Overcoming Zener Knee for BMS

Sunder

10 MW
Joined
Sep 6, 2011
Messages
3,054
Location
Sydney, Australia
Hi All,

I briefly brought up the idea before of using a Zener Diode as a crude BMS, when a proper BMS can't be found.

I'm still hoping to build a 55S LTO battery, but finding any BMS that is programmable for LTO is rare, and up to 55S, as hens teeth.

I was told before that low voltage Zeners can't really act as BMSes for batteries, because the knee is too soft. Since then, I've thought - The knee may be soft for the diode, but it's very hard for the battery, and LTOs tolerate over-voltage very well. I've had one at 5v for weeks, with no long term damage.

So my new thought is to pick a Zener diode which has the reverse voltage over the max charge voltage, and use the leakage as a FEATURE, not a bug. Hear me out.

My base Zener would be 3.0v

According to the spec sheet, at 1v, the leakage is in the micro-amp range. So unless I am leaving the bike uncharged for weeks, no chance of completely flattening the battery.

At 3.0v, the LTO battery is still within safety limits, but may be very slowly losing its life. But at this point, it's fully open, with an impedance of 20ohm. 3.0v @ 20ohm = 0.15a - about the same as most BMSes bleed off, and adequate given there's almost no capacity above 2.55v. A drain of 0.15ma would bring the pack down fairly quickly: 5% of a 30Ah pack = 1.5Ah, @ 0.15A, = 10 hours, or basically overnight.

Hopefully, above 2.8v, the leakage is starting to get serious. Whatever it is, doesn't matter that much, as long as it's somewhere between the 0.15ma above, and nothing. So, we can actually use the knee as part of the design.

But will it keep draining during the bulk of the nominal voltage range. Being pessimistic, if the voltage leak at 2.6v is half of its rating, I'd be leaking 0.065A - it'd take me over two weeks to drain the battery flat. I ride my bike almost daily, and recharge very 2-3 days.

So... I've half convinced myself that while not ideal, it's more workable than with other battery chemistries. What do you think?

The only concern I have is that Zeners fail closed (short circuit), but a 1 amp fuse is easy and cheap enough, so if the Zener fails closed, the fuse blows.

Does anyone have any comments on this idea?

Thanks in advance.
 
Its a stupid idea.

Even if you got the cells for free it is still a very expensive thing. A programmable bms dedicated to charging duty only is about 100 bucks for a 32S model. Just spend the money for 1 part, have the battery in the correct voltage range and everyone lives happely ever after.
 
I appreciate the reply, but I don't believe you understand the problem.

The only BMS I can find is about $1400USD, but unfortunately is way too large to fit on a bike.

The only question is whether my solution works, not whether you can buy smaller BMSes for more common batteries for cheap. Trust me, even if the BMS was $1400, if it fit on my bike, I'd still get it.

This is about solving a very specific problem.
 
this one works from 2.6V, probably even lower:
https://www.aliexpress.com/store/product/16S-to-32S-2018-new-DIY-Lifepo4-li-ion-50A-80A-100A-110A-120A-smart-bms/1821822_32847205125.html
for the price it can be worth the trouble of ordering one.

another option is to double up. have one BMS connected to all the even cells and the other one to the uneven cells, so each bms balance line is doing 2 cells instead of one. that way you still get even balancing on all cells but can only use it for balacing, not other stuff as you need to keep the rails of both bms separated.
 
Look, mate, I don't mean to be rude, but I would rather stick to the original topic.

This battery is buried quite deeply inside the motorcycle, and managing a 36S LiPo battery with 3 x 12S BMSes, has been a nightmare. Two 12S packs will be shunting off energy, while the last one is struggling to balance. Once you have even one cell out, the three packs become unbalanced... And imagine that with an extra 19 cells.

If anyone can comment on whether my solution is workable, I would appreciate on-topic comments.
 
This isn't a Battery Management System, you are really talking about an overcharge balancer only. A BMS provides more features like low voltage disconnect that will be missing. LTO batteries may not mind being fully discharged, but how do they respond to being charged in reverse when the load current continues after one cell is discharged, effectively charging in reverse at high current? If such protection is needed then perhaps different approaches should be considered.

Each diode you buy will have somewhat different parameters, and temperature sensitivities. You may have to select them from a large batch to get units that match.

Zeners are not very good at dissipating heat. At full charge voltage you need to insure that the heat is managed. Current and heat rise quickly as voltage increases. The rising temperature will shift the operating parameters of the zener as well. Keeping the zener within the safe operating area may be difficult.

Have you found good data for the zeners you are considering? You may have to buy a few and set up a measurement system to get the data you need.
 
I thought about using some big blue LEDs instead of a zener. Most of them come on around 2.8v. You could also visually check balance by looking that the LEDs. You would still need some mechanism to prevent charging with more current than the LED or zener can handle.

