Okay, time for an update.
Richard and I have been working on a major new revision to the BMS, which adds several new functions and features, is more modular and expandable, can support shunt currents as high as 2A with active cooling, has an improved PWM-based current control and has an auto power on/off power detection scheme that eliminates the need for a "3rd wire" to power the control section. The new layout has multiple completely independent 3" x 4" 6s sections that are now stackable. The new control section is on one more 3" x 4" board that goes on the top of the stack. The cell circuits are laid out better and use fewer parts per channel. The number of connections between the 6s sections and the control board have been reduced to just two, by having the LVC and HVC detection functions share the same opto outputs. This makes total sense as these functions are never used at the same time. The LVC is used during discharge and the HVC during charging/balancing. This also eliminates the need for the resistor/diode network that was used to generate the ALL SHUNTS ACTIVE signal that was used in the old auto power off logic. The new control logic uses current sensing to detect when the cells are full. Finally, this new version supports both LiFePO4 and LiPo-based setups with two part changes per channel. For existing LiPo setups that already have separate LVC circuits, this new configuration can also be used as a standalone balancer but one that can be left permanently installed on the bike and left connected to the pack. Without the charger/supply connected, there is absolutely zero current drain in the control section.
Although the new PCB has one control board and three 6s sections, organized in the familiar "tearoff" configuration, you will be able to order the control section and however many 6s sections that are required. The 2-wire opto lines are left connected, which allows flexibility in how the overall BMS is physically laid out. They can all be in one stack, or all laid side-by-side, as with the current BMS PCB, or a combination of both. When stacked, the opto lines are connected between boards with two board-to-board headers. The spacing between boards is 15mm. Here's what the new control and 6s sections look like:
- 18s LiPO BMS-v3.9d-02.png (162.91 KiB) Viewed 1063 times
As you can see, the connections all exit one end of the boards and the LEDs are all at the other end of the board. The LEDs can be installedat a 90-degree angle so that they are at the edge of the board, facing out. There has been a lot more attention paid to thermal considerations this time, with onboard pads/heatsinks with thermal vias between layers. The traces for the shunt transistors and the shunt resistors have been beefed up to support higher shunt currents. Also, in addition to having a flat mounting for the "standard" BD136 TO-226 packages, mounts are also provided to use a larger TO-220 - based TIP105 darlington power transistor in place of the BD136. Also, to make assembly a bit less tedious, the parts are all oriented in the same direction, wherever possible. The parts used are pretty much the same as before, just fewer of them. There is one difference. This version needs to use the non-complementary open drain version of the TC54, which has the designation TC54V
N-xxx, vs. TC54V
C-xxx. This saves having to add a diode to isolate the HVC and LVC functions.
The new control section is most of the changes have been made, and new features have been added. The first is a replacement for the old "throttling" logic. The new circuit now uses the HVC signal to control the duty cycle for a PWM circuit, much like a controller does. This sounds like a subtle difference, but what it does is allows the charge current to be higher, for a longer period, before the "throttling" of the current starts, as the cells get close to being full. The two-color red-green LED remains (the one on lower right...), which starts out red during the CC phase, and starts to transition to green as the PWM duty cycle is reduced. It will gradually turn from red to orange, then yellowish, light green and finally fully green, when the cells are full.
There are now provisions for two FETs to be used to control the charge current. This is primarily to be able to support higher voltage EV setups. Normally, the FET just sees the difference between the pack voltage and the charge voltage, so a single 100V IRFB4110 works for most setups. A higher voltage FET, like the 150V IRFB4115 can be used, but these have a higher on resistance, so provisions for adding a second one in parallel are provided. This should support much larger configurations with pack voltages up to 400V, or so.
There is now also a spot for another FET, which is used to control a fan, or multiple fans in series. The fan(s) are connected to the main pack/charger positive connection so that we can keep the current down in the 12V control section to logic levels. Otherwise, we would need to beef up the 12V regulator.
To eliminate the need to have a "3rd wire", to keep the control logic off until the charger is connected, Richard has come up with a clever SCR-based scheme to not turn on the 12V regulator until suddenly there/s at least a volt difference between the charger voltage and the pack voltage. Once on, the SCR keeps it latched on. There are two conditions that will cut off the power to the control section. The first is if the current suddenly drops down close to zero, which would happen if the charger/supply was disconnected. The second condition is via the new end-of-charge detection logic. A current sensing circuit is used to monitor the charge current. When the current gets down to the level that the shunts can bypass, that means at least one cell is full. There is now a switch-selectable option that will either stop the charging process at this point, or will continue on to balance the cells but for a selectable period of time. This balance time can be for 15 minutes, 1 hour, 2 hours or 4 hours, selectable via a jumper block. This range of balance time elections can be adjusted up or down by changing the value of one resistor. At the end of the time period, the charge current is cutoff. There is another LED (on the bottom left...) that will initially be red, during the basic charge phase, and then "blink" changing between red and orange while the balancing mode is on and the timer is operating. So, during the initial CC charge mode both of the LEDs on the control board will be red, and the cell circuit LEDs will be off. Once the shunts start operating, the LEDs for those channels will come on (orange...). As the cells get fuller, and the PWM logic kicks in, the control board LED on the right will start transitioning towards green, while the one on the left remains red. Finally, if the balancing mode is enabled, the LED on the left will start blinking orange. When complete, the whole system shuts down, and all the LEDs go off. Power to everything remains off, until the charger/supply power is recycled.
