Exactly how much battery current are we talking about? (MSX electrical test)

[RagingGrandpa]

Answer: 50 amps max while stationary
(and higher currents are possible while at speed)
 

Who cares?

• You could use this number as a basis for retrofitting a safety fuse to an existing wheel.
  (e.g.: I'd feel comfortable riding my MSX with a 40A-rated fuse, based on this data.)
   
• This reference is useful when evaluating new wheel designs with regard to battery stress.
  (e.g.: the new KS S18 pack uses 3p M50T cells, for a rating of only 22 amps sustained! I expect cell degradation if that pack were frequently used to climb long steep hills.)

Test setup:

• Gotway MSX 84V: 2000w-rated motor, 1600wh pack (6p 18650 Sanyo GA), 2019 production, 2500 miles on wheel, 20% SOC.
• DC clamp ammeter (w/ 2020 cal cert) installed at the power input to the ESC. (Voltage also monitored, at the charging port.)
• Test performed by jamming the wheel against a wall and pulling on the shell firmly until overlean occurs (this happens when the stall torque is exceeded).

 

Test results:

Below shows a baseline for 'normal use' - a firm forward lean like I regularly use to accelerate from a stop:

And next, an overlean created by grabbing the shell and forcing the wheel to dip:

 

Also, below is the relationship between Gotway-reported 'motor current' vs my pack current measurement:
(using recordings from EUC.world for Android, app version 1.4.2)

If you like raw data, find it here.
 

Cheers

Edited November 8, 2021 by RagingGrandpa
fuse clarification

[RagingGrandpa]

@Chriull I like making you squirm, please click below

p.s. I performed this test near an exterior door, with plans to drag the flaming wheel outside using a floor mat at the first sign of trouble...

 

[joca]

10 minutes ago, RagingGrandpa said:

This reference is useful when evaluating new wheel designs with regard to battery stress.

• (e.g.: the new KS S18 pack uses 3p M50T cells, for a rating of only 22 amps sustained! I expect cell degradation if that pack were frequently used to climb long steep hills.)
   
• You could use this number as a basis for retrofitting a safety fuse to an existing wheel.
  (e.g.: I'd feel comfortable riding my wheel with a 75A fuse, based on this data.)

75a seems a bit high for these cells in 3p?  Max continuous discharge (recommended) is just short of 15a at 10-25degC (and only 2.5a below 10degC) and capacity starts to drop off above 10a (test here). 50a is right at the limits of their capabilities, much more than that would put severe stress on the batteries and risk fireworks.

[Chriull]

51 minutes ago, RagingGrandpa said:

Answer: 50 amps.

Peak could be much more?

Depending on the ?corner? frequency of the clamp amperemeter. And the scope log in your link has a time resolution of 2ms. So sampled values at 500 Hz. The scope screen showed 2.5 kSamples/s. The "current" fluctuations should be with some higher frequency (PWM?). Or here is already some kind of current limiting active?

If one looks at the average battery current it's somewhere about 35A:

59 minutes ago, RagingGrandpa said:

• Test performed by jamming the wheel against a wall and pulling on the shell firmly until overlean occurs (this happens when the stall torque is exceeded).

Imho stall torque was not reached by then (firmware current limiting?)

At "real" stall torque motor current and battery current would be equal. ( ... and the wheel burning...)

Here we still have some factor of ~35/210A = 17% - should be the duty cycle of the controller. ... depending on the reliability of the reported motor current, of course...

Especially as 210A should be above the TO-247 package current limit?

If this test would not be extremely risky to the health of your EUC i would ask you to repeat this and measure a phase current...

... and after having the value by the clamp amp meter measure some voltage drop over battery wires/connectors and motor wires/connectors with higher timely resolution so one gets the "exact current spikes/waveform" ...

Would be great to have enough EUCs in a fireproof lab with test equipment!

 

51 minutes ago, RagingGrandpa said:

Who cares?

• This reference is useful when evaluating new wheel designs with regard to battery stress.
  (e.g.: the new KS S18 pack uses 3p M50T cells, for a rating of only 22 amps sustained! I expect cell degradation if that pack were frequently used to climb long steep hills.)

 

I assume the battery current measurement was already the whole current of both packs?

So although a 3p m50T should only burdened continously with 22 amps such a 35A burden would be ~12A per cell.