Another option would be some regular rectifier diodes in series. About 4 in series would be close to 2.8v. They are pretty cheap and can handle pretty good current with proper cooling.
 
Have you looked into something like a HCX-D170V1 from BestTechPower? They can adjust the dividers on most of their BMSes and advertise typical ranges for different chemistries. They may be willing to run the numbers to make you an LTO variety if you spec the target voltages. The D170's are claimed to be programmable with comms, but it's unclear exactly what is going on there. They have simple balance boards as well.

You might use a couple of their other BMSes to do the job, but the D170 uses 16S daughterboards and they claim it can be extended to 96S so you should only need one of that model. I have never used the D170.

An Alibaba price I found for stock units was about $30 per piece (but the quantity was probably large) The minimum order is qty 2. I've found BTP nice folks to deal with.

http://bestechpower.com/27Sto32Spcmbmspcbforlifepo4batterypack/BMS-D170V1%20with%20Relay.html

Just a thought...
 
Thanks All,

Alan, yes, I'm aware I'm only getting the most basic function. But... I would actually rather damage my pack, than have power instantly cut at 70km/h. I've had that happen with the existing set up. A bit scary. At least you cruise a fair bit and it doesn't regen by default.

The reverse polarity I hadn't thought about. I really hope it never gets that far out of balance. What I am now thinking of doing is running a ribbon cable or something similar, to the helmet compartment. 56 Wires... Maybe use LEDs like Fetcher said, and when the battery is on full charge, not all LEDS light, I know out of balance, and I can hand balance it.

@Fetcher - As above, I think I will look into the big blue LED. At first, I thought because the battery is so deep in the bike, no point, but I forgot some of the bigger ones burn quite a few milliamps,

@Teklektik I was committed to getting one of those, but they said the bleed off voltage didn't go down to 2.8v, so I had to cancel the order.

I'm now thinking I'm going to get 5 cells to start off with, test everything systematically, before I commit. There's still life in the LiPo, but after 18 months of some pretty spirited riding, it's starting to see some increases in internal resistance.
 
I have some LTO cells I want to use for backup power, so the application is quite different. A real BMS made for that voltage would obviously be much better, but they seem to be pretty rare or too expensive. I looked at various shunt methods and found the blue LEDs were pretty close. A comparator with a voltage reference driving a power transistor can make an accurate one.

You can run into trouble if one cell goes high early when the charge current is still high.

If there was a photocell looking at all the LEDs, it could trigger a current rollback.
 
Try Energus TinyBMS, it supports very low cell voltages, as long as overall pack voltage is above 9V. Also compatible with ultracaps.
https://www.energusps.com/shop/product/tiny-bms-s516-150a-750a-36

Up to 750 amps with external sensor.
Up to 16s though.
 
Yeah, I had seen those.

I had considered it, but given the difficulty keeping my current 3 x 12S LiPo packs in balance, I ruled that out early.

I think LiPo is a bit more finicky than LTO though, so it's a backup plan if the Zeners/LEDs don't work. If I run it hard, I have to do a manual rebalance once a month or so. If I don't, it seems to mostly stay in balance. But who builds an electric motorcycle to ride like a grandma?

One potential is to run 3 x 19S as balancers for a 57S pack, and run 4 cables to a water proof connector at the front of the bike to perform manual inter-pack balancing.

I've ordered a metre each of 2.7, 2.8 and 3.0v zeners, and 100 blue LEDs. (I've got to say, I've never bought electronic components by the metre before. Feels like buying ammunition... I'll have the whole 9 yards please!).
 
Sunder said:
So my new thought is to pick a Zener diode which has the reverse voltage over the max charge voltage, and use the leakage as a FEATURE, not a bug. Hear me out.
A few problems:

One is that leakage isn't controlled well across devices. So some will leak a little and some will leak a lot - thus unbalancing the pack.

Another is that they aren't very accurate, so you won't get accurate top balancing.

Yet another is that that they are very temperature dependent, so you will see different balancing voltages at different temps. (And when they get hot because they are working, of course, they change even more.)

A final one is that most of them aren't good at dissipating heat. You can get the chassis mount type but they take up a lot of room.

If you want to do this, get an accurate shunt regulator (like a TL431) and use them. You can make a high power shunt regulator with a TL431, a cheap transistor and a few resistors. Put one per cell.
But... I would actually rather damage my pack, than have power instantly cut at 70km/h.
OK. So do balance during charge only. You could do it with a regular charger, 9 quad op-amps and a voltage divider. (Each op amp drives a battery junction.) Or with 36 $5 switchers, each set to 3.0 volts.
 
Thanks Bill.

I'm still going to give it a try - I've bought all the parts anyway... None of that worries me. E.g if some cells are 2.7 and some are at 2.8, I don't care, because that's effectively saying "If some cells are at 99% full, and some are 100% full...". That's what I meant by the cells having a hard knee also - 2.2v, it's empty, 2.5v, it's full, 3.0v and it might start to very slowly shorten life. So I have a 0.5v range to play in, without losing much capacity.

