Inmotion V12 HT board: "More durable MOS"

[Tawpie]

24 minutes ago, conecones said:

They seem to be fine, so it does show the design is solid

I can't agree! All electronics are subject to random failures so a self balancing system is already gambling, but when you design a part into a circuit that exceeds the absolute maximum ratings for that part, you move the odds from "will most likely be just fine" to "do you feel lucky". You may very well be lucky and never have an issue, but you are living on borrowed time.

Think of it as russian roulette... the first bullet is a random failure somewhere in the system and can't be removed. You do however, have a choice of whether or not you load a second bullet. Operating outside of the absolute maximum ratings is a second bullet.

[conecones]

1 hour ago, supercurio said:

@conecones please note that on the public cutout tracker, batch 1 represents 25% of the wheels affected.

@Nick McCutcheon also reported a cutout followed by burned MOSFET on an early production V12 EUCO got as demo wheel.

It's a common misconception, I believe initially driven by Inmotion that only some batch 2 V12 are affected, blaming bad components from Infineon in the process.

However the data which was collectively gathered contradicts this view.

I think you're right mentioning the many other board failures, some of which have been particularly harmful affecting Gotway wheels as well. I can't help but wonder if That Black Cobra's Monster Pro would have survived the type of stress test which were invented to identify the worst V12 units.

I know some will never even contemplate the idea, afraid of destroying a fine wheel but I would recommend anyone getting a new wheel (any manufacturer, any model) to try the same stress tests before riding.

The spreadsheet is great in showing that the cut out symptoms are all so similar but I am a bit hesitant to put too much weight on the overall batch percentages with that sample size. We should wait for the dealer numbers to get released/leaked. 

I know of only one wheel that can absolutely pass the spin test without issue (and as tested on my own), the RS-T. It's generally agreed that this is the most reliable jumping wheel because it is the only one that can take the punishment of landing horribly from free-spinning in the air, and especially from heavy riders. Doing the spin test on this wheel is almost a joke when I have messed up jumps so badly, that the wheel is in the air spinning backwards hitting 80% beeps before landing and somehow it is able to keep me balanced and let me ride it out. It is able to take this punishment over and over and over without fail.

[Planemo]

14 hours ago, conecones said:

While some are minor issues, there are also reported instances of "random" faceplants on Gotway wheels. If you frequent the Gotway groups, you'll see examples of (usually new) riders cutting out "randomly" because of riding in the rain. I don't know how they missed the memo but many thought it was waterproof .

I don't think you can even introduce water-induced failures to support your argument. We all know the wheels aren't waterproof. If people choose to use them in rain then theres not much we can do if it all goes belly-up for them.

14 hours ago, conecones said:

One of the early RS19's in our local group cut out at high speed from a fried mosfet. Smoke started coming out the wheel and the batteries were disconnected immediately as a safety precaution. This is a first gen large bearing RS with green board. As a result, everyone in the same group buy got free replacement black boards from Gotway.

And theres a huge amount of RS19's that haven't failed. I mean HUGE.

14 hours ago, conecones said:

Because Gotway purposely doesn't tell you when they screw up.

Don't think for a minute I support this stance. I have also been very vocal about the 900Wh pack issues, so I'm no GW fanboy either.

14 hours ago, conecones said:

 This Inmotion problem is only getting more attention because Inmotion officially addressed it.

Don't ply them with too much backslapping. They only addressed it because the problem is more catastrophic and more widespread than any other recent wheel issue (aside from maybe the 900Wh pack problems). Inmotion knew that the ratio of V12's failing was going to cause a lot of broken bones, so they had no choice but to formally announce it. They didn't do it out of generosity I can assure you.

14 hours ago, conecones said:

There's many Gotway cutouts if you look around.

I'm not blind to GW cutouts or problems. Don't forget I ran MSX's in both 84 and 100v guises so I followed a lot of chat. The point I am making is that totally random (not water induced etc etc) GW/Begode mainboard failures are in no way at the same ratio of V12 board failures, and so I still don't agree with your stance.

And Black Cobras MSPro - I have no idea why it failed. But again, there will always be random failures. It's the ratio thats important. And again, I don't need to say how many MSPros have been sold and how many of those haven't dumped the rider.

