New 16-cell Battery Management System (BMS)
- Thread starter GGoodrum
- Start date
GGoodrum
1 MW
Interesting... 
Now, how about going the other way? What would the circuit look like if I wanted to do a separate protection board that could be used with existing LVC boards. I have a bunch of the original ebrake-only LVC boards that are on some of my early a123 packs that I would like to add active cutoff to. A couple of these I'm letting my brother-in-law use, and he is certainly capable of the type of bonehead moves I spoke of, like leaving the controller on (he's done it already...)
-- Gary
Now, how about going the other way? What would the circuit look like if I wanted to do a separate protection board that could be used with existing LVC boards. I have a bunch of the original ebrake-only LVC boards that are on some of my early a123 packs that I would like to add active cutoff to. A couple of these I'm letting my brother-in-law use, and he is certainly capable of the type of bonehead moves I spoke of, like leaving the controller on (he's done it already...)
-- Gary
A comment about the high power dissipation in the cell balancing clamps brought up another benefit of the revised charge controller I posted.
In the original controller the situation arose that during the Constant voltage charge phase when the charging current was being regulated to 1 amp ( for arguments sake) that all the cell clamps were on and dissipating power. In the worst case when the battery was fully charged all the current was being shunted by the clamps and each clamp was dissipating 3.65 watts (1 amp at 3.65v cell voltage). For 16 cells that about 65 watts. added to that the heat dissipated by the charge controller FET produces a lot of heat not easily gotten rid of. At higher charge currents there would be even more heat, 2amps would be up to 135 watts.
In the revised controller the charging current is reduced so that at least one cell clamp is almost completely off dissipating little heat. If the other cells are fairly close to this their clamps will also be shunting little current and producing little heat. Since during this process the cells are being balanced they will all tend to be fairly close and the overall heat dissipated should be a great deal less than with the original system. This may make it possible to increase the Max charge current setting to allow faster charging without running into overheating problems with the system. Only if one cell is substantially out of balance, or if a number of cells have substantially reduced capacity from others in the pack will you start to dissipate more heat.
In the original controller the situation arose that during the Constant voltage charge phase when the charging current was being regulated to 1 amp ( for arguments sake) that all the cell clamps were on and dissipating power. In the worst case when the battery was fully charged all the current was being shunted by the clamps and each clamp was dissipating 3.65 watts (1 amp at 3.65v cell voltage). For 16 cells that about 65 watts. added to that the heat dissipated by the charge controller FET produces a lot of heat not easily gotten rid of. At higher charge currents there would be even more heat, 2amps would be up to 135 watts.
In the revised controller the charging current is reduced so that at least one cell clamp is almost completely off dissipating little heat. If the other cells are fairly close to this their clamps will also be shunting little current and producing little heat. Since during this process the cells are being balanced they will all tend to be fairly close and the overall heat dissipated should be a great deal less than with the original system. This may make it possible to increase the Max charge current setting to allow faster charging without running into overheating problems with the system. Only if one cell is substantially out of balance, or if a number of cells have substantially reduced capacity from others in the pack will you start to dissipate more heat.
GGoodrum said:Interesting...
Now, how about going the other way? What would the circuit look like if I wanted to do a separate protection board that could be used with existing LVC boards. I have a bunch of the original ebrake-only LVC boards that are on some of my early a123 packs that I would like to add active cutoff to. A couple of these I'm letting my brother-in-law use, and he is certainly capable of the type of bonehead moves I spoke of, like leaving the controller on (he's done it already...)
-- Gary
I'm not sure exactly what you are asking. I also don't have a schematic of the original ebrake only LVC board you are referring to. Can you post that or direct me to it?
GGoodrum
1 MW
It is the same as what is in the BMS. Here is the schematic for the original 10-cell version:
All the opto outputs are simply ganged together. I would simply like to run two wires from one of these boards to the cutoff board.
Thanks -- Gary
All the opto outputs are simply ganged together. I would simply like to run two wires from one of these boards to the cutoff board.
Thanks -- Gary
Gary, I think that should be fairly easy.
I also like the "modular approach" thinking. Just get the modules you need.
The trick is to make the FET switch fast enough that it doesn't overdissipate and limit the gate voltage to around 10v. I'd be tempted to use a CMOS inverter chip to get a fast risetime, but there is probably a cheaper way to do it with just a couple of transistors. I'm not sure what the power consumption of an inverter chip is.
