Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me.
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 09:27am 19 Jun 2018
Copy link to clipboard
Print this post
Number of turns (18) was determined by the maximum flux density, voltage, frequency and core cross section in the usual way. The gap sets the required inductance which was something like 140uH if I remember, although it was a very long time ago since I actually worked all this out.
I did measure core saturation on an inductor tester, something like 50 amps at max duty cycle I think it was. There is some margin there, but not a big margin.
There is a current transformer on each of those PCBs, blue in the top picture, and black in the lower picture, each has a red wire going through the hole. Control chip is a UC3825 run in current mode, so it has a very effective cycle by cycle current limit at 20Khz.
Input reservoir capacitor is a 2,200uF 500v Evox Rifa low ESR electrolytic with a 40A rms ripple current rating. There are two smaller Evox Rifa capacitors on the dc output rails, each 470uF 385v. Those are small, but there are two on each dc/dc converter, and more in the inverter, and still more in the mains rectifier, which are all effectively in parallel, but there are long cable runs between
These electrolytics are bolted direct to the circuit board from underneath, so it was simple to use crimp lugs as well for input and output dc connections under the same nuts that secure the electrolytics.
Each 225v dc rail has something like 4,000uF total capacitance to ground, so its all pretty well tied down at 100HZ, so power drawn from the solar panels has fairly low ripple current at 100Hz.
Edited by Warpspeed 2018-06-20Cheers, Tony.
zaphod Regular Member Joined: 03/06/2018 Location: United KingdomPosts: 93
Posted: 06:34pm 19 Jun 2018
Copy link to clipboard
Print this post
Ahh I had those CT's down as big ferrite beads :) Thats a lot of capacitance, guess I ended up with about 1000uF/Kw as a comprimise on cost & space vs loss of solar efficiency. I see you have an MPPT pot there, do you run your panels at constant voltage or do you have an MPPT finding algorithmn ? I use perterb & observe running at 30hZ that I find is fast enough to track rapid cloud edges.
Regarding your question of how the unipolar rectified transformer output becomes a bipoler signal I use an H-bridge that switches at 100hZ. Originally this was constructed of Mosfets but I found them unreliable and now use IGBT's that are much more robust IMOP. Driving the H-bridge from the control circuit at ground means there has to be an isolation barrier, I use standard opto-couplers but the edges are cleaned up by schmitt input gate drivers to ensure fast switching. Power for the H-bridge drive comes from a small ferrite ring transformer having several secondaries all wound with thin TIW wire to withstand the voltages. The primary is driven by a standard mosfet driver from the main ground referanced 12V auxilery power rail. I hope I can find a picture
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 10:31pm 19 Jun 2018
Copy link to clipboard
Print this post
Yes, 1000uF/Kw is pretty much the industry standard for energy storage at around 360v to 400v for most things. The inverter presents a pretty savage pulsing load, and if there are only solar panels as a power source (and no battery) there needs to be provision for a lot of energy storage somewhere.
Agree totally that IGBTs are a lot more rugged than mosfets and above about 200 volts they win hands down. My current new inverter project uses half bridge IGBT power blocks, still need some transformers wound for that, and when its complete all this flyback hardware I have been telling you about will no longer be required.
A very long time ago, I did some testing of solar panel output with a watt meter and a continuously adjustable load, and discovered that the power does not fall off as much or as quickly as many people think it does. I also have far more solar capacity than I need, so trying to squeeze the last half a percent out of it is just not worth the effort involved.
I just use a PID system that holds the panel voltage constant to prevent falling off the broad hump under low light conditions. It just overpowers the voltage feedback to reduce loading on the solar source as required.
At sunrise the solar panel voltage springs quickly up to 220v or whatever the setting is, and the solar panel current slowly rises up from zero as the sun gains strength. Its just an op amp integrator, but it seems to work perfectly well.
My system is very crude and totally in hardware. Its definitely a real Frankenstein compared to what you are doing in every respect.