 

[tonto_euc]

On 2/4/2022 at 9:15 AM, Waulnut said:

Let us know the result if possible. I'm hitting over 400 miles on my wheel so far. I'd pay for half the cost to get a new controller. 

Contacted eWheels and they are getting several hundred controller boards for the V12. These will have the new MOSFETs. Since my controller is still good, I don't know if InMotion would pay for my replacement or not. If they won't, I'll pay out of my own pocket to either buy a new controller or send the wheel in and have eWheels service it. The latter option may be more expensive, but I would feel more comfortable having eWheels open up the wheel and do any necessary calibrations and such with the new controller board. For now, I'm not going to risk riding the V12 until I get the new controller replaced. I got plenty of other wheels I can ride in the meantime. 

[nosamplesplease]

1 hour ago, Tan Ho said:

...send the wheel in and have eWheels service it. The latter option may be more expensive, but I would feel more comfortable having eWheels open up the wheel and do any necessary calibrations and such with the new controller board. ...

For me having a qualified technician reassemble would be worth it. Not worrying about getting water seals right is worth every penny.

[Paradox]

10 minutes ago, nosamplesplease said:

For me having a qualified technician reassemble would be worth it. Not worrying about getting water seals right is worth every penny.

Exactly.  Thermal pads and thermal paste?  Like all I have are a few bandages and some tooth paste. I doubt they would be good substitutions.   I would send it in too.

Edited February 5, 2022 by Paradox

[Steve Evans]

3 hours ago, Tan Ho said:

Contacted eWheels and they are getting several hundred controller boards for the V12. These will have the new MOSFETs. Since my controller is still good, I don't know if InMotion would pay for my replacement or not. If they won't, I'll pay out of my own pocket to either buy a new controller or send the wheel in and have eWheels service it. The latter option may be more expensive, but I would feel more comfortable having eWheels open up the wheel and do any necessary calibrations and such with the new controller board. For now, I'm not going to risk riding the V12 until I get the new controller replaced. I got plenty of other wheels I can ride in the meantime. 

did they have any idea when these boards would be received?

steve

 

[tonto_euc]

no hard date at this time other than March/April timeframe. 

[RagingGrandpa]

On 1/7/2022 at 3:21 PM, RagingGrandpa said:

(My expectation of 100A @ 40us is based on my observations of stall currents just above 200A in Gotway EUC's, with that 200A supplied by two FETs in parallel. This is overlean at zero speed, truly the highest output current that can occur during normal riding.)

39 minutes ago, Chriull said:

The operating condition for these fets is either conducting or switched off.

Just while switching there is some, very short time while the mosfet is in the "linear region". 40μs would be very slow switching.

So the points Vds from 1-2V with max 100-200A, just limited by rdson are of interest.

(Better to nerd out in this topic, than buried in the S20 megathread...)

Right, calling into question "where on the map do we really operate?"...

Vds:
I agree that when the FET is ON and conducting current, Vds becomes <1V because the voltage drop across the FET is tiny. 
But I think these "safe operating area" maps are meant to indicate Vds in the resting state. (If not, how could we ever reach 100V Vds-on??) 
Did you have another idea about when during a pulse we should find the Vds to index the map?

100A @ 20kHz @ 80% duty:
+ Frequency: I had to assume the PWM frequency of the controller, which I guessed was 20kHz. I think modulating the output slower than 20kHz is bad because the motor becomes noisy (and also vibration, if frequency is very low); and higher frequencies imply higher-spec components and timers (more cost, or less PWM resolution). A high-frequency recording of phase voltage could confirm it... Or perhaps you know what is typical of other ~1kW-class hobby ESC's? 
+ Duty: Clearly 100% duty with a stalled motor is too much, and the controller will rely on firmware-based current limiting to keep the system from melting. Once the motor is rotating, we'll have a commutation period that makes the control never "continuous duty" for any single FET... but our 20kHz PWM is much faster than commutation, so we'll have many periods of the PWM output representing the peak phase current while the motor moves past one phase of windings. At max output, these handful of PWM periods seemed likely to be a very high duty on their own... so I suggested studying 80%. Then 80% @ 20kHz = a 40us activation. 
And I completely agree, 40us cannot be the turn-on time of the FET. Turn-on rise-time is much faster (49nanoseconds for HY5012). 
+ Current: We do see >200A instantaneous phase current being reported by the EUC, borne by a pair of FET's in parallel. The controller measures phase current directly, using a shunt. So we're fairly confident that 100A pulses are real.

