Even Newer 4 to 24-cell Battery Management System (BMS)

[phatburner]

Hi to all on the list.

I am new here and appologize for not looking thoroughly through the archive before posting this question if this has already been discussed here.

I keep reading that the nanophosphate chemistries are relatively less prone to going out of balance. I was wondering whether it would be possible to build a battery box which had as part of it's physical design a large multipole switch. I am imagining a sliding switch which is designed to isolate between switching states and it would in one position connect all the cells in series and in the second position all in parallel. Then all the cells could be charged together by a single output low voltage high current supply. A cell monitor would still be needed to cut off discharge when any cell gets too low. The charge plug could be cleverly keyed to the mechanism so it can only engage when the switch is in the charge position. Similarly the full output voltage could only be delivered to the speed control when the switch is in the use position. The cells would still need to be soldered to the switch contacts, but all of the sliding contacts could be of a large blade type and could be gold plated to keep resistance to a minimum, probably comparable to a soldered connection. Is there something I am missing here which would prevent this from working?

Joe

 

[amberwolf]

Is there something I am missing here which would prevent this from working?

Only that the size of the switch you propose might be greater than that of the pack itself for an average ebike pack. The wiring alone would probably be quite a ratsnest, and might be nearly as big as the switch.

And it would certainly be more expensive, as you would probably have to have it custom-made. Remember that in addition to being rated to handle the full pack voltage at *DC* levels, not *AC*, it must also be able to handle the full pack series/parallel current, as well as the massive fully-parallel current that would be required during charging.

None of this stops the idea from working, but it does kind of move it out of the practical realm.

 

[fechter]

Pat, unfortunately, you haven't missed much. We, on the other hand, have missed deadline after deadline with the v4.x BMS. :x
After god nows how many iterations, I think we have the most stable design yet. No more oscillations and no more voltage drop issues and no more weirdness.

Well, the old, unsophisticated V1.5 boards continue to work well for me (or is there something I need to know?). Even when charging at 16A by doubling-up the chargers the masonite cover over the board only gets a bit warm over the charger control FET. They may not be capable of shutting charging off upon completion of charging, but they have held up to at least 4 hours of forgetting to unplug the charger when fully charged - basically putting the packs of "float" over this period, so I haven't worried too much.

What new features will the new boards have?

The only oddity is that the last cell at the positive end of the pack seems to get to a higher voltage (2.72 to 2.73) compared to the others (2.67 to 2.69), and this happens on both boards I assembled, so it isn't just some random variation in the LM431 or resistor values. This is still within charging tolerances.

At 3 years of calender age, (about 2 years of service age) the Thundersky cells themselves are degrading in the same manner in the scooter with 2500 miles on the pack as the one with 6000 miles on the pack - so calender life, not cycle life seems to be the governing factor. At 2.3C discharge, they now sag to about 2.9 volts at a temperature of 60F. Still plenty usable for another season, but probably not very good for winter use anymore. But I'm wandering off topic...

Glad to hear the long term report.

The voltage elevation on the end cells is due to what I call the "end cell effect". The voltage drop in the wires going from the end cells to the board will add to the clamping voltage, resulting in a slight increase. The cells between have the shunt current balanced on either side, so there is no voltage drop in the tap wires if they are all up to voltage. To minimize the end cell effect, use extra heavy wires to go from the board to the ends of the pack. Even with heavier wire, you will still have a little voltage elevation due to resistance in the traces on the board. When the charger is removed, there is no more current in the middle shunts, and you may see the end cell LEDs stay lit a while longer as those cells bleed back down to the level of the others. If you measure the cell voltages 15 minutes after disconnecting the charger, you should see much less variation.

