• Howdy! we're looking for donations to finish custom knowledgebase software for this forum. Please see our Funding drive thread

Thundersky cells and BMS

ryan_lirui

10 mW
Joined
Oct 15, 2009
Messages
30
Location
Shanghai
I'm looking into what kind of BMS i'd need for my scooter project. I've already decided on buying some Thundersky cells, 20 of the TS-LFP40AHA in series - for the Kelly Contollers 72V 8KW regen kit. Looking at the thundersky website, it looks like they have a "Mini BMS" that could work with this: http://www.thunder-sky.com/products_en.asp?fid=95&fid2=96

Does anyone have experience with this? Do you think there is a better alternative? Also, do you think that 20 cells are going to be enough (that's the limit of the MINI BMS)?
 
By the way, I forgot to mention that Thundersky quoted the MINI BMS at 2800 RMB, or $410. I'm beginning to think that maybe they got mixed up an quoted the fully featured color screen BMS rather than the mini one because that sounds really extravagant when you compare it to the 8-12 cell BMS on www.bmsbattery.com (even if it can handle less batteries).
 
The guy who made Voltblochers (Brian) had had the ThunderSky BMS in his Honda S2000 with ThunderSky cells. As far as I remember he said that the BMS sucked. It was reporting bad cells and cutting the current off even dough they where fine
 
You could always build one of my BMS circuits.

I wish we could offer a pre-built version. Maybe someday....
 
watching this guy from evtv.me makes me wonder if even having a BMS is neccesary. Basically saying that if it's battery management, then it generates too much heat and creates a fire hazard. If it's battery monitoring, then it's too inaccurate for practical real world usage since there's lots of noise, etc, etc.

I was really hoping for something that could be pre-built since I'm not that experienced in creating circuits. While I don't mind using my computer to set things up on the software side, I am hoping I can have some sort of set up that doesn't obligate me to micro-manage it after i set it up the first time.

if your'e interested in the video i'm referring to, you can check out http://www.evtv.me, scroll down and play the video entitled "Porsche Battery Upgrade and Battery Management Issues". Be prepared, because it's a long video. I'd suggest scanning to the middle of the video if you want to get straight to BMS.
 
ryan_lirui said:
watching this guy from evtv.me makes me wonder if even having a BMS is neccesary. Basically saying that if it's battery management, then it generates too much heat and creates a fire hazard. If it's battery monitoring, then it's too inaccurate for practical real world usage since there's lots of noise, etc, etc.

I think that he should keep his opinion for himself.

When I was charging my ThunderSky pack for the first time without a BMS some cells had 4,25V, others had 3,9V. Imagine what would happen after 100 charges/discharges.
 
I don't have to imagine it. I've done it. Nothing happens. In fact, this entire concept of active balancing of cells during charging not only does NOT prevent damage to your cells, generally it does just the opposite - damages them irreperably.

I've just this week completed a fairly exhaustive series of "active balancing" tests and the results were disastrous - 3 totally dead cells. And NOW I finally know why. It is the active balancing itself that kills them. I JUST posted this weeks FRIDAY show with the results and you'll be very surprised at what we found, and what we figured out regarding the entire concept of battery management, but specifically the issue of active balancing and shunt regulators.

And I'm rather given to spreading my opinions as widely as possible. I notice you're not inclined to keeping your opinions to yourself.

Check out the video. It's a kind of a detailed story. http://evtv.me.

Jack Rickard
 
jrickard said:
I don't have to imagine it. I've done it. Nothing happens. In fact, this entire concept of active balancing of cells during charging not only does NOT prevent damage to your cells, generally it does just the opposite - damages them irreperably.
...........
Jack Rickard

Thank you for going to the trouble of presenting your factual findings, your approaches, your methods. The video was great!

