Kingsong s20 Cut out FIRE (New York)

[fryman]

1 hour ago, U-Stride said:

23 hours ago, fryman said:

Its unfortunate they did not check the smart BMS after the accident, then we would know for sure.

Here is what happened

Sorry, I was referring to the crash Hsiang had 3 weeks ago.

[supercurio]

Hi @U-Stride, thanks for sharing with us all the details here! 

@Mike Sacristan made a quick demo today how what type of voltage bump happens near full voltage when braking, not even hard on his Nikola+ with custom packs equipped with Smart BMS

I can only imagine how high the voltage shot up in your max/overpower breaking tests. Hopefully Kingsong will be able to extract the data from this mysterious chip we heard about and share with us transparently.

IMO there's a part of user error because I have can predict what happen in that case (very high voltage peaks) but not to put the blame on you at all @U-Stride I must say.

The blame is 100% on Kingsong who didn't establish sufficient protection, which should be reducing the amount of regen braking at higher battery voltage, and beep to alert the Rider "you're about to reach a limit", just like it'll beep when reaching a torque or current limit on accelerating.

Some might think, wait but my wheel never beeps while braking, including when overpowering the motor/controller. True! But it should instead of dumping us on our asses or damage the board or cells in the process (room for improvement on that on all wheels)

Then all safety mechanisms failing: Boards safeties, BMS soft limits, BMS fuse, fire propagating to the whole thing: 0 excuse on any of that.

So even though it was risky to test, thanks again @U-Stride for finding out what must be improved to make the S20 a safe wheel hopefully at some point. And phew, that was close (before shipping a 1000 of them apparently)

 

If you're reading that wondering what could happen to your wheel when braking near full battery: most likely some accelerated cell degradation. Could it contribute to battery fires? Not impossible. How to mitigate that? Enable "over voltage alarm" in your favorite app (and keep braking as much as you need to avoid hitting this pedestrian/bike/car)

Edited March 30, 2022 by supercurio

[techyiam]

32 minutes ago, supercurio said:

@Mike Sacristan made a quick demo today how what type of voltage bump happens near full voltage when braking, not even hard on his Nikola+ with custom packs equipped with Smart BMS

He braked while traveling at about 10 kph, certainly below 16 kph, and saw voltage going higher than his initial fully charged voltage. Didn't @RagingGrandpa said that there should be no regen below 10 mph?

Even if you don't live on a hill, how can you not over-voltage on braking with a fully charged battery?  Are we damaging our cells every time we go ride with a fully charged battery?

Edited March 30, 2022 by techyiam

[.....]

1 hour ago, U-Stride said:

That would be ideal. I would love to try it again and I'm pretty confident it will happen again. It seems to be a component quality issue or even lack thereof i.e fuse

 

Here is what happened 

1. I made several successive Hard accelerations with minor to hard braking initially to feel out the motors power delivery 

2. Hard acceleration but gradual still with some torquing 

3. 1st Hard braking attempt at high speed where I felt the motor slipping. I then let off so it catches itself 

4. About 2 seconds after it catches itself I attempt another hard braking at speed where the motor slipped again and right after the 2nd slip, I felt/heard a loud pop. 

5. The wheel looses balancing ability 

6. While on it's side the wheel shakes jittery like 

7. Few seconds later sound of seething 

8. Smoke appears 

9. Explosion of fire

So, what happened? I didnt quite catch that....

[supercurio]

23 minutes ago, techyiam said:

He braked while traveling at about 10 kph, certainly below 16 kph, and saw voltage going higher than his initial fully charged voltage. Didn't @RagingGrandpa said that there should be no regen below 10 mph?

The braking algorithm in our current wheels will remain a black box since we don't have access to its source code, but maybe we can get a better sense for it when looking at the source of projects like VESC.

@Mike Sacristan's quick demo is particularly valuable since it's the Smart BMS itself, a separate system measuring the voltage (as well as regen current) and not some kind of current estimate or voltage measurement by the board which could be affected by its algorithm.

A few months back Mike has been capturing more braking data in order to tune the BMS settings. One problem with this system is that if you have the BMS quickly cut-out beyond a certain voltage as protection, you might easily be dumped to the ground by the wheel on hard braking, which might also fry some components if the battery packs as energy sink gets suddenly disconnected (TBD)

23 minutes ago, techyiam said:

Even if you don't live on a hill, how can you not over-voltage on braking with a fully charged battery?  Are we damaging our cells every time we go ride with a fully charged battery?

Short answer: yes, maybe not much, maybe more than we think.

