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

[pm_dawn]

Back to the topic of the BMS design:

There's probably a whole new thread worth of discussion to be had on this, but what about making the LVC adjustable for cold-weather usage?

My understanding is that LiFePO batteries can tolerate a slightly lower LVC when they are close to freezing, and they also have more voltage sag. An electric vehicle that operates fine in warm weather might, under cold-weather conditions, hit the LVC while the batteries still have a reasonable charge level and under reasonable current load. It seems that lowering the LVC slightly might make sense.

Is this at all practical? Could the BMS circuit accommodate two LVC paths, and a switch to toggle between them? Or perhaps send out two different signals when different thresholds are reached? Or is this just a bad idea, and we should always base our LVC on warm weather thresholds?

Hi !

The MiniBMS has this feature.
http://www.cleanpowerauto.com/MiniBMS.html

I'm pretty sure that the circuit is pretty much open source and can be found on DIY here:
http://www.diyelectriccar.com/forums/showthread.php?t=38863&highlight=minibms

I would love that feature i the V4.x but I guess It will not be in.

Regards
/Per

 

[AndyH]

Temperature compensation for LVC...A Step Back for a Moment

I would like to know from a battery expert whether it's best to make LVC a 'hard limit' or allow the cell to sag below the LVC under load.

As I see it, again - still from an attempt at a 'big picture overview' - the real point of LVC detection is to try to keep a cell from dropping below a state of charge limit by monitoring a second-order voltage. This is just an approximation

The benefit of a simple LVC detection that provides a 'hard' minimum in any operating condition is that it will keep a cell from over discharge and it's cheap/easy to implement. The down-side - at least for larger and/or low-quality/inexpensive cells is that these puppies sag. A lot. So we can loose performance over all states of charge when the pack is cold, and during the last 1/2 of charge even with a warm/hot pack.

Temperature compensating an analog detector - like the TC54 - can lower the trip voltage a bit when cold. That will give us a bit lower 'hard floor' but might not be worth the tradeoff in added parts. For example - Dimitri added temp compensation to his Mini-BMS. It only moves the set point 200mV, and measures air temp, not cell temp. It costs one thermister per channel.

One feature the Mini-BMS does have is a "2-4 second" LVC delay. It still detects LVC activation, but 'hides' it for some time just in case the voltage drop is sag under acceleration. I don't know how the delay was determined so can't say how effective it might be at stopping 'nuisance' LVC alerts (provided one thinks that it's OK for a cell to sag below the LVC set point - otherwise it's not a nuisance...).

I can see both sides, as my range drops in the winter and the hard 2.7V LVC I have on my TS pack can start tripping under hard acceleration rught off the charger (my max accel load is 1.6C). .2V wouldn't be worth my time to install - it might delay winter LVC trips about 100 feet. I'm sometimes tempted to swap the 2.7V TC54s for 2.1V detectors, but then would have to accept that voltage not only under acceleration but as my new '0% SOC approximation' as well.

So..battery pros...please tear this apart. And for a conservative user - is it better to have a hard-limit for LVC or allow voltage to sag below LVC on accel?

Andy

 

[webfootguy]

As I see it, this is a case of nice in theory but fails in practice. Different cells sag differently (some a lot some not so much) and it varies with temperature. So you measure the temperature and allow a lower cutoff voltage at low temps as a start. To do even better, you monitor the current draw so you can allow lower cutoff voltage at high current.

What happens when you are near the end of your capacity and you try to accelerate. If one cell is empty, its voltage will drop but due to the current draw, you ignore it. You could reverse / damage the cell. Even worse if you let low temps lower the cutoff. Maybe a micro can attempt to sort this stuff out but even then the battery makers do not seem to tell you how to set the cutoff versus temp and versus current draw. If you had those specs, you could do something about (even in the analog world) but clearly this would be easier to have a micro making such decisions.

Maybe the best thing is to design for the most popular cells and recommend heating your cells in winter (only works if you charge before every use). A "good solution" rather than a perfect solution might work well enough.

I think I would prefer a solution that uses PWM/FET to reduce current draw until any cell was above the cutoff. This would gradually cut back until the pack could not supply any more voltage/current and remain above the cutoff.

 

[chroot]

@Gooddrum - The PCB arrived from the tppacks. Thank you and am going assemble it soon.

 

[chxs]

Any update ???

Tanks
Chris

 

[fechter]

We are still experiencing problems with certain aspects of the boards. Not quite ready to roll out yet

One thing is for sure, these are much more complex than the earlier versions and Murphy's law is taking its toll.