In the end, I think we'll never precisely know "where on the map," but it doesn't matter so much if our objective is only to compare existing EUC's relative to eachother. Because the maps are all an increasing characteristic where current rating increases with pulse-shortening and decreases with voltage, all that matters is that we choose the same datapoint for all the FETs we compare. Thusfar, my conclusion about pulse-current ratings has been: 
- HY5012 (Gotway; Leaperkim)   600A
- IPP023N10N5 (original V12)      30A
- HGP039N15M (updated V12)   180A
- JMSH1504 (S20)                        50A
(And even if you choose a different operating point, this trend will remain)

If we were choosing chips for our own controller and didn't have the luxury of comparing to an existing EUC, I agree, we would need to sharpen our pencil for these assumptions about the worst-case condition.

Comments appreciated

[Chriull]

16 minutes ago, RagingGrandpa said:

(Better to nerd out in this topic, than buried in the S20 megathread...)

16 minutes ago, RagingGrandpa said:

Right, calling into question "where on the map do we really operate?"...

With your above example numbers one point (mosfet on) is at Vds=1V and Id=100A. (1) This is appropriate for this Mosfet forever (DC burden) - as long as cooling is sufficient.

The other point at PWM off is at ID = 0A and Vds=?126V? - no problem neither.

22 minutes ago, RagingGrandpa said:

Turn-on rise-time is much faster (49nanoseconds for HY5012). 

For the JMSH1504AS more in the 60-90 ns Range. Even if they switch "slow" with some 200-400ns that should be no problem.

Interesting is the way during switching where the "worst point" is somewhere about 126V/2=63V and 100A/2=50A. The voltage rise and current drop is not really linear, but should be ok as a first approximation.

For this 63V and 50A one is somewhere around the 100us limit for the JMSH1504AS - seems not somehow bad, too. As this operation point will just be "visited" for a couple of ns.

37 minutes ago, RagingGrandpa said:

- JMSH1504 (S20)                        50A

As this operation point will never be reached and so is irrelevant for S20 wheel operation

2 hours ago, RagingGrandpa said:

That is: if we were expecting "Gotway performance," I look for a maximum output condition of 100A per FET @ 40us on-time, at full battery voltage. 
But as shown above, this JJM 150V FET is only meant to supply about 50A for that condition. 

No way ever 100A with 150V with 40us pulses ever will happen. That would be 1.5kW power dissipation! Just for a short time, but still. Afair from SOA diagramms this pulses are not guaranteed to be survived if they come periodicly - one has to let the junction cool down again, before the next pulse is allowed.

Imho the most important points are how nicely switching is designed (2) and the thermal parameters (heatsink, thermal resistances, removal of (excess) heat,...)

 

 

 

(1) They nicely show this SOA diagramm not for 4mOhm which the mosfet has at 25°C but at 10mOhm for higher temperatures (?TJ(max)=175°C - the worst case?)

(2) Enough gate current for fast switching, nice dc-link capactors to support the drain currents, pcb layout,...

[enaon]

49 minutes ago, RagingGrandpa said:

- HY5012 (Gotway; Leaperkim)   600A
- IPP023N10N5 (original V12)      30A
- HGP039N15M (updated V12)   180A
- JMSH1504 (S20)                        50A

The IPP023N10N5 is smaller in physical size, and rated at 100 volt for a 100 volt system, is it not important?

[RagingGrandpa]

19 hours ago, enaon said:

IPP023N10N5 is smaller in physical size

Smaller in size, lower in ratings, yes and yes.

19 hours ago, enaon said:

rated at 100 volt for a 100 volt system, is it not important

Very important... but a different problem than "too much current," which is what the 'safe operating range' map shows, that we're discussing.

Say for a moment the V12 100V FET were magically improved to survive 150V, but all its other ratings remained the same. It would still be very weak in terms of safe current output, relative to the FETs in these other similar EUC's. The outcome would be a durable EUC with weak torque; or a powerful-feeling EUC that burns up controllers easily.

[edit: maybe not so weak actually... the 23N10N5 100V FET was quite efficient, if you respect its voltage limits.]