I like Randomly's Vbe multiplier clamp; I've never seen that arrangement before. You could use a zener diode too.
OK Randomly, I think I follow your LVC circuit now.
Any reason Q12-17 couldn't just be diodes?
On the charging side, I agree that the total dissipation needs to be limited.
I was thinking if the voltage detector activated a resistor shunt that would limit the shunt current to a reasonable heat level (~200ma), then you would drop the charging current whenever any shunt became fully activated (which you could detect my measuring the voltage across the shunt resistor).
When the cells are charging, you can pound them with big amps until one of the shunts becomes fully on, at which point the charge current is reduced to prevent any shunt from being swamped.
This way the charging current is maximized until the first shunt gets fully on.
From this point, continued charging will serve to equalize the cells. If the cells are well balanced, the balancing phase will be quite short. If the cells are out of balance, it will take longer.
It would be nice if the charge current limiting was not dissipative (buck regulator?) but that would probably be too expensive. Another possible cruder approach would be to use a FET switch with a resistor across it. When the FET is on, full charging current and little dissipation happens. When the switch opens, all the current goes through the resistor. The resistor is chosen to limit the current to a level slightly lower than the shunt current. The switch would be either full on or off to minimize dissipation on the FET. In operation, I would expect the switch to oscillate on and off through part of the charging cycle.
I still think it is good to avoid needing heat sinks on the transistors. I'm also thinking eventually we would want the whole thing made from surface mount parts.
I also like the "modular approach" thinking. Just get the modules you need.
The trick is to make the FET switch fast enough that it doesn't overdissipate and limit the gate voltage to around 10v. I'd be tempted to use a CMOS inverter chip to get a fast risetime, but there is probably a cheaper way to do it with just a couple of transistors. I'm not sure what the power consumption of an inverter chip is.
I like Randomly's Vbe multiplier clamp; I've never seen that arrangement before. You could use a zener diode too.
OK Randomly, I think I follow your LVC circuit now.
Any reason Q12-17 couldn't just be diodes?
On the charging side, I agree that the total dissipation needs to be limited.
I was thinking if the voltage detector activated a resistor shunt that would limit the shunt current to a reasonable heat level (~200ma), then you would drop the charging current whenever any shunt became fully activated (which you could detect my measuring the voltage across the shunt resistor).
When the cells are charging, you can pound them with big amps until one of the shunts becomes fully on, at which point the charge current is reduced to prevent any shunt from being swamped.
This way the charging current is maximized until the first shunt gets fully on.
From this point, continued charging will serve to equalize the cells. If the cells are well balanced, the balancing phase will be quite short. If the cells are out of balance, it will take longer.
It would be nice if the charge current limiting was not dissipative (buck regulator?) but that would probably be too expensive. Another possible cruder approach would be to use a FET switch with a resistor across it. When the FET is on, full charging current and little dissipation happens. When the switch opens, all the current goes through the resistor. The resistor is chosen to limit the current to a level slightly lower than the shunt current. The switch would be either full on or off to minimize dissipation on the FET. In operation, I would expect the switch to oscillate on and off through part of the charging cycle.
I still think it is good to avoid needing heat sinks on the transistors. I'm also thinking eventually we would want the whole thing made from surface mount parts.
jeffkay
100 W
- Joined
- Sep 14, 2007
- Messages
- 115
Can a TC54 device (trimmed to a higher voltage) be used for the top-of-charge detection and activation? Would it be high right away if it never went above the detection threshold, or does it need to see the higher voltage before it can work? If so, could it be "primed" by a switch or something. I am asking this because in Bob/Gary circuit they use the different detector for high...
Jeff K.
Jeff K.
disadvantage
100 W
- Joined
- May 23, 2007
- Messages
- 126
jeffkay said:
I think I can answer this question.
The TC54 is used to detect discharged cells. It pulls its output low when the voltage on the input falls too low.
The LM431 is used to detect fully charged cells. It pulls its cathode low when the voltage on the ref pin rises too high.
I think you meant using an LM431 to detect end-of-charge. From the datasheet, the LM431 can be used in a voltage divider up to 36V.