That's one of the theories @Jason McNeil had in mind regarding the Gotway battery fires: sudden high voltage & high current charging.
It should be explored further IMO - and in the meantime: app alarms!

Also, high drain cells will help a lot here: thanks to their lower internal resistance and higher charging capacity: they'll take the energy in much more easily, avoiding to push the voltage up as much.

Some e-scooters dial their regen braking way down on high battery voltage BTW.
Described by Adam's video here (with timestamp)

 

Edited March 30, 2022 by supercurio

[Roadpower]

2 hours ago, Rawnei said:

Another big difference is that military hand funding from the state while Chinese manufactures rely on funding from us.

Both are us.

[Paul A]

Perhaps KS will/should replicate Ustride's ride again, capturing data, and try to have it explode again.

[Grimm10]

I'd like to see Wrongway, EEvees, EUCUpgrades, Chooch, or anyone else that has a pre-pro try to replicate the same hard brake/acceleration test.  

Although, Kingsong did ask everyone to stop riding until the next update...maybe after that.

[Planemo]

10 minutes ago, Paul A said:

Perhaps KS will/should replicate Ustride's ride again, capturing data, and try to have it explode again.

Exactly. It's not difficult. ALL wheels should be ridden to destruction by the manufacturers IMO, not because all buyers will do the same, but because without a benchmark to base the stress testing on you are pi$$ing in the wind with regards to how well your unit has been designed.

I get that this is the purpose of demo wheels to some extent, but if I was CEO of an euc firm it's too late by the time even demo units are in the wild. look at the reputational damage that the S20 fire has generated. As CEO I would want to be avoiding that sort of situation at all costs, not to mention the very real chance of my demo riders smashing themselves to bits.

De-buggers trying to break software have been around for donkeys years, getting paid well to do whatever they can to crash apps before anything goes live. I accept that euc manufacturers do 'test' their wheels before releasing demo units, but I don't think they test them nearly hard or long enough.

[supercurio]

@Paul A @Planemo I suspect that KS already experienced burned boards on full battery regen or other similar high voltage condition, but so far dismissed it as "some % of the boards burn anyway / random QC issue".
But now that it lead to a fire, they have to fix it.

It would be my guess from an odd confidence and short delay in promising preliminary analysis right away.

[.....]

7 minutes ago, Planemo said:

Exactly. It's not difficult. ALL wheels should be ridden to destruction by the manufacturers IMO, not because all buyers will do the same, but because without a benchmark to base the stress testing on you are pi$$ing in the wind with regards to how well your unit has been designed.

I get that this is the purpose of demo wheels to some extent, but if I was CEO of an euc firm it's too late by the time even demo units are in the wild. look at the reputational damage that the S20 fire has generated. As CEO I would want to be avoiding that sort of situation at all costs, not to mention the very real chance of my demo riders smashing themselves to bits.

De-buggers trying to break software have been around for donkeys years, getting paid well to do whatever they can to crash apps before anything goes live. I accept that euc manufacturers do 'test' their wheels before releasing demo units, but I don't think they test them nearly hard or long enough.

SECRETLY hire an outside contractor that will ride these wheels into obvlivion? Don't stop until the rider finds the fail point. Rinse, repeat, until the fail point is either WELL above expectations of ride, or it at the least is NOT causing catastrophic consequences. Perhaps make the weakest link to be something more arbitrary, on purpose? Of course, you would need a lock tight NDA, private testing grounds and to probably pay the person. I would assume you would hand them more than one wheel to try, and perhaps hire a few riders. Allowing the public youtubers to do this testing, is scary for PR and its not a very effective manner to conduct precise data. But hell, letting first batch riders do the testing, is also one way to handle it.

[supercurio]

6 minutes ago, ShanesPlanet said:

SECRETLY hire an outside contractor that will ride these wheels into obvlivion? Don't stop until the rider finds the fail point.

I think this contractor might be Kuji, but it's possible he said "no more" after a first S20 crash.

Then we've seen more Chinese riders do extreme stuff but not necessarily as much aggressive street riding (if any)

Edited March 30, 2022 by supercurio

[Tawpie]

43 minutes ago, Planemo said:

I don't think they test them nearly hard or long enough

Fine balance here... test too hard and you end up with something that's over built for your intended audience. Test too long and... you end up being NASA with SpaceX taking your launch market and in spite of your caution you can still have Apollo 13/Challenger/Columbia.