Gary, Andy and I are also getting pretty burned out trying to rush this thing along to the point where it can get the Fechter Seal of Approval. No seal yet, but all the problems seem solvable.

 

[fechter]

Another weekend of bizzare behavior and burned up parts. At one point I had cell circuits cranking out a couple of watts at right around 100MHz. Cool if you want to make an FM radio jammer

Hmmm... what if we just radiated the excess power instead of turning it to heat in a resistor? That could solve the heat issue.

OK, the FCC would probably frown on that approach.

I also found some impressively large (>50v) voltage spikes on the input to the control cicuit during throttling which I assume is largely due to the inductance in the wires. My guess is that you could get equally large spikes from a high power motor controller during discharge. Some serious filtering is needed to keep this stuff out of the control circuit. Ringing is not too bad but I was surprised at the amplitude of the spikes. I'm pretty sure spikes had something to do with component failures during testing. All the sections seem to work so nicely alone, but blow chunks when you put them all together. The complexity of the interactions is not something that is easy to calculate or predict and has to be validated by testing.

I haven't given up, and I feel like I'm getting close to a stable, reliable configuration, but it's just not quite there yet. I think I need to burn up a few more parts first.

 

[ejonesss]

fechter if you can reduce the frequency to the 73 to 74 mhz range then you would be in the radio astronomy (hunting for et and alien signals) probably non essential so if seti gets false hits you should not have problems.

however you then are close to frequency used by rc hobbyists and may cause problems with hobbyists operating in the area.

http://www.ntia.doc.gov/osmhome/allochrt.PDF

also i think in an emergency all frequencies will be commandeered for the message for example an evacuation order or a terror alert may be broadcasted in redundancy across the entire spectrum so you would only lose fm 100

of course who knows what else would be affected.

 

[velias]

Dude, just go ride your bike and quit thinking too hard.

 

[amberwolf]

Hmmm... what if we just radiated the excess power instead of turning it to heat in a resistor? That could solve the heat issue.

Hmm...beam the power to the accessories on the bike.

 

[ejonesss]

@fechter pm sent

 

[fechter]

Well, some strategically placed capacitors seems to have the oscillations under control but they only seem to happen at certain load/wiring conditions and it's hard to test all possible conditions. At 1A shunt current, the regulator just seems inherently unstable and I've had to resort to kind of a brute force approch to keep it happy. I think there must be a more elegant way of creating stability. This is really a servo loop stability kind of problem that is typically solved by using some kind of PID arragement. A real PID loop would be overkill for a shunt circuit, but the basic principle still applies.

One thing I've noticed is the loop gain seems very high with the boosted LM431 setup. Stability is usually improved by reducing the loop gain. I'm going to try to figure out how to put a little negative feedback into the 431 input to lower the gain.

 

[methods]

Sounds like black magic to me
For my test setup lately I have been using 15' of 10awg charge cable and 50cm of balance tap extensions to try and get some worst case noise.
Not seeing much difference - but then my PWM is running at 3hz ... not 20khz... so I guess it is not much of a test.

On V3.0 (Boards will be back next week) I am trying out some radial mounted wire-wound resistors electrically similar to the parts you are using.
I have some 6.1V Zeners in parallel that I hope will do double-duty as freewheel diodes for the inductance.
I seem to remember a ways back where someone suggested free-wheeling the resistors in your design....
Did you get a chance to try that?

(someone also said in my LTC thread that they used freewheel diodes with WW resistors for their design)

Also (and I know this is cost prohibitive) did you ever get a chance to try a set of non-wire wound resistors? Would be interesting to see the change.

I could not solve the 30khz PWM noise issues in my design (didn't even really try to tell you the truth) so I eliminated the problem area by opto-isolating each part of the circuit.
Any chance of opto-coupling the PWM?

-methods

 

[fechter]

The wire wound resistors definitely have enough inductance to make a nice spike on turn off. A freewheel diode or resistor, or even a cap can snub it. I used metal film resistors and saw no problems like with the wirewounds. Since my circuit is analog, the feedback from the input to the output is pretty fast, so it was possible for it to set up an oscillation. If I put one wirewound in parallel with a metal film, it worked OK. I think some kind of additional negative feedback would improve the stability to the point where it wouldn't ever oscillate.

Seems like with a microprocessor, you would have a lot more delay and you could gate the sampling to avoid picking up the output.

I'm not shure what you mean by opto coupling the PWM. If the charge control circuit is throttling the current with a 30khz PWM, there is going to be some ripple on the cell voltages. Since the frequency is fixed and pretty high, it should not be too hard to filter it out of the measurement circuit. The PWM on the cell shunt might be what you mean? That could run way slower than 30khz.

