Worst-case fault current (Gotway battery destructive test)

[RagingGrandpa]

Why is this useful?

This can be used as a basis for pack, harness, interruptor, and controller design and analysis.

Caveats:

• There was only one 900wh 100V pack present in this test.
  In a real EUC, there would be 2, 3, or 4 of these packs in parallel. It means higher controller current, or lower cell current, would be likely for these faults in a real EUC.
• This was the 1st batch (green color) MSP control board. It had a glitch related to headlamp and LED operation, but did run a motor normally.
• My current probe is this one, rated only for 100A pulse current. It reports values past 160A, but accuracy is not guaranteed in such a high range. Perhaps currents above 18C were present.
• I was glad to see the 100V MSP board showed similar current limits as my previous 84V MSX test, for the jammed motor condition.
• The cells were rather cold at 50°F / 10°C.
   

Pack history in the concept thread.

Multimedia here.

Data recordings here.

Cheers

Edited October 2, 2023 by RagingGrandpa
(img url)

[Denny Paul]

Nice! Finally some actual testing! Thanks for taking the time and sharing!

[RagingGrandpa]

My followup thoughts:

I had no idea that M50T could pulse at 80A per cell.

Had I known, I would have sought out a larger-range current probe... a lesson for next time.

 

The pack is a fuse.

Currents >140A will quickly melt the strips and disconnect the pack.

(This 'strip fuse' melted in about 700ms with 150A, so perhaps consider an equivalence to a ~60A label rating.)

But are cells instantly damaged from this? No immediate ill effect was seen, but long-term monitoring was not done.

 

Controller failures can flow massive current.

I was surprised that the briefly-shorted FET test caused a high-current fault, near the level of a bolted short, and did damage the pack (melted strip).

It increases my desire for a fuse in the harness. I would want the fuse to melt before the pack strips could melt, so the objective is 300ms @ 150A, per pack. I think that means the rating should not be more than 30A per pack for typical melting fuses, or 60A at the controller input for 2-or-more-pack EUC's.

Last year, Gotway put an inline fuse in some new EUC's, a mini-ANL fuse with an absurdly high label rating >100A. And later I did see one report of that fuse being blown during a real controller failure. Current >200A didn't feel plausible before my test, but now it seems quite so (for an EUC with 2 or 3 packs). I still worry that such an oversized fuse could pass a middle-value fault current...

It also drives home the point that fuses aren't there to protect the controller, since controller failures could be mild or severe, with widely varying fault current.

 

Inrush is ugly.

We really do get the full short-circuit current of the pack, when connecting the controller. Imagine Monster Pro with more capacitors and 4 packs: 500A pulses?!

It makes me want to retrofit a precharge-type connector like XT90S, whenever boards are being installed...

(XT60 was a fine connector choice for the total current, but they don't make an XT60 variant with a builtin resistor like XT90S.)

 

Desoldering isn't the problem.

(I don't mean that bad joints are OK; I mean that good joints probably won't be damaged during faults.)

Given the presence of the cell-strip "fuses" (and hopefully, additional replaceable fuses in the harness), it seems unlikely that a solderjoint would heat and melt, before something else opens the circuit.

 

The controller protects itself quite well, when there are no electrical faults. (jammed motor test)

I was astounded to watch it survive at its ~220A peak phase current limit rather indefinitely (>20sec sustained).
In an EUC, this is overlean, an unsustainable riding condition. Or perhaps a jammed motor- which seems not so instantly damaging, considering you have many seconds to recognize the problem and flop the EUC onto its side, which turns off the motor.

We know from other real-world examples that controller overloading is still very possible, usually through sustaining a high but less-than-maximum current, for a time duration >60sec, allowing more heat buildup. But, wow, mine sat there for 25 seconds screaming and buzzing without a whiff of smoke.

 

The controller can't always protect itself from a motor fault. (shorted motor test)

This makes me even more cautious about motor wire connections and motor cable damage, knowing that shorted phases could mean instant death to an otherwise-good controller.
I would also beware that any EUC with a severed motor cable could likely have controller damage also.

 

Why did the 3 caps burn instantly? (shorted motor test)

This test had the worst overvoltage spikes. Did the caps become internally shorted from overvoltage? Or did their leads melt only because of the extreme pulse currents?

 

During controller current limiting, massive HF pulses are making it back to the battery.

Any logic-controlled pack interruptor (e.g. newest Gotway BMS, or inline Gotway shutoff board) must be carefully designed to not shutoff accidentally, in the presence of these huge spikes. Melting fuses sound more robust and less vulnerable to such nuisance trips (but are also less fast-acting).