I for one, think that your observations and deductions regarding balancing Thundersky's, towards the lower end of the cell voltage, rather than at the uppermost level, may well prove correct. I certainly can't fault the logic. Actually from what I understand about the chemical reactions of LifePO4 , and the anode and the cathode, and where the chemical reaction during over-discharge is no longer reversable, your insight just makes sense.
Edit:
http://en.wikipedia.org/wiki/Lithium_iron_phosphate
http://en.wikipedia.org/wiki/Lithium_iron_phosphate_battery
http://pubs.acs.org/cen/science/84/8424sci1.html
 
jrickard said:
I JUST posted this weeks FRIDAY show with the results and you'll be very surprised at what we found, and what we figured out regarding the entire concept of battery management, but specifically the issue of active balancing and shunt regulator

What kind of battery chemistry are you using? I am looking into building an a123 pack and just wonder if it would work for a setup where you have a bunch of smaller cells in a pack would work the same way. Fridays post says 1 of 6, where are the other five?
 
gestalt What kind of battery chemistry are you using? I am looking into building an a123 pack and just wonder if it would work for a setup where you have a bunch of smaller cells in a pack would work the same way. Fridays post says 1 of 6, where are the other five?

watch the video. What he says has applicability to Thundersky cells.

Edit: followed your blog link and a comment about taking appart what looks like an Axial motor
....... before attemting to open your motor - watch this u-tube video for clues.
http://www.youtube.com/watch?v=xbVTqkE_F1U
I'm gonna say that this needs "tooling" to come appart safely.
 
gestalt said:
jrickard said:
I JUST posted this weeks FRIDAY show with the results and you'll be very surprised at what we found, and what we figured out regarding the entire concept of battery management, but specifically the issue of active balancing and shunt regulator

What kind of battery chemistry are you using? I am looking into building an a123 pack and just wonder if it would work for a setup where you have a bunch of smaller cells in a pack would work the same way. Fridays post says 1 of 6, where are the other five?

At http://evtv.me there are no parts. It's one hour long video. On YouTube, I am limited to 10 minutes per segment. So I post the first 10 minutes and a link to http://evtv.me.

Thundersky and Sky Energy are the cells I work with. This test was done with Thundersky LFP160Ah cells.

Jack Rickard
http://evtv.me
 
Great video. Made the whole problem much clearer. The max voltage problems have definitely been very hyped and I've been worrying about getting my cells balanced at the top when it's not a serious problem.

Interesting link discussing some of what you talked about.

http://liionbms.com/php/wp_lovtg_cutoff.php

I think the simplest way to help mitigate this problem is to match cells in strings. Let's say you have a 24S 4P setup. Split the cells into 4 groups of 24 with similar capacities. Then put cells with the same capacities together in strings. The stronger strings will essentially pick up the slack as the weaker strings fade and prevent them from dropping too low.

Of course, if you only have one string you might need to work out some kind of Zener diode system. I'm trying to figure it out in my head, but no idea on how to negatively bias the voltage from two leads of a battery so it will activate a short circuit to the next battery. The gains don't seem worth the price and complexity.
 
I haven't watched jacks video yet, but I think there's a difference between just any BMS and a really good one. I'm curious to know how you managed to kill three cells with a BMS as I thought the BMS would cut power to the pack if one cell reached LVC? Without LVC I think you're just playing russian roulette and waiting for a cell to die.

If it was me I'd go for a 24s pack rather than the 20, and then get one of the fetcher/goodrum BMS's made to look after it. If you can't build one yourself get andy to build one for you, I have one of his at home and it looks great. I haven't tested it yet though as my headway BMS is working fine at the moment. I'd probably go with 500ma balacing and 2.7LVC for 40AH cells.
 
Jack,

Have a squizz at...... if you haven't done so already.

http://www.rcgroups.com/forums/showpost.php?p=13427812&postcount=1

Could a LVC & alarm be this easy? and this cheap?

http://www.hobbycity.com/hobbycity/..._Name=Cell-Log_Cell_Voltage_Monitor_2-8S_Lipo

I find it has software settable low voltage setpoints. It has an output to an "alarm" which can drive a latching relay or perhaps shut down the controller. Not only that but it logs everything.

Yes, you would need one cell logger for every 2x 4 pack of cells.
EDIT: Also a good thread going here http://www.endless-sphere.com/forums/viewtopic.php?f=14&t=12815

Also I think that USB opereational saftey would need to be checked (ie how "isolated is the USB to pack voltage, as each goup will be in a different part of the pack).
 