At https://lygte-info.dk/review/batteries2012/LG 21700 M50T 5000mAh (Gray) UK.html they show discharge graphs for 10A and 15A which do not look too frightening temperaturewise...

... but easily possible to overheat climbing long steep hills!

I'd feel better if EUC manufacturers would have stayed with 18650 cells.... A similar 18650 pack would stay within it's specifications!

But it's not only the average burden - as one sees from your measurements the bigger pulsed peak currents! Could not find something specific in regard to this till now, but it's rumoured that li ion cells don't like such burdens? ;(

1 hour ago, RagingGrandpa said:

p.s. I performed this test near an exterior door, with plans to drag the flaming wheel outside using a floor mat at the first sign of trouble...

Good idea! And very brave to perform such a test! Thanks!

If i got it right the highest temp recorded from the flir was about 61.3°C? And should be something white? So mainly the area between the two motor phase wires beside the capacitor?

[RagingGrandpa]

1 hour ago, RagingGrandpa said:

I'd feel comfortable riding with a 75A fuse

 

1 hour ago, joca said:

75a seems a bit high for these cells in 3p?

No - I meant 75A fusing for my MSX. (Not S18.)

My biggest concern with fuses is unwanted opening. Normally it would happen during lowspeed hillclimb, and result in a walk-home nuisance. But if the fuse opened at higher speeds, it's a injurious crash. (I'm not going to add a fuse to my Gotway... but some have considered it. Because I have no fuse, I'm risking pack damage and fires if my controller were to fail.)

 

23 minutes ago, Chriull said:

The "current" fluctuations should be with some higher frequency

Yes- there was lots of high-frequency in the current measurement. In a prior screenshot of only mild torque I used a higher sampling rate which revealed even more noise, but the log exports were too large for the recording duration I needed, so I turned down the sample rate. Induced noise perhaps? The probe is this one.

 

36 minutes ago, Chriull said:

At "real" stall torque motor current and battery current would be equal.

I don't trust the Gotway-reported 'motor current'-  we have no basis to believe it.

I agree there is a system limit that happens outside the motor- the motor 'stall torque' resulting from a pure voltage source of 84V would be much, much higher. But the EUC is a system, and the system reached it's stall torque limit (and the pedals dipped). Where the bottleneck is, I can't exactly say...

 

39 minutes ago, Chriull said:

I assume the battery current measurement was already the whole current of both packs?

Yes. Hopefully it is clear from the picture of my setup.

 

40 minutes ago, Chriull said:

the highest temp recorded from the flir was about 61.3°C

The IR camera was the least accurate (and most expensive!) device in the room. I did not adjust its emissivity calibration for this test. Please don't trust its absolute readings- I like it for heat location spotting only.

My board was in a very normal temperature range throughout the test, with the fan just barely at its 100F turn-on threshold. The cover was removed (no trapped heat) and I had rest periods between tests. I would expect much higher temperatures during real-world use.

43 minutes ago, Chriull said:

mainly the area between the two motor phase wires beside the capacitor?

The whole left side of the board warmed quickly during loading. Depending on which motor phase was aligned during each test, it was biased either to the top or bottom. The capacitors hide the camera from certain portions of the board, so we can't see everything...

My conclusion from the thermal imaging was: the battery cables stay cool; the motor cables get hot; and the bottom right corner of the right-side pack will suffer notable convection from all of this

[RagingGrandpa]

51 minutes ago, Chriull said:

i would ask you to repeat this and measure a phase current

I think it would only be meaningful if all 3 phases were recorded, or if the motor position could be locked precisely during the test... sorry I don't have a practical method available here.

[Chriull]

3 hours ago, RagingGrandpa said:

Yes- there was lots of high-frequency in the current measurement. In a prior screenshot of only mild torque I used a higher sampling rate which revealed even more noise,

Maybe. Imho most probably this noise is the ?5-10kHz? Pwm.

3 hours ago, RagingGrandpa said:

but the log exports were too large for the recording duration I needed, so I turned down the sample rate. 

Yes. Excel gives up on such amount of data...

3 hours ago, RagingGrandpa said:

Theprobe is this one.

Wow. Great specs! Should be (more than) enough to see all switching ?transients?!

3 hours ago, RagingGrandpa said:

I think it would only be meaningful if all 3 phases were recorded, or if the motor position could be locked precisely during the test... sorry I don't have a practical method available here.