 

Exposed cap leads maybe don't matter.

It's something I've always been paranoid about, having two always-live, non-insulated conductors poking out there at the top of the controller...

If a short happens here (e.g. from loose screws bouncing around inside the EUC) the small leads blow apart so quickly, that there is minimal opportunity for pack damage.

This 'debris vulnerability' is perhaps the (only?) benefit of the plastic controller cover box which Gotway has been installing in recent EUCs.

 

You can overcharge to 6V cell voltage without fire. (Maybe not always, but sometimes.)

I previously thought the cell would spit its guts before reaching 6V. But 2 of my M50T's endured this calmly.

(More overcharging discussion here.)

 

1-pack operation of a fullsize EUC should be discouraged, both because of the cell rating, and the pack construction.

Upsizing the controller and motor in future EUC's, raising the current limits, should also include improvements to the packs and cells.

If a higher-current cell were used (e.g. P42A), a minimalist 1-pack lightweight EUC might also be possible, but heavier pack construction would be needed along with careful selection of fuses or interruptors.

 

Comments welcome

 

Edited March 15, 2022 by RagingGrandpa
(fix url)

[supercurio]

Amazing job @RagingGrandpa on the whole thing 👏

I also hate inrush so I made a XT60 cable with an inline 3K 10W resistor to charge board caps first before connecting the packs directly.
Connecting 5p Sanyo NCR18650GA to a capacitor (Sherman) is not a nice experience and the spark damage the connectors every time.
Likewise, I wish this was taken care of out of the box by the connector or the BMS.

The only additional test I could think of is to observe the failure when maximizing the current drawn from the pack while riding. Like at medium or high speed, with something like 100A phase current and 60A+ battery current (or probably a lot more from your last figures)
There's no doubt it would be harder harder to replicate however, either by:

• simulating an actual load mechanically on a fully built wheel
• simulating realistic back-EMF signal and hall sensor output

Maybe the board would fail in a different way, when taking even more current in without being limited by the max phase current.

Edited December 30, 2021 by supercurio

[RagingGrandpa]

40 minutes ago, supercurio said:

maximizing the current drawn from the pack while riding. Like at medium or high speed, with something like 100A phase current and 60A+ battery current

The peak power point.
I think the peak power point is the speed when back-EMF is just high enough to allow the controller to use 100% duty without triggering its phase current limiter. So it means the same ~200A phase current, on the verge of overlean, at some mid-range speed, perhaps ~50% of freespin speed (such as 25mph?).
Maybe possible to simulate by burning out a tire with the EUC supported in a jig? (Not practical for me, sorry.)

Because of the ~200A phase current limit, we know an upper limit for operational input current: the same 200A. That's certainly achievable (in a non-continuous manner) from 2 or more packs, based my test data.

Your question of "what is the actual pack current at peak power?" is appealing, I agree.
But bear in mind, it's an inherently short-lived condition for an EUC: you'd be at the verge of overlean with no overhead for additional maneuvering, at speed, sustaining an aggressive and intense acceleration command. I think it would need a heavy rider hauling ass up a paved 50% grade (an unrealistic terrain) to sustain this condition for more than a moment.
So why would the number matter, if we knew it... perhaps it's an additional consideration for interruptor design, especially with peak-detecting logic (and not melt-type fuses).
 

40 minutes ago, supercurio said:

Maybe the board would fail in a different way

Resistive heating is a function of current, not output power.
Admittedly, there will be changes in thermal dissipation during high-speed operation (think: capacitor cycling and FET turn-on times), but I think resistive heating dominates the story.
So, I expect failures would be similar to the many other during-riding overloads we've seen (such as my MSX chugging through deep sand, here).

[supercurio]

Yes @RagingGrandpa exactly.

I have a special interest in this data and modeling in order to write algorithms for better software alarms which would protect from each of the known failures without alerting unnecessarily.

It won't be trivial to build an accurate model, but it's an interesting challenge for useful potential result 😃

Which was the failure with the EX.N here for instance? From your tests today I would imagine it would survive.

Apologies tho for the off-topic on battery testing.
I wish Gotway wheels would measure and report battery current directly.

Edited December 30, 2021 by supercurio

[Elwood]

I have no idea what you guys are talking about. Is my Nik+ gonna blow up in my basement if I'm the kind of rider who never tops 30mph street riding? I never let the batts go lower than 40 %. It rides so darn good 

[supercurio]

12 minutes ago, Elwood said:

I have no idea what you guys are talking about. Is my Nik+ gonna blow up in my basement if I'm the kind of rider who never tops 30mph street riding? I never let the batts go lower than 40 %. It rides so darn good 

Have you heard about Schrödinger's cat... 

