weber

I'm not sure I understand the question. Are you proposing that as a solution, @phil.g00? Because it certainly isn't one. And it's amazingly asymmetrical.

Sorry @phil.g00. I should have explained that my particular denial of reality consists in every jumper having the same resistance. That could be taken to mean that all the jumpers are of the same length (e.g. they might be supplied by the battery manufacturer and are capable of skipping a battery, as in the real-life case that prompted this), or it could be taken to mean that varying the lengths of the jumpers has no effect because the cable has zero resistance and all the resistance is in the crimps, all of which are the same. The resistance of the bolted joints has to be zero, so that lug stacking order doesn't matter. I realise that's different from your original criteria.

So now, who can balance 3 batteries (with some doubled-up links), and 5 batteries (without)?

Here's a corrected version of TTTs 8-way. The green numbers are the first pass I mentioned above, and the purple numbers are the second pass.

OK. Now that we've got reality out of the way. Back to the fun puzzles.

Unfortunately, these puzzles, while fun as geometrical and mathematical exercises, abstract away too much of the real world. First, we have to recognise that most of the resistance is not in the cable, and not even in the crimp. The bolted joints generally have higher resistance than the cable or crimp, and this is highly variable. It is close to inversely proportional to the bolting pressure, and also depends on thickness of oxide layers. All cells and lugs had better be equally clean, and you had better use a torque wrench to set them all the same. You also need to pay strict attention to the order in which the lugs are stacked on each terminal. In a simple 2-way diagonal takeoff, the takeoff lugs must be stacked on top (i.e. furthest from the terminal). But it gets worse. The internal resistance of the cells themselves is usually much greater than that of the cables plus crimps plus bolted connections. If not, then you probably haven't designed the cabling correctly. And cell internal resistance depends strongly on temperature. Lower temperature means higher resistance, due to chemical reactions slowing down. And notice the vicious cycle here. Heat generated inside a cell is inversely proportional to its resistance P = (Vcell - Voc)²/Rint, which means the amount of heat generated increases as temperature increases, which increases the temperature further ... The temperature profile of a cell on the end of a row will be very different from one in the middle of a row. So you either need a way to keep all your cells at the same temperature, such as the liquid circulation in the Tesla car batteries, or you need to arrange it so that every string has equal numbers of cells near the middle of the pack, and equal numbers of cells near the outside of the pack. And then you have manufacturing variation between cells, so you have to sort them and group them correctly. Hopefully now you understand why balancing the current between multiple strings of cells is really really difficult, and should be avoided if at all possible.

Some more such puzzles: 1. There is no theoretically-perfect sharing arrangement of this kind for 3 strings, so you are forced to either use Chris Gibson's method 3, or parallel some jumpers. By "of this kind" I mean where all connections are made at the cell terminals, and all jumpers have the same length and cross-section (or all jumpers in symmetrically-equivalent positions have the same length and cross-section). How can we parallel some jumpers to make a theoretically-perfect arrangement for 3 in parallel? 2. You might, by now, think that such arrangements only exist for powers of 2. But some years ago I discovered such an arrangement for 5 in parallel. I sent it to Chris Gibson and he agreed it was legit (although I'm not sure if it meets your rather strict "no link longer than it has to be" requirement, Phil). What is this arrangement for 5 in parallel? I have not found any such arrangement for 6, 7 or 9, although I can't say I've exhausted all possibilities, as I have with 3. But since 5 and 2 both work, we can do 10 as 2 x 5, and of course any number whose prime factors are only 2 or 5.

These battery paralleling puzzles are fun. I've studied them extensively. Chris Gibson deserves to get credit for those diagrams posted above. He also gives your solution. So it isn't new. But good on you, since you came up with it independently. Here's his web page. Yes, each battery must see the same voltage. But that doesn't mean that the path through each battery must have the same resistance, i.e. the same lengths of the same cross sections. Nor does it mean you have to "equalise the current through the various cables", in any way I can parse that. What is required is that the sum of the volt drops be the same along the path for every battery. i.e. the sum over all hops, along the path through each battery, of the resistance times the current for each hop. If you make all jumpers the same length and the same cross-section (and therefore same resistance), it's easier to check any proposed solution. On your first pass, you note against each jumper, how many cells it has to carry current for. Then on a second pass, you note against each battery, the sum along the path through that battery, of the numbers you wrote on the first pass. This also shows that TTT's 8-way is wrong. Coulomb knows the correct 8-way solution, he just hasn't spelled it out, presumably because he thinks it's obvious once you've seen the correct 2-way and 4-way, and realised that the correct 4-way scheme is just the correct 2-way scheme (known as "diagonal takeoff") applied at a higher level, to two 2-ways instead of two batteries.

