Gotway MSX battery safety margin

[Chriull]

1 hour ago, RagingGrandpa said:

My question:
In this example, does (GW) BMS keep the S4 load resistor engaged after Charge Stop? (I hope so...)

This would discharge the cells after overcharging it.

So maybe it's an optimization to discharge the cells downto the 4.2 V again...

I have no idea if it stresses the cells more to leave them a bit above 4.2V or have this overcharge/discharge cycle.

From what i remember from a (?simple?) BMSs schematic it could work this way - "stupid" comparators switching the bleeding resistors in parallel if above the threshold and "away" if below...

1 hour ago, RagingGrandpa said:

And, what condition resets the Charge Stop event?
(Has anyone demonstrated this reset behavior with an EUC pack? @Chriull?)

The charge stop by single cell overcharge could reset automaticly once this cell goes voltagewise below the threshold again.

An "intelligent" BMS should only reset the charge stop if the charger is dis- and reconnected! So repeatedly overcharging this cell is avoided.

I sincerely hope second "intelligent" version is normally implemented...

Did not see/read/do enough detailed analysis till now of an BMS.

[RagingGrandpa]

We need a guinea pig with a multimeter and a damaged pack.

I hope never to own the 2nd one, but if that day comes, I promise to post results

[WI_Hedgehog]

I feel smarter. (I may not "be" smarter, but I "feel" smarter.) This is a great thread, thank you!

Edited May 5, 2020 by WI_Hedgehog

[mrelwood]

7 hours ago, WI_Hedgehog said:

The biggest problem seems to be lack of proper balancing (the 80% group largely neglecting balancing), however, there are plenty of reports of failed cell groups on the forum by 100% chargers. I want to learn how & why things work and maximize pack life, but that's just me.

Absolutely, me too! But since I don't believe that 80% charging or immediate-green-disconnect -charging would have helped the failed cell groups from forming, to me concentrating on prolonging battery lifetime seems like a waste of time and energy.

7 hours ago, WI_Hedgehog said:

Maybe I'll build this out later, right now I need to finish the miniPRO Speed project.

Which I must say I'm waiting eagerly for as well! Thank you for taking the effort in making a clear guide for the rest of us!

7 hours ago, WI_Hedgehog said:

EUCs haven't been out long enough to age the cells to death, but we should start seeing that in a year or so. The other thing is early EUC adopters that ride 10,000 miles seem to buy several EUCs, so don't tend to notice if the batteries wear or not since they are riding newer, faster technology.

Another point that can be seen to claim battery babying to be futile, since the battery packs do not form a noticeable drop in performance for the lifetime of the vehicle. Surely this could change if the expected lifetime of an EUC would grow much longer. But I don't think that EUCs are reaching maturity in a sence that would change the basic concepts any time soon: 1) Active riders want to upgrade in three years at the latest. 2) Occasional riders will never reach the amount of charge cycles that would clearly show battery degradation. Both being cases that are immune to battery degradation.

What I'm interested in instead is if we can figure out guidelines that would minimize the prematurely failing cell groups. Because that is an issue we are facing today.

[RagingGrandpa]

15 hours ago, mrelwood said:

guidelines that would minimize the prematurely failing cell groups

Life can be simple...

"Charge overnight before a ride"

Don't charge after a ride, unless you're going to ride again immediately. (Minimizes storage SOC.)

I truly think this is all that would be needed for most EUC owners to keep their packs happy. If they knew this and nothing else, I think we'd be in a better place than today.
Perhaps some Z10's would self-discharge during long-term storage and need manual recovery... but if you put a Gotway in the basement at 3.3V/cell (unridable), come back 6mos later and I think you'll find it happily resting at 3.0 and ready to recharge.

The only remaining concern is fire safety... I'm comfortable charging indoors overnight (with me sleeping upstairs), for unmodified EUCs, that aren't soaking wet, using OEM chargers.
Start adding jenky untested custom stuff, and 'sleeping upstairs while charging' would stop.

