 762W (reduce the duty for lower loads...). Suppose this is a bit simplistic as it's not really taking into account the larger phase offset at lower loading (at idle the apparent load is near 100% capacitance, the 5uF).

Increasing to 90% duty, maintaining a maximum 15V ripple the load tops out at 3.8kW, that is if all of the PWM smoothing were to be handled by the cap on it's own..... 4V ripple and 98% on duty the cap can handle 5568W.... This is partly why i am suging the board at maximum 6kW, even though it is 'rated' (chinese deffinition) for 10kW as I don't want a high ripple on the output under high loads as I think this would lead to failure quite quickly.

Adding a separate filter in I had looked into that, specifically something around a 8kHz low pass rather than the higher frequency units I have ended up with. For a 8kHz low pass you then need something like an inductance of 850uH and a 0.47uF cap.
The 850uH would then need to remain unsaturated for currents upto say 25A (minimum) for a 6kW load at 235V, which can then make for a rather large choke. Using a single T-400-26D core (AL262) you then need 57 turns on a single T400 and that saturates at 21A so....

Stack two T400's and 41 turns then saturation is raised to 29A. I only have one T400 and did try it on the output with 56 turns but the wire I used for the test was 0.75mm2 (5A) and did not go back to re-winding it.
Some serious currents there....
I'm thinking such a filter might be best with two inductors (one each leg) to reduce the RF noise?

(06-08-2019, 08:29 AM)completelycharged Wrote: I'm not an electronics engineer by profession
You've had us fooled the whole time :-)
completelycharged and Korishan like this post
Running off solar, DIY & electronics fan :-)
Reply
#55
Appart from finding out my power sockets have a design flaw where the switches partly melt when used below the rated value, the system is otherwise still running fine and getting quite impressed with the whole setup.

The most beneficial part of it is the overall efficiency as it appears to be around 10% more efficient than my grid tie units for serving up energy as long as the overall load is below about 4kW.


Above 4kW I^R losses start to create a lot of heat and efficiency drops away.

The grid tie units I have in comparison seem to be about 80% at best (will try to do some more accurate figures) which is a bit dissapointing, so the preference is for the off-grid inverter to supply any loads.


Most horendous appliance attached award goes to the lawn mower with a horrible brushed motor, which surges over 4kW on start and makes the fridge look like a kitten

Most energy used award is probably a tie between the washing machine and kettle, although the kettle is older and soon to be over taken by the washing machine

Most enjoyable appliance award goes to the slow cooker, sometimes just the smell of the food is enough, amazing what a few Watts can do

Most useful appliance award goes to the vacume cleaner as it is great at collecting all the stripped wire ends while nowhere near as brutal as the lawn mower in the process



Next up - making a box. The inverter should be able to fit into a box the same size as the battery packs so the thought is to create another box or three, which then server as inverter and power distribution boxes to add to the stack and remove all the shiny bare conductors....
RikH, smaresca, 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 ......
Reply
#56
Ok, so with the box building I have a separate 1kVA transformer that I want to build a smaller unit (even lower idle Watts) for remote site where the wind and solar output is quite limited, so conservation a maximum.

Rather than buying a smaller board I ended up ordering one of these
 

The board is effectively the same as the existing board I have (so in theory a spare) but they have added another set of FET's into the mix so each leg has 7 FET's instead of 6. This should help drop the temperature above 4.5kW (the point it starts to warm up) and provide for additional surge resiliance way above the horrors of the lawn mower.

The existing two chokes will have the air gap widened slightly to drop the overall inductance down to 50-55uH and this should also increase the maximum (satruation) current handling a bit at the same time.

The existing board will then get the 1kVA toroidal unit attached (way under/over sized combination) as a bit of an experiment and will use the T400 toroidal ferrite ring as the choke (with a few windings to get 50uH).

Back to cutting wood up...
Korishan likes 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 ......
Reply
#57
Bit of an update, 7 boxes still under construction with one of them to be a small self contained central busbar and breaker box for portable use. After a bit of thought the busbars (aluminium) will go vertical for a few reasons, better space use, dropped wires will not cross bars, the +ve and -ve sides are separated with the breaker rail in the middle, incomming battery box wires enter each side and take up minimal wiring space. Will post some pictures when the area is a little bit less of a combination of circus balancing act and an assualt course.

Having used the inverter for a few months now and done a bit more research and calculations, I think I have come up with a kit of parts that is all quite cheap and easy to put together (appart from the toroidal transformer if your not in the UK). For a basic 1kVA inverter that can handle "any" 1kW load I will hopefully create a separate thread in a few months. Will be trying a mini £41 board (https://www.aliexpress.com/item/32976783366.html) in a few months with the 1kVA transformer I have to see if a near £100 build is possible, which will cope with a whole lot of horrible electronic devices plugged in, possibly even the lawn mower for a few minutes even though it would overload shock everything.


1. Inverter board (same size or smaller than the one I have) which have the juimper wire from the EGS board and have separate push pull FET pair to control the FET gates, will add a picture pointing this out (they are the pair on the bottom left corner of the board in the picture in the previous post and missing from the smaller board linked above). This is possibly critical for high loading and high number of FET's in the board because without the additional push pull to the gates of the FET's the switching is slowed down and under high load that means a lot more heat.