Two would be enough. Two coils are commuted together - currents through borh have different sign and the third phase has zero current.

But yes - while this test the commutation most probable switches between (at least) two states, so one won't get sensefull numbers with just one phase. 

[Chriull]

15 hours ago, RagingGrandpa said:

I don't trust the Gotway-reported 'motor current'-  we have no basis to believe it.

 

If you have a setup to safely measure at the spinning wheel and feel the urge to do so, one phase current measured should enable one to calculate the effective motor current.

The three currents are symetric - phase shifted by 120° - so if one takes one commutation cycle of one phase, does the two phase shifts and take the root mean of the sum of their squares one has the effectiv motor current and some "factor" to come from the rms of one phase current to the effective motor current.

Edit: or one takes simply the rms of one phase current times the square root of 3... 

This one could compare to the reported motor current. If it's hopefully proportional to effectice motor current...

Edited July 31, 2020 by Chriull

[RagingGrandpa]

Status update on the MSX board I tested in the OP

I melted it.

A 3 mile ride through deep sand sustained an average load that was too high.

 

Observations from the failure:

• Average current reported was less than half of the maximum possible current.
  This ride was not "max output"; but the average load was too high.
   
• I was lucky to avoid the "popcorn sound" total board shorting and arcing that other MSuper's have experienced, so I was able to examine the damage more conclusively than usual.
  One of the drive phases burned open, and then the shorting stopped. Afterwards, the board would still turn on and communicate via bluetooth.
   
• Typical from other EUC failures, the motor was 'cogging' notably when rolling the EUC after the event.
  Disconnection of the motor leads from the control board resolved it, and the motor spun freely again.
  (This indicates abnormal shorting of the motor phase circuits in the controller.)
   
• The visibly damaged components were:
  • One pair of power FETs had their 'source' leads blown open.
  • The small, gate drive IC IR2184 for those FETs had its 'lowside' output lead blown open, and a pit melted in its package.
  • A loose solder blob was found at the bottom of the shell near the board, which melted a small pit into the plastic.
  • The label was blackened from heat.

 

Failure mode hypotheses:

1. Gate drive semiconductor failure
   The IR2184 could have failed first, with its lowside output stuck ON, commanding full duty from the power FETs and destroying their leads.
    
2. Power FET semiconductor failure
   The power FETs could have failed first, sinking so much gate current that the IR2184 became overloaded. And stuck-on with so much output current that their own leads blew open.
   This mode seems improbable to me, as the MOSFETs are HY5012W, with truly massive current and temperature ratings. The package is the limit (90A each device), not the semiconductor.
    
3. Interconnect failure
   The leads and solderjoints melted open first. But this doesn't directly explain the damage to IR2184.
   Perhaps molten solder formed a bridge between traces near the power FETs, overloading the gate driver.
   
    

Conclusions:

The 'old' 2018-era 84V MSX boards like this do not measure temperature well enough to alert you to a problem before damage occurs.
Using a smartphone alarm around 155°F is highly recommended for those older EUCs.

(The 'new' boards, which are identified by having separate DC-DC daughterboards, are much more sensitive to temperature, and are more likely to alert the rider with beeps and tilt before a problem gets this bad.)

 

Comments appreciated

 

Edited April 19, 2021 by RagingGrandpa

[Chriull]

3 hours ago, RagingGrandpa said:

I melted it.

Condolescence! I hope/assume you took no harm?!

Seems the thermal pad was applied nicely - cleanly mounted between mosfets and heatplate?!

The thermal pad seems a nice forensic excess heat indicator.

 

3 hours ago, RagingGrandpa said:

Comments appreciated

An important point in 

3 hours ago, RagingGrandpa said:

The package is the limit (90A each device), not the semiconductor.

is the side note "Tc=25°C" which means absolut perfect cooling, about not reachable with (practical) real world measures - if at all!

So with far from perfect cooling, ?some minutes? motor current in this range (however accurate the reported values are) the mosfets have no chance...

Afair often for GW wheels (motor) current alarm somewhere around 80-100A is used and one should stop such burdening not to slowly...