In the newer version of the BMS, I'm trying to implement a better end of charge detection that shuts down the charging current. Ver2.6 did a forced balance, then shut down. This is OK if all your cells are healthy and well matched. It is thought that the charge should terminate as soon as all the 'healty' cells reach target and not be held on float while lower cells are brought up on every charge cycle. We have looked at several ways to do this, and most likely we will end up with a current based end of charge detection that terminates the charge when the current drops below a certain level, indicating full charge.

Another feature we're trying to work in is a fixed frequency PWM. Ver 2.x boards oscillated by hysteresis, so there was no real control on the oscillation frequency. Some power supplies and battery chargers go nuts when this hits a certain frequency. Around 1khz seems to be particularly troublesome. With the fixed frequency PWM, we can make sure the frequency stays out of the trouble zone. The frequency can be changed if needed. Your charger must not be the type that is bothered by this (as most are).

Yet another feature is some form of automatic switch to eliminate the need for more than 2 wires going to the charger connection. This feature has been particularly troublesome to implement and presently is configured for a manual start (push button) and automatic shutoff. I still want to do a fully automatic switch, but need to come up with something more bulletproof. Alternately, if the control circuit power consumption was low enough, it could just stay on all the time.

As mentioned in recent posts, some kind of protection features may be added. TVS diodes and fuses for each cell would be fairly robust, but somewhat expensive. At the minimum, zener diodes across each cell are needed to prevent failures caused during connection and disconnecton between the board and the pack. If you leave the board permanently attached, this isn't a real issue, but you always have to connect it in the beginning at least once. Zeners won't really protect against misconnection or shorts, but will prevent overvoltage on the cell circuits during normal pack connection.

Yet another feature for newer versions is some kind of automatic fan switch for cooling fans. With 500ma shunts, most folks got away without using fans, but at 1A, fans are a necessity. Seems like this ought to be easy, eh? But not! I did find some spiffy little thermostatic switch chips that are inexpensive, but somehow they can raise hell with the LM431 regulators. Makes no sense, but happens repeatably under certain conditions. These are nice since they will keep the fans on as long as it's hot, whether or not the charger is still connected.

There were several other nice features that I flat gave up on for now so that we could try and get at least something that works. I think eventually those things will be worked out.

I'm getting about as toasty as a runaway shunt resistor trying to get all this stuff working right.

Somewhere down the road, I hope to try out my capacitor coupled balancer, which promises to get rid of all those heat producing resistors and wasted power. Preliminary tests and calculations are encouraging, but I just don't have time lately to work on it much. There is a thread about it here: http://endless-sphere.com/forums/viewtopic.php?f=14&t=18392

 

[LI-ghtcycle]

Thanks for the Up-Date fetcher, I eagerly await the release of the boards, but I also appreciate that your getting things sorted out before they are released.

Do you have any specs for the size of the boards? I understand the current config will be 6 channel groups, and that later on there might be a larger dedicated single board for 16-18 ch.

I am wondering how I might tuck these boards inside my battery case, and what kind of space will be needed to allow them to have proper cooling. I might have better luck with 3 6 channel boards than one 18 ch board depending on how I have to configure things for cooling.

Is it doable to have the BMS in a separate box on the back rack and the batteries in a front triangle case? I am wondering if having the batteries too far from the BMS might be detrimental to it's function.

Thanks!

 

[ejonesss]

the end cell effect i have been noticing is not in charging so much as discharging.

i noticed that on inductive loads like motors i get the 4 negative cells in a 48 volt pack goes lower.

and i think due to the design of the 2.2 bms they never get fully charged.

if i connect the charger hours after the bms shuts off then all the rest of the cells light up first while the 4 negative cells continue charging.


however a resistive load like a light does not seem to cause the effect.

 

[GGoodrum]

the end cell effect i have been noticing is not in charging so much as discharging.

i noticed that on inductive loads like motors i get the 4 negative cells in a 48 volt pack goes lower.

and i think due to the design of the 2.2 bms they never get fully charged.

if i connect the charger hours after the bms shuts off then all the rest of the cells light up first while the 4 negative cells continue charging.


however a resistive load like a light does not seem to cause the effect.