I will indeed have a squizz. I think you've found it man. These RC guys are ALWAYS way out ahead of the pack.

But I have to believe this is almost too good to be true. Correct me if I'm wrong. EIGHT cells, you can SET the upper and lower voltage alarm BOTH of them to anything between 1.3 and 4.5 vdc, it actually makes a current switched output to trip a relay on hitting an alarm voltage, it DISPLAYS all eight voltages, and it SAVES the last 36 hours in a 16MB memory you can download with a PC? And it's $27?

I ordered five of them. I'll smash one with a hammer. Burn one up with a power supply. And probably use the other three. It sounds WAY too good to be true.

You may have found the BMS of the year. As I say, it looks WAY to good to be true.

I could do the ENTIRE Mini with 112 cells with 14 of these and less than $400.

Jack Rickard
http://evtv.me
 
I have just bought 5 of these to put in my TS-LFP 160 ah converted Renault Clio.
I have the Fechter/Godrum BMS today.

Planning to put it in as soon as possible.
Regards
/Per
 
Lapwing said:
Jack,

Have a squizz at...... if you haven't done so already.

http://www.rcgroups.com/forums/showpost.php?p=13427812&postcount=1

Could a LVC & alarm be this easy? and this cheap?

http://www.hobbycity.com/hobbycity/..._Name=Cell-Log_Cell_Voltage_Monitor_2-8S_Lipo

I find it has software settable low voltage setpoints. It has an output to an "alarm" which can drive a latching relay or perhaps shut down the controller. Not only that but it logs everything.

Yes, you would need one cell logger for every 2x 4 pack of cells.
EDIT: Also a good thread going here http://www.endless-sphere.com/forums/viewtopic.php?f=14&t=12815

Also I think that USB opereational saftey would need to be checked (ie how "isolated is the USB to pack voltage, as each goup will be in a different part of the pack).

A friend is using these to monitor a 20 cell TS pack. The USB output is NOT isolated from the input. While they do work to monitor and log cell data, do not connect more than one to a USB connection at a time lest thou smite thy monitors, thy USB hub, and thy laptop.

Andy
 
I read all ten pages in the other forum, and I understand, and agree, with the notion that you can't just balance cells at the top end, and not do cell level LV protection. You might be able to get away with the latter if you charged until the lowest capacity cell was completely full, but this doesn't take into consideration the case where you simply have cells of equal capacity, but they just happen to be out-of-balance. In that case, the first cell to hit the cutoff is simply the fullest one. One solution would be to do like the RC balancers all do which is to balance to the voltage of the lowest cell, and then charge only to the point that the first cell that hits the HV point gets full. You can't just shut the charger off when that first cell hits the HVC point, as that would leave that cell less than full, but if you hold the voltage and let the current taper off, it will be as full as it is going to get, and the rest of the cells are going to be at various lesser voltages, based on capacity differences. All the cells should have roughly the same amount "gas in the tank", though, so assuming the IRs aren't too far off, and the loads aren't crazy high, they should all reach the empty point at roughly the same time.

I would still not be comfortable without cell level LV protection, though, but I can definitely see how you may not want to unilaterally cut the throttle in larger EV, like a car or a motorcycle/scooter. In a bike it is not such a big deal because you still have the pedals. :roll:

Anyway, figuring out how to make this automatic might be tricky, but doing the part where the first cell that hits the HV point starts the CV mode is pretty easy, I think, simply by using the "throttling", or PWM control of the charge current, that is in our current BMS design, but without the shunts.

-- Gary
 
As I said rather clearly in the video, I'm not only not opposed to battery management, now I have some inkling as to the kind I want. And it's basically as you describe. What I have been opposed to was these blind current shunt balancers and this concept of "top charging" which is a carryover from the lead acid guys. If you get your pack balanced at the top, where you're trickling in 10 amps of current, you UNBALANCE at the bottom, where you're draining 100-500 amps driving the car. Inevitably, the lower capacity cells hit the knee of the curve first and at a couple hundred amps are violently driven to zero volts.

If you HAVE to balance (and for example, marrying a new cell into the pack, you do) do it at the BOTTOM, so that all the cells hit the knee of the DISCHARGE curve more or less simultaneously. The knee of the CHARGE curve at the other end really doesn't matter as long as you don't overcharge the least capacity cell.