 

[amberwolf]

Regarding the v2.6 BMS, I have some parts substitution questions AndyH referred me to GGoodrum for. Rather than PMing, I figured I'd ask in the thread, in case anyone else might also know, and so the answers might help others out, too.

As far as I can tell, the following parts might be usable substitutes for certain v2.6 parts. But I may have missed something important, so before I order any or use parts I already have and let the smoke out, I figured I should ask. I've been known to forget to ask certain important questions, to my later regret.


Avnet shows an alternate part for the ILD2 that costs only 17 cents each, the LiteOn LTV-829 (which despite what their site shows is an 8-pin PDIP, not a 6-pin).
ILD2:
http://pdf1.alldatasheet.com/datasheet-pdf/view/45330/SIEMENS/ILD2.html
LTV-829:
http://pdf1.alldatasheet.com/datasheet-pdf/view/167406/LITEON/LTV829-V.html
They appear to be pin-compatible; not sure if any of the other spec differences matter in this application?


Instead of the BD136, might these work?
http://www.onsemi.com/pub/Collateral/MJE521-D.PDF
http://www.onsemi.com/pub/Collateral/MJE371-D.PDF
I have some of these already, so it'd save significant money. Same form factor, Vce 40V, Ic 4A. Not sure if there's enough for the whole BMS but I'll have to dig them out and see. Also might not be high-enough Vce.

For the FK20X7R1C106K (10uF 16V MLCC cap) does it *have* to be MLCC? Or can a same-value non-polarized cap of another type be used?

Same question for FK28Y5V1C105Z.

Same question for SA105E104MAR.


Can the IXDF404PI
http://ixdev.ixys.com/DataSheet/99018.pdf
be replaced with just a regular push-pull discrete PNP/NPN transistor driver, wired into where the chip would be?
Or would either of these sub for it?
http://www.onsemi.com/pub/Collateral/NCP5106-D.PDF (not pin compatible, but I can dead-bug wire it, and have both A and B versions)
or
http://www.micrel.com/_PDF/mic5011.pdf (I don't think this one will)
http://www.micrel.com/_PDF/mic5013.pdf (maybe?)
(again, they're not pin compatible but I can wire them in, and have them already).

Also, if anyone (preferably one person, because of shipping cost) has available the following parts I could buy, since shipping in a little envelope has got to be cheaper than what Mouser or Digikey, etc. charge.
2N5060 (2)
6.8ohm 5W (32)
IXDF404PI (2) assuming I can't sub it
BD136 (32) assuming I can't sub it
KSP94TA (4)
KSP44TF (2)

 

[fechter]

Most of your substitutions should be fine.
The transistors look great.

The opto coupler does not need to be anything special, but I did not check the pin out. If the LTV-829 works, that would be good to know, since I recall they are a bit cheaper.

The caps can be anything, but I preferred the MLCCs simply for mechanical reasons. Aluminum electrolytics need to be glued down or they could be prone to vibrating and breaking off.

The gate driver is another story. The IXDF404 is a dual low side driver. The outputs are complimentary, so when on is on, the other is off, but they are both low side. One half is pretty much wasted driving the LED. You could use any low side driver to run the FET gate, and use a dead-bug inverter made out of a transistor and resistor to drive the other LED half. There may be some other drivers configured like the IXDF404.

I don't think you could use a simple PNP/NPN totem pole setup to drive the gate since the input is not digital. We need the driver chip (or something) to square up the pulse so it's either full on or full off. The signal going to the input is pretty mushy.

 

[sandman]

Is it possible to connect one of the 4-cell breakoff sections from V 2.6c to the much older V1.5 (16-cell) such that I can have LVC/HVC for 20 cells? I am using LiFePO4's.

Thanks,
Dick

 

[mwkeefer]

Gary,

I will be receiving the LVC detection IC chips tommorow... and some other component from digikey from one ofthe BOMs - JST connectors I think... They had a good price break at 100 units so I figured what the heck - once I break out what I need for my 4.0.6 to 4.0.7 boards (one for a 15/18S lipo pack and one for a 24S lipo pack with a 12-16S LiFePo to follow) I will be happy to offer the remainder to those trying to build these at my cost + s&h (I'll post to a new thread of "Components and Small Parts for tppacks 4-24 cell BMS" (if that's okay with you)... this way people getting enough for say an 18S pack can share the discount I can get at 500-1000.00

I was wondering if you would be so kind as to send me
3 x more PCB plates of LVC/HVC breakouts (24 total)
3 x more PCB plates for the Charge Controller / Battery Medic Booster (which I have managed to adapt to some low (I mean low) cost 6S balancers... When the 8S version of your HVCLVC is available... I will be moving onto that since it will require only 2 of those and 2 8S balancers with the booster circuit.