GGoodrum
1 MW
disadvantage said:
Actually, the LM431 is basically an adjustable zener diode. It has an internal 2.5V reference, so a voltage divider is used to scale 3.65V down to 2.5V. The 100k pot is used to adjust this voltage. I think Bob said the adjustment range is somewhere between about 3.0 and 4.0V, with the parts used. Anyway, once the voltage reaches this point, the LM431 starts conducting, which turns on the TIP105 darlington pair power transistor. This combination will absorb what ever current the cell can't take in.
The TC54 voltage detector comes in various voltages, from 2.5V to 6.0V, but most places like Mouser only stock a few. As I remember, they have 2.1V, 2.7V, 3.3V and 4.0V, but I could be wrong. I know they have the 2.1V and 2.7V, because I have those.
In any case, you can use either of these with a voltage divider to set pretty much any input voltage. As for using it for top-of-charge detection, this won't really work. What needs to happen is to have the voltage held at 3.65V, and let the cell taper off the current it absorbs. If you just shut things off when the cell hits that voltage, the cell would only be charged to about the 85% level.What the existing BMS design does is turn on an opto-coupler, once the LM431 starts to conduct. All of these opto outputs are logically "OR" together, via a diode matrix, to create the ANY SHUNT LOW signal, which goes low if any of the shunts are operating. These same opto outputs are also logically "AND"-ed together, via a 2nd diode matrix, to create the ALL SHUNTS LOW signal, which goes low when all the shunts are on. What Randomly's mod does, is to keep cutting off the current, once one of the cells is at the cutoff point, which, if I understand it right, causes the cell's voltage to drop below the cutoff, which allows the current to conduct again. The current keeps dropping as the cell keeps getting fuller, to the point that the current required to keep it at the cutoff goes below some adjustable threshold, and everything shuts off. What I haven't quite got my head around is how this works with multiple cells, and what happens if one, or more, are out-of-balance with the rest, by some significant amount (i.e. -- >.1V, for instance...).
Richard, the existing circuit will let the charger "blast through", until the ANY SHUNT LOW signal first goes low. Then the current limiter kicks in. What I think Randomly needs to add to his mod is the temp controlled "throttling" back of the current. Bob added two thermistors to the original design, which will keep the heatsinks from getting over 150F.
-- Gary
Yes you could also put a Zener from the collector of Q8 to ground, but I had to have the transistor anyway so I just added R9 instead of a zener to clamp the voltage. The turn on time maybe slightly faster with a zener but switching times here are completely unimportant.fechter said:
What do you mean? they ARE diodesfechter said:
yes you could replace them with diodes, BUT you'd have to find some low leakage diodes to do the job. The CB junction of most transistors has much lower leakage than most standard diodes, and the EB junction is even lower leakage (but has a low breakdown voltage). In this case I'm using the EB junction since the voltages are low enough, and I'm connect the Collector to the base to get less voltage drop. Note that switching diodes like the 1N4148 are gold doped to make them turn off very fast, however it gives them substantial leakage. If you look up the data sheet it's specced at 25nA at room temp, but goes up to 50uA! at 150C. A great many diodes are specced at 25na leakage at room temp but this is because the automated test equipment in production can't test lower than 25nA without slowing down, the actual diodes may have leakages in the pA range. So if you want a low leakage diode you have to pay more because of the testing. Or you can use a transistor junction, and at 3 cents transistors are a good solution. Avoid the high speed transistors as they can also be gold doped. Even some 2N3904s are.
So why do they have to be low leakage? Consider if the bottorm TC54 trips, pulls it's output low and turns off it's transistor. Now no more current flows up through the chain to the PNP at the top. EXCEPT the reverse leakage currents of those diodes, and since they all add together if you have 100nA leakage in each( if the diodes are warm) , you end up with 1.5uA leakage into the base of Q7 which is enough to start turning it on. I guess that was the long answer.
fechter said:
The modified charger circuit I posted should go a long way toward reducing heat. In the best case where all the cells are in balance the shunts will barely be dissipating any power. If the cells are almost balanced then they will all be absorbing almost all the charge current and only the small differential between cells will need to be shunted.
The original charger never reduced charging current during the CV portion so as the cells absorbed less and less current the shunts had to dissipate more and more power. Towards the end of the charge the heat dissipated was = Charging current x Pack voltage = lots. With the modified charger as the pack nears end of charge all the cells should be pretty much in balance and the heat dissipated drops to 0.