[RagingGrandpa]

On 4/19/2021 at 12:14 PM, Chriull said:

I hope/assume you took no harm

Slow progress through soft sand... I can't think of a better environment to step off a wheel into

On 4/19/2021 at 12:14 PM, Chriull said:

current alarm somewhere around 80-100A

Sure, but I think in the beginning of the thread we demonstrated that sustained current is never more than 50A battery current... and so having two of the 90A-package TO247's in parallel should easily tolerate those 50A, even with generous temperature de-rating. And in the case of my board, it had done so successfully on many occasions.
Edit: it was 50A average battery current. Pulsed motor phase current, which the FETs see, can be much much higher than 50A.

The difference in this situation was: heating of components over many minutes, at moderate loads. Current was not excessive; temperature was.

Edited May 10, 2021 by RagingGrandpa

[Chriull]

26 minutes ago, RagingGrandpa said:

Slow progress through soft sand... I can't think of a better environment to step off a wheel into

Quote

Sure but I think in the beginning of the thread, we demonstrated that sustained current is never more than 50A battery current... and so having two of the 90A-package TO257's in parallel should easily tolerate those 50A, even with generous temperature de-rating.

Battery current is almost constant. Motor current is proportional to speed (plus a motor current voltage drop along the coils).

The ESC/inverter/controller works as DC/DC converter transforming the voltages according to the needed values and the currents vice versa.

So a motor cureent 2-3 times the battery current is very normal at lower/medium speeds at higher burdens.

It's not a real transformation like with a transformer - while the "on" phase of the duty cycle battery current equals motor current (as there is a closed circuit), while the "off" phase battery current is zero and the motor coils keep the current flowing through the body (freewheeling) diodes. (More advantageous if the mosfets are switched on instead of (mis) using their body diodes == "active freewheeling")

So motor current is more or less flowing constantly through (various) mosfets. Battery currents average corresponds with the invers factor to motor current as average battery current to motor current:

https://forum.electricunicycle.org/topic/7549-current-demand-versus-battery-voltage/?do=findComment&comment=104078

So, depending on the real details which current flows through which mosfets the safe area of operation chart could be restricting, too.

Quote

And in the case of my board, it had done so successfully on many occasions.

The difference in this situation was: heating of components over many minutes, at moderate loads. Current was not excessive; temperature was.

The chart on the euc world website is a bit small and not too detailed, but motor currents seem to be in the 100A range for some minutes before meltdown?

[RagingGrandpa]

6 hours ago, Chriull said:

The chart on the euc world website is a bit small and not too detailed, but motor currents seem to be in the 100A range for some minutes before meltdown

Correct. Raw logfile here.

[Boogieman]

@RagingGrandpa Very nice test and after that really sad failure 
 

Quote

 

Conclusions:

The 'old' 2018-era 84V MSX boards like this do not measure temperature well enough to alert you to a problem before damage occurs.
Using a smartphone alarm around 155°F is highly recommended for those older EUCs.

 

Would that temp 68 deg C / 155 deg F also be valid for a Gotway Tesla v 1.2 (bought 2018-06).
Also being a 84V GW wheel, but with with smaller pack (1KWh) and motor (1.9KW), but I guess the base setup is quite similar?
Nice info here to fine tune the settings in EUC World 

Edited May 12, 2021 by Boogieman

[RagingGrandpa]

16 hours ago, Boogieman said:

<phone alarm at> 155 deg F for a Gotway Tesla v 1.2 (bought 2018-06)

We can't be sure, but it sounds like a nice conservative place to start.
If the 155°F alarm does sound during otherwise 'normal' riding, take it as a reminder to inspect inside the shell to confirm the fan is still working.

With MSX, it took a long, arduous, 20 minute ride struggling through deep sand to melt it... nothing felt 'normal' about that torture-test ride

 

Edited May 13, 2021 by RagingGrandpa

[EMA]

interesting topic tnx 

[alcatraz]

Cool study. Thank you for taking the time. I'm curious...

If 100A reported current is only 7A at the pack and a 1600Wh pack has ~22Ah. Then we're supposed to be able to ride at 100A reported for 2, maybe 3 hours before running out of juice. Can that really be the case?

I don't think I can get 2h of 100A reported out of my msx84. If I had to guess, maybe 30min perhaps? Not sure really. I rarely see numbers over 40A reported on mine. I'm ~70kg.