Late discovery that there is a reasonable use case that you didn't test is actually more common than you might expect. U-Stride was able to demonstrate one that not only caused a primary failure but that primary failure cascaded in a way that wasn't "good". NASA would call this a "critical oversight" (at least I bet Mark Robbler would classify it like that) and they are never good. Sure, it's better optics when the faults are found in your lab, behind closed doors. But it's also good that they're found before broad distribution so either they can be fixed or consumers are (more) aware of the risk they are assuming.

KS has work to do. Meantime, I'm back on my S18 because I'm not expecting an S20 anytime soon.

Edited March 30, 2022 by Tawpie

[Robse]

2 hours ago, techyiam said:

He braked while traveling at about 10 kph, certainly below 16 kph, and saw voltage going higher than his initial fully charged voltage. Didn't @RagingGrandpa said that there should be no regen below 10 mph?

Even if you don't live on a hill, how can you not over-voltage on braking with a fully charged battery?  Are we damaging our cells every time we go ride with a fully charged battery?

Upcoming supplements to all EUC user manuals:
Regarding riding on a newly 100% charged wheel:
Please observe the following rules:
1: If you plan to start at the top of a hill,
please carry the wheel down to the foot of the hill
before turning it on. Then ride up to the top of the hill,
turn off the wheel and carry it down again.
Then you can turn on the wheel and ride.
2: If you start riding in a horizontal area, remember that
you must not brake for the first 8 km.
Enjoy :-)

Edited March 30, 2022 by Robse

[U-Stride]

6 minutes ago, Robse said:

Upcoming supplements to all EUC user manuals:
Regarding riding on a newly 100% charged wheel:
Please observe the following rules:
1: If you plan to start at the top of a hill,
please carry the wheel down to the foot of the hill
before turning it on. Then ride up to the top of the hill,
turn off the wheel and carry it down again.
Then you can turn on the wheel and ride.
2: If you start riding in a horizontal area, remember that
you must not brake for the first 8 km.
Enjoy :-)

Hahahahahaha I enjoyed this 🤣 😂 

[Tawpie]

31 minutes ago, Robse said:

Upcoming supplements to all EUC user manuals:
Regarding riding on a newly 100% charged wheel:
Please observe the following rules:
1: If you plan to start at the top of a hill,
please carry the wheel down to the foot of the hill
before turning it on. Then ride up to the top of the hill,
turn off the wheel and carry it down again.
Then you can turn on the wheel and ride.
2: If you start riding in a horizontal area, remember that
you must not brake for the first 8 km.
Enjoy :-)

Good one! Add those to...

Max speed: 20 kph
Do not make rapid acceleration or deceleration or over incline forward or lean backwards
Do not ride on rough or bumpy surfaces, such as muddy or pebbled roads
Do not ride on wet, slippery surfaces, especially snowy roads
Do not ride in bleak winter conditions (I guess a bright sunny winter day is ok though)
Don't jump with the EUC
Do not charge or store your wheel indoors
Replace your batteries at least annually

Edited March 30, 2022 by Tawpie

[Mike Sacristan]

I didn't have a GPS lock at the beginning of the clip which is why it shows 10kmh.
I really wasn't going fast though and just did a small acceleration to braking and shot +1V over my resting voltage.
The SmartBMS in my Nik is set to stop charging if any cell goes higher than 4.22V for 2 seconds. I know from experience that this setting prematurely ends the charge cycle.
However since the BMS balances early I don't lose anything except that I usually end up 4.15V per cell instead of 4.2V which is just a plus for longevity. That top 0.5V disappears as soon as we get on the wheel anyway.
If I change that setting then the charger will just charge the wheel to full until it can't charge anymore and the wheel will show values as high as 101.5V during the charge cycle.

Gotway/Begode wheels have a 102V overvoltage alarm.

As has been pointed out before in several forums, groups, etc... voltage during charge is usually a bit higher than the actual voltage. This should be true for voltage during regen.

I owned a Koowheel electric longboard a few years ago. It would overcharge a bit so when I rode it for the first 1km it would hit the brakes when I wanted to coast. I eventually sold it... the guy who bought it from me started off on a slope (despite my warnings) and the motherboard got toasted.

I imagine that hard braking at some point should cause the motor to want to spin in the other direction and that should be an application of drawing current instead of regen. Excess regen could be dissipated as heat... or maybe goes into a capacitor, which pops, and the remaining excess goes back into the cells.

Still a lot of speculation and guessing. If I were Kingsong I would hire U-Stride to do more riding on another S20 of the same batch/style. With logging enabled. One could even multilog with François app as a proxy. Perhaps monitor cell behavior with the Kingsong app as well.