Also could you make each of these the kits without the mouser parts (assuming that includes the mcmaster hardware)... I will once completed order all the required bits from mouster, all electronics and so on... build up an inventory of the BOM parts and replacements / substitutes if it's okay with you... We could maybe work something out in PM

I was hoping to get the newer .7 version of the boards, the changes were easy enough but time consuming...

I could also use some of the shrink wrap for doing packs you offer... I am building 2P 5 and 6S clusters then series from there... would your wrap be sufficient to handle that dimension?

A suggestion for your next run of the LVC/HVC boards... I would eliminate the perf edging and the 2.59mm clearance... even the worst hacksaw would need only 1mm of clearance in a miterbox a dremel and guide woudl require far less... The thought behind this is an 18S pack would require much less room (well not much less but it would be cleaner looking than I can acheive now without independing each like vPower BMS does.

A question - once these boards reach the HVC for the cell (at cell level)... instead of just kicking off the PWM which isn't working 100% right now (10A current hit with a 10% duty cycle due to a cell at 4.16 instead of the rest beign at 3.9) modulation will toast the speced fets...why not add a simple resistor bridge similar to your booster and instead of worrying about any potential hysterisis... just engage the drain/load for x milliseconds on timer when triggered. This would basically be the same as the booster board (but could be done so much smaller and cleaner) but would mate with the hvc/lvc via jumpers/pin headers of some sort (perhaps the .5mm gold plugs would be best) and when the HVC is triggered it would cause the drain to go active for say 2000ms, then disengage... on next cycle if HVC is still an issue... again 2000ms and so on.

I know this would compicate issues but that complication would be in the drain module and would only need a trigger of 5v from the HVC trap?

You seem to know the extruded cases well... I have some PCBS I want to mount nicely and in smaller enclosures than they currently are mounted in... is there some easy way to define the edge parameter of a PCB (length, width and thickness/guage) and the minimum spacing above and below the PCB to get a proper enclosure recommended? I assume it exists (the calculator / engine) but don't know where to look!

Final question of the night... Do you have ANY of your older or newer BMS (full BMS) available (even just PCB form that I must correct)... I am dying to get a preview of one = )_

Regards,
Mike

As always, name a price (PM) and I will remit paypal...

 

[amberwolf]

The transistors look great.

Good--that saves significant money right there.

The opto coupler does not need to be anything special, but I did not check the pin out. If the LTV-829 works, that would be good to know, since I recall they are a bit cheaper.

A *lot* cheaper, actually--about 1/3 the price! At mouser the ILD2 is $1.32 each and the LTV-829 is $0.40, in qty of 1. In quantities like you'll be using, like 1000 at a time would be $0.613 vs $0.236 each.

Pinout is identical, AFAICS. Specs appear the same for the important stuff; you'd have to verify that. I'll chance it since the price difference is huge; I can't afford to buy enough if I had to get the ILD2; that'd be over $42 just for those parts, for the 32 I'll need. Even at 1/3 the price, it's almost $13, which is a lot for me.

I had originally thought I'd try to order from Avnet, but they won't sell less than 2000/2500 of them at a time per the parts pages, and I cannot get anyone there to answer any of my requests to buy anything (even though they sent me a coupon for free shipping!?!? and I've gotten stuff in tiny quantities thru them before). So wherever I order from it'll cost me a lot of shipping, too. Pretty sad since I think everything I need would fit (as separate parts not in tubes and stuff) into a small bubble pack envelope, at most.

Now that I know for sure what I need, I'll repost the parts list of what I still need as a want-ad in the marketplace section.

The caps can be anything, but I preferred the MLCCs simply for mechanical reasons. Aluminum electrolytics need to be glued down or they could be prone to vibrating and breaking off.

Ok. I'll deal with that as I need to.

The gate driver is another story. The IXDF404 is a dual low side driver. The outputs are complimentary, so when on is on, the other is off, but they are both low side. One half is pretty much wasted driving the LED. You could use any low side driver to run the FET gate, and use a dead-bug inverter made out of a transistor and resistor to drive the other LED half. There may be some other drivers configured like the IXDF404.

Ah; for some reason this did not sink in till now.

I don't think you could use a simple PNP/NPN totem pole setup to drive the gate since the input is not digital. We need the driver chip (or something) to square up the pulse so it's either full on or full off. The signal going to the input is pretty mushy.