As your system undoubtedly uses a microprocessor to do all that it does, a proper perturb and observe algorithm is not going to be much extra work to do and makes perfectly good sense to have that.
The reason I built these flyback monsters originally, was because the inverter I am still using has no internal voltage regulation. It requires a constant dc input voltage to produce a constant ac sine wave output voltage degraded only by some very minor conduction losses. Its actually four separate half bridge square wave inverters with four low frequency transformers, each having different secondary voltages.
The secondaries are all connected in series, and by switching these inverters at the appropriate times in a particular sequence, I can build up a sine wave shape of many fine steps with very low harmonic distortion. This too is a bit Frankensteinish, but its simple, very robust, and handles nasty power factor loads and huge inrush currents with ease. No high frequency PWM haha, it all switches at a few hundred Hz at most, so its all very non critical.
Edited by Warpspeed 2018-06-21Cheers, Tony.
zaphod Regular Member Joined: 03/06/2018 Location: United KingdomPosts: 93
Posted: 07:13pm 20 Jun 2018
Copy link to clipboard
Print this post
Yes I should have mentioned the voltage as C alone is a bit meaningless! Well my bus voltage is just a nominal 180V so about half the energy storage you envisage so at full power some of the 100hZ ripple is reflected back to the panels with a consiquent loss of efficiency however for axample doubling the energy storage is simply not worth the gain in efficiency IMOP I vexed a lot about this and somewhere there are some graphs and figures to justify my choice.
I dont know why I didnt use IGBT's in the first place, all I remember is I did not want thyristors due to drive consumption, I had not used IGBT's anywhere before and as high voltage mosfets are available they were just a natural for me. I still don't know clearly why they were failing and it seemed random and intermittent. The problem with a CSI is zero cross timing is extremely critical but at the same time (pun) difficult to derive as it is an extremely noisy environment. My best guess is despite all my fixes occasionally it just went wrong and whereas an IGBT can swallow the resulting surges a mosfet hasn't got quite the same overload capacity, at least the very high voltage ones (I was using 800V). So far the 1200V IGBT's I am using have survived (touch wood)..........
There have been many discussions about the pro's n con's of true MPPT (the Tracking bit), lot's of misinformation and misselling especially amongst Chinese charge controllers. Realistically it depends if you want to spend a few more bucks optimising your output, there are quite large changes over temperature and output current that can quite easely tip you off mpp especially on the voltage increase side where the slope is steeper as in this picture for example.
I don't think totally hardware is frankenstein!! It can and is very effective, I avoided a total DSP approach for the CSI as advocated by many because in conversion from analogue to digital domains and vice versa you introduce many problems requiring much extra work on top of analogue only solutions so I prefer a mixed aproach each where it's best. I have seen analogue mppt's using sample & hold techniques too
One of the big problems with the CSI that the use of an MPU helps to solve is that of a sequential start and shutdown, at the end of the day it could be done discretely but would involve many many more components. Analogue integration in some MPU's is now quite advanced with not just ADC's n DAC's but also op-amps n comparitors so the number of chips (and space) for a given function can be reduced, very usefull in a high noise environment.
That inverter sound simple and effective and I guess is run solo (not in tandem with any other power source) ? I think this is where things get really difficult, I would say a very high proportion of my development work was/is being able to run in tandem with another power source and inject current synchronously with it. Cheers Roger 1Kwp DIY PV + Woodburner + Rainwater scavanger :)
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 11:48pm 20 Jun 2018
Copy link to clipboard
Print this post
The way I look at the whole MPPT issue is to just have something very simple that works reasonably well without needing to strive for absolute perfection.
These days, solar panels are so cheap that its dead easy to overpower the system with additional panels, rather than trying to chase a few extra percent of output with something complicated.
Running from more than one dc power source really involves just a diode or two, which is how I plan to do it.
