Icon-YouTube(250x93px).png
Icon-LinkedIn(250x93px).png

© 2017 Ian Watts

To Fuse ?... or Not to Fuse ?

...that is the question, I hope to answer here...

Yet another topic of much debate !

Li-Po cells are well known for their high energy density - hence the reason they're employed so readily in many projects I guess !

 

The idea, as far as I'm concerned, is to isolate a defective cell form the rest of the circuit in case of excessive current flow due to, say, a short circuit / cell failure. So, a blown, cell-level, fuse should prevent a cell from continuing to pass excessive current which results in higher temperatures which, in turn, may result in fire or explosion... well... right or wrong, that's my take on the matter.

There must be some credence to my thought process as I note that Tesla employs the very same idea, albeit with a more unstable battery chemistry (they use cobalt in their cell chemistry for even higher energy densities).

Accepting that the idea isn't new and that large EV companies, such as Tesla, do employ the idea, the task was to investigate the best method to fuse each cell but without adding (too much) additional expense - ignoring the expense of the research and subsequent tests, trials and re-tests I carried out in my quest !

 

I began 'Googling' and 'YouTubing' in an attempt to get some answers.

It quickly became apparent that, for the most part, Li-Po battery packs were not fused (eBay, Amazon etc.) either because of the difficulty or cost (or a mix of both) in so doing.

 

Whilst it seems to be generally accepted that fuses are a great idea for mains appliances, distribution boards and general house-hold / vehicle applications, we don't have the same safety concerns for our battery packs... batteries that have a proven track history of behaving poorly when mis-treated. It's more concerning that most of us DO understand the consequences of our mis-use of them but DO NOT understand enough about how to look after them ! 

 

I kept reverting to the same initial question - "Why not fuse when, for the most part, we are aware of the potential devastating consequences of cell / pack failure?"

Why would large, commercial, EV companies, like Tesla, go to the added trouble of fusing a battery pack, at cell level, if there really was no benefit ?

Close-up of a Tesla fused cell : anode

Part of a Tesla fused cell-pack

It then occurred to me to look at this from a different perspective; if it were cheaper or easier to fuse then, surely, every pack would be fused ?

 

I think I may have stumbled across the 'problem', or one 'reason' at least, for non-fused cell-packs !

So, having decided that fusing is, 'obviously', the better way to proceed, it came down to two distinct steps :

  • what fuse rating and where to get suitable fuse-wire ?

  • how to undertake the inter-cell connections with fuse-wire as opposed to the more normal / standard nickel strip ? 

Taking these in order and accepting that the aim is to increase battery pack safety without decreasing battery pack efficiency or increasing battery pack cost (too much) let's start with :

Suitable fuse-wire:

I approached this from two angles :

1: Fuse-wire to suit intended / designed load:

The idea with this method was to calculate my power requirement and maximum instantaneous current required by the system I was to place the battery pack into. The 'problem' or, rather, 'limitation' with this approach was that I wasn't convinced (given the cost of a battery pack) that I would only use my designed battery pack in one specific, pre-planned application​. So, to design a fused system around a specific project's load requirement seems a little short-sighted. None-the-less, this didn't occur to me until after I'd started my work ! So let's investigate the math here before moving on to my second, now adopted-as-standard, approach - so much more obvious (with hindsight).

The battery design was to be for a 50cc scooter conversion to electric.

My motor of choice (for various reasons, not least of which was governmental) was of a 3KW BLDC design (5HP).

My voltage of choice (again, for various reasons) was to be 72V nominal (a 20S4P pack)

The available space turned out to be sufficient for TWO small-ish 20S4P packs, side-by-side. Sure, I could have designed and built a single 20S8P pack but I felt that two smaller packs in parallel would suit my intended use pattern better, making them small enough to be re-tasked to other projects and sited in different locations should I find my intended location was not in-fact suitable... Planning for all eventualities is good... I think ! ;)

The downside with this approach is that the two packs effectively double the protection and management systems I decide to employ which doubles the cost of so doing but the upside is more flexibility and even greater safety...

At the end of the day, I guess it's all down to personal choice and risk acceptance / aversion levels...

Anyways, to the math :

Power = 3Kw = 3000w

Power = Volts x Amps

Volts = 72 (nominal / 20S pack)

Current = 3000 / 72 = 42A (give or take)

Now then, taking the 42A current requirement and dividing by the number of cells will give me the required current from each cell.

42 / (2 x 4) = 5.25A per cell

So, we need 6A fuse-wire (at around 72V) to suit the 3KW BLDC requirement.

Around this time it hit me that a different project, using the same battery pack, may require more (or less) current and that basing my fuse requirement upon intended current drain (upon intended at-this-moment-in-time use) was too simplistic and in-flexible (especially given the price of Li-Po cells ! ).