I could use a comparator if I have to, that then triggers the totempole. Or perhaps a darlington totempole, which ought to switch harder, IIRC. Eh, I'll figure it out. I only need to make two of them. Just have to check the IXDF404's spec for the trigger voltage. If one of the other drivers I have will work for that part, I'm all set.

I *think* the second Micrel driver will do it, but I have to dig deeper to find out.

 

[fechter]

The IXDF404 trigger threshold is around 1v. You need something with a fast rise time to square the pulse. Just about any logic chip could do this. If you used an inverter chip, this could also be used to invert the signal for the other LED. I've also thought about using a TC54 to square the pulse and then you could drive a simple totem stage.

Good to know on the LTV-829 pricing. I don't see any reason why it shouldn't work. I know I looked at using these before but didn't want to risk that it would screw something up, which is still a remote possibility. The ILD2 was inhereted from the very early designs and I just never got around to experimenting with alternte parts.

In the Ver.4 designs, we're sharing the HVC and LVC lines, so the number of devices is cut in half.

One other thing to consider would be to add some .1uf caps across each cell circuit to suppress voltage spikes. These could be stuck directly to the board or possibly attached to the legs of the power resistors. I think the best location would be close to the LM431 and TC54, since those are the parts that will blow with a spike.

Alternately, unidirectional TVS diodes could be placed across each cell. I haven't experimented with these yet, but from reading a lot of datasheets, it seems the lower voltage units (5v) have a fairly high leakage current when off. The 10v devices are way lower, like maybe 10uA, so they may be preferrable. I think all the parts in the cell circuit can withstand a 10v spike.

 

[fechter]

Ver.4 Update:

Preliminary testing with the latest revisions looks promising.
Gary had a capacitor blow up in his face and nearly took out an eye while testing one theory. As a result of this test, we've concluded that on board current limiting to allow using a 'really dumb' Meanwell-type power supply is not going to work with the present topology. The other functions appear to be working correctly, but we still need to do more testing.

We'd still like to have something for using the hiccup mode Meanwells, but I think we will make this a separate unit that goes in line with the output of the supply. Basically, this would be a buck converter that is regulated by current rather than voltage. Since I'm not much of a switching power supply designer, I have to do a lot of reading to figure out the values needed.

I think the spec will be for something that can handle up to 20A and be adjustable down to something like 2A with a pot. 100v rating for now, as higher rated FETs don't perform as well and most of us won't need more than 100v. You could always use multiple units in tandem for larger setups.

The main thing to nail down is the inductor and switching frequency. 20A inductors tend to be a bit pricey, but if I can get away with something in the 10uH range, it's not to bad. Most of the time, the voltage differential will be quite small and as the cells approach full, it should go to zero.

Maybe one of you professional power supply designers out there could give me some hints on the design...

 

[ejonesss]

wow that was close maybe a good idea to use goggles or have a sheet of clear plastic over the board.
we users would hate to see this project to be put on hold because of a hospital trip and months of recovery time.

i had nearly poked my eye out once stripping wire for recycling.

i was stripping telephone wire by drawing it between my teeth to weaken the coating and the end came across my eye.

fortunately it was just an abrasion and took a couple years to heal to the point of no reoccurring phantom pain.

Ver.4 Update:
Gary had a capacitor blow up in his face and nearly took out an eye while testing one theory

 

[GGoodrum]

Luckily, I'm old enough to require reading glasses, as that probably saved my eye. As it was, I still got some of the inside bits from a 10uF electrolytic cap in one eye, but they washed out. The cheap $3 Costco reading glasses are pitted though. Good thing I got the 3-pack.

Anyway, Andy has now had some successful tests with the latest changes, using real cells, and he made it through with none of the weirdness that has plagued this effort for the last several months. As Richard says, we are going to move the current limiter function to a separate board that can mount to the terminal block on the front of the MeanWell supply. I successfully tested changes yesterday to the standalone Charge Controller that makes use of the trim pot used in the current limiter to now control the low current set point for the auto-shutoff function.

-- Gary

 

[amberwolf]

Maybe one of you professional power supply designers out there could give me some hints on the design...

You might page Solcar:
http://www.endless-sphere.com/forums/memberlist.php?mode=viewprofile&u=14958
he might be able to help.

 

[fechter]

Thanks Amberwolf, I'll check that out.

Here's Gary's picture of the exploded cap:

Dang, I just melted the end off of a nice grabber hook on an overheating resistor. This project can be measured by the number of destroyed components, which is considerable at this point.