Adam, How has your pack building come along? Have you done anything with those 125 cells?
My pack construction is moving along slowly. Life has its wonderful little interruptions like work sending you out of town for a week at a time and the occasional vacation that keeps the girlfriend happy. I've gotten more time to work on the pack some more in the last two weeks. I guess it's time for me to contribute to the 38120S pack building thread.
Time to write a novel!
At this point I've inventoried my cells and all of their initial voltages in a spread sheet. After that each cell got balanced-charged in a 5p block using a voltphreaks charger. I built two charging harnesses for balance-charing the cells. Right now all of my cells are fully charged and seem to be holding charges in the 3.4x voltage range. Even the one cell out of all 125 that arrived at a dangerous 1.77 volts seems to be holding a charge well after some time on a VP charger. I consider a 1/125 weak cell “rate†of less than 1% to be a very acceptable failure rate. I've specially labeled that weak cell and use it as my guinea pig cell for experiments.
My end goal for my first pack is a ((5p) x 24s) configuration. That's nominally a 76.8 volt 50 amp-hour battery. It will use 120 of my 125 cell inventory since I budgeted for failures and goofs during construction in my original purchase. I usually call it a 72v battery pack to keep things in the same general Pb-A terms that on-road EV conversion folks are familiar with. I've already built my Goodrum/Fechter BMS kits for managing the cells. I have two of them built – partially for redundancy and partially because I plan to expand to a 144v system later. That's all I will say about BMS systems in this thread
Here is a screen cap of an Auto-Cad drawing representing the full 144v pack that I will eventually get to. (AutoCad services courtesy of engineer and Houston EV enthusiast Bill S.). (attachment 1). The packaging represents a battery box that I will cut-out of my VW's trunk and drop in where the spare tire well used to be.
View attachment pack144v-hwc38120s.png
The initial pack will be broken up into 6 segments of ~12 volts each. Each segment is (5p) x 4s. That's half of what is shown in the auto cad drawing. There will be fuses and bus bars across the ends of each segment to protect it and to connect it to the other segments and to the leads to the contact-ors and to the controller.
I've been working with a few “real engineers†and getting their advice on my pack. I say “real engineers†because some people make jokes about computer scientists like me not being engineers
. I just like to pretend I'm a EE on the weekends. Much of the advice I've gotten has been good, some contentious as I mention below.
The plan for the pack is to build the cells in “sheets†or “cards†wired cell-to-cell in a way very similar to how “otanet†(JD) laid out the pack in his VW van conversion. My pack will differ from JD's in a couple of ways. JD used an (8p) x 30s configuration. Like JD's layout, I am using M6 set-screws to mate the positive end of one cell with the negative end of the other. Using set screws makes a very solid connection between two cells that I think will withstand road vibration nicely.
JD's pack uses bus-bars across each of of the 8p cells. Mine will not. Here's why: I've had one electrical engineer tell me that this design is a bad idea for a couple of reasons: (a) under-load cross currents and (b) cell-failures to closed (short) circuits. A couple of other engineers agree with reason (b) and disagree with reason (a). (I mean not to insult JD's layout here, of course, as his design is what is inspiring mine!)
The problem with (b) is obvious – using a huge bus bar to connect cells in parallel will lead to melt-down in the event that one fails to a short circuit. Albeit a rare circumstance, in this situation the other N-1 cells in the parallel group will dump all of their current through the shorted cell until either the cell melts and fuses open, the currents melt the bus bar or other cells, a fire starts, or all of the above. It gets even more dire when you realize that the previous and following parallel groups ahead of and behind the group in the series pack will drive even more current through the failed cell.
While I think this failure scenario is unlikely I still want to account for it in the pack. This can be mitigated by not using bus bars to parallel the cells in a parallel group but to use a smaller gauge conductor that will fuse if serious amounts of current are thrown down it. Using individual “per column of cell†fuses every 4 cells (~12volts of cells) is another step in mitigating this risk.
If a cell fails to short the low-gauge balancing wires will short like a fuse. This fixes the immediate neighbors issue when a cell shorts. In my pack I am using a a 0.75†x 0.6†x 0.020†copper tab in between each cell that is connected in series. In a 5p group these copper tabs have a relatively low-gauge balancing wire that connects them. Here is one of my in-progress diagrams of how the tabs mount to the end of the cells along the set screw. (attachment 2).
View attachment tapWire-38120.png
I plan to use 5 x 100 amp fuses to interconnect the output from one segment's cells to the bus bar at the input of the next. Fuses at the end of the ~12v segments will handle the current from the previous and next group of cells if a cells fails to short.
The contention among the engineers I have spoken with has to do with issue (a) – the cross currents under load. One engineer I have spoken with strongly believes that cross currents for LiFePO4 cells among parallel cells when under load is an extremely bad thing to have in a pack and that it will damage if not destroy the pack entirely. He suggests that I use a special small gauge wire as the balancing lead between the tabs that increases its resistance as more current flows between the cells under load.
I disagree with this engineer's suggestion. I have a feeling that if LiFePO4 cells had problems being massively paralleled with a large low-resistance bus-bar between paralleled cells then I would have heard about it on the ES forums by now. Does anyone here have an opinion about under-load cross currents in paralleled cells? (Experts/EE's opinoins welcomed – Doc/Fechter/Gary/anyone?
).
I originally planned to use single-strand 24 awg Bell Wire to connect my balancing tabs in the 5p groups in order to meet the low fusing current requirements suggested by the guy who was fearful of the cross currents. I have changed that plan to use 16AWG stranded wire to connect the balancing tabs. Another engineer suggested that stranded cable worked better versus single-strand under automotive vibration circumstances and I agree with that assessment. That and if a cell fails to short there will still be ample current to fuse 16AWG wire and I would prefer there to be as little resistance between the cells as possible when charging and discharging.
Here's a picture of the balancing tabs cut from the copper strip. (attachment 3)
View attachment balancingTabs.png
I've got more to say about packaging and mounting the cells on a back-plane of metal but I'll save those for another post. I'm interested in checking out both the headway and PSI-lego end caps for being end-caps to these cells. Both of those look like they are preferable to other design options that are candidates for how to lay out the cells in an EV car's battery box. The decision on which kind of packaging brick to use effects what length the wires between the balancing tabs have to be, so I'm currently stuck there, experimenting with what end-block design works best. Anyone who has a good source for affordable 0.75" x 0.25" copper bus bar I would sure love to give them some business!
Cheers,
--Adam E. Hampton
). I assumed that the ecity BMS (looks like the exact same one KAE is selling) would start doing PWM on the charging FET to limit current once one of the channels reached the shunt voltage threshold.