That's because the 12V regulator is tapped off of the first four cells on the pack, on v2.x boards. In the v4.x versions, the regulator works off the whole pack voltage.

 

[ejonesss]

how then do you explain if i charge the 4 cells using a car charger first before charging the pack with the normal charger the effect not showing?

also i have seen this effect with nicads in other devices too.

i suspect that is why rc'ers like to use light bulbs to drain their packs.

not so much as memory as overcharging the other cells while the negative cell(s) are still charging.

as an experiment i discharged a bit using a resistive load like a shop light then recharged and repeated a couple times and it all balanced out ok.

the end cell effect i have been noticing is not in charging so much as discharging.

i noticed that on inductive loads like motors i get the 4 negative cells in a 48 volt pack goes lower.

and i think due to the design of the 2.2 bms they never get fully charged.

if i connect the charger hours after the bms shuts off then all the rest of the cells light up first while the 4 negative cells continue charging.


however a resistive load like a light does not seem to cause the effect.

That's because the 12V regulator is tapped off of the first four cells on the pack, on v2.x boards. In the v4.x versions, the regulator works off the whole pack voltage.

 

[phatburner]

Yes I did envision the switch being rather large actually. My thought was to build the switch as part of the battery containment box. I am imagining a box being made from an insulating material like fiberglass circuit board stock. The copper would have to be thicker than common circuit board material and would probably be copper sheet epoxied to the insulator but imagine that a custom box is made such that one complete side is a big sliding switch. Done properly it doesn't have to be very bulky or heavy at least not significantly more than a box made of the same material anyways. The slide would have to be pulled to the parallel position in order to expose the charge jack. Inserting the charge plug would prevent the slide from being moved to the series position until the charge plug is removed.

Joe

 

[PJD]

Fechter wrote:

Ver2.6 did a forced balance, then shut down. This is OK if all your cells are healthy and well matched. It is thought that the charge should terminate as soon as all the 'healthy' cells reach target and not be held on float while lower cells are brought up on every charge cycle.

In the case of my Thundersky cells, it is the less-healthy cells that come up to the shunt voltage sooner while the healthy new new cells lag behind. I learned this when I had to replace a couple bad cells in the pack with new ones. The cause of this is simply that the mode of declining cell-health in Thunderskys is increasimng internal resistance, particularly charging resistance, so the voltage drop across them will be greater than the healthy cells in the string. This gets particularly bad at cold temperatures.

- Paul

 

[fechter]

In the case of my Thundersky cells, it is the less-healthy cells that come up to the shunt voltage sooner while the healthy new new cells lag behind. I learned this when I had to replace a couple bad cells in the pack with new ones. The cause of this is simply that the mode of declining cell-health in Thunderskys is increasimng internal resistance, particularly charging resistance, so the voltage drop across them will be greater than the healthy cells in the string. This gets particularly bad at cold temperatures.

- Paul

Interesting. My abused A123 cells generally behave like they have a very high self discharge and run lower in voltage than the surrounding healthy cells.

Just like lead-acid or any battery chemistry, failures can go both ways.

 

[GGoodrum]

Fechter wrote:

Ver2.6 did a forced balance, then shut down. This is OK if all your cells are healthy and well matched. It is thought that the charge should terminate as soon as all the 'healthy' cells reach target and not be held on float while lower cells are brought up on every charge cycle.

In the case of my Thundersky cells, it is the less-healthy cells that come up to the shunt voltage sooner while the healthy new new cells lag behind. I learned this when I had to replace a couple bad cells in the pack with new ones. The cause of this is simply that the mode of declining cell-health in Thunderskys is increasimng internal resistance, particularly charging resistance, so the voltage drop across them will be greater than the healthy cells in the string. This gets particularly bad at cold temperatures.

- Paul

Actually, this makes sense. The less-healthy cells will have less capacity, so they will get "full" first, and will hit the HVC point first.