The modified system will only dissipate significant power if cell capacities in the pack are substantially different. If that situation arises you already have problems because the pack capacity will be substantially reduced (Pack AH is = AH of the smallest capacity cell). With LEDs on the shunts for people to see, if one or more shunts comes on long before the others you'll know those cells are going bad. As someone posted earlier, this is for enthusiasts, not mass market.
I would try it and see how it works out.
Here is an implementation of a Multipack scenario. The first schematic is the LVC modified to put out an LVC signal that can be combined with other pack LVC signals on a main board that implements the power disconnect.
The next schematic is the mainboard that combines all the LVC signals and controls the output FET. Quiescent power consumption is kept fairly low.
Each pack LVC draws less than 40ua when active, about 1 uA in cutoff. For a 48V pack the main board draws about 50ua plus 10ua for each unused pack connector. The system is designed so that if any LVC signal from a pack is disconnected from connection or wiring problems the pack will shutoff. Any unused LVC connector on the main board needs to have a shorting plug on that connector to enable the output. Q1 keeps quiescent current at a minimum for any unused LVC connectors by supplying just enough current for the shorting plug to do it's job. The FET driving circuitry is the same as the earlier posted LVC schematics.
Operation -
A correctly functioning LVC output from a pack will supply 20-40ua into the gate of it's associated 2N7000 FET pulling it up and turning it on. The gate voltage is clamped through a FDH300A low leakage diode to a shared 12V zener. The FDH300A are cheap and have low enough leakage for this job. When all the 2N7000 are on, they pull the base of Q3 to ground and turn it off. R15 is then allowed to pull up Q9 and the output circuit turns the Load FET on.
If any LVC output stops supplying current because it is tripped, or the wires become disconnected etc, the gate will be pulled to ground by one of the 470K resistors and the FET will turn off. This allows Q3 to turn on and pulls the base of Q9 to ground shutting off the Load FET.
Bothe Q1 and Q9 need to be able to handle the maximum full pack voltage with all packs installed. D2 should be an appropriate low power (low current) zener. The 20K resistors on the connectors are for ESD protection.
The next schematic is the mainboard that combines all the LVC signals and controls the output FET. Quiescent power consumption is kept fairly low.
Each pack LVC draws less than 40ua when active, about 1 uA in cutoff. For a 48V pack the main board draws about 50ua plus 10ua for each unused pack connector. The system is designed so that if any LVC signal from a pack is disconnected from connection or wiring problems the pack will shutoff. Any unused LVC connector on the main board needs to have a shorting plug on that connector to enable the output. Q1 keeps quiescent current at a minimum for any unused LVC connectors by supplying just enough current for the shorting plug to do it's job. The FET driving circuitry is the same as the earlier posted LVC schematics.
Operation -
A correctly functioning LVC output from a pack will supply 20-40ua into the gate of it's associated 2N7000 FET pulling it up and turning it on. The gate voltage is clamped through a FDH300A low leakage diode to a shared 12V zener. The FDH300A are cheap and have low enough leakage for this job. When all the 2N7000 are on, they pull the base of Q3 to ground and turn it off. R15 is then allowed to pull up Q9 and the output circuit turns the Load FET on.
If any LVC output stops supplying current because it is tripped, or the wires become disconnected etc, the gate will be pulled to ground by one of the 470K resistors and the FET will turn off. This allows Q3 to turn on and pulls the base of Q9 to ground shutting off the Load FET.
Bothe Q1 and Q9 need to be able to handle the maximum full pack voltage with all packs installed. D2 should be an appropriate low power (low current) zener. The 20K resistors on the connectors are for ESD protection.
Gary,
Here's how to hook up your opto LVC boards to a FET cutoff.
The wire to the pack positive can be small since it carries little current to the circuit.
The negative wire going to the source of the Power FET must be large to handle the load current. The wire from the negative pack terminal should be connected very close to the Source of the Power FET. The ground connection for the rest of the circuit should be a trace that comes right off the source lead of the Power FET, no load current should ever flow through that trace. That's why I drew it that way to make it clear.
Here's how to hook up your opto LVC boards to a FET cutoff.
The wire to the pack positive can be small since it carries little current to the circuit.
