You make some good points. So far I've been very brief with my sketchy plan, but what I've written is not representative of the thinking I've done....I just have a habit of writing reams of detail and indecisive waffle sometimes, which I'm trying to leave behind.
In your case it is interesting that you had a whole layer of 20 parallel cells fail. I'm unusre what chemistry was involved, usually for a pack voltage of 12V three 3.7V li-ion cells are employed. Four 3.7V cells have a nominal voltage of 14.4V, so perhaps your 12V batteries were LiFePo-type cells, with cell voltages of 3.2V. The chemistry should be the clue. Anyway, it sounds very much like one cell shorted and took the others with it. Kindof unusual for that to happen from what I've read, but not unheard of. A more common incidence is for the failed cell to go open circuit eventually. (again, only going by what I've read, not conclusive lab tests)
So I have blocks of parallel cells, similar in idea to yours. Each layer is going to be made of an equal mix of cells of various capacities, to hopefully produce layers of roughly equal capacity and cells of similar age characteristics.
Instead of a nice flat repeating capacity measurement curve during the first few dozen cycles, I expect to immediately see a curve representing a fall in performance. I already have a battery monitoring device to hand which I'm probably going to install at each 3.7V block level, and adding thermal monitoring is definitely on the cards anyway. One of the inherent benefits to using this particular battery holder design is that it leaves a small air gap around each cell through which I could feasibly blow cooling air, right to the heart of the pack, if required.
The thing which is going to exacerbate ageing and increase the risk of early cell failure is going to be high currents, both during charge and discharge. One of the main reasons for spending weeks testing each individual cell, in order to roughly characterise it, was to weed out the obviously weak cells. I've tried to do that by using a combination of decision emptors, rejecting any cell which reached a temperature of even a few degrees above ambient during charge, and again rejecting any cell which got too hot during discharge. Any cell which refused to complete a charge cycle due to low maximum voltage (even if it didn't get warm) was also rejected. I haven't posted a picture of the 'failed' cells, but most of them have the words 'too hot' written on them.
The theory behind going 'massively parallel' is to spread the current demand over as many cells as possible. I have as a reserve a small pack of lead acid gell cells which I can add in parallel to provide the deeply damaging high current pulses, of the sort which I know the SCR chopper based speed controller is going to produce. But my first tests of the lithium blocks have looked quite promising....but the acquisition of the supercapacitor has got me feeling very comfortable (yet still touching wood, just in case)
The supercap has 18 3000F caps in series. Each cap is individually monitored for over voltage and there is an inbuilt balancing mechanism and alarm circuit (plus pack centre temperature monitoring) , making it pretty much 'fit and forget'. It's going to really ease the load on the lithium cells, so that as they become more prone to failure through normal use they aren't exposed to the high peak currents of the sort which could be just enough to take one into thermal runaway.
http://cgi.ebay.co.uk/ws/eBayISAPI.dll? ... t_82wt_987Anyway, that was why it was so important to me to build a charger/discharger which had identical characteristics for each cell station, because I knew that I wasn't going to read the cell capacity in units of mAh, but in terms of time spent on the discharger before reaching the lower limit. This form of comparative testing is frowned upon by some people because the general belief is that you must use 'correct' units, but ultimately the units don't matter where the aim is to produce a comapratively meaningful result. In this context AFAIC being able to say that one cell ran for 30% longer than another cell is perfectly acceptable...short of resurecting the Duracell bunny test I don't know what else to do !
My whole approach to this project is pragamatic. I am testing this source of lithium cells to see if there would be any point pursuing another similar project, I am building a long dreamt of electric motorbike and have built in the versatillity of design layout to allow me to play some more should either the motor, controller or indeed the cells themselves be inadequate. This isn't an attempt to get it right first time, but to leave the door open for some iteration of the design later on. (All on a very tight (but not safety corner cutting) budget.
Today I priced up a 48V AC drive system, with a controller and 11kw motor, for $900, which is around £600- £700. To my mind, if I find the cells aren't ageing too quickly and work well for the overstretched DC drive system then the next step will be low voltage AC drive. (or even higher voltage AC drive for about the same money). If the cells are letting the design down then I could spend the same £700 on some Kokam or Headway cells and improve the design that way. I can't afford both at once so I have to suck it and see.
I don't feel it is beyond me to try to design a really superb electric motorcycle in one go....it wouldn't be perfect but I have the skills to fix any problems, or know people with the specialist skills to finish it off.....but what I lack are funds, as always, and so what I'm unable to afford I'm having to build for myself.....it's like the beginners EV kit bike.
By the end i'll have done enough to put an electric bike on the road and get it through an MOT, and then afterwards it's all down to funds, but I'll have something for puttering about on at the weekends in the meantime.
Chris