Edited May 6, 2020 by RagingGrandpa

[WI_Hedgehog]

Unless modifying the BMS, I agree with @RagingGrandpa, other than thinking to charge in the garage, or basement if necessary (cement floors), with a smoke detector/alarm. Generally LiPo fires produce small flames, it's the EUC plastic & rubber tire gasses that are toxic and produce larger flames.

I charged non-EUC LiPo stuff in a fire resistant bag, the charger having a temperature sensor that went in the bag too. Temperature sensors linked to cutout circuits are a huge safety gain

[RagingGrandpa]

Oh wow, stumbled upon lots of useful tidbits in an old thread.

It supports that the Gotway BMS will connect the balancing resistor at 4.20 V/cell, and disconnect it at 4.19 V/cell, without regard to connection of a charger. (I like this.)

 

On 12/14/2015 at 9:38 AM, zlymex said:

Yes, battery packs continue to "charge" for as long as the charger is in power. In another world, the charging current will be approach to a small but none zero value. The reason for this is the BMS balancing circuit which consume certain amount of current at charge end. It is my suggestion to charge this way(continue to charge 3 hours after the Charger LED turns green) once every week so that the batteries remain balanced.

Those small components (the 16 rows of transistors/MOSFETs, resistors, etc) are main part of the balancing circuit and consume this amount of current.
The main component of a typical charge balancing circuit for a cell is the charge balancing IC represented by HY2213-BB3A
I cannot find the data sheet in English so I just explain a little.

The IC itself consumes very little current(3.5uA max) and monitors the cell voltage continuously. Once it reaches 4.200V it will turn on the MOSFET and bypass certain amount of current from the cell.
The amount of current bypassed depends on the resistor Rbn and is 4.2V/Rbn (neglect the on resistor of the MOSFET).
For Gotway MSuper2(I had dissembled the battery pack), the resistor value is 75 ohms(two 151 in parallel) making the balance current of 4.2/75=56mA. Therefore, the total power consumption of a battery pack is 16*4.2*4.2/75=3.76Watts. This will make the pack very warm if you continue to charge for several hours after the light turns green.

I don't have a 170Wh battery pack from Gotway and I haven't seen any inside photo. I guess the balancing circuit for 170Wh is the same as that of a 340Wh. I have several other BMS boards or BMS photos and they look very similar except the value of the bypass resistors may be different.

As for the remaining 0.17A current, it's quite normal for a 850Wh Gotway owning to there are three packs and each pack consumes 0.056A.
However, if the EU is a 680Wh, it's a little higher than normal. The reasons might be:
--- error reading by Charge doctor
--- the current consume by Charge doctor also counted in
--- faulty BMS
--- EU leakage current
--- battery leakage

BTW, IPS adopts different balancing circuit(MAX14921).

Edit: Add a photo showing balancing circuit of a BMS board. The balancing current would be 4.2/68=0.062A.

 

[RagingGrandpa]

@Chriull and others familiar with circuit analysis-

I'm trying to rationalize my observations with my 1600wh 84V MSX (two packs of 20s 3p 3500mAh; healthy packs, no complaints):

1. If I leave it on the original slow 1.5A charger connected 4 hours after the light turns green, I observe 84.3V and 60mA flowing to the wheel.
   (This continues practically indefinitely.)
   The app-reported voltage is 84, but this value cannot read above 84.0 and is likely clipped.

Test equipment: I'm using a Charge Doctor for voltage and current indication. I don't have a calibrated voltmeter to check it against today; I did baseline its current measurement with a 6000count DMM in 600 mA scale, and found the Charge Doctor current to be reading ~15mA higher than the actual output to the pack. It does properly report 0.00A when the output is open-circuit.

My question: do we think some or all of the balance resistors are connected, in this continuous 'float' charge of 40-60mA?

I think if all balance resistors were in-circuit, they alone should pass 56mA per pack, so 112mA at the charge port. This is based on the MSV2 data posted above (75ohm bypass resistance for each parallel group). My math shown here (better if downloaded & opened in excel; web viewer shows the diagram poorly).

An additional small current should be consumed by the cells themselves ('float' charge current, which is something we try to minimize). Honestly for this 6p wheel I was expecting something like 6x30mA = 180mA to the cells, but in reality I observe far less current.