The negative wire going to the source of the Power FET must be large to handle the load current. The wire from the negative pack terminal should be connected very close to the Source of the Power FET. The ground connection for the rest of the circuit should be a trace that comes right off the source lead of the Power FET, no load current should ever flow through that trace. That's why I drew it that way to make it clear.
GGoodrum
1 MW
Randomly said:
Thanks for posting that. I'll give this a shot, and see how it does.
What are the part numbers for the transistors? Can something like 2N3904/3906 parts work, or do they need to be something different? I guess I can go back to one of the earlier diagrams, and try to find them. I think you mentioned them already.
-- Gary
rkosiorek
100 kW
Randomly
how many IRFB4110s can your opto load cut-off circuit control? i would like to use 4 FETS in parallel to be able to handle very high discharge rates and to keep the Total resistance of the switch very small
rick
how many IRFB4110s can your opto load cut-off circuit control? i would like to use 4 FETS in parallel to be able to handle very high discharge rates and to keep the Total resistance of the switch very small
rick
PJD
100 W
There is certainly a lot of good discussion here, but it should probably be considered a separate topic.
Bob,
Any change in the status of the original design? There are a lot of people waiting for word on it here...
Bob,
Any change in the status of the original design? There are a lot of people waiting for word on it here...
rkosiorek
100 kW
jeffkay said:
have you priced 100A relays recently? have you seen the size of one? or checked out how much current is used to hold the relay closed?
plus remember that the maximum number of FETS does not have to be used. the module could be sized to match the system size. for my system i would prefer to use a cutoff built into the pack. and as long as it is there i may as well connect a power on/off switch to it.
rick
There was a resistor in the original Opto cut off circuit I posted that shouldn't have been there. Following is an updated version of the schematic and I have also included hooking up multiple FETs. I have also included part numbers as appropriate.
If you are connecting the Battery pack to a solid state motor controller you will probably not need diode D1 because the motor controller will already have a freewheeling diode to handle the inductive current of the motor. If you are connecting this directly to a motor, or a non solid state controller then you will probably need D1 to keep your FETs safe. D1 is cheap insurance, if in doubt put it in.
The problem arises when you have a lot of current flowing through your motor and you suddenly cut it off. The current through the motor inductance has stored a large amount of energy in the magnetic field, as that magnetic field collapses the energy has to go somewhere. Without D1 the inductance kicks the voltage up on the drains of the FETs until one of them starts to avalanche and conduct. The total energy stored in the motor inductance then dissipates inside that FET and heats it up tremendously.
I have a 1 hp 24V motor that kicks out about 2000 millijoules worst case when you turn it off. If you look at an FET data sheet you will see that they have a rating for maximum single pulse avalanche energy allowed. This varies a lot between parts. An IRF3205 is a 55volt 9milliohm part that can only absorb 20 millijoules of energy, but they only cost about $0.85. An IRFB4110 is rated for 190mJ. A Fairchild FDP038AN06A0 is a 60V 3.8 milliohm part that can handle 625 mJ (these are pretty good value parts for 48V packs and under). Another good part is the STP80NF55-06 a 55V 6 milliohm part that can handle 1300 mJ.
IRFB4110 are premium 100V parts. If you are using 48V packs, a 60V part will give you lower Rdson for less money. The lower voltage the FET, the lower it's resistance for the same size die. the FDP038AN06A0 is about 1/2 the price of the IRFB4110 for the same on resistance. Don't try to cut the voltage rating of the FET too close to your maximum battery voltage, give them some safety margin. Another thing to consider is that the wiring in TO-220 packages are usually only rated for 75 amps maximum, regardless of what the FET inside is rated for. For high currents use one FET for every 50-60 amps of maximum current that will flow.
Paralleled parts don't share avalanche energy, one part always avalanches before the others and absorbs all the energy.
A way to avoid the avalanche problem is to put a high current diode from the drain of the FETs to the positive supply. This shorts the motor current to the positive terminal and lets it flow back around to the motor. The FETs never avalanche.
So where can you get an appropriate high current diode? You just use the body diode in one of the same type of power FETs you are using in the output switch. Short the gate to the source and connect as shown in the schematic. Since this is only a transient current, one FET is all you should need for the job unless your motor currents are over 200A. In that case you can parallel another diode in.
