Capacitors on motherboards...any tech gurus?

[Planemo]

I have come across an issue with capacitors on an MSX board. I wont go into any detail right now but is there anyone who can enlighten me on exactly what they do?

I am of the understanding that they store energy to 'smooth out' demanding power requirements because a direct feed from the batteries wouldnt be man enough, but this seems odd to me, given the sheer size and capability of the packs we use these days, plus what I feel are relatively small caps. Do they really hold enough energy for this purpose? Am I barking up a totally wrong tree?

Any advice would be greatly appreciated 

[Chriull]

10 minutes ago, Planemo said:

I have come across an issue with capacitors on an MSX board. I wont go into any detail right now but is there anyone who can enlighten me on exactly what they do?

I am of the understanding that they store energy to 'smooth out' demanding power requirements because a direct feed from the batteries wouldnt be man enough, but this seems odd to me, given the sheer size and capability of the packs we use these days, plus what I feel are relatively small caps. Do they really hold enough energy for this purpose? Am I barking up a totally wrong tree?

Any advice would be greatly appreciated 

I tried to simulate a bit what happens around the mosefts (simplified) in https://forum.electricunicycle.org/topic/7549-current-demand-versus-battery-voltage/?do=findComment&comment=104078