[RagingGrandpa]

15 hours ago, alcatraz said:

ride at 100A reported for 2, maybe 3 hours before running out of juice

Not "ride"...
because these tests were at a stall condition (zero speed). There is a high torque but no mechanical power being produced. A standstill.
So theoretically, the battery will allow you to press your EUC into a wall with medium pressure (100A motor current), for 2 or 3 hours.
(But in reality, the controller will overheat with 100A sustained for more than a few minutes, like I found with my sandy beach.)

[Tawpie]

Some wheels, and I've personally observed this with the pre-production S20, may detect the stall condition and either reduce or stop providing current to the motor. One would do this to protect the controller from overheat. I can imagine a lot of normal situations where you wouldn't want to shutdown after a short stall, but for one that continues for several seconds it's fair though. By definition you're not going fast (the motor has not turned for x seconds) so you shouldn't get killed very badly when it 'cuts out'.

[supercurio]

12 minutes ago, Tawpie said:

Some wheels, and I've personally observed this with the pre-production S20, may detect the stall condition and either reduce or stop providing current to the motor. One would do this to protect the controller from overheat. I can imagine a lot of normal situations where you wouldn't want to shutdown after a short stall, but for one that continues for several seconds it's fair though. By definition you're not going fast (the motor has not turned for x seconds) so you shouldn't get killed very badly when it 'cuts out'.

From rider's feedback and seeing this video from Adam / Wrong Way, it almost sounds like the stall condition detection is currently too aggressive.

Maybe shown here?
https://www.instagram.com/p/CafF2hhFqvV/

On going up the steps at 0:30. I wonder if the wheel dips forward due to the late bonking & loss of traction, or due to this aggressive current limit on stall condition

[Tawpie]

8 minutes ago, supercurio said:

From rider's feedback and seeing this video from Adam / Wrong Way, it almost sounds like the stall condition detection is currently too aggressive.

There's definitely a need to tune this based on real life use (off road mode? stair mode? YT review mode?). You do have to start somewhere though—my sense of the delay was that it was roughly a half second when pushing the wheel into a wall (not measured, and I was only watching, I was not performing the test). That's certainly aggressive, perhaps too quick for going up stairs and superman WW testing—seems to be ok for hitting jumps a little short though.

During the testing I saw, the wheel did recover without any additional interaction, balance was smoothly reengaged as soon as the wheel tipped forward (the cutout).

[alcatraz]

11 hours ago, RagingGrandpa said:

Not "ride"...
because these tests were at a stall condition (zero speed). There is a high torque but no mechanical power being produced. A standstill.
So theoretically, the battery will allow you to press your EUC into a wall with medium pressure (100A motor current), for 2 or 3 hours.
(But in reality, the controller will overheat with 100A sustained for more than a few minutes, like I found with my sandy beach.)

Oh ok.

It would be cool to see that Oscope/Eucworld graph at a fixed speed to see if the curve shape changes.

This may be offtopic but do you think there's a reported current "safe threshold" for the msx84? I wonder what I should use. 200A seems absolutely insane to me. I can't believe that's only 50A at zero speed. It must look way different at speed.

[RagingGrandpa]

18 hours ago, alcatraz said:

do you think there's a reported current "safe threshold" for the msx84?
I wonder what I should use.
200A seems absolutely insane

Well... "safe" like avoiding an overlean crash? 
Or, "safe" like avoiding controller damage?

Avoiding overlean requires some estimate of the controller's effort. EUC World created one (called "safety margin") - you could create alarms based on it. Seba attempted to set the estimate, based on speed and motor current, where 0% means the EUC is at its torque limit and losing balancing control. It's nice.

"Safe for the controller" is not about a peak current (since the controller limits the peaks already), but more about a medium-level current sustained long enough to cause overheating. 
Ideally, you get alarmed about this problem by the EUC's built-in temperature threshold, not a current threshold.
(Like I mentioned earlier in the thread, if you have the original-generation MSX84 controller with its poorly-located temperature sensor, perhaps a lower temperature threshold in a smartphone app is appropriate, e.g.: 155°F.)

[thendless]

Wasn't sure where to post this but this seems like a start. It appears EX30 is the first GW wheel that has proper measurement of battery current

[supercurio]

3 hours ago, thendless said:

Wasn't sure where to post this but this seems like a start. It appears EX30 is the first GW wheel that has proper measurement of battery current

@Freestyler @enaon @Seba maybe some hope for battery current data ?