My inverter will run direct off the solar panel voltage which will have an MPP voltage of roughly 120 volts (four 24v panels in series) and an open circuit voltage of up to about 150 volts maximum. There will be a large energy storage capacitor of 36,000uF at that voltage, that hopefully should be enough to tame the 100Hz pulsing load of the inverter. During a normal day that should all just run by itself directly from solar, requiring no additional power source.
There will also be a 100 volt lithium battery (30 cells 3.3v) so that when there is insufficient solar, or at night, the battery gradually takes over the inverter load from solar as the voltage falls away at dusk. Minimum battery voltage will be around 90 volts (30 cells 3.0v).
This requires an inverter that will run with a very wide input voltage range 90v to 150v. But it has the advantage that during the day the battery gets a slow constant steady charge, without seeing any discharge at all except on really bad days. Even in mid winter the battery will see a seven hour slow steady constant charge during the day, and a seventeen hour slow discharge at night.
Its totally different to feeding solar into a battery through a high power MPPT solar controller, then connecting the inverter direct onto the battery. Everyone does it that way, but the solar controller needs to be powerful, and the battery is constantly having high currents shuffling back and forth in and out during the day.
My system should be much kinder to the battery, and a long slow steady charge should enable a higher rate of final charge state. The battery charger only needs to be tiny. To fully charge a 50Ah battery over seven hours only requires 7 amps. My system is not that large, 50Ah at 100v = 5Kwh storage. Its mid winter now and my night time load is 2.9Kwh. That falls 1.2Kwh in mid summer.
The inverter needs to be able to cope with an unusually wide input voltage range, this new design works over a 2:1 input voltage range, 90v to 180v directly, without needing the flyback monsters as pre regulators.
Anyhow, I have been thinking about your scheme of feeding PWM through a transfomer, then a rectifier, and then through an LCL filter to generate unipolar half sinusoids. What discharges the capacitor on the falling half of the sine wave if there is little or no load on the inverter ?Cheers, Tony.
zaphod Regular Member Joined: 03/06/2018 Location: United KingdomPosts: 93
Posted: 08:22am 21 Jun 2018
Copy link to clipboard
Print this post
The thing that discharges the capacitor is my neighbours but actually my house load never falls below ~20Watts anyway. Remember this C is small, it just provides a return path for the ripple component in the first L so in my case is just 1uF, smaller than the PF correction capacitors used in flourescant lights. But I think I should also mention a CSI always needs to operate in tandem with a VSI to provide voltage regulation whereas I think you are designing a VSI, my design is NOT a VSI.
I use a pre-regulator to remove one of the many variables from the equation and simplify the design but then I already had the pre-regulator it was simply the immersion boost converter arranged to provide a constant output voltage when not driving the heater IMOP it is already hard enough to get enough effective PWM resolution for the sine wave and power control.
I can assure you that not using MPPT can loose you a lot more than a few percent but if you have the space and money to overpower the system why not, it's just as valid as any other solution Cheers Roger 1Kwp DIY PV + Woodburner + Rainwater scavanger :)
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 09:29am 21 Jun 2018
Copy link to clipboard
Print this post
I am not suggesting to not use any MPPT at all, just that a the difference between a constant voltage system and a peturb and observe algorithm is probably not large enough to worry about.Cheers, Tony.
zaphod Regular Member Joined: 03/06/2018 Location: United KingdomPosts: 93
Posted: 06:23pm 22 Jun 2018
Copy link to clipboard
Print this post
Well we will just leave that one as a differance hahaha Each day for testing I have to wait for the hot water to reach temperature before I can begin testing the CSI as I have now moved from bench test to real installation. I am going to put you off MPPT and digitally controlled converters for life as 99% of my problems result from two digital converters fighting each other that fight is over the buss voltage they share. When not heating the MPPT controller/boost converter tries to maintain a bus voltage of 180V however due to both using IRFB4227 mosfets (very robust) rated 200V it overvoltage trips at 190V. Meanwhile the CSI (that also uses IRFB4227's on the primary) regulates power to maintain a target bus voltage a little lower, maybe 175V. Ok so far so good BUT when a fault of somekind is detected by the CSI it shuts down and due to the low C on the bus it's voltage rises rapidly (due to sudden loss of load) and hits the overvoltage trip! (of the booster). Unfortunatly I already discarded some of the best scope shots of what happens but in this pic you get the rough idea, red trace is bus voltage, yellow is CSI demanded power.
