New 16-cell Battery Management System (BMS)

[PJD]

Fecher and Bob,

Thanks for the suggested work-arounds. Both look simple and elegant. The throttle signal solution looks like the simplest from my perspective.

I originally got into e-cycles strictly from an environmentalist point of view (I'm a civil engineer) but I'm slowly getting an electronics education too.

 

[kbarrett]

Waiting on a kit ... any notion on price yet?

 

[GGoodrum]

I'm sorry for the delay. We found two more issues. The first is that we ran into a problem where the one-shot timer wasn't triggering reliably, due to smal leakages in the diodes used to encode the two signals for when at least one of the shunts are on and the other for when all the shunts are on. The fix required adding a pullup resistor on each channel.

The second issue we have is with heat. There is a tradeoff between heat sink size and the amount of current that the shunts are limited to, during the final phase of the charge process. What we found is that even limiting the current to 1/2A, the present heat sink arrangement could still get pretty hot, if it took longer than about 15 minutes for the final CV portion of the charge cycle, when the cells are gradually reducing the amount of current they let in. The "Catch 22" is that if you use higher than 1/2A, it takes less than 15 minutes, but if you use 1/2A, it takes over 20 minutes. Before I got too far into this chasing my tail-type of problem solving, Bob suggested we use the other half of the op-amp chip we are using, to actively control the current limiting, based on heat sink temps. We're going to use a couple of inexpensive thermistors, one on each heat sink. This way, the current can start out higher and only throttle back as it needs to. This will actually shorten up the final phase charge time. It also gives us a lot more flexibilty in what size heat sink is used.

Since these changes are going to require a board layout change. I've decided to put in one other change/upgrade we've been playing with. Instead of just turning on a green LED, when the charge cycle is complete, we are now going to use a two-color, red-green LED. When the charge process first starts, the LED will be red. As soon as the shunts start operating, meaning the cells are starting the CV phase, both colors will be on, so it will be an orangish, lighter color. When everything is done, the LED will be green.

Anyway, new boards are in the works. I should have them back by early next week. Bob says he can make the existing boards work, for those that want an assembled and tested unit, so we will make those available, as soon as he gives me the go-ahead.

-- Gary

 

[Deepkimchi]

Hey Gary,

If the existing boards are at a discount :wink: , make mine with a final phase of 30-40 minutes. As long as the overall charging process is less than 4 hrs, doesn't bother me. I get home about 5:30 pm - go to bed at 9:30 pm. I'd rather have reliability than speed of charging.

Right now with my present commute I'm only using 6.5 ah.

DK

 

[efreak]

That's yeoman's work guys thanks a million and I am in, for my wife's batts ( did that come out ok?!) at least for sure

what a swift progression thanks again.

this is like the early days of anything i guess..... like the electric guitar ...... and you guys are like les paul (C) ... hendrix


efreak

 

[fechter]

Limiting the current based on temperature seems like a pretty optimal solution if it can be easily implemented. It will charge as fast as it can based on ambient conditions.

Another possible approach might be to add a CPU fan to the heatsink that turns on when it gets hot. A little bit of air flow does a lot of cooling. The fan would only run during the balancing phase. You could possibly trigger the fan based on being in the balancing phase vs. an actual temperature sensor.

I'd still propose using big resistors to dissipate the heat and use transistors in the switching mode to turn them on. Resistors are happy at 150C. I guess the heat still has to go somewhere though.

Do you have any plans for a LVC only board? I think there are a few folks that are planning to use separate cell chargers but still need a LVC for discharge.

 

[GGoodrum]

Limiting the current based on temperature seems like a pretty optimal solution if it can be easily implemented. It will charge as fast as it can based on ambient conditions.

Another possible approach might be to add a CPU fan to the heatsink that turns on when it gets hot. A little bit of air flow does a lot of cooling. The fan would only run during the balancing phase. You could possibly trigger the fan based on being in the balancing phase vs. an actual temperature sensor.

I'd still propose using big resistors to dissipate the heat and use transistors in the switching mode to turn them on. Resistors are happy at 150C. I guess the heat still has to go somewhere though.

Do you have any plans for a LVC only board? I think there are a few folks that are planning to use separate cell chargers but still need a LVC for discharge.