It suggests most of the balance resistors are 'not engaged'... but if they're all truly off, it means the pack voltage cannot be more than 4.19x20 = 83.8V. So, a logical conflict.

Comments appreciated

 

Edited May 12, 2020 by RagingGrandpa

[mrelwood]

9 minutes ago, RagingGrandpa said:

It suggests most of the balance resistors are 'off'... but if they're all truly off, it means the pack voltage cannot be more than 4.19x20 = 83.8V. So, a logical conflict.

Where do you get the 4.19V value?

For demonstration purposes I will use 84.0V charger output voltage. The way I see it is that if the balance resistors are switched on at exactly 4.20V, even after a few hours of balancing some of the cell groups can be at 4.201V and some at 4.199V. The former balancing resistor is activated, the latter is not. The total pack voltage is already the same as the charger output voltage, so the tiny current that should flow into the 4.199V groups is mostly turning to heat in the connectors and cables. Hence the situation remains.

 Then there are the tolerances of every single resistor, transistor and other component that is in the circuit.

[RagingGrandpa]

22 minutes ago, mrelwood said:

Where do you get the 4.19V value?

Previous post described the logical behavior of the IC that engages the balance resistor.

22 minutes ago, mrelwood said:

some of the cell groups can be at 4.201V and some at 4.199V. The former balancing resistor is activated, the latter is not.

Agreed.

22 minutes ago, mrelwood said:

the tiny current that should flow into the 4.199V groups is mostly turning to heat in the connectors and cables

Eh I don't think that plays for me... all the cell groups are passing some amount of current. Even those that are balance-connected, are by definition >4.2V. Holding 4.2V across a cell will cause some amount of current flow and associated chemical activity or heat dissipation. If the cell is fully charged and then left open-circuit (truly zero current flow), it will relax to a slightly lower voltage. But here it cannot, because we're forcing it to stay at 4.2 using a charger.

~40-60mA (10mA per cell) seems to indicate the cells are extremely saturated and passing almost no current, and that few or no balance resistors are engaged... it just happened faster and at a higher voltage than I expected for a huge 6p design. Before the test, my expectation was >200mA indefinitely.

Edited May 12, 2020 by RagingGrandpa

[Chriull]

 

1 hour ago, RagingGrandpa said:

1. If I leave it on the original slow 1.5A charger connected 4 hours after the light turns green, I observe 84.3V and 60mA flowing to the wheel.
   (This continues practically indefinitely.)

I don't know the exact circuitry used in GW wheels. Is it like with KS that charge input goes over the mainboard, which gets activated by plugging in the charger? Or is it directly connected only to the battery packs and the mainboard to the output of the packs?

1 hour ago, RagingGrandpa said:

1. 
   The app-reported voltage is 84, but this value cannot read above 84.0 and is likely clipped.

The wheel can report to the app just any value it measures. And this not very accurately.

Never heard of a firmware clipping voltage readinfs to 84V, but could be. The 0.3V difference could just be the voltage sag of the input protection curcuitry.

1 hour ago, RagingGrandpa said:

 and found the Charge Doctor current to be reading ~15mA higher than the actual output to the pack. It does properly report 0.00A when the output is open-circuit.

I never had any real details in regard to the charge doctor. Afaik @hobby16 calibrated them quite accurate. Depending on the age the CD could need to be calibrated again.

Is this a linear error or some quite constant offset (beside the zero value beeing just)?

1 hour ago, RagingGrandpa said:

My question: do we think some or all of the balance resistors are connected, in this continuous 'float' charge of 40-60mA?

Depends on the first part i don't know.

1 hour ago, RagingGrandpa said:

I think if all balance resistors were in-circuit, they alone should pass 56mA per pack, so 112mA at the charge port. This is based on the MSV2 data posted above (75ohm bypass resistance for each parallel group). 

With all balance resistors enabled, assuming the cells consume negligable current and the motherboard is not active, that seems sound.

1 hour ago, RagingGrandpa said:

It suggests most of the balance resistors are 'not engaged'... but if they're all truly off, it means the pack voltage cannot be more than 4.19x20 = 83.8V. So, a logical conflict.