Jesus wept! Don't you guys know about aluminium oxide! It's invisible, it's very hard, it's a very good insulator and it grows back within 4 minutes of being removed (assuming the aluminium is exposed to air). This transparent natural oxide layer is what stops further corrosion and keeps aluminium looking shiny. It's such a good electrical insulator, it is used in electrolytic capacitors where it can withstand 500 volts. And it is so hard it is used as grit on sandpaper. The natural oxide layer is only 4 nm thick, but stock aluminium from the hardware store may be "anodized". That's where they deliberately increase the thickness of this oxide layer by a factor of several thousand, typically to 10 μm (or 25 μm for marine use). Even if you start with natural aluminium, both contact with dissimiliar metals, and electrical voltage, will tend to "anodize" the aluminium over time. So while it may be fine at first, joint resistance will slowly increase over time until one day you may have a fire on your hands. While anodized aluminium cannot be used for electrical connections, non-anodized aluminium can be used, but there are two important precautions you must take: 1. You must remove the natural oxide layer from the contact area, e.g. using silicon-carbide paper. This abrasive paper is dark grey or black and is also known as carborundum (not corundum) paper, or wet-and-dry paper. Be sure to brush any grit from the surface when you're finished. 2. You must, within a few seconds, smear the clean aluminium with electrical jointing compound (google it -- there are lots of brands). Most of it will be squeezed out of the joint when the bolt is tightened, thus allowing metal to metal contact, but what remains is essential to fill any voids and exclude air and water vapour. Other electrical metals such as copper, lead and tin do not have this problem as their oxide layers, when thin, are sufficiently conductive. In your specific case, the heads of the bolts appear very small relative to the contact area, and because aluminium is much softer than copper I suggest some large steel washers (preferably stainless), between the bolt head and the aluminium, to spread the pressure more evenly over the whole contact area.

We have just bumped patched firmware 73.00e (the first one with kettle compensation) from beta to release. Many thanks to all who tested the beta version.

I hope you're not referring to my blowup, that happened to occur when a vacuum cleaner was turned on. http://forums.aeva.asn.au/viewtopic.php?f=64&t=4332&p=66045#p66045 As it says two posts later, it was not the fault of the vacuum cleaner. The same vacuum cleaner has since been used many times with the repaired inverter.

I'm kind of surprised that no one has mentioned the fact there is no inverting going on here. The Axperts are in bypass. So they have no way of controlling how load current is shared between them. Their current-sharing cables make no difference in bypass mode. So it's not the Axpert's fault, unless the master has a failed relay, which seems unlikely. There has to be a high resistance, or open circuit, in that part of the path from the master's AC output to the loads, that is not shared by the slave. There is a load switch on the bottom right of each inverter. Turning it off will disconnect that inverter from the loads. If everything was working correctly, you could turn off either one and the loads would still be powered via the relays in the other inverter. But if the above is correct, turning off the load switch on the bottom of the master will make no difference, however turning off the load switch on the bottom of the slave will cause the loads to brown-out, or lose power completely. If they brown out, you should turn the load switch back on immediately. And in any case, consult an electrician.

@plonkster, You really should have a winky-face on that, not just a smiley face. People who don't actually watch the video might think you're serious. The guy in the video is either the best deadpan comedian of all time, or he doesn't realise what he's saying. "We'll fit it all in this box. <only fits the capacitors>" "It can run this LED for half an hour" "Of course you can't run anything with it at night, but we could use it to charge a battery ..." If he'd used the $70 he spent on supercaps and instead spent it on a battery, he'd be way ahead.

I'm sorry @Hercules Weyers, but you appear delusional when you claim: (a) there was no over-current protection (thanks @plonkster), and (b) there was no fire—only a DC arc. You also imply that the owner is certifiably insane, by claiming they had the system redone "using exactly the same modules". Can you provide any evidence for this? Or for any other of your claims that weren't already disproved by the available evidence before you made them?

It's not that you need the bootloading function of the bootloader. You need its initialisation code and its reset and interrupt vectors. And you said the microprocessor was "burned". I've never see a microprocessor burned without practically everything else on the same power supply burned too. What makes you think replacing and reflashing the micro (it it was possible) would fix it? Given the version number, I found what I assume was the web page it came from: https://translate.google.com/translate?hl=en&amp;sl=cs&amp;u=http://www.ostrovni-elektrarny.cz/index.php%3Fpage%3Dpodpora&amp;prev=search You said yours was an Axpert KS 3K. You seem to have the firmware for an Axpert MKS 3K. Both are linked from the above page.

Hi @MalanT. No, you should definitely not set the float voltage the same as the so-called "bulk" voltage (which is really absorb voltage). From your battery manufacturer's datasheet, you should set it somewhere between 54.0 V and 55.2 V. If you're in a warm climate you should use a voltage in the lower half of that range, i.e. 54.6 V or less.