Actually, we have half an op-amp that wasn't being used (the chip we are using has two...), so the impact is pretty minimal. Using resistors would still generate the heat, which has to go somewhere, and big resistors take up a lot of real estate. We've also thought about a fan, which could easily be triggered using the signal that the shunts have startedto operate, but this is a last resort. I'd really like to keep it fanless.

I do have some 16-channel LVC boards, with active cutoff FETs (two 4110s...), if anybody is interested. These can come with 18-pin Molex connector, with a matching plug, that can be used to conect to individual chargers, which is what I was originally going to do. Anyway, if somebody wants one, just PM or email me (ggoodrum@tppacks.com). I was going to add these to my website, once I got the BMS boards up there, but they are ready in any case.

-- Gary

 

[bobmcree]

the design is greatly simplified by using the TIP105 darlington pair, and we are only asking them to dissipate a few watts when they can handle 80W with the proper heat sink. These parts cost about 50 cents, and with a resistor you would still need a fet to control it and a way to dissipate the heat. The TO-220 is a very efficient package in terms of conducting heat from the die to the heat sink, and a device that physical size would still be needed if a resistor were to be used to dissipate the heat. Because the system is designed to operate with a shunt voltage as low as 3v, a fair amount of the heat would be dissipated in the drive transistor anyway and the power resistor would just be more complexity and higher cost with no appreciable gain.

i did consider a fan, but i think most users will enclose their system and a fan would be of little value. the amount of current we can use during the final charge phase when some shunts are conducting is limited only by the heat sink, so it will certainly be possible for users to crank up the current and add a fan if it works for them. i feel that with an adaptive system that cuts the charge current as the heat sink temp rises we can please almost everyone, and if you want it to charge faster you will just need to let it run cooler, if you dont care about an extra hour or so and you want to enclose the system it will automatically compensate for the environment where a fan would be of little utility anyway.

The design provides for a variable time duration final charge phase during which all the shunts are conducting and all the cells are held at 3.65v. Gary tells me his experience with A123 cells is that they continue to accumulate significant charge energy for quite some time after reaching this peak voltage. In the LifeBatt cells i have seen some extra capacity from cells i maintained at 3.65v until the charge current dropped to a few ma., but this is just a few %, and if my LifeBatt pack is charged just up to the cutoff voltage of 3.65v/cell x n cells , I am seeing at least the full 10 Ah delivered on subsequent discharge tests. It appears that with these cells it is not quite as important to hold them at terminal voltage for awhile. In the present state of the design, this phase of the charge process is adjustable by changing an RC time constant from milliseconds to hours. By shortening or eliminating this time the heating can be greatly reduced and charge time reduced as well.

 

[disadvantage]

Thank you Gary and Bob, for taking the time to work all the bugs out of the design.
I'm anxious to start soldering a new kit together, but I can wait for quality.

If you don't mind me asking, what's the part number for the high-gain dual optic isloator in the schematic?
It's labeled U102A and U102B.

I've been studying the schematic thoroughly.

disadvantage

 

[bobmcree]

we are using the Avego ILD2, it has a minimum gain of 100% at 1 ma, typically 2-3x that. they are costing us about .85 each and i just found what may be a cheaper substitute, i have to test them but they have the specs we need.

if you have been studying the schematic, can you find the missing resistor? there is one missing that keeps the thing from working. i fixed the drawing but left the old one posted because i wanted to see if anybody was paying attention. so far nobody has caught it.

 

[disadvantage]

Now that you mention it, there was some weirdness around one of the optoisolator outputs that I couldn't understand.

I attached a Windows Paint picture of the changes I think need to be made.

In the upper portion of my picture, LVC Power Cut should be an output from the optoisolator, but it's connected to the base of a phototransistor like it is an input. Also, there appears to be no collector load to the pack positive on the phototransistor; I would expect one if the emitter is grounded.

View attachment BMS-miss-steak1.png

In the lower portion of my picture, I disconnected LVC Power Cut from the base of the photo transistor and connected it to the collector. I let 100K R4 serve as the collector load to pack positive voltage, with D4, the 6.2 zener diode, protecting the gates of MOSFETs Q2 & Q3 from blowing on full pack voltage.