So you can see the bus voltage drooping (bloody UK clouds) and the demand power (yellow) trace starting to reduce but then BAM the CSI primary overcurrents (due to reduced voltage) and unfortunatly hits the power demand to low so the bus voltage uncontrolably rises till the booster hits overvoltage and cuts off! The CSI is still soft starting at this stage and desperatly tries to hold regulation but the booster is in shutdown and the CSI finally gives up when the bus falls to low. Well the fix was not to unload the booster so fast just for one little meany overcurrent event Of course another equally valid (but expensive and space consuming) fix would be to employ a much larger DC-link capacitance.
This picture shows them co-operating perfectly
I enjoy this sort of debugging, it's hard to capture the actual events and work out the sequence, there were quite a few more including switching noise creeping into the heatsink temperature measurement causing random shutdowns but today its basically working hooray
Sadly however after adding half a ton of extra ferrite the EMC problem remains and needs to be dealt with. The hard part of EMC is localisation and meaningful comparisons when changes are made As everybody is at home for the weekend watching football I don't think I can do anything except theorise till Monday
Edited by zaphod 2018-06-24Cheers Roger 1Kwp DIY PV + Woodburner + Rainwater scavanger :)
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 09:35pm 22 Jun 2018
Copy link to clipboard
Print this post
I doubt if adding extra bus capacitance would help all that much anyway. Its basically an unstable load sharing situation, and slowing down the catastrophe is not really a solution.
Stable load sharing is probably the key to this, and the two systems may need to be linked somehow either in hardware or software, or a third system arbitrate. Cheers, Tony.
zaphod Regular Member Joined: 03/06/2018 Location: United KingdomPosts: 93
Posted: 07:16am 24 Jun 2018
Copy link to clipboard
Print this post
Nope it's not unstable as it employs feedback to control it. The two systems do not need to be linked, thats like saying a series regulator must have feedforward to compansate for line variations, not required.Cheers Roger 1Kwp DIY PV + Woodburner + Rainwater scavanger :)
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 07:25am 24 Jun 2018
Copy link to clipboard
Print this post
I assumed that meant that the two inverters were fighting each other.
But if the system is stable as you now say it is, then there is no problem eh ? Cheers, Tony.
zaphod Regular Member Joined: 03/06/2018 Location: United KingdomPosts: 93
Posted: 07:51am 24 Jun 2018
Copy link to clipboard
Print this post
Sorry my fault for my expressive language! That was happening during fault conditions that were not tuned to synchronise properly having never worked together before. I think as long as the energy consumer has more feedback bandwidth to react to changes than the energy provider has to change availability the system will work After all that's also how MPPT works. Of course if the consumer has to much feedback bandwidth especially with a load profile of 100hZ it incites itself to oscillate!! so the CSI was initially heavely damped but in practise the responce time had to be shortened or bandwidth increased to maintain stability with source changes. The other thing affecting system bandwidth is the link capacitance and increasing it would reduce the rate of source energy change requiring less bandwidth in the CSI feedback loop but fortunatly that was not nececcery.
I am busy working on EMI.....I hate the requirement as it is annoying to have a design that functionally works but is emitting to much magic radiation that cannot be seen and is very difficult to quantify. I think the next step is to reduce power for testing (thus reducing annoyance) whilst I try to tune the snubbers onsite (they were developed in simulation) as I suspect the actual circuit parameters are different from the model (of course) causing the snubbers to be off-tune! At least try and kill as much at source as I can. These damm IRFB4227's are very fast due to low Qc and once again I am wishing maybe I had used slower IGBT's....... Another possobility is to use spread spectrum clocking but that is gonna reduce processing power as the CPU wont be able to clock multiply with such a clock source and it comes at a price $$
Meanwhile the family is on my case so priorities lie elsewhere Edited by zaphod 2018-06-25Cheers Roger 1Kwp DIY PV + Woodburner + Rainwater scavanger :)
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 08:18am 24 Jun 2018
Copy link to clipboard
Print this post
Being retired I am out of the EMI compliance loop thank goodness. A good spectrum analyser might be worth looking into. I bought a 9Khz to 2.6Ghz dinosaur for not that much cash, and it has opened up a whole new world for me.