Keep your high current traces and connections short. If you are paralleling FETs in the switch, keep traces short, wide, and the layout symmetrical so trace lengths from the battery connection to the FET sources are the same length. Do Not have any load current flowing in the traces that go to the FET drivers, run a separate trace directly from the last point where the source traces split up to go to the individual FETs back to the drive electronics.
It will probably handle at least 10 FETs in parallel just fine, although 10 FETs is overkill in most cases. 4 FETs should handle up to 200A comfortably. Note from the schematic that you should connect 47 ohm resistors from the gates to the drive circuit. These resistors should be placed close to the Gate pin.They keep the multiple FETs from differentially oscillating with each other.rkosiorek said:
If you are connecting the Battery pack to a solid state motor controller you will probably not need diode D1 because the motor controller will already have a freewheeling diode to handle the inductive current of the motor. If you are connecting this directly to a motor, or a non solid state controller then you will probably need D1 to keep your FETs safe. D1 is cheap insurance, if in doubt put it in.
The problem arises when you have a lot of current flowing through your motor and you suddenly cut it off. The current through the motor inductance has stored a large amount of energy in the magnetic field, as that magnetic field collapses the energy has to go somewhere. Without D1 the inductance kicks the voltage up on the drains of the FETs until one of them starts to avalanche and conduct. The total energy stored in the motor inductance then dissipates inside that FET and heats it up tremendously.
I have a 1 hp 24V motor that kicks out about 2000 millijoules worst case when you turn it off. If you look at an FET data sheet you will see that they have a rating for maximum single pulse avalanche energy allowed. This varies a lot between parts. An IRF3205 is a 55volt 9milliohm part that can only absorb 20 millijoules of energy, but they only cost about $0.85. An IRFB4110 is rated for 190mJ. A Fairchild FDP038AN06A0 is a 60V 3.8 milliohm part that can handle 625 mJ (these are pretty good value parts for 48V packs and under). Another good part is the STP80NF55-06 a 55V 6 milliohm part that can handle 1300 mJ.
IRFB4110 are premium 100V parts. If you are using 48V packs, a 60V part will give you lower Rdson for less money. The lower voltage the FET, the lower it's resistance for the same size die. the FDP038AN06A0 is about 1/2 the price of the IRFB4110 for the same on resistance. Don't try to cut the voltage rating of the FET too close to your maximum battery voltage, give them some safety margin. Another thing to consider is that the wiring in TO-220 packages are usually only rated for 75 amps maximum, regardless of what the FET inside is rated for. For high currents use one FET for every 50-60 amps of maximum current that will flow.
Paralleled parts don't share avalanche energy, one part always avalanches before the others and absorbs all the energy.
A way to avoid the avalanche problem is to put a high current diode from the drain of the FETs to the positive supply. This shorts the motor current to the positive terminal and lets it flow back around to the motor. The FETs never avalanche.
So where can you get an appropriate high current diode? You just use the body diode in one of the same type of power FETs you are using in the output switch. Short the gate to the source and connect as shown in the schematic. Since this is only a transient current, one FET is all you should need for the job unless your motor currents are over 200A. In that case you can parallel another diode in.
Keep your high current traces and connections short. If you are paralleling FETs in the switch, keep traces short, wide, and the layout symmetrical so trace lengths from the battery connection to the FET sources are the same length. Do Not have any load current flowing in the traces that go to the FET drivers, run a separate trace directly from the last point where the source traces split up to go to the individual FETs back to the drive electronics.
Nice!
That looks like a good design.
I think D1 is a good idea too, but the controller should absorb everthing from the motor.
That looks like a good design.
I think D1 is a good idea too, but the controller should absorb everthing from the motor.
GGoodrum
1 MW
I'm doing a new "hybrid" 4-cell version that uses the opto-based circuits from the original design, with Randomly's modified cutoff circuit. These will go on top of a new 12V/9.2Ah a123-based "stick" pack that I'm going to do as a kit. Here's what the PCB looks like:
This version uses a single 4110 FET, which should be good for 50A bursts, or so. You can also use these without the active cutoff, like the existing boards.