You have some information about the exact voltage threshold for bslancing resistor enabling for your specific GW BMS?

Sorry - did not look by now really in detail at @zlymex's post from 2015.

 

[Chriull]

So if there is no motherboard interference, and presumably there are "synchronized" BMS. So if one cuts of charging the other cuts off, too.

On 5/12/2020 at 10:26 PM, RagingGrandpa said:

It suggests most of the balance resistors are 'not engaged'... but if they're all truly off, it means the pack voltage cannot be more than 4.19x20 = 83.8V. So, a logical conflict.

Comments appreciated

We know that no cell has reached 4.28V (or any other value thi BMS cuts of for single cell overvoltage).

So 19 could be just below 4.28 and one at around ~84V-19*4.28=2.68V...

One cell has to be below 4.2V (no balance resistor.4.19 is the switch of threshold).

If the other pack uses with all resistors 56 mA there would be 4mA for this cell group (3 cells in parallel) which sounds way to low?

So in every pack should be at least one cell group below 4.2V.

The comparators have an accuracy of +/- 0.025V (they say ), so every cell just below this on threshold would be a pack voltage from 83.5V to 84.5V. (+/- 0.6%)

The wheel reported value can confidently be ignored (if you did not check the accuracy/offset). The CD accuracy could be in about a similar range?

There is some tenth of volts voltage drop over the input protection circuitry.

So imho not much more infrmation to conclude from this. The balancing resistor can have any state - the cells are endlessly trickle charged and have some cell voltages between 4.28V and ~2.68V.

Most probably they are very ok, maybe one (or more) are about to die.

You'll know this very soon - if one/two cell groups are such outliers the will be 0V or even be forced by the others to negative polarity soon.

So without single cell voltage measuring quite nothing can be said...

Imho there could/should be range/performance drop to be noticed before? Or this comes to "soft", slowly over time and too less in effect to really be noticed?

Batteryuniversity.com suggests to stop charge at 3% rated current. Should be about 3.5A/2*6*3%~300mA in your case.

Edit: stroke out wrong statements.  

Edit: they seem to be true, unfortionately - but not the only possibility...:(

 

Edited May 14, 2020 by Chriull

[RagingGrandpa]

17 hours ago, Chriull said:

I don't know the exact circuitry used in GW wheels.
  Is it like with KS that charge input goes over the mainboard, which gets activated by plugging in the charger?
  Or is it directly connected only to the battery packs and the mainboard to the output of the packs?

It's the 2nd one- diagram here might help.

17 hours ago, Chriull said:

firmware clipping voltage readinfs to 84V

It's a Gotway BT protocol limit. It's easy to view in data recordings of riding... if you ride immediately after charging overnight, you can even sustain significant motor currents and wheel speeds before it comes away from the 84.0V reading.

17 hours ago, Chriull said:

Depending on the age the CD could need to be calibrated again.

Is this a linear error or some quite constant offset

Oh, but my comments implied the CD is actually quite good at measuring current. If anything, it reads a few ~15mA above reality. That's still very useful considering the behaviors we're looking for.

(I did not check it at higher currents.)

9 hours ago, Chriull said:

There is some tenth of volts voltage drop over the input protection circuitry.

So imho not much more information to conclude from this.

That really sums it up for me too.

I have a pretty good measurement of current, but no trustworthy measurements of voltage (no 100mV accuracy @ 84V), and no per-cell voltages; and there are losses in the CD and in the BMS.

I still intend to teardown and experiment with an old Gotway pack, some day in the future...

9 hours ago, Chriull said:

Batteryuniversity.com suggests to stop charge at 3% rated current.

Yep, and I prefer cell-specific datasheets- Panasonic (Sanyo) says my GA cells meet their design life for charge-stop at 100mA / cell (which conveniently matches the 3% guideline).

Because I think balancing is more important than minimizing float charging, I typically set my CD to disconnect at 100mA total, since it's no hassle to enable this mode. I won't know if it matters until I finally wear out the pack years from now; I think it certainly allows plenty of top-balancing.