Just last week I bought a second spectrum analyser, another dinosaur this one 5Hz to 50Khz that I plan to use as part of a Bode plotter system for measuring closed feedback loops of switching supplies. Quite difficult to do well on a limited budget.
Anyhow, getting back on topic to "current source inverters".
My approach to this particular problem is a direct feed forward system. This assumes the actual inverter itself has a very low output impedance with good enough output voltage regulation. All the trouble comes from the highly variable solar panel source.
So what I am doing is measuring the inverter dc input voltage with a dual slope A/D synchronized with the inverter. That will average out any 100Hz ripple voltage and give a pretty fast and accurate update of solar source voltage cycle by cycle.
My inverter then uses the digital output of the A/D to directly switch between 256 different lookup tables at the zero crossings. Very fast response to load changes, and no stability problems. It covers a 2:1 input voltage range, and works like magic. Very pleased with how it works.Cheers, Tony.
zaphod Regular Member Joined: 03/06/2018 Location: United KingdomPosts: 93
Posted: 08:03am 25 Jun 2018
Copy link to clipboard
Print this post
I am retired too but I cannot have it killing my broadband!! Anyway I was very interested in your lookup table system, how often do you update the power selection part of the address, every zero-cross, more or less frequently ? I guess with your huge DC-link capacitors the proportion of 100hZ ripple on the bus voltage is very low ?Edited by zaphod 2018-06-26Cheers Roger 1Kwp DIY PV + Woodburner + Rainwater scavanger :)
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 08:53am 25 Jun 2018
Copy link to clipboard
Print this post
The Intersil 12 bit dual slope A/D integrates over the first quarter of its operating cycle which is 10mS in this case. That averages out one complete ripple cycle of 100Hz ripple. The whole thing takes up to 40mS to take one dc input voltage reading or 25Hz, but the de-integrating time is variable. The lookup table address is updated every second mains cycle via a latch synchronized with the zero crossings.
I probably could run it faster, but then I lose the wonderful noise immunity that the synchronous integration provides. Twenty five corrections per second is in practice quite sufficient. Updating the waveform right at the zero crossing means I can make massive step amplitude changes without any waveform discontinuity. Also, the positive and negative half cycles always match up, so there can be no staircase saturation of transformers, or other oddities that might progressively grow and possibly cause dramas.
Its pretty simple, the solar dc input provides power to the control board and does all the amplitude controlling. Optically isolated gate drivers then just drive four separate bridge inverters. 4.5Kw, 1.5Kw, 500W and 167w. And that is all it is...
Measured distortion is 1.0% with 81 steps at minimum input voltage. Max measured distortion is 1.7% at 41 steps with twice the minimum input voltage. That is the broad band THD measured without any output filtering.
There are 256 lookup tables that provide better than half volt amplitude steps at 230v, so its pretty seamless in operation. Its a quite unique design approach as far as I know, and that is the fun part of it.
This is the actual mains waveform of my current inverter at 100v per division with 81 steps peak to peak. This is actually powering my house with all the mixed reactive and non linear loads, so the steps here look a bit more ragged than testing with a pure resistive load.
Edited by Warpspeed 2018-06-26Cheers, Tony.
zaphod Regular Member Joined: 03/06/2018 Location: United KingdomPosts: 93
Posted: 10:45am 25 Jun 2018
Copy link to clipboard
Print this post
Very nice idea indeed but a bit thirsty on components (4 inverters, trafo's etc). Did I miss something, are you talking about your ADC sampling the DC-link voltage or the output AC waveform ?