Edited March 25, 2022 by RagingGrandpa

[RagingGrandpa]

1 hour ago, RagingGrandpa said:

- HY5012 (Gotway; Leaperkim)   600A
- JMSH1504 (S20)                        50A

@Chriull - theory aside, do you share my conclusion that the peak motor current of the S20 126V controller is substantially lower than the peak motor current available with MSuper RS 100V? 

(Or, was there some reason the component specs didn't convince you?)

[enaon]

6 minutes ago, RagingGrandpa said:

Say for a moment the V12 100V FET were magically improved to survive 150V, but all its other ratings remained the same. It would still be very weak in terms of safe current output, relative to the FETs in these other similar EUC's. The outcome would be a durable EUC with weak torque; or a powerful-feeling EUC that burns up controllers easily.

I can see what you say, I only commented because not everyone can, your post on the s20 thread was transferred like "s20~=v12 disaster" on our telegram group.

 

Edited March 24, 2022 by enaon

[RagingGrandpa]

4 minutes ago, enaon said:

"s20~=v12 disaster"

Ah well that depends on parsing syntax... in some circles, ~= is "not equal"  

But right, S20 is certainly not the same disaster as V12.

Could it be a brand new sort of disaster? Only time will tell; fun to watch  

[enaon]

9 minutes ago, RagingGrandpa said:

Ah well that depends on parsing syntax... in some circles, ~= is "not equal"  

But right, S20 is certainly not the same disaster as V12.

Could it be a brand new sort of disaster? Only time will tell; fun to watch  

 

I typped == and changed my mind, spelling error

50A x2 at 150V is very good, assuming the heat dissipation works as good as on the s18, it is the most powerful machine that can be ridden on a summer day in Greece uphill. It will not have the bite of gotway/sherman self destructing abilities, but numbers follow a balanced powerful design I think. 

 

Edited March 24, 2022 by enaon

[Chriull]

12 hours ago, RagingGrandpa said:

@Chriull - theory aside, do you share my conclusion that the peak motor current of the S20 126V controller is substantially lower than the peak motor current available with MSuper RS 100V? 

(Or, was there some reason the component specs didn't convince you?)

From the datasheet figures the (theoretical) continous Drain Current the JieJie JMSH1504AS (S20) is about ~65% of the Hooyi HY5012W/A (?RS 100V?).

The JMSH1504AS has a by packacke rating limited train current of 180, the current limit from package rating for the HY5012W/A is just 90A! 

These values are given for case temperatures of 25°C and 100°C - Hooyi has no further footnotes for continous drain current, while JieJie states that "Computed continuous current assumes the condition of TJ_Max while the actual continuous current depends on the thermal & electro-mechanical application board design."

This could mean that Hooyi has some theoretical value specified which leads to impossible cooling needs (not possible to keep the case at 25°C, like many manufacturers market their numbers), JieJie computed some maybe a bit more realistic,conservative value which still could be further limited on the actual design.

The SOA diagram for the JMSH1504AS is for 175°C junction temperature - the absolut worst case! So Rdson is worst case high temperature ~10mOhm (4.9mOhm * 2.x) instead of room temperature typical 4mOhm (Edit: 2.9mOhm was a typo)

The HY5012W/A SOA is for an rdson of 0.5V/200A~2.5mOhm, which is about the typical rdson at room temperture (2.9mOhm). Max value would be 3.6mOhm times ~1.9 for 175°C, leading to ~6,8mOhm. Thus the currents in the SOA diagram have to be reduced by the factor 6,8/2,5(?2,9?) - some factor 2.3-2.7 lower. As this is an SOA diagram the higher resistance and temperature limit could lead to further limitations.

And as written above - the operation points at which the Mosfet have to perform are for the S20 100A/~1V and 0A/126V - for both operation points he's a perfect choice with enough safety margin. And for the switching transition every operation point is very well within the SOA, too.

The HY5012W/A seems to have a bit better values - imho a portion of the advantage is by a better/ not so shy marketing department. However, the package limit of 90A makes him the worse choice for 100A burdens - although this absolute limit currents come seldom and short enough that he seems to survive this burdens good enough.