I looked and looked for two hours for a missing resistor, but I couldn't find it. I think the circuit mistake is the misplaced connection that I described. I must go home now from work; my car is parked in an expensive ramp.

disadvantage

 

[fechter]

I think he's right. I noticed that too.

 

[bobmcree]

that was a typo on the drawing that obviously goes to the collector of the opto and not the gate. there is still a missing resistor for proper operation. the path through the diodes that create the and/or logic out of the opto signals looks fine when you consider one opto, but when all the channels are in parallel it is not possible to tell the difference between some shunts being on and all to a degree adequate to fire the one shot, without an individual pullup resistor in each channel on pin 6 of each opto. R9 pulls up the resulting signal "ANY SHUNT LOW" but once any of the parallel optic isolators conducts there is no longer a pullup to drive the input of U1 through any of the diode pairs. The missing resistor is 51k that goes between the collector of each opto and the 6.2v supply. This way when any opto is off its 51k will provide a pull up path through its diodes to the one shot, and R10 provides a path for the current to bias the diode on. R10 should be 150K so between the 50k and a couple of .5v diode drops it will still be a good solid high going low, to trigger the input of U1, thus initiating the second charge phase.

Gary and i have devised a simple way to add a sort of adaptive current limiting to the circuit that cuts the charge current after the first cell hits the desired peak, to help us control the amount of heat generated. i will use a couple of common 10k ntc thermistors to lower the voltage supplied to the pot that sets the reduced current limit as temperature increases. This will permit setting the current to a nominal value for operation in a cool ventilated environment and if the environment changes the current will drop to avoid excessive heating. this is a very simple function and it will not be perfect, but it should be possible to 'dial in' the component values to suit both the pack getting lots of air and one inside a bag, without a need for the user to change anything. If the heat sink exceeds 160F a backup thermal switch will cut off charge current until it cools, then the cycle will resume.

 

[stevero2001]

I know that lifepo4 is supposed to be charged to 3.65v per cell to get a full charge. If it were instead charged to 3.6v, how would that affect the charge level and lifespan? I believe that from reading other posts that charging to less than full could extend the life of a pack. Would it make sense to get a larger pack than what you need in order to allow a less than full charge? Would the increased longevity make this more economical in the long run?

 

[bobmcree]

I know that lifepo4 is supposed to be charged to 3.65v per cell to get a full charge. If it were instead charged to 3.6v, how would that affect the charge level and lifespan? I believe that from reading other posts that charging to less than full could extend the life of a pack. Would it make sense to get a larger pack than what you need in order to allow a less than full charge? Would the increased longevity make this more economical in the long run?

i think 3.65v is a conservative figure, most bms systems cut off when any cell hits 4v. i would be surprised if there was a significant longer cycle life between 3.60 and 3.65v. What i have seen with the LifeBatt cells is that if i charge them up to 4.0v i can get an extra 5% or so capacity, and i expect that might shorten the cell life. the bms gary and i developed is adjustable from about 3v to 4v so you can set your cells terminal voltage to any level you like; the shunts prevent any cell from reaching a higher voltage, so 3.65v seems like a reasonable cap if all the cells are capped there instead of in the typical systems where they can reach 4v or more.

 

[PJD]

Bob,

You have noted on your schematic "charger voltage 3.65 volts per cell + 0.5V"

You also have stated that the design intent of this BMS is to use an ordinary CC-CV SLA battery charger. Couple questions:

1. Do you mean "3.65v/cell + 0.5v" or "3.65 v/cell + 0.5v/cell"?

2. If the latter, than few SLA chargers will meet this requirement, as the cutoff voltage for a typical 48 volt pack charger is 59 to 60 volts, while your requirement would be 66.4 volts. So, I assume you mean the former. Can you clarify?