Additionally, the 72V is a nominal voltage. In fact, fully charged, this 20S pack will be at 84V (plus or minus) and fully discharged it will be around 56V (plus or minus). No matter what voltage the battery pack supplies, the 3KW BLDC motor may still require 3000W. Some more math :

Fully charged / 84V pack :

Power = 3Kw = 3000w

Power = Volts x Amps

Volts = 84

Current = 3000 / 84 = 36A (give or take)

36 / (2 x 4) = 4.5A per cell

No great problem here, the motor requires less current when fully charged compared to nominal / part charged. The fuse-wire would remain intact.

Fully dis-charged / 56V pack :

Power = 3Kw = 3000w

Power = Volts x Amps

Volts = 56

Current = 3000 / 56 = 54A (give or take)

54/ (2 x 4) = 6.75A per cell

More of a problem here, the motor requires more current when fully dis-charged compared to nominal / part charged. The fuse-wire would need to be slightly larger, perhaps 8A to 9A to handle the 3000W draw toward discharge.

Furthermore, the idea of the fuse-wire is to isolate one (or more) catastrophically damaged cell(s), not necessarily to isolate the whole battery pack should, over time, one (or two or three) cell(s) become isolated. So, the next problem with this approach is that with just one cell being isolated on a 4P (4 cells in parallel) pack, the current requirement through the remaining 3 (non-isolated / intact) cells rises to compensate (assuming the same load scenario). So, back to some more math :

Power = 3Kw = 3000w

Power = Volts x Amps

Volts = 72 (nominal / 20S pack)

Current = 3000 / 72 = 42A (give or take)

42 / 7 (one cell isolated) = 6A per cell or 8A nearing fully discharged.

During the above process, I began to think of other uses for the battery pack I was designing. It occurred to me that the new projects I began to line up as being suitable for the 72V / 20S packs I was planning would need both more and less power (different projects).

2: Fuse-wire to suit individual cell characteristics (as opposed to an individual project)

I began to think how that would suit my logic re: fuse ratings and decided that the above logic was flawed. Surely it would be better to select a fuse-rating based upon the selected battery chemistry / type : LiFePo418650 cells. Looking at the cell datasheet, it became apparent that these Panasonic NCR18650GA cells are quite happy at a maximum continuous discharge of10A (around 3C) taking care 'handle' any temperature.

So, looking at the 20S4P packs I was designing;

20S = 72V nominal (84V fully charged) and

4P (4 x 3340mA = 13.3A), so

3C equates to 3 x 13.3A  =  40A.

All-in-all  72V x 40A  =  2.88Kw

The scoot SHOUD run OK on a single pack BUT I feel this is on the 'risky' side (for me) so I opted to build a PAIR of 20S4P packs connected internally (for the scoot project) in parallel.

The motor is to be a 3Kw unit : 3Kw at 72V = 40A - although during acceleration this could quite feasibly double.

So, TWO individual 20S4P packs in parallel effectively make one 20S8P pack albeit with the additional cost of individual BMS and output protection/connection units. The up-side here is the individual packs with be lighter and should offer increased fitting flexibility not to mention easier re-tasking to other projects as / if required. 

In the end, I decided that an inter-cell fuse-rating of around 12A would suit my purposes well whilst a BMS and external, final / system voltage, thermal circuit breaker would provide my specific-to-project isolation switch and pack-level protection.  

The next step was to select a suitable fuse-wire and test it for that rating - i.e put a 10A load through it with no obvious heat / visible degradation, however, put a 12A - 13A load through it and... it should blow within 30 seconds. Well... that was my logic and it seems to have worked out OK so far...

The challenge here then, was to select a fuse-wire rating that would :

  • be suitable to handle all the 'expected' currents (think of the short-term - acceleration - ones too)

  • limit voltage drop across them (smaller fuse rating = higher voltage drop across it),

  • protect the pack by isolating one (or more) rogue cell(s),

  • not cause isolation run-away / cascade fuse blowing, should only a 'reasonable' number of cells be isolated due to failure, thereby compromising the supply in its totality.

In other words :

  • too low / conservative a fuse-rating stands the risk that one cell failure (and subsequent fuse blowing) could cause multiple fuse failure as the load is increased and the remaining cells provide greater current to compensate for the defective (blown fuse) cell(s).

  • too high a fuse rating stands the risk of making the addition of the fuses pointless.

In the twin battery pack case, if too low a fuse-rating is specified it would also preclude running on ONLY one battery pack at any time - TWO packs would HAVE to be in circuit ALL the time to ensure the current draw from any one cell would max out as calculated for that particular / designed load.

Having said all that and accepting that the project was designed for two packs I would prefer to select a fuse rating that would allow (maybe on reduced load / speed) operation on ONE battery pack if necessary.

In addition to all the above we must remember that fusing a cell only protects the connected cells from uncontrolled current DUMPING that is to say current flow that's way out of tolerance. Fusing does NOT protect from slow discharges by faulty cells but then that type of fault would not cause an increased safety risk due to increased temperature and eventual fire... or worse...

