wrobinson0413 wrote:For short circuit, I was thinking of the protection for all cases not just phase to phase. After thinking about it, I think that phase to phase is the minimum that you would need since most problems occur with phase wires shorting during spin outs. The other two areas where you could have potential failures are phase to rail shorts, but since the bus power is on a separate set of connectors, there is very little possibility of those cases occuring during run time. If someone wires up the battery input to a phase input accidentally, your startup diagnostics should see that condition and not enable the bridge.
If they put the battery input on one of the phase wires, then the controller will detect an under-voltage condition and not activate the FETs. As for the phase-to phase protection, I am already using phase current sense, so all that is needed is a very fast response time. I am not quite sure how fast this reaction time would need to be. Anyways, the main positive fuse would blow in such an event and provide backup protection if the current limit doesn't kick in.
Reverse voltage protection is done using mosfets on the negative return to the battery. A diode is not practical because the power dissipation is too high for large battery currents. If you wanted to make a bullet proof design, you would want out of saturation detectors on each mosfet bank as well as the reverse voltage protection. Too bad the IR2125 is so much money.
What would pull the FETs out of saturation?
As for the reverse battery protection, a fuse would be enough to provide simple protection, don't you think?
Waterproofing is a very tricky problem. No case will remain waterproof indefinitely when you need to have connector openings or cabling exits. Soft potting is one of the more secure ways of doing it. When I mean soft, it is like cured silicone but easier to remove. If the box is not hermetically sealed, then you will have condensation forming inside over time, which will eventually lead to a failure.
Any links or names of easily available good soft potting stuff? This will be a very important aspect for me, specially because of the particular FET mounting I have.
Datalogging is not as difficult as you might think. The idea is to have a sample window allocated in ram that you fill on a trigger, then blast the data back to your GUI. National instruments has very nice sope tools that allow you to display your data in a nice format. The only issue with that method is that you need to have a laptop connected during your testing phase. For longer time period datalogging of slow signals, you could add a large serial EEPROM to store items like RPM, voltage, current, etc, locally.
I don't see the first type of datalogging you describe (using the PC user interface) as complex as the long-term storage solution using EEPROM or SD cards. I've never used external EEPROMs before, but I'll have to learn. And if I find that the SD card solution would be easy enough to use, imagine how much we could log with 1 gig of flash?
Some of the BMS features are nice to add to your controller, but they add cost. The basic stuff that is free like LVC and OVC shutdown limits is easy to add to your diagnostics. The rest of the BMS functionality is more difficult to add in a cost effective way. One key functionality of a BMS for LiFePO4 would be the individual cell LVC shutdown command. If you had that local to the battery, then the rest of the run-time functionality could be incorporated in your controller. I think all you would need is a reliable measurement of the battery current to complete the protection of the battery. You would still need a ballancing method for charging your battery pack. If you were using SLA, then all the battery management could live on your controller.
What I meant when I was talking about linking the BMS to the controller, was more like just taking an existing BMS design like the one made by the guys here on ES, and just mod it's assembly to suit the special controller. The simplist thing to do, and most important is just sending the BMS's cell-level Over Voltage Cutoff signal to the controller. This combined with the controller's current limiting behavior would make regen always a safe experience for the battery. If the Under Voltage Cutoff signal would also be routed to the controller, then the BMS's discharge cutoff FETs become useless, and could be possibly removed to reduce system cost and losses.
Have you started to tally up your cost on the material for your controller yet? I imagine that you may get a bit of a surprise when you finish your tally. I have gone through a small costing exercise myself, and I wondered at just how much money the chinese actually make per unit on their higher end controllers.
I have checked individual prices out quite a bit when comparing different possible parts to be used, but I've only counted up the whole total as an estimate. As for the cheap chinese offerings issue, I guess I'm hoping to start off by offering something quite different, mostly for the higher-end savvy ebikers such as we find here on ES. I am hoping also to do a bit like Justin does with his Cycle Analist and build the controller here in North America, altough I know that avoiding using cheap labour makes costs go up. Anyways, there are still too many unknowns for me to have a good idea of total costs just yet. To name them: PCB costs, assembly costs, case, connectors, waterproofing and finishing. I'm thinking something like a bit under 100$/unit probably in low quantities, and I hope less with higher volume and more optimized assembly methods later on.
Thanks again for your thoughts, Wayne!
All my ebikes have gone up in flames (with my whole house) --> Many lessons learned (like that insurance companies can be cold hearted a$$holes; and to read and understand your contract before your house burns down...)