[Chriull]

25 minutes ago, RagingGrandpa said:

Because I think balancing is more important than minimizing float charging, I typically set my CD to disconnect at 100mA total, since it's no hassle to enable this mode. I won't know if it matters until I finally wear out the pack years from now; I think it certainly allows plenty of top-balancing.

I've read lately a new paragraph at batteryuniversity.com

"Passive balancing bleeds high-voltage cells on a resistor during charge in the 70–80 percent SoC curve"

which i do not really comprehend.

70-80% SoC while constant current stage would correspond to 4.0-4.1V cell voltage 

I've never seen any schematics of passive balancers doing this. Just ones like @zlymex showed...

My personal gut feeling was, that more or less in the beginning of the constant voltage/saturation stage most balancing happens. So prolonging this stage with trickle/floating charging could have no(t too much) chance of any positive effect.

However, gut feelings and specilations take noone any further!

My idea of a "gentle" way for "intensifying" the balancing is to discharge the battery just a bit (to ~80% SoC) , charge them again. And repeat this some times.

[Chriull]

@RagingGrandpa after getting some more grey cells activated and rethinking this, the following 

On 5/13/2020 at 9:00 AM, Chriull said:

So 19 could be just below 4.28 and one at around ~84V-19*4.28=2.68V...

needs correction. From the reports here its mostly/always just one cell group "prematurely aging" (in some more seldom cases two?). This weakest cell group will be the first to be charged to 4.28V and cause the BMS to cut off charging. So the remaining 19 cell groups will have ~4.19V (in average).

This weakest/aged/bad cell group will be the first to "loose" voltage by riding (below 3.3V!), and the first to reach 4.28V while charging again. So it gets extremely stressed until in some first stage looses (?more or less?) its capability to hold a charge and stays at 0V. This is the moment most riders come here posting that there wheel cannot be charged to 100% anymore.

A next step should be, to my (limited) understanding that this weak cell will be sooner or later forced while riding to change polarity (keep negative voltages). Which rises (?significantly?) the danger for thermal runaway.

So as your packs charge quite "forever" without the BMS cutting off the balancing is very good and the cells are ok!

Edit: This is one possibility, but the abovementioned is still viable

Edited May 14, 2020 by Chriull

[mrelwood]

 

Quote

This weakest/aged/bad cell group will be the first to "loose" voltage by riding (below 3.3V!), and the first to reach 4.28V while charging again.

If the weak group would indeed be very aged and lost a notable portion of its capacity, that does makes sense. But since some reports of unbalanced cells are before even 1000 miles of riding and have seemingly been fixed by just charge balancing, I have thought that the group with the low charge hasn’t lost it’s capacity but is just constantly at a lower voltage than the rest. Both before and after charging. As the wheel is ridden, the low group has a larger voltage drop, gets stressed a bit more, and the voltage difference gets slightly bigger at every cycle even much before the capacity of the cells decrease very much.

 The few times I have measured individual cells after the charge, the group that deviates has always been at a lower charge than the rest, not higher.

Edited May 14, 2020 by mrelwood

[Chriull]

1 hour ago, mrelwood said:

The few times I have measured individual cells after the charge, the group that deviates has always been at a lower charge than the rest, not higher.

A good point! That the imbalances do not solely come from capacity mismatches of some cells, but also from "perfectly" matched cells if one (or some cells) were not at the same state of charge (voltage) from the beginning!

By having no low cell voltage protection/alarm cutoff (to protect the rider) this cells sooner or later get too stressed and prematurely aged.

Raises the question if this can only happen in conjunction with a charger delivering not enough max voltage, so the balancing function is hindered. Or only by not charging to short to not balance not enough this low cell(s).

Or this imbalances can get started by other external influences like temperature differences of the cells, peak burden currents,...

Seems like really not much can be said about a battery packs state without knowing the individual voltages...

ps.: Seems the scenario of my first post should be reedit as it can happen too and no real reassuring deductions exist, as long as one does not open the pack and measures the voltages...

LiIon cells seem to take quite some beating and fires/explosion are possible but very seldom? With the amounts and age of EUCs by now there are pleasently low numbers of desaster reports!

So the next big leap for EUC development should be smart BMS reporting individual cell voltages! Hopefully already with the next announced wheels 

[RagingGrandpa]

21 hours ago, Chriull said:

So as your packs charge quite "forever" without the BMS cutting off the balancing is very good and the cells are ok!