The "normal", non-balancing mode can be used by those concerned with "top-balancing" issues that have been discussed in detail, here on E-S and elsewhere. I do not want to discuss this issue in this thread, so please don't. Use the other threads. What I will say here is that Richard and I believe that if balancing is not to be done, the way we have it here is a better solution than simply letting the high cell get to a much higher HVC point, and then simply cutting power. In some of the schemes I've seen the HVC set point is at something like 3.85V, while the charge voltage is set lower, to like 3.60V per cell. The theory is that the lowest capacity cell is allowed to go higher than the normal CV point, until it hits 3.85V and then the charge process is stopped. I have a couple of problems with this approach. Forst of all, what I've found is that the higher the charge current is, the greater the reverse voltage sag there is going to be between the cells resting voltage, and the voltage the detection logic will see. What this means is that the higher the charge current the sooner the cell is going to be pushed up to 3.85V, so if you stop at that point, the cells are going to end up not getting as full a charge. Another concern I have is that although the LiFePO4 cell manufacturers, like Thunder Sky, might tell you it is okay to repeatedly let cells hit 3.85V per cell, the jury is still out, IMO, on what this will eventually do to the longevity of the cells. To make matters worse, the only cells that will be allowed to overvolt will be the weaker, less capacity cells, so my gut tells me this technique will further hasten their eventual demise. Also, for LiPo-based setups, with the more volitole Litium-Cobalt-based chemistries, this sort of planned overcharging of cells, weaker, or otherwise, is simply not an option. Your overvolt these cells and you are going to end up with an early 4th of July.
What our design does differently is to not
ever let any one cell go over a safe point, which is set just above the per cell CV set point, byt using PWM control on the charge current. The shunt turn-on is set just below this HVC set point. When the charge current drops to the amount the shunts can bypass, like 1A for instance, at least the low capacity cell will be completely charged to whatever the desired level is set to, so in the above example 3.60V. If you don't want to balance the rest of the cells, that's fine, the "normal" mode setting will stop the charge process, but at least the first/lowest cell has been safely charged full. If balancing is allowed to continue, the full cells don't get "cooked" by trying to stuff more current into them, as it has been suggested (
). It doesn't work that way. Full cells simply can't absorb any more current at all, unless the voltage is allowed to keep climbing. Once full, that's it, the shunt circuit absorbs all the current. Eventually, all the cells are at the point they can't accept any more current and the shunts are bypassing it all.
One other point. You can't just assume that the first cell that hits the cutoff every time is the lowest capacity cell. Ths doesn't take in the possibility that cells still have the same capacities, but just happen to be out-of-balance, or at different states of charge. Whether at the "top", or the "bottom", at some point the cells need to be balanced. Richard and I both feel strongly that so-called bottom balancing is dangerous, and just not all that practical. Balancing at the top, but without letting cells overvolt, is a much better solution, in our opinion.
Okay, soapbox mode off...
Although the cell circuits and the new PWM logic have been tested, and the new auto power detection circuit has been breadboarded, we still need to test the whole system together, using the new boards. There's still a few resistor and capacitor values that need tweaking, but that shouldn't take long. I'm hoping we can get that effort down by Christmas, or the end of next week. I will the make at least the PCBs and BOMs available shortly after that. As before, Andy Hecker will also provide fully assembled and tested units, for those that are soldering skill-challenged.
This version should be a lot easier to assemble though, and over all, uses fewer parts. Because of a larger order for customer doing a new electric motorcycle, we are finally going to be able to do a surface mount version that can be automatically assembled. These we will make available sometime in the 1st quarter of 2010.
Finally, there will also be an option to order the control section alone, for those who want to upgrade their exiting v2.x BMS units, with the new control features. All it will take is a two wire connection to the existing BMS boards.
When we get some more info, and I have something more to report, we will retire this thread, like we did after the last major revision/update, and start a new thread for this one, which will have the initial version designation of v4.0a.
-- Gary
26" Townie: Crystalyte 5304; 18-FET (4110) 100V/100A; 24s3p 88V/15Ah Turnigy 20C LiPo
26" Townie: Crystalyte 5304; 7240v2; 24s 72V/10Ah PSI
20" Dahon Mariner: 9x7 9C; 18-FET 100V/100A 18s2p 67V/10Ah Zippy 25C LiPo
20" West Marine Port Runner: AF 3220-7t; HV110; S-A 3-sp hub; 12s3p 44V/15Ah Turnigy 20C LiPo
....