-- Gary

 

[PJD]

Actually, this makes sense. The less-healthy cells will have less capacity, so they will get "full" first, and will hit the HVC point first.

That is another way of viewing it, although if they were balanced to start off, they should be discharged the same amount and should reach "full" at the same time as well. The few cells that I actually had to replace behaved as Richard described - as if they were self-discharging and so got prematurely discharged - and in tests their capacity had declined from 42AH when new to about 32AH (at 16-18A discharge). But, the majority of the cells are showing their age by simply producing more voltage sag during discharge and more voltage rise during charging, while still maintaining most of their capacity.

 

[Mike B]

What do you recommend for a good trade off between quality, efficiency and economy in a 48 volt charger? I know I could find it here on ES, but i just did a search for that and it retuned +1000 instances (nuff said)

Also, I have been following the threads on these boards for a while. Great craftmanship, and simplificated too. I wonder if you have realized how these boards and a few k$ of A123s could be a game changer in renewable energy systems? Lead batteries just have too many problems, and the main problem is having to overdesign the PbA array to meet the (ideally less than) 30% discharge window, the nasty losses in charging, and also the peak amps available without sag. A few kwh of LiFeP will do fine, as long as the total kwh requirement is kept low over the dark/no wind windows and a generator is available--and this is exactly what the PbA systems do already.

 

[thedarlington]

hi i was checking out the site to order one of these bms"s but they show up as out of stock, whats the scoop can we buy these to build ourselves then?

 

[hydro-one]

If i did a search for this string "I eagerly await the release of the boards" --- I would have 1000's of customers for my bms?

just a joke....(i wish i had a bms)
mike

 

[wookey]

It's gone a bit quiet here. Where are we at? Maybe everyone's gone on holiday.

I was under the impression that there was more to come than the "5/6/8-Channel LVC/HVC + 4p Parallel Adapter" and "Charge Current Limiter/Controller" listed on http://www.tppacks.com/products.asp?cat=26, but perhaps I am confused and those are it and all is good and I should just go shopping?

Could someone clarfy for those of us who don't spend every waking hour reading E.S. and thus are a bit vague on the details?

 

[GGoodrum]

It's gone a bit quiet here. Where are we at? Maybe everyone's gone on holiday.

I was under the impression that there was more to come than the "5/6/8-Channel LVC/HVC + 4p Parallel Adapter" and "Charge Current Limiter/Controller" listed on http://www.tppacks.com/products.asp?cat=26, but perhaps I am confused and those are it and all is good and I should just go shopping?

Could someone clarfy for those of us who don't spend every waking hour reading E.S. and thus are a bit vague on the details?

Yes, wookey, there is more to come. We are working parallel paths on two slightly different BMS approaches. One is a modified version of the more "traditional" BMS design, like we've been doing, and the second is a digital/analog "hybrid" variant, that uses the popular CellLog units to generate the HVC and LVC signals. I posted details about this in the other thread here: http://endless-sphere.com/forums/viewtopic.php?f=14&t=17168&p=289655#p289655.

More later...

 

[Indubitably]

Goodrum:

Perhaps, since there seem to be a lot of us laymen who are not so electrical engineering savy, but who are interested in employing your bms in our projects, some sort of simplified application oriented guide would be useful. You know, something that would classify each of your products (or various combinations of your products) according to application rather than capability. For instance there are going to be a lot of people who are doing small ebike projects using a bunch of a 123 batteries in parallel, or people like me doing light motorcycle or heavy scooter projects using thunder sky cells in series, but we don't know how many channels we need, or what the advantages of an analog/digital hybrid are vrs a more traditional bms. What we want to know is which products are going to work best for our type of application, and what exactly we will need in order to go from raw batteries to something we can slap on a bike and hook up to a charger. Try as I might, I just haven't been able to find this kind of information in any one place, or presented in an easily digestible format, and I think it would make life a lot easier both for you and your customers if such a guide was easily accessible.