I should apologise!! 4 component thirsty inverters that WORK are a hell of a lot better than one that doesnt Think I just lost an IGBT just as I was making progress on EMC suprisingly by moving the system earth point! Left it running while I traced the earthing of the house (turned out to be a spike under a concrete raft) and by the time I got back it had blown it's fuses at a measly 100W.
Frankly I think thats it for a long time as I have so much other work piling up, chainsaw to fix (horrible carb with EPA jets), next winters wood to cut and wife to talk too!! After a year at this I might consider I am no good at direct online grid stuff must have cost me a few hundred to work that out At least I have a commercial unit to fall back upon even though it has annoying foibles. Worst are it won't connect if it thinks grid is to high (common rural occourance) often thinks there is not enough solar power and for any fault whatsoever insists on staying offline for fully 5 minutes but at least it stay's working and doesnt blow up like my efforts Edited by zaphod 2018-06-27Cheers Roger 1Kwp DIY PV + Woodburner + Rainwater scavanger :)
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 09:17pm 25 Jun 2018
Copy link to clipboard
Print this post
The ADC measures the dc input voltage to the inverter. That selects the correct lookup table to generate 230v on the inverter output. There is no voltage feedback, its entirely a voltage feed forward correction system.
I can tweak the inverter output voltage up and down with a potentiometer on the ADC, and there the voltage will pretty much stay over a massive 2:1 input voltage range. Below the low voltage limit, it just runs out of voltage regulation, but keeps on going. At the high voltage end it shuts down, but restarts if the voltage falls back.
The reasoning behind all this is that the solar panel voltage is a highly variable power source, with horrible voltage regulation under varying load.
Inverter output voltage variations will be almost totally due to very large input voltage swings caused by load changes, and a highly variable solar input.
If I can correct completely (and very fast) for all the dc input voltage variations, the ac output voltage should hardly change at all under changing load conditions. And this definitely does work !!
The same concept should work equally well from a battery power source. We only really need voltage regulation of our inverters because the battery voltage goes up and down with load and state of charge, not because the inverter itself is "weak".
Output voltage regulation does not need to be perfect, the grid goes up and down all the time, and everything works just fine.
As you will already know, the really big advantage of feed forward correction is its very fast acting, and very stable. No chance of hunting or instability. Response to load changes will be much faster than any feedback system could be, and by making corrections at the zero crossings, there are no abrupt waveform shape changes.
Ah yes, four inverters.... Only two need to be really large, and all four can use identical circuit boards. But these are very simple low frequency square wave inverters that only need slow IGBTs with plenty of dead time and very low gate drive power. Much easier to get going without all the blow ups the PWM guys are having.
Only significant disadvantage is the need for four special output transformers that need to have fairly exact turns ratios. That is a lot more work, but its fairly straightforward. Everything else is much easier to build and get going.
Have to agree though, four very simple low tech square wave inverters that work, have to be better than one high tech inverter that keeps self destructing.
Anyhow, its been a lot of fun, very instructive, and something rather different that nobody else has.
Four inverters on the same heatsink. The two high powered ones on the right use Semikron 200amp IGBT half bridge power blocks. The two small inverters on the other side use individual 50amp mosfets (which I already had) but which will probably later get replaced with suitable IGBTs. I have had all this powered up, and running with the final drive waveforms from the control board. So far so good. Still have no transformers for it yet. That comes next.
Edited by Warpspeed 2018-06-27Cheers, Tony.
johnmc Senior Member Joined: 21/01/2011 Location: AustraliaPosts: 282
Posted: 12:04am 26 Jun 2018
Copy link to clipboard
Print this post
Warpspeed Been watching your inverter system, very interesting . Am I correct that each of your 4 transformers will produce different output voltages from the 48V battery supply, These voltages, are then under microprocessor control which will add or subtract the sum of the voltages to produce the sine wave. The capacity of the inverter is governed by the physical size of the transformers.