 

More concretely, it's hard to tell.
From the impressive figures demonstrated in this testing so far, find a bit more likely that fires are primarily linked to defective cells than usage, with usage acting as a trigger or multiplier of the existing risk.
It's only an hypothesis tho.

[RagingGrandpa]

1 hour ago, supercurio said:

Which was the failure with the EX.N here for instance? From your tests today I would imagine it would survive.

[board failed after multiple ascents of a steep hill at ~5mph]

 

Right, low-speed high-torque. Same failure as my MSX. Motor current is high, but pack current and system power are low. Board overheating and failure happened, but the battery packs weren't especially stressed.

To predict such a failure using software during riding, I would suggest some time-weighted / leaky-bucket estimation of "FET temp," based on phase current, sensor temp, and speed (e.g.: cooling rate); and without regard for battery current.
Calibrating the rise and fall rates of the estimate does sound tricky, but I think you could do it in a stationary setting. You probably need to sacrifice a few boards, to find the right calibration...
 

Hopefully we won't mix in "defective boards" with this story (as the video thumbnail of a fire-spitting Sherman reminded me)... I think a few Shermans (and also a small % of most modern Gotways) had board failures without high loads, which I attribute to defects. Some defective boards just shut off quietly; others die with smoke and arcing. But we should ignore defects here, since there is practically nothing software can do to fix them.

 

1 hour ago, Elwood said:

Is my Nik+ gonna blow up in my basement

Ah. Not the topic of this thread.
Here I am studying performance of a "normal" pack in the presence of faults outside the pack.
I consider my test pack "normal" despite its history, because I repaired it before the test (by simply removing the dead cell pair from the circuit).

Regarding faulty packs: rather thorough discussion of Gotway's pack fires, their potential causes and remedies, is available below:
https://forum.electricunicycle.org/topic/25541-begode-900wh-battery-recall-options/

To answer your question:
Based on my experience, your Nik+'s electrical design is very reasonable, and I do not expect it to become unsafe during normal use (including aggressive riding!).
Things are sized appropriately, and simple faults (such as those I created) do not start a battery fire. Crashes and collisions are a much greater risk to the rider, than the batteries. Gear up and go ride!

But like any modern EUC, all users should be cautious about the most-frequent themes:
- Water: immersion cannot be tolerated. If any water gets inside the pack, it begins a process of degradation that often results in fire (sometimes many days later).
- Incomplete charging / heat during charging or storage: Internal cell defects are a random occurrence, and often cause irregular behavior during charging as the first symptom (like my test pack did). (Always recharge to 100%, since partial charging defeats top-balancing.)
With any irregularity, the EUC should be moved to a fire-tolerant location, and diagnosed.

And, your "used" EUC is more trustworthy than a new one, in my opinion. If the cells had defects, I would have expected trouble within the first 50 cycles (like my test pack had).
Newly-manufactured packs (and controllers) are the most suspicious to me, given the ever-present possibility of manufacturing defects. An EUC must earn your trust over time.

 

Edited December 31, 2021 by RagingGrandpa

[alcatraz]

6 hours ago, RagingGrandpa said:

An EUC must earn your trust over time.

I like that. 👍 Buying 2nd hand could actually be wise to not become the guinea pig. 

[Bizra6ot]

@RagingGrandpa Is this com on the exn fail part can be a significant point to add to your suggestion?

"Looking to the wall of the pylon of that bridge (the "pythagorean triangle" at about 4:20), the mathematician in me :P says that the slope is way over 30°. You measure the first slope, but not the second, after changing the location"

[Freestyler]

On 12/30/2021 at 10:19 PM, RagingGrandpa said:

Inrush is ugly.

We really do get the full short-circuit current of the pack, when connecting the controller. Imagine Monster Pro with more capacitors and 4 packs: 500A pulses?!

It makes me want to retrofit a precharge-type connector like XT90S, whenever boards are being installed...

(XT60 was a fine connector choice for the total current, but they don't make an XT60 variant with a builtin resistor like XT90S.)

Sorry for the necro-bump.

Together with my capacitor I had ordered an anti-spark cable, which is essentially two xt60 connectors with resistors in the middle.

This cable proved invaluable as I frequently had to reconnect the batteries during firmware development and I completely avoided sparks and the bad things they bring.

In my opinion it is a must tool for anyone that does work on his wheel!

 

Edited June 8, 2022 by Freestyler