The kits will group four blocks of four cells each, which are paralleled together first, and then the blocks placed end-to-end, in series. Here's roughly what that will look like:
There are G10 plates in between each block with nickel-plated springs on eather side, to make contact with the cells. A threaded rod goes down the middle and is used to hold everything together. This technique actually works quite well, and allows solderless packs to be put together.
Anyway, I will start a new thread when I get the first of these together. Basically, the idea is to combine multiple 12V "sticks" together in order to make larger packs. The opto circuits from each sub-pack can be "daisy-chaned" together in parallel, and connected to the first pack in series, which will be populated with the cutoff parts.
-- Gary
This version uses a single 4110 FET, which should be good for 50A bursts, or so. You can also use these without the active cutoff, like the existing boards.
The kits will group four blocks of four cells each, which are paralleled together first, and then the blocks placed end-to-end, in series. Here's roughly what that will look like:
There are G10 plates in between each block with nickel-plated springs on eather side, to make contact with the cells. A threaded rod goes down the middle and is used to hold everything together. This technique actually works quite well, and allows solderless packs to be put together.
Anyway, I will start a new thread when I get the first of these together. Basically, the idea is to combine multiple 12V "sticks" together in order to make larger packs. The opto circuits from each sub-pack can be "daisy-chaned" together in parallel, and connected to the first pack in series, which will be populated with the cutoff parts.
-- Gary
i am glad to see you guys going forward with variations on the original design, and new ideas. when i was working as a biomedical engineer we always had design reviews so when somebody made an error in a design there were other sets of eyes and other peoples' experience to draw from. the design was released on this forum long before i was confident in it, and i did not mean to react unfavorably to criticism, some of which is indeed valid. i can blame the errors on the high doses of narcotics i take, lack of sleep, etc, but the truth is i just did not take enough time to analyze everything before gary and i yielded to the pressure to release something, as it seems there are very few options available.
people will not quit bugging me to get the bms working, and i appreciate that. sometimes i need a little motivation to get out of bed and get to work. i have several customers who are willing to do their own heat sinks, so i am working on getting the bugs out of the design so we can start there. my original idea was to have the shunts all active for about 15 minutes then have a long time-out just to keep the cycle from restarting, so it did not matter if the long interval timer ever terminated; in fact it is just as well if it does not, and the cycle only restarts if power is removed. removing the cap completely from the second one shot and replacing it with a resistor will make the pulse infinite, so that is one possibility. It also turns out that some people want an hour or two at the terminal cell voltage, for thunder sky lifepo4 cells, so i need to replace the one shot with a timer that can reliably produce the longer pulses. i am debating whether to just put in a PIC to do that, but it means significant revs to the board. i have used some one shots in the past with built in counters to produce long delays, like the MC14541B; i will be looking at that as an option along with a small PIC next.
i am committed to get the design fully functional, as i have been promising people long enough and there does seem to be a need. constructive criticism is welcome. i will publish progress reports here over the next week or two, and hope by then to have something ready to ship. i really do appreciate the patience of the people who still have faith i will deliver something working, and i feel it is the best way i can contribute to making electric bikes more accessible, something which i feel is if critical importance in the process of weaning ourselves off our addiction to oil. when gas hits 6 bucks i predict ebikes will get a big boost
i do not plan on providing a high current cutoff; i think a fuse is adequate for most users. i do plan on making the time at terminal voltage adjustable from 15 minutes to a couple of hours, balancing current adjustable up to several amps, and having the bms cut off charge after that and not restart until power is cycled. i also plan on restoring the individual leds for each cell. i think one use for the bms may be for the occasional balancing of packs that are otherwise just bulk charged. in this configuration size is not as important, so a heat sink capable of handling balance cycle charge current of several amps will be possible. One bms could then be used to balance several ebike systems.
thanks again for everyone's interest, patience, and kindness.
people will not quit bugging me to get the bms working, and i appreciate that. sometimes i need a little motivation to get out of bed and get to work. i have several customers who are willing to do their own heat sinks, so i am working on getting the bugs out of the design so we can start there. my original idea was to have the shunts all active for about 15 minutes then have a long time-out just to keep the cycle from restarting, so it did not matter if the long interval timer ever terminated; in fact it is just as well if it does not, and the cycle only restarts if power is removed. removing the cap completely from the second one shot and replacing it with a resistor will make the pulse infinite, so that is one possibility. It also turns out that some people want an hour or two at the terminal cell voltage, for thunder sky lifepo4 cells, so i need to replace the one shot with a timer that can reliably produce the longer pulses. i am debating whether to just put in a PIC to do that, but it means significant revs to the board. i have used some one shots in the past with built in counters to produce long delays, like the MC14541B; i will be looking at that as an option along with a small PIC next.