 

[GGoodrum]

Goodrum:

Perhaps, since there seem to be a lot of us laymen who are not so electrical engineering savy, but who are interested in employing your bms in our projects, some sort of simplified application oriented guide would be useful. You know, something that would classify each of your products (or various combinations of your products) according to application rather than capability. For instance there are going to be a lot of people who are doing small ebike projects using a bunch of a 123 batteries in parallel, or people like me doing light motorcycle or heavy scooter projects using thunder sky cells in series, but we don't know how many channels we need, or what the advantages of an analog/digital hybrid are vrs a more traditional bms. What we want to know is which products are going to work best for our type of application, and what exactly we will need in order to go from raw batteries to something we can slap on a bike and hook up to a charger. Try as I might, I just haven't been able to find this kind of information in any one place, or presented in an easily digestible format, and I think it would make life a lot easier both for you and your customers if such a guide was easily accessible.

You know, this is a very good idea. I will attempt to do this, just as soon as the dust settles a bit.

-- Gary

 

[rscamp]

Sorry if I've missed this elsewhere, but what value R101 resistor would be required to change the HVC from 4.16V to 4.20V? R101 is currently 73.2k in the Rev. 4.0.8 HVC/LVC board and it would need to be a bit higher.

Thanks!

 

[GGoodrum]

Sorry if I've missed this elsewhere, but what value R101 resistor would be required to change the HVC from 4.16V to 4.20V? R101 is currently 73.2k in the Rev. 4.0.8 HVC/LVC board and it would need to be a bit higher.

Thanks!

Try 75K.

 

[MitchJi]

Hi,

Sorry if I've missed this elsewhere, but what value R101 resistor would be required to change the HVC from 4.16V to 4.20V? R101 is currently 73.2k in the Rev. 4.0.8 HVC/LVC board and it would need to be a bit higher.

Are you sure you need/want to do that?

Actually, what many have found is that it does seem to prolong the life of the cells if you pick a "full" point a little below the generally recommended charge voltage for each type. For LiPos, the recommended number is 4.20V and for LiFePO4 it is 3.65V. What I'm using for LiPo is 4.15V and for LiFePO4, 3.60V. This is what we are fine-tuning the fully on shunt set point to be at.

For my 12s LiPo test setup, the total would be 12 x 4.15V, or 49.8V. What you want to set the charger voltage to is just a hair over that, to get the best results.

In your application is it really worth finding a non-standard part that will reduce the life of your cells to squeeze every last drop of range from your pack?

 

[mwkeefer]

Mitch,

I have to assume that sooner or later that 4.2v ceiling will be the conservative cutout point with all the new Lipo nano this and flexible pouch cells... Just saying while it may not be a "super" idea, knowing how to do it precisely is important I suppose.

Personally I prefer to charge to 4.18v but I find that sticking with 4.15/6 gives me a better safety margin in case of drift, etc. So I stay below to provide some measure of protection for my cells... when parallel bulk charging at 6S - I push them to 4.18 and even 4.19v sometimes... since I am charging at about 1/3C (into 6S30AH) and using a fairly tightly calibrated balancing charger (iCharger) there is little to no chance of exceeding 4.2v per cell

-Mike

 

[rscamp]

Try 75K.

Thanks, Gary!

 

[rscamp]

Hi,
Are you sure you need/want to do that?

In your application is it really worth finding a non-standard part that will reduce the life of your cells to squeeze every last drop of range from your pack?

Mitch.

Yes and yes, but not for the reason you assume. Without mods, I can't get the charger I am using to terminate lower than about 4.18V/cell. This trips the HVC of the BMC off the charger.

But there's more than one way to skin a cat. If anyone knows how to (easily!) lower the termination voltage or increase the fast charge termination CV current on a TP1010C I'm all ears!