Cheers johnjohnmc
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 12:39am 26 Jun 2018
Copy link to clipboard
Print this post
Yes John, that is pretty much exactly it.
Each inverter has three output states, +ve, -ve, and zero. Two inverters with voltage ratios of 3:1 can produce 3x3 output steps. Thats four up, four down and zero volts in the middle.
Four inverters with voltage ratios of 1, 3, 9, and 27 can create up to 81 steps (3x3x3x3).
The actual transformer secondary voltages are 225v, 75v, 25v, and 8.3v. If all the inverters switch on together with the same polarity, we get 225+75+25+8.3 = 333.3v peak, or 235v rms.
My very first attempt at this used a single 1K lookup table, and the eight data bits provided eight gate drives for four inverters. One very important detail of this is that the inverters in the zero or "off" state MUST present a dead short across the particular transformer, as the secondaries are all in series. You cannot just open circuit a primary winding. Its pretty easy to do with a bridge, just turn on either the two upper, or the two lower devices together, instead of diagonally.
I did then build a microprocessor version which crunched numbers real time, and it worked fine, but I quickly realized I could simplify things a great deal by having just a huge ROM. So using the same processor and number crunching software, I added an EPROM burner routine to it, and it made me a very nice 256K x 8 EEPROM in about two minutes of running time which I now use.
The lower 1K is just incrementally addressed forever by a crystal driven counter. The upper 256 blocks are addressed by a A/D converter. It could not be simpler.
The idea is that as the dc input voltage rises, each step is made very slightly narrower. Not much happens down near the zero crossing, but the top step rapidly narrows until it disappears completely. In this way you can control the ac output voltage by very fine increments.
Its still a form of very slow motion PWM, which has to work at only 50 Hz.
My inverter runs at 100 volts nominal, which is compatible with the thirty 3.3v Lithium cells that I already have. Transformers will be wound with 90v primaries that correspond to about 3.0v per cell end point. Inverter operating range will be 90v to 180v dc input. Max solar open circuit voltage is never more than 150v (four 24v panels in series)
The idea is that during the day the inverter runs straight from the solar panels. At dusk the solar panel voltage fades away and the battery gradually takes over the load.
Yes power is only limited by temperature rise in the transformers. Not only that, power flow is completely bi directional, so nasty reactive loads are no problem. If you use large enough switching devices, an ordinary circuit breaker will protect against overloads. This makes for a really tough inverter.Edited by Warpspeed 2018-06-27Cheers, Tony.
zaphod Regular Member Joined: 03/06/2018 Location: United KingdomPosts: 93
Posted: 04:12pm 09 Jul 2018
Copy link to clipboard
Print this post
Shame about the thread hijacking here!!
Just to conclude the story the Hbridge IGBT's are fine but a surge arrestor (P6SMB550CA) failed short circuit and was responsible for blowing the grid fuses. As I think I mentioned before I live in a rural area on a pretty long supply line with some pretty hairy farm equipment around (grain dryers etc) so just have to toughen up the energy capacity on the surge arrestors.
Anyway as I am definetly not off-grid I don't think my work is of interest to you guys. I should have mentioned my long abscence from the project was due to fruit tree pruning etc, fortunatly I can put it on one side as I have another grid tie I can use in emergencies :)Edited by zaphod 2018-07-11Cheers Roger 1Kwp DIY PV + Woodburner + Rainwater scavanger :)
Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me.
Print this post
but actually my house load never falls below ~20Watts anyway. Remember this C is small, it just provides a return path for the ripple component in the first L so in my case is just 1uF, smaller than the PF correction capacitors used in flourescant lights. But I think I should also mention a CSI always needs to operate in tandem with a VSI to provide voltage regulation whereas I think you are designing a VSI, my design is NOT a VSI.
As everybody is at home for the weekend watching football I don't think I can do anything except theorise till Monday 
but at least it stay's working and doesnt blow up like my efforts 
