Actually, there are two different BMS units we are testing, one that makes use of the CellLogs, and one that is standalone, and that does not use the CellLogs. Both will use the same basic shunt balancing circuits, and both will use the same charge control section. The difference is where the LVC/HVC signals are generated. The standalone unit will have the same sort of detection logic in with each shunt circuit that all of our previous BMS units have had, and will have an LED on each channel so that you can see when a shunt is operating. The CellLog-based BMS eliminates the LVC/HVC detection logic and the LEDs from the cell circuits, as the CellLog's adjustable LVC and HVC feature provide these functions, as well as individual voltage readouts for each channel to monitor what is going on.
By removing the LVC/HVC and LED logic from the cell circuits, it frees up a lot of room, which allows a full 24 channels of shunt circuits to fit inside one fairly small box. All of this "extra" logic is in the CellLogs, which mount on top of the box.
View attachment 24s CellLog BMS-01.jpg
In order to make use of the CellLogs, there are a couple of things that need to be added. First is an optocoupler for the CellLog "Alarm" outputs, which are not isolated. This allows multiple Alarm outputs to be combined into a single combined LVC/HVC signal "bus", just like we do with the standalone BMS design. Another function we add is a relay to cut the ground connection (pin 0...) so that the CellLogs can be turned off, when not in use. This is done automatically. The CellLogs are turned on whenever either the motor controller is on, or when a charger/supply is connected. There is now also an "override" switch that lets the CellLogs be turned on anytime, for a quick cell check, for instance. Finally, there are some diodes and a single resistor that basically make sure that the standby and operating current drain is completely even, across all cells, regardless of how many cells are used in each pack (i.e. - 5s, 6s, 7s or 8s...). Here's the schematic for the CellLog-related logic we've added:
View attachment 24s CellLog LVC-HVC-Relay Logic-v4.1.7a.png
For the new CellLog BMS, I put all this CellLog-related circuitry on the underside of the custom lid for the extruded aluminum Hammond box. The traces, etc., on the top side re hidden by the CellLogs themselves.
View attachment 24s CellLog BMS-03.jpg
View attachment 1
This initial version of the CellLog BMS has four 6s shunt/balancer sections, divided into two 12-channel boards. These are connected to the four 6s groups via four sets of seven 18-gauge wires. The CellLogs, however, have 8-channels each, so there is a 4x6-to-3x8 adapter built into the lid that remaps the four 6s connections into the three 8-channel CellLogs.
After we get this initial 24-channel version going, I plan on also doing a shorter, two-CellLog version that has two 8-channel shunt/balancer boards, in a shorter (4.7" vs. 6.3"...) version of the same box. This will be used with 12s and 16s setups.
The standalone BMS format will be similar to the existing v2.x-style boards, with the control section on the left, and the shunt/cell circuits on the right. These will be done in two versions, one with three 8-channel sections, and one with four 6-channel sections.
As I said, both BMS variants will share the same charge control logic. Basically, what the charge control circuit does now is simpler than some of the v4.x variants we've been trying for some time now. The basic function it performs is to monitor the charge current and to cutoff it off when it drops below an adjustable set point, which has an adjustment range of 0-2.5A. The charge control logic also monitors the HVC signal and temporarily cut the charge current when it trips. This is a lot different than the problematic PWM-based "throttling" logic that we used previously. The reason is there is a big difference in how we use the HVC signal now, due to a change in the whole shunt balancing scheme. Previously, what we did was fix the HVC voltage at a point the the shunt circuits were fully on, bypassing 1A+ of current. The HVC signal line was then used to modulate the PWM duty cycle to throttle back the charge current in order to keep the cell voltages from trying to go over the point that the shunts couldn't handle. Without this throttling, if the imbalances are large enough, the shunts will swamp/overload, and the cell voltage will keep rising. This sort of managed shunt control is the basis of virtually all existing shunt-based BMS designs I'm aware of, but has the disadvantage of having all the shunts cooking away at full bypass at the end of charge. The idea is that you set the charge voltage a bit above the sum of the HVC set point voltages, and let the throttling logic keep the voltages in check. While the low cells are catching up, the shunts on the high cells, that are full already, are in full bypass. Once the low cells catch up, their shunts will also be in full bypass so in the end, there is no more current going into the cells and all shunts are blasting away. On some of the less-capable BMS designs, the shunts may only be able to bypass 200-300mA of current, so the heat generated is manageable. With our 4.x designs, however, the shunts are capable of bypassing 1A+, which means there's quite a bit of heat to deal with at the end, so big fans are mandatory.
One other problem with this approach is that it is impossible to tell for sure when the last low cell is full. You can't just monitor the current drop, because it will not go below the max shunt bypass current level (i.e. -- 1A...). When the current first gets down to that level, all of the current will still be going into the cell. It will take some time, but eventually the shunt will be bypassing the full 1A, and no more is going into the cell. The problem is knowing when that happens. In our early 4.x designs, we waited for the current to get down to this level, and then we started a timer, to let the low cell(s) have enough time to get full, before shutting things down. This timer was adjustable from 30 seconds up to about 4 hours. Again, this worked, but I was never happy with this approach, especially having the shunts cook away needlessly.
The new shunt/balancing scheme, which we will use on both the CellLog and standalone BMS variants, on the surface looks like a subtle change. Basically, all we did was lower the charge voltage to be right at the point the shunts start to come on, instead of at a point above where they are in full operation. Now, if the cells are in perfect balance, the cells will all hit this set point at the same time. This will cause the charger/supply's CV mode to kick in, and the current will start to drop. In this scenario, the shunts don't come on at all. With the case where cells are imbalanced, what will happen is that the high cells will have their respective shunt circuits come on, but only enough to keep the voltage "clamped" at the charge/balance point. When all the cells catch up, the high cell shunts don't have to work as hard, and eventually as the cells all become balanced, the shunts will go off completely. Now, the current can be monitored and when it drops below a preset point, we know that all the cells are as full as they are going to get, so the charge process can be shut down.
We now also do something different with the HVC signal. It is still used to keep any one cell's voltage from going over a preset limit, but how that is done is different now. Before, the charge voltage was set higher than the desired balance point, so the shunts have to be held right at the point they start "swamping". With the charge voltage set lower, it takes quite a bit of imbalance before a shunt will be overloaded. For example, in one of Andy's recent tests, the shunts were set to come on at 3.60V, but the charge voltage was set much higher, at 3.73V. To make matters worse, this was a real "worst case" setup, using a "tired" TS 40Ah 21-cell pack that has one cell that takes a lot linger to get full than the rest. He let the charge process go for some time, with no throttling/control at all, and the shunt circuits themselves were still able to keep the 20 other high cells from going over 3.93V, or about 200mV above the charge point. The shunts got hotter than hell, with no fans, and he ended up melting some plastic on his scooter, but the cells were still protected.
Anyway, what we will do now is set the HVC point higher than the charge point, like maybe 50-100mV higher. For healthy packs, even with pretty significant imbalances, this higher HVC trip point won't ever be tripped. If it does occur, however, what will happen is that the charge current will be cutoff for about 15 seconds, but the shunts for the high cells will continue to operate, pulling these cells down closer to the low cell. The charge current will then come back on and the low cell will resume its catchup. For really screwy imbalances, or with packs that have "tired" cells, like Andy's, this process might repeat a few times, before the cells are close enough that the shunts can keep things in check on their own, but for the vast majority of setups, this "failsafe" mode won't ever be engaged.
All the above applies for both LiPo and LiFePO4 setups, and as I said, this "shunt-by-exception" scheme is now what we will use for both the CellLog and standalone BMS variants.
-- Gary