 

[bobmcree]

Bob,

You have noted on your schematic "charger voltage 3.65 volts per cell + 0.5V"

You also have stated that the design intent of this BMS is to use an ordinary CC-CV SLA battery charger. Couple questions:

1. Do you mean "3.65v/cell + 0.5v" or "3.65 v/cell + 0.5v/cell"?

(3.65v/cell) + 0.5v i am just figuring that .5v is needed to allow for some voltage drop in the wiring, and a few millivolts of error in each channel. There needs to be enough voltage to turn on all of the shunts, the reason i used a .004 ohm fet in the charge path was to allow for the lowest possible charger voltage and still let me have control over the current. Operating the fet in analog mode means more heat is produced, but it makes it easier to turn the fet full on or have it drop relatively low voltage only when needed to limit the charge current in the bms.

with the latest rev of the design we added a pair of thermistors to the heat sink, so that when the shunts are active the current will start out at the value set by VR1, and then as the temp rises the current will be reduced, to about half by the time the heat sink hits 140F. This will allow the bms to adapt to the degree of cooling available by reducing charge current. if this control mechanism is not adequate there is a bimetallic switch on the heat sink that cuts the charge current completely at 60C, and when the heat sink cools the charge cycle will resume, unless the open charge circuit causes the charger to cut off.

2. If the latter, than few SLA chargers will meet this requirement, as the cutoff voltage for a typical 48 volt pack charger is 59 to 60 volts, while your requirement would be 66.4 volts. So, I assume you mean the former. Can you clarify?

thanks for asking for the clarification, if you were not sure i am certain there were others. I am figuring on realistically needing a volt more than the total of the shunts, just to allow for wiring drops. The final phase of the charge cycle is completed at limited current, so the voltage drop will be much less when the need to measure the cell voltage is most important. This compensates for most of the voltage drop in the wiring.

For customers who have a supply that needs a few more volts on the output i have some 3.3v 66w isolated dc-dc converters that can be powered by your 48v supply then the output wired in series to boost the voltage 3-4v for each module added. These $170 modules are going for 10 on ebay, i have posted on them before. they still have hundreds. talk about a solution looking for a problem

 

[PJD]

Bob,

Are you talking about the TYCO# KW015A0F41, ebay item no. 110234884035 - comes in a through-hole and SMD version?

I must admit, the attractiveness of sixteen LiFePO4's cells was originally that they could be charged exactly like a 48 volt lead-acid pack. But we are now throwing complications into the mix. I suspect the word "elegance" doesn't get used a lot among EEs!

But, these modules are small enough to be wired in a little enclosure right onto the charger output cord - like the GFI's on a hair dryer, so, that's elegant.

 

[bobmcree]

Bob,

Are you talking about the TYCO# KW015A0F41, ebay item no. 110234884035 - comes in a through-hole and SMD version?

no, this is a PowerOne and is in brick format about the size of a matchbox, adjustable from about 3-4.5v 66w by using the remote sense inputs and trim input

i have a batch and have used them as single cell chargers, they are in digikey for $170 so for $10 i wish i had a use for more

http://tinyurl.com/2b7ayo

 

[PJD]

Bob,

Another question.

Some SLA chargers, for some reason, need to detect a certain minimum amount of voltage at their output before they turn on. This is often discovered when attempting to charge a pack that is completely dead. How will this affect the usage of such a charger with your BMS? Any work-around?

 

[GGoodrum]

Let me be clear about one thing, 3.65V x 16 + 0.5V = 58.9V, so most SLA chargers WILL work fine. Some, like the Soneils, normally have cutoff voltages a bit lower, like around 58V, but they have at least a 10% adjustment range, so that they can be "tweaked" up a bit, to around 59V. the ones I'm getting will come "pre-tweaked", otherwise they are exactly the same.

-- Gary

 

[PJD]

Can anyone answer the other question? What about 48 volt chargers that require a certain voltage to their outputs (not sure what it is) before they turn on?

 

[GGoodrum]

Can anyone answer the other question? What about 48 volt chargers that require a certain voltage to their outputs (not sure what it is) before they turn on?

This won't be a problem, as long as the pack isn't completely dead. The charger is connected directly to the main pack leads, through the charge current FET, which comes full on, as soon as the charger is connected, even if the charger is not yet plugged into the wall. As soon as the actual charger connector is plugged into the BMS connector, the charger control logic is then powered up via a connection to the pack positive.

 

[bobmcree]

As of 3/22/08 I am ending my association with LifeBatt. They have a good product but there is no room for me to make an honest profit and i no longer wish to be associated in any way with them. I made some completely reasonable and factual remarks about the lack of a LifeBatt bms for a system i sold 5 months ago, and don decided to turn it into a personal attack against me. Gary will be their distributor from now on. I cannot work with Don.

 

[kyakdiver]

Bob, That sucks...

Life is short, so follow your gut and wallet.

kyakdiver