From what i got from @Marty Backe ?extended? overheat hill test of the S20 in comparison to other actual wheels it stayed very cool during the ride - so also in reality the mosfet choice, controller, pcb and heatsink design seems good!

So i think your table with 600A vs. 50A does not regard many important points is is choosen for an operating point that is of no relevance during operation.

Edited March 25, 2022 by Chriull

[supercurio]

Thanks @Chriull for looking into it.

Where do you see that the SOA diagram for the JMSH1504AS is for 175°C junction temperature? I missed that from the datasheet.
I also read the max junction temperature to be specified at 150°C

[Chriull]

47 minutes ago, supercurio said:

Thanks @Chriull for looking into it.

Where do you see that the SOA diagram for the JMSH1504AS is for 175°C junction temperature? I missed that from the datasheet.
I also read the max junction temperature to be specified at 150°C

Yes - the 175°C i mixed up with some other number ;(.

It's  a note in the soa diagramm saying Tj_max=150.

And in the SOA diagram ~10mOhm are used as Rds on which is for 150°C.

[supercurio]

I got confirmation that the V12 HT board will use TO-247 MOS.
From previous statements, I guess it would be HY5012W then.

It means it'll be a significant departure from the original V12 batch 1&2 board. On paper it would be certainly capable of matching the "Increased max power output by 23%" and "More durable MOS" claims thanks to the change of components.

[Steve Evans]

3 hours ago, supercurio said:

I got confirmation that the V12 HT board will use TO-247 MOS.
From previous statements, I guess it would be HY5012W then.

It means it'll be a significant departure from the original V12 batch 1&2 board. On paper it would be certainly capable of matching the "Increased max power output by 23%" and "More durable MOS" claims thanks to the change of components.

is it likely then that the v12 HS revised driver boards will match those of the V12 HT?

[Tawpie]

1 minute ago, evans036 said:

is it likely then that the v12 HS revised driver boards will match those of the V12 HT?

Maybe? In many ways it would make sense to only have 1 board for all wheels, but if the HT motor isn't different than the HS one you lose the marketing value of two variants. There's no way for me to guess—but new HS boards are rolling in CA at any rate, so we should know more shortly.

[RagingGrandpa]

On 3/25/2022 at 5:47 AM, Chriull said:

i think your table with 600A vs. 50A does not regard many important points

Fair, ok, so if we move away from the SOA table (because it's more meaningful during times the gate is being modulated in its linear region)...

To compare FET's for our 200A short-duration phase current objective, how about this?

My objective is an apples-to-apples comparison of maximum phase current capability of each design... to set expectations for how highly stressed the controller is. Shown this way, S22 stands out again, going more than twice as high into its FET ratings as Gotway. (And to deal with this deficit, I'm suggesting that S22 won't command as much current.) 

We don't know enough to say the load condition exactly, so we can't use this to declare "how much is good enough"... but I still think it's very relevant to compare existing EUC's against eachother, in terms of design margin.

p.s. that new 150P220 FET looks awesome!

 

Edited April 27, 2023 by RagingGrandpa
(added Patton)

[Tawpie]

6 minutes ago, RagingGrandpa said:

S20 stands out again, going more than twice as high into its FET ratings as Gotway

assuming the same cooling setup from the part's heat tab to ambient... 

[Chriull]

3 minutes ago, RagingGrandpa said:

To compare FET's for our 200A short-duration phase current objective, how about this?

Looks much better! 

4 minutes ago, RagingGrandpa said:

My objective is an apples-to-apples comparison of maximum phase current capability of each design... to set expectations for how highly stressed the controller is. Shown this way, S20 stands out again, going more than twice as high into its FET ratings as Gotway.

Don't know if such an interpretation makes sense?

There is no reason to not use a mosfet within it's specification plus some safety margin. There is no benefit to be overly precaitious - sometimes overdesigning the mosfet gives disadvantages like higher needed gate charges, etc...

7 minutes ago, RagingGrandpa said:

And to deal with this deficit, I'm suggesting that S20 won't command as much current

Kingsong historicly always performed a bit less than gotways

Firmware/hardware design always tended to be a bit more "prudent".

In regard to current the S20 should be able too take more - twice as high package current limit and still 20% too go with max power dusdipation

The a bit high rdson hinders maybe a bit, bit as martys overhear hill showed very low temperature increase the design seems more than sufficient.