What does worry me now that you've pointed it out, is that Zeners have a NTC... The hotter it gets, the more current it passes. If I was way out of balance, I could cause an avalanche condition. One fails and shorts, and boom, there goes the rest.

I am going to do a lot of testing, and I will document everything and post it here. The thing is, everyone has been saying "it's a bad idea" on paper. But they've been saying that in the context of LiFePo4 and LiPo, which have much lower tolerance for overcharge, and termination voltage is much closer to the nominal voltage.

It's possible that this is a $1500USD mistake. In fact, I could fireball my house. But I doubt it, and I think it's worth trying.
 
billvon said:
OK. So do balance during charge only. You could do it with a regular charger, 9 quad op-amps and a voltage divider. (Each op amp drives a battery junction.) Or with 36 $5 switchers, each set to 3.0 volts.

Sorry, missed this bit. Can you expand on this?

Not sure I understand how this would work.
 
Sunder said:
billvon said:
OK. So do balance during charge only. You could do it with a regular charger, 9 quad op-amps and a voltage divider. (Each op amp drives a battery junction.) Or with 36 $5 switchers, each set to 3.0 volts.

Sorry, missed this bit. Can you expand on this?

Not sure I understand how this would work.

That wouldn't work unless each op amp or switcher had an isolated input supply (or you disconnect all the series connections between cells). It would work with a bunch of isolated dc-dc converters.

Overheating the shunt element is a concern but you can do the math on it. One possible workaround for this might be to use a PTC thermally attached to each shunt element and have all the PTCs in series and monitored by a simple circuit. Similar to the way Honda built the Insight packs. At normal temps, the PTCs are very low resistance. If any one gets hot, the resistance rises and the monitor circuit senses this and drops the charge current.

If you have a series string of shunts and you measure the voltage where they first start to conduct, you can set the charging voltage to slightly below this so when everything is charged and balanced, the shunts are close to zero current. Once balanced, maintaining the balance shouldn't take much current.
 
Sunder said:
billvon said:
OK. So do balance during charge only. You could do it with a regular charger, 9 quad op-amps and a voltage divider. (Each op amp drives a battery junction.) Or with 36 $5 switchers, each set to 3.0 volts.
Sorry, missed this bit. Can you expand on this?
Not sure I understand how this would work.

Two options.

One, use a regular charger. Then do a voltage divider with 36 precision resistors between + and -. Each node represents a "perfect" balance voltage, one per battery tap. You can connect them directly to the battery but that results in weak balancing. Or you can use op-amps as voltage followers at each node to amplify the current. You will likely need 2 "banks" of op amps, since you can easily get op amps that work to 60 volts but it's hard to get them to work to >100 volts.

Two, use 36 3V output AC to DC supplies, and put them in series to charge the battery. (About $5 each.) One way to look at that is that they enforce the 3 volts per cell. Another way to look at it is you have 36 individual battery chargers. This works well if you are charging to 100%, less well if you regularly charge to 50% or so - because charge current is not well controlled in these chargers.
 
billvon said:
Another way to look at it is you have 36 individual battery chargers. This works well if you are charging to 100%, less well if you regularly charge to 50% or so - because charge current is not well controlled in these chargers.
might i recommend using fully isolated chargers in this case? it has been noted here already that people dont like it when their house burns down.
 
flippy said:
might i recommend using fully isolated chargers in this case? it has been noted here already that people dont like it when their house burns down.
Yes, any AC/DC supply is going to be fully isolated due to UL requirements.
 
Just thought of another option last night.

What if I had 55 (or 36) isolated 2.8v buck converters, and a common 5v rail for balancing?

Can you tell without physically testing whether a buck converter is the isolated or non-isolated type?

I was looking at this:

https://www.ebay.com.au/itm/1PCS-LM2596-DC-DC-buck-adjustable-step-down-Power-Supply-Converter-module/301924607468?hash=item464c1be5ec:g:5zEAAOSw~bFWKKJO

s-l1600.jpg
 
An isolated converter needs to have a transformer to provide the isolation.

Put your ohmmeter across the photo and measure the resistance between the negative input and output terminal. :)

The pictured unit has a coil, which is needed for bucking the voltage down, but does not appear to have an isolation transformer. An isolated converter will say so, if it does not say isolated it won't be.
 
Okay thanks.

If it has a transformer, it has a fixed voltage (as far as I know). So I'd have to be pretty lucky to find a 3v one, and even luckier to find a 2.8v.

Back to the original plan of trying Zenners + Leds.

My first "destructive" test, will be to put a pair of zenners across a single cell, then put an increasing voltage over it for 10 hours at a time, and see what voltage it finally fails.
 
Oh really? I thought the ratio of the transformers windings fixes the voltage.

I may have to look a bit harder.
 
Back
Top