A LV protection circuit is clearly necessary. Now what does that look like? I like the cost structure of these little 8 cell units with the alarm. But isolation is indeed an issue with 112 cells. Do I need to isolate the alarm line? I have some on order so we'll see.

Linear makes a chip that will "monitor" 12 cells and talk on a bus with isolation. I would like to see this with an Arduino or something minding the string. But my fear is the communications bus. When driving a car with a 9 or an 11 inch motor, you're basically hauling a 50kW AM radio station around town. The EMI issues are pretty hard on TTL signals carried on little wires running through your car acting basically as "antennas." I can't tell you how many people have designed excellent battery "monitoring" systems in the lab and garage, who were just crushed to learn after a lot of hard work that they were fantastic in operation - as long as you didn't drive the car. Taking off down the road, the whole bus was swamped and went totally useless.

EVPOwer actually makes a shunt balancer that has some interesting characteristics. There is one circuit per cell and they are daisy chained with a single wire. This forms a loop that is open or closed. If any cell hits the low limit, this loop (isolated) changes state. Not digitally TTL style, more like a relay. It has shunts built in, but you could set it up to shut off the charger when the first one goes into conduction (I think). A module per battery is kind of cost prohibitive on a string of 112.

So the BMS issue is a wee bit harder than it looks. What it ISN'T is this concept of simple current shunt balancers. They're a fire hazard in the first place if done at larger current levels, and they are basically UNbalancing the pack in any useful sense. Active charge pumps are generally not a fire hazard, but they are terrifically expensive, and they still unbalance the pack rather than balancing them.

So we're left with a super fuel gage, that detects the state of the cell with the least capacity, and at least notifies that it's reached the knee of the discharge curve. If you want to be able to let your daughter take the car out, it really needs to disable the car or put it in limp mode or something similar. Ideally, if the car was limited to 15 miles per hour, she would have some indication that something was terribly wrong, but would still be able to creep along and get out of traffic, parked somewhere, or if close enough, creep home.

Jack Rickard
http://evtv.me
 
Hi !

Does anybody know if the Cell monitor can be daisy chained in the alarm port?
Or do I need to go Opto on the alarm port?

Jack:
I think a couple of the larger (better) controllers have the posibilty to run in "limp home mode". Ideally the BMS would pull that switch on the controller as soon as a cell comes below the fixed value.


I think that the active shunts could be useful acutally even for you Jack.
Set it up to shunt down to 2.95v and connect to the cell. Now you will have your simple balance at the bottom tool.
Connect and let it bleed the cells down to the 2.95v once and you will not have to spend that much time in manually bleed, measure, bleed, measure .........
I would rather put a couple of dollar in buying a RC charger that has a good logger and a fairly good discharge function. Pretty much like your West Mountian CBA.

Regards
/Per
 
What I'm thinking about would be to use the same basic building blocks we have now, but in a different fashion. Right now each channel uses two opto circuits, but they are both setup as outputs (one for the LVC and one that turns on when the shunt comes on...). What I want to do is use one as an input that enables the shunt logic to bring down the cell's voltage to some common low point (2.8-2.9V...). This will also turn on the output opto for that channel, which can be shared with the LVC circuit. Once the voltage reaches that point the shunt will go off, and the circuit will also turn off the opto output. The charger control logic will then watch the ganged output lines and wait for all of them to go off. It will then turn on the charge current. The cell circuit will monitor the cell's voltage and when it reaches the set point (3.65V?) it will turn on the otpo output again. The existing PWM/"throttling" logic will keep the cell at this voltage, just like it does now. The difference is that it won't turn the shunts on, so that the lower voltage cells can catch up. The current will taper off as this cell gets full, which also limits the current to the rest of the cells.

What this process should do is first get all the cells to a common "empty" point and then charge them until the lowest capacity cell is completely full. The rest of the cells should have roughly the same amount of stored energy so that on subsequent discharges, they should all hit the "drop off the cliff" point at pretty close to the same time. This discharge first procedure wouldn't have to be done with every charge, if you aren't hitting the LVC point on discharge. In fact you'd probably want to run down to close to this anyway, so it won't take as long for the shunts to get all the cells to the common low point. For normal charging, just the throttling portion would be active which simply stops the charge process when the first cell is completely full.