i am committed to get the design fully functional, as i have been promising people long enough and there does seem to be a need. constructive criticism is welcome. i will publish progress reports here over the next week or two, and hope by then to have something ready to ship. i really do appreciate the patience of the people who still have faith i will deliver something working, and i feel it is the best way i can contribute to making electric bikes more accessible, something which i feel is if critical importance in the process of weaning ourselves off our addiction to oil. when gas hits 6 bucks i predict ebikes will get a big boost
i do not plan on providing a high current cutoff; i think a fuse is adequate for most users. i do plan on making the time at terminal voltage adjustable from 15 minutes to a couple of hours, balancing current adjustable up to several amps, and having the bms cut off charge after that and not restart until power is cycled. i also plan on restoring the individual leds for each cell. i think one use for the bms may be for the occasional balancing of packs that are otherwise just bulk charged. in this configuration size is not as important, so a heat sink capable of handling balance cycle charge current of several amps will be possible. One bms could then be used to balance several ebike systems.
thanks again for everyone's interest, patience, and kindness.
Doctorbass
100 GW
GGoodrum said:I'm doing a new "hybrid" 4-cell version that uses the opto-based circuits from the original design, with Randomly's modified cutoff circuit. These will go on top of a new 12V/9.2Ah a123-based "stick" pack that I'm going to do as a kit. Here's what the PCB looks like:
This version uses a single 4110 FET, which should be good for 50A bursts, or so. You can also use these without the active cutoff, like the existing boards.
The kits will group four blocks of four cells each, which are paralleled together first, and then the blocks placed end-to-end, in series. Here's roughly what that will look like:
There are G10 plates in between each block with nickel-plated springs on eather side, to make contact with the cells. A threaded rod goes down the middle and is used to hold everything together. This technique actually works quite well, and allows solderless packs to be put together.
Anyway, I will start a new thread when I get the first of these together. Basically, the idea is to combine multiple 12V "sticks" together in order to make larger packs. The opto circuits from each sub-pack can be "daisy-chaned" together in parallel, and connected to the first pack in series, which will be populated with the cutoff parts.
-- Gary
13.2V 9.2Ah hummmm! very similar to my boost pack
Did you know i've boosted my car with it this winter at -15 celcius!!?
[youtube]JgJ-BHUXQeQ[/youtube]
PJD
100 W
Does anyone have an update on Bob's BMS? If no one has any news, I'll probably give him a phone call and report back.
PJD
100 W
Bikeraider,
I mentioned this in the "BMS's that are available NOW" thread, but I called them last Sunday, Bob was out and I spoke with his wife. Due to health issues (non life threatening, but painful and activity restricting) with both him and his wife, I got the strong impression that, when the modifications and design a new board are factored in, the BMS would not be available on any guaranteed time scale.
If you want to call Bob, I'm sure he wouldn't mind - his phone number is listed (Oregon). He needs to know how many people are interested in the eventual development of a final product.
But as for me, I caved in and ordered the LVC and TP210 balancers that Gary sells - more expensive, but I got tired of waiting. But, I am still interested in at least one of Bob's BMS's since I have a second scooter to convert, or if Gary's products don't perform satisfactorily on larger motorcycle-sized packs.
I mentioned this in the "BMS's that are available NOW" thread, but I called them last Sunday, Bob was out and I spoke with his wife. Due to health issues (non life threatening, but painful and activity restricting) with both him and his wife, I got the strong impression that, when the modifications and design a new board are factored in, the BMS would not be available on any guaranteed time scale.
If you want to call Bob, I'm sure he wouldn't mind - his phone number is listed (Oregon). He needs to know how many people are interested in the eventual development of a final product.
But as for me, I caved in and ordered the LVC and TP210 balancers that Gary sells - more expensive, but I got tired of waiting. But, I am still interested in at least one of Bob's BMS's since I have a second scooter to convert, or if Gary's products don't perform satisfactorily on larger motorcycle-sized packs.