 

[supercurio]

1 hour ago, Tawpie said:

Good one! Add those to...

Max speed: 20 kph
Do not make rapid acceleration or deceleration or over incline forward or lean backwards
Do not ride on rough or bumpy surfaces, such as muddy or pebbled roads
Do not ride on wet, slippery surfaces, especially snowy roads
Do not ride in bleak winter conditions (I guess a bright sunny winter day is ok though)
Don't jump with the EUC
Do not charge or store your wheel indoors
Replace your batteries at least annually

Wait, that sounds a little too realistic 

[chanman]

I'm sure kingsong will be doing their own repeated harsh acceleration / deceleration and maximum regeneration tests, either on a real wheel or on a hardware bench where things can be better monitored. The questions that arise for me are 1. Why was this sort of stress testing not previously done? 2. If it was done, were there board failures? 3. If there were board failures, why were the problems not addressed and will they be addressed in the production version? 4. How did the battery fire occur?

Edited March 30, 2022 by chanman

[meepmeepmayer]

The big issue is that any board failure should never never never be able to do anything to the battery, and a fuse or some other protection mechanism must kick in. That's a board-unrelated problem with the design of the battery packs and how they are connected to the rest of the wheel.

The battery was possibly shorted for the ~25 seconds between the board popping and the start of the fire. That is a long time for no protection to happen. No fuse? No BMS shutdown?

-

Assuming the board caused the battery to go up and not the other way round. In that case, the board was collateral damage and something is seriously wrong with the battery. No matter what the (here assumed to be working) board does, like sending regen current to the packs or whatever, the battery should protect itself.

-

So either way, what was wrong with the battery?

Edited March 30, 2022 by meepmeepmayer

[Planemo]

3 hours ago, Tawpie said:

Fine balance here... test too hard and you end up with something that's over built for your intended audience.

 'Over built' seems a bit of an oxymoron to apply to EUC's coming out of China. 

3 hours ago, Tawpie said:

Test too long and... you end up being NASA with SpaceX taking your launch market and in spite of your caution you can still have Apollo 13/Challenger/Columbia.

We are light years away from any comparison whatsoever with regards to what NASA would have done with with Apollo 13 etc.

3 hours ago, Tawpie said:

But it's also good that they're found before broad distribution so either they can be fixed or consumers are (more) aware of the risk they are assuming.

I'm not suggesting that demo programs should not exist. They should, but in addition to thorough manufacturer testing. The V12 mosfet fiasco is a prime example of not doing enough manufacturer testing.

I'm not trying to be flippant, and I do get where you're coming from (how much testing is tooo much..?) but I just think we are a long way off from that. Certainly too far for me to be confident of buying a 1st batch wheel. I shouldn't feel like that. It doesn't happen with anything else I buy.

[RockyTop]

Large industrial motors use “starters”. They replaced the old wye delta start. They basically control the amperage going to the motor using IGBTs for a “soft start”. The problem is that IGBTs can short out with spectacular destruction at 480 plus volts. As a safety, we series a mechanical contact that drops out when the circuit detects a problem. ( contact - great big relay) I don’t think a mechanical relay would be a good idea but a secondary holding circuit might be a good solution. Ultimately if the IGBT shorts the circuit drops the power from the batteries. 

[Tawpie]

1 hour ago, supercurio said:

Wait, that sounds a little too realistic

Max speed is from the S20 marketing literature. Everything up to the last two lines is verbatim from the KS 16X user manual... the last two lines are advice from Begode, not currently in my KS literature.

[RockyTop]

13 minutes ago, Planemo said:

 The V12 mosfet fiasco is a prime example of not doing enough manufacturer testing.

You are too kind. I remember arm chair critics predicting the problem way in advance.

[Planemo]

24 minutes ago, meepmeepmayer said:

So either way, what was wrong with the battery?

Thats pretty much it really. Even if the board/board fuses failed in some way causing a short in the packs, there should be additional safeguards downwind. A bit like the fuses on the Sherman which are separated from the board enough not to be affected even if the board is in flames.

Some might suggest that this fire was due to a manufacturing defect with the pack builds from the outset but the chance of any defect occurring just around the same time as the board failure would make this option seem like it has impossible odds.

Therefore we can only be left with one thing - the board failure caused the pack failure - which as you say should never happen.

Many wheels don't have any safeguards on the pack outputs at the BMS, I assume this may be the case with the S20, especially if the only fuses were the SMD ones on the board. And as has already been said, they seem way too confined to be relied on for the voltage/current present and in any event are in a vulnerable location should the board go nuclear.