@Coulomb, you really should stop encouraging @vladgon to waste his time. You should have stuck to what you said at first. His only option is to buy a new board. Hi @vladgon. It is simply not possible, for the reasons that plonkster gave. Because we do not have the main firmware and we do not have the bootstrap loader. And it is not possible to read these out of an existing chip because they are protected.

You can find a schematic (traced by Coulomb) showing the IGBT drivers, here: http://forums.aeva.asn.au/viewtopic.php?title=pip4048ms-inverter&amp;p=64476&amp;t=4332#p64476

They seem to have got to a point in their development, where there is no single common point of failure, so it is not worth upgrading anything. Just accept that this is a very cheap inverter, about a third of the price of the nearest competitor, so if you have to buy two of them, for every one of the other, you shouldn't complain. However, if it is connected to the grid, Gnome's reason number 2 above may be worth considering, and so it may be worthwhile installing a modular surge suppressor across the grid, in your switchboard. Such as these: https://www.hagerelectro.com.au/e-catalogue/energy-distribution/modular-protection-devices/surge-protection-devices/surge-protection-devices/14161.htm

I suspect both @Javi Martínez and @Coulomb are confused here. There is no way that an Axpert can inject reactive power, or any kind of power, into the grid. The Axpert draws reactive power from the grid, even in battery mode, because it has capacitors connected across the grid. This is perfectly normal. 0.4 A is not surprising. I suspect Coulomb mixed up his true, reactive and apparent powers when he wrote: "Axperts will never admit that their measurements result in more reactive power than real power". First, Axperts don't attempt to display reactive power (VAr) at all, and second, real power (W) is always the smallest of the two numbers that it does display. I suspect he really meant to write "Axperts will never admit that their measurements occasionally show more real power (W) than apparent power (VA)". And neither should they, since it is physically impossible for real power to be greater than apparent power. But this seems, to me, completely irrelevant to either of the two phenomena that Javi reports. The second phenomenon that Javi reports is that the apparent power (VA) number suddenly drops to becomes the same as the true power (W) number, simply by connecting the grid to the Axpert's grid input, with no change to the load, and while still powering the load from the battery, not the grid. This is indeed strange. But I don't believe it has anything to do with the fact that the Axpert draws a little reactive current from the grid, even when in battery mode and not charging from the grid. My PIP-4048MS (equivalent to the 4 kW Axpert MKS 5K-48) continues to show apparent power (VA) greater than true power (W) no matter whether the grid is connected or not, and no matter whether the loads are powered from the battery or the grid. This sounds like a bug in the firmware. Javi, what make and model is your inverter, and what U1 firmware version does it have?

The capacitor is rated at 350 V AC (presumably that's volts RMS at 50 Hz), so it is operating within its voltage rating. Whether it is operating within its high-frequency ripple-current rating is another question.

Kind of. PWM (Pulse Width Modulation) is one part of the process, and filtering is the other. This combination can be used for DC-DC conversion as well as DC-AC inversion. The top of this diagram shows the output of the PWM stage (in the Axpert, a full bridge of 4 IGBTs operating from 400 Vdc), and then at the bottom it shows what it looks like after it has been through a filter (in the Axpert, a series inductor (L) followed by a parallel capacitor (C)). PWM operates at a fixed frequency. I think it's around 30 kHz in the Axperts. So there are around 600 cycles of PWM for every cycle of the 50 Hz sine wave (as opposed to 40 in the diagram), and so the output of the filter is way smoother than that shown in the diagram. Solar charge controller manufacturers have confused people by using the term "PWM" for what are in fact simple on-off controllers. It is the MPPT SCCs that actually use PWM.

Oh sure. It is the C of the LC filter that filters the PWM from the output of the inverter. The associated L is huge and out-of-shot-above in the photo above. It can be seen here. There is a LEM hall-effect current sensor in between the L and the C, also out-of-shot-above.

That big black film capacitor is rated at 20 uF 350 Vac. I wonder if less-complete breakdown of this capacitor could explain some of the Error 51s.

There's also the Voltronic premature-float bug(s) to be considered. They almost certainly exist in the InfiniSolars as they do in the unpatched Axperts. They can cause the inverter to go directly from bulk to float without going via absorb, i.e. without ever going above the float voltage. So @Johandup, I made the assumption that your absorb setting was 52 V, the same as your apparent float setting, but given that you can't read your cutoff setting, I assume you can't read your absorb setting either. I see that you can't read the InfiniSolar settings from the LCD as you can with the Axperts. So maybe your absorb setting is higher than 52 V but it is not getting there because of the premature-float bug.