The LVC section of our BMS couldn't be simpler. If the voltage gets too low, it trips the opto. Since all of these are completely isolated they can be ganged together in parallel. On the bikes we simply use this signal to trip the brake input on the controller. I've done other versions that trip an active cutoff FET (or FETS...) and that trip an automotive relay. In any case, hundred of these can be paralleled and they are extremely reliable.

-- Gary
 
Using the LVC detection leads in parallel is exactly what I plan on doing. In my finished car I'll have two 24 channel G/F BMS systems and two 24S packs attached in series to make my car a ~144 volt car. The LVC circuitry will be wired in parallel between the two BMS systems. When charging I am going to split the pack into two separate ~72 volt packs and charge them in parallel. This is how I'm working around the voltage limitations of the FET (STP160N75F3) that the kit recommends. It was also easier to configure a ~72V charger for LiFePO4 than a higher voltage charger.

The G/F BMS-es will be wired in to the packs at all times. When the LVC trips I will do something with that signal. I'm not exactly sure what that something is yet; my options are from something as simple as trip a bright idiot light on the dashboard to placing it in the control circuitry for the main contactor and interrupting the controller current. Either way, or something in between, will work fine. If I go the interrupt-the-controller route I'll have to use an MPS 2222 + 7805 or some other gate or Darlington to drive a relay that interrupts the contactor's coil. There's still some thought and design to be done there.

It looks like the ILD2 likes a 5 volt signal and ~50 milli-amps so it is very much just a logic switch. They appear to be completely isolated from the rest of the BMS circuitry, so grounding them to the 13.8v chassis electrical system shouldn't be an issue.

Cheers,
--Adam
 
GGoodrum said:
What I'm thinking about would be to use the same basic building blocks we have now, but in a different fashion. Right now each channel uses two opto circuits, but they are both setup as outputs (one for the LVC and one that turns on when the shunt comes on...). What I want to do is use one as an input that enables the shunt logic to bring down the cell's voltage to some common low point (2.8-2.9V...). This will also turn on the output opto for that channel, which can be shared with the LVC circuit. Once the voltage reaches that point the shunt will go off, and the circuit will also turn off the opto output. The charger control logic will then watch the ganged output lines and wait for all of them to go off. It will then turn on the charge current. The cell circuit will monitor the cell's voltage and when it reaches the set point (3.65V?) it will turn on the otpo output again. The existing PWM/"throttling" logic will keep the cell at this voltage, just like it does now. The difference is that it won't turn the shunts on, so that the lower voltage cells can catch up. The current will taper off as this cell gets full, which also limits the current to the rest of the cells.

What this process should do is first get all the cells to a common "empty" point and then charge them until the lowest capacity cell is completely full. The rest of the cells should have roughly the same amount of stored energy so that on subsequent discharges, they should all hit the "drop off the cliff" point at pretty close to the same time. This discharge first procedure wouldn't have to be done with every charge, if you aren't hitting the LVC point on discharge. In fact you'd probably want to run down to close to this anyway, so it won't take as long for the shunts to get all the cells to the common low point. For normal charging, just the throttling portion would be active which simply stops the charge process when the first cell is completely full.

The LVC section of our BMS couldn't be simpler. If the voltage gets too low, it trips the opto. Since all of these are completely isolated they can be ganged together in parallel. On the bikes we simply use this signal to trip the brake input on the controller. I've done other versions that trip an active cutoff FET (or FETS...) and that trip an automotive relay. In any case, hundred of these can be paralleled and they are extremely reliable.

-- Gary


I like this thinking. The important point is to balance at the bottom, and fill to the capacity of the least capacity cell. In reality, we're seeing 5-7% differences in capacity. They're not all over the place. But by top balancing, you maximize the variations at the bottom, which is NOT the place to have them. Your plan would automate bottom balancing, and fill to the capacity of the least cell. I like it.

Jack RIckard
http://evtv.me
 
Back
Top