Agreed. BMS charge-stop should not be happening for packs in good condition.

(My float current data was from my main wheel, 1600wh MSX, which is about 8 months old and 1,400mi traveled, and still in great shape.)

12 hours ago, Chriull said:

Seems like really not much can be said about a battery packs state without knowing the individual voltages...

But actually, doesn't this charging event demonstrate a practical "basic check" of a pack?
If it reaches [4.2V x series-count] and the BMS does not disconnect, and it remains at 100% indicated SOC hours after the charger is removed... well I'd trust that used pack without opening it

12 hours ago, Chriull said:

With the amounts and age of EUCs by now there are pleasently low numbers of desaster reports!

Funny you mention... I've now acquired a 5-year-old 67V Gotway pack that was damaged in two ways: severe corrosion; and short-circuit at the charge port.
It had 2 other healthy packs in parallel, so quite lucky that it did not form a short to absorb the energy of the other two packs!
I believe it has no overcurrent protection of the charge input... the short-circuit at the charge port blew open a trace on the PCB and fire was avoided (thank god!).

p.s. - I will now perform destructive testing of this Gotway BMS- anything specific you'd like data on?

Edited May 14, 2020 by RagingGrandpa

[Chriull]

25 minutes ago, RagingGrandpa said:

I believe it has no overcurrent protection of the charge input...

No. Just the

25 minutes ago, RagingGrandpa said:

the short-circuit at the charge port blew open a trace on the PCB and fire was avoided (thank god!).

fuse trace .

26 minutes ago, RagingGrandpa said:

p.s. - I will now perform destructive testing of this Gotway BMS- anything specific you'd like data on?

Normal operations already yield enough "unknown".

You could test cell overvoltafe threshold. 

And gave fun with whatever crazy ideas come to your mind. Stay safe and document (at least the best moments) on video!

[RagingGrandpa]

On 5/5/2020 at 12:28 PM, RagingGrandpa said:

My question:
In this example, does (GW) BMS keep the S4 load resistor engaged after Charge Stop? (I hope so...)
And, what condition resets the Charge Stop event?

Tested this-

Balancing load resistors will stay connected without regard to Charge Stop.

Charge Stop is a 'latching' condition- even if the offending cells become in-range again, the charging will not resume until the charging input voltage goes away (cycle the charger).

 

[mhpr262]

I think it hasnt become clear in this thread that charging to 100% is mostly bad because of calendarial aging.  Lithium cells age in two ways - by electrode degradation and by electrolyte degradation.

 

Electrode degradation happens because the electrodes inside the cells physically swell and shrink as billions and billions of electrons migrate from one to the other and back during charging and draining. You can mitigate that by keeping the cycles shallow, that is, charging at every opportunity, not just when the battery is completely empty. Two 20km charges will stress the battery much less than one 40km charge.

 

Calendarial aging happens because the electrolyte inside the cell degrades. And that happens at vastly accelerated speeds at very high and very low states of charges. But it takes time! As long as you dont let your wheel sit fully charged for a long time then charging to 100% will have very little effect on the battery. Just put a timer on the charger so it will be full an hour or three before you start riding and you can always charge to 100% without any fear. Btw high temperatures have a huge effect on calendarial aging. With temperatures around freezing you can basically let it sit at 100% all the time and you will notice barely any degradation, but leave it at 100% in a hot car (45°C-50°C) for just a day and you may easily lose 10% capacity or more. I store all of my expensive RC lipos in the fridge for that reason.

Edited June 9, 2020 by mhpr262

[RagingGrandpa]

10 hours ago, mhpr262 said:

I think it hasnt become clear in this thread that charging to 100% is mostly bad because of calendarial aging.

But in this thread we're making the point that "avoiding balancing" is a far worse, and far more common outcome than packs aging out.

So:

On 5/6/2020 at 11:41 AM, RagingGrandpa said:

Life can be simple...

"Charge overnight before a ride"

Don't charge after a ride, unless you're going to ride again immediately. (Minimizes storage SOC.)