 

To sum up; fusing is, in my mind, an additional safety feature against catastrophic failure. That is all ! It is one of a number of safety features I add to my larger battery packs; others being :

  • a BMS,

  • temperature verification / over-temperature cooling and subsequent shut-down (in design, build & test sequence - late 2018) and 

  • a thermal circuit breaker on the output / system voltage side.

Cell-level fusing itself must not, for me, add more than a minimal amount to the final battery pack cost - both in terms of assembly time and monetary outlay.

In the end I decided to put a 250W / 20A dummy load to good use. Ideally the fuse-wire would 'blow' / melt at a continued current of around 12A - 13A... ideally... I was looking for a maximum fuse-time of 30 seconds at 15A max.

Testing for a suitable fuse-wire :

The next challenge was to find a simple way to test potential fuse-wires.

In the end I decided to put a 250W / 20A dummy load to good use. Ideally the selected fuse-wire would 'blow' / melt at a continued current of around 12A to13A... 

ideally... 

I was looking for a maximum :

Current : 15A

and

Fuse-time : 30s

What I found, after quite a lot of experimentation, was :

1: 1/4 watt resistor leads (0.6mm ball-park diameter / 23AWG):

easy to spot-weld BUT melted / fused / blew (after 15+ seconds) at a continuous draw of high-end 36A. Too high !

2: 1/8 watt leads (0.5mm ball-park diameter / 24AWG) :

easy to spot-weld BUT still too high at around 28A.

I reverted to some Kynar wrapping wire I had in the cupboard :

3: 30AWG (0.3mm ball-park diameter) :

hard to spot-weld (VERY small dia) and fused at between 9 and 10A. Too low !

4: 28AWG (0.35mm ball-park diameter) :

still hard to spot-weld but easier than the 30AWG. Fused at 12.8A.

Bang on !!

I double checked the dummy load readout values with a DC clamp meter. They matched, although it's pretty tough to set my clamp meter to zero. In most of my tests I gave up and let the meter 'zero' point rest at circa -0.9A.

Anyways, I checked the results on numerous (upwards of 20) tests and always the same result :12.5 to 12.9A - like the short video clip above and to the right.

Whilst I'm aware that the above is neither overly scientific nor exhaustive, it was proof enough, for  me, that fusing LiFePo4 cells was possible with little fuss once the fuse-wire had been selected.

For sure, there will be better, more exhaustive and more 'complete' tests but my aim was to create a safe LiFePo4 battery pack within a reasonable period of time and at a reasonable price. These test results were more than 'good enough' for my requirements...

It will, obviously, be up to you, as an individual, to decide should you want to undertake additional / more in-depth tests, for your specific use, before employing cells in this manner... but, like I say, this is how I did it. What you do, is up to you... You can never be too safe !

One drawback of using the insulated 28AWG wire was with the wire-stripping process :

I noticed that if I stretched the wire during the stripping process / removal of the outer sheath, it fused at a much lower current rating (pretty obvious when I stopped to think about it).

If I was too zealous (impatient) when removing the outer sheath the whole wire stretched during the process. In stretching the wire it becomes thinned (again, obvious I guess) and as it becomes thinner its current handling capacity is reduced. Ideally I'd not have to strip the wire but I have yet to find a simple un-insulated 26 / 28AWG wire source... Until I do, I'm careful and test quite a few of the fuse-wire 'links' I make. To be fair I have only had 3 failures in 20+ battery packs / 400-odd 18650 cells but I'd still like to find a better / less finicky wire source.

Panasonic NCR18650  20S4P battery pack.

Each cell, top and bottom fused to a common nickel strip.

The fuse-wires are spot welded to the battery caps and soldered to the common strip.

The common strip also caters for the BMS sense wire connections without having to solder to a battery cap.

It takes around 2 hours to get to this stage.

I then have a BMS to connect, the output and charge cabling, the thermal circuit breaker and the final packaging to complete - another 2 hours, or thereabouts.

Inter-cell connections :

In a previous section, I mused over whether it was better to solder or spot weld inter-cell connections. You can read about it <HERE>.

All-in-all then, once a suitable gauge of fuse-wire had been determined it was simply necessary to spot weld short lengths to the cell caps and the interconnecting 'busbar' / nickel strip... and that was when the next set of challenges arose. You can read all about my attempts (and final successful solution) to spot-weld wire to cell-caps <HERE> .

Anyways, with the above in mind, what I REALLY WOULD like to do is to find a way of knowing that a fuse has blown... but that's something for the future...

Whilst fusing 18650 cells is most definitely not mandatory, I prefer to do it.

The first time round, it most definitely was not as quick or easy as simply connecting each cell cap to its neighbour (in the desired format) with nickel strip BUT subsequently it adds very little to a battery pack build and does give me confidence to string together large packs of serial / parallel cells, with associated capacity (= range) benefits, without worrying if a faulty cell will ruin my day a year or two down the road.

For me, that's reason enough !

To sum up :