2. "Magic loop" Inductor. T400-26 (https://www.aliexpress.com/item/32880955248.html) with 15 turns (spaced evenly) should give 58uH and saturate around 78A. The 58uH is higher than the 45uH typically indicated by sellers, however the higher inductance with a toroidal transformer seems to work ok. If you can, set it up with thinner flexible wire and try out 12-17 turns to see what happens with the idle current and output waveform. I have one of these ferrite cores, seems to be ok so far and have another on the way to try a few more experiments. This is where multiple transformers come in for higher capacity.

3. Toroidal transformer (https://airlinktransformers.com/category...dard-range) and depending on the working voltage range of your pack the voltage may be different (will try and create a lookup table) and would not recommend going above 3kVA for a single transformer, even after I built with a 6kVA toroidal transformer. If you need 6kVA capacity then use two 3kVA units in parallel (input choke(s) per toroid) or 3x2kVA and switch them in/out to reduce idle losses if needed, similar to some of the Victron units. Idle losses on a 3kVA unit should be around 10W and 7W on 2kVA at a guestimate. The smaller VA / current allows the input chokes to be a lot smaller and way less problematic in construction, which means the T400-26 can be used off the shelf with just a few turns through with way less than 100A and smaller wires. My 6kVA unit (4x30V) might end up connected to a reconfigured 96V pack arrangement and the toroid wired 2x60V or stay as a cool running unit.

4. Output filters, use the CW4L2-20A-T and use more in parallel if you need the higher current. Will be having a go at soldering to the terminals rather than using spade connectors, so will post an update if they melt badly.

5. Circuit breakers. I can't really recommend the DC brekers I have used, they are cheap and not very cheerful, although the good point is that none of them caught fire even after the internals burnt out, which is not a great selling point. The smaller 63A and lower devices seem ok, but once they have tripped the contacts seem to burn and impact the rating at high loading (higher heat generation then trips the device), so, looking to test some alternatives which do not have a massive brand premium.
Korishan likes 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 ......
Reply
#58
Boxes built and the inverter was taken appart, along with the output filtering board.

The inverter board was then re-housed into a box along with 14 turns on the T400-26 ferrite core (to give 51uH) and the 1kVA tranformer and powered up (supplying 105W at the moment for the screen and computer) after testing by supplying it through a boost converter attached to a 30V power supply as this saved me from a fireworks display previously. The board has just registered it's first 1.05kWh of power throughput.

Output filtering is with a single CW4L2-20A-T filter and connected to ground via the house earth connection (which has three additional earth spikes in the garden) so that the residual current circuit breakers will also function properly.

The idle power consumption is around 16W, which is quite a bit higher than I was hoping for (near double) and an interesting comparison for the lack of linear reduction in idle losses vs the transformer. The transformer losses for the 6kVA unit is just under 15W and the whole 6kVA inverter build was around 20W-25W, so a small reduction for the loss of a lot of capacity. If the inverter board has losses around 6W (as per the 6kVA build) the idle losses of the 1kVA transformer would then be around 9-10W so again a relatively small gain in loss reduction for a lot of capacity loss. This is making me think, something like a 2.5kVA tranformer, beefy enough for the kettle, yet still light enough to avoid a hernia moving the "portable" inverter. Hopefully small enough to fit in the new box.

Need to decrease the output capacitor from 5uF down to around 3.5 - 4uF range to bring the resonant frequency up a bit so the transformer does not hum after a few seconds of stable load (build up of resonance) which may hopefully reduce the idle losses a fraction (ok slim chance....)


Tested the build with just a single a grid tie inverter plugged in and set the grid tie inverter to export 300W, which was then fed back in through the 1kVA inverter and back into the battery, so that was a brief test of backfeeding (charging) through the inverter. The test was listening for noises / issues as the power reversed flow because the PowerStar unit is noticable when the power level crosses zero to backfeeding. Interestingly the AC power meter connected to the 1kVA inverter output registered 298W flowing, which gave some comfort the readings on the grid tie export unit is somewhere near ok.

Now waiting on the new board to arrive to rebuild the 6kVA unit again.
Korishan likes 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 ......
Reply
#59
So, fresh morning (ok, from going to bed at 4:30am...) and soldering iron at hand, swapped out the capacitor.

5uF for 3.35uF, still humming,

3.35uF for 2.75uF, still humming but a lot less than the 5uF

2.75uF for 1.55uF, silent on power up. Waveform seems ok, so put the fan heater on, 1kW voltage ok, not much noise. So then for a bit of overload and the 2kW setting. Voltage dropped to 221V and as a result it only reached 1.8kW output, still not too bad for a small  Big Grin overload on a 1kVA transformer !!

Noise re-appears with some loads connected but suspect that this is due to the additional capacitance that the power supplies may add to the line, which alters the resonance again.

Will post some pics later on after a coffee.
Korishan likes 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 ......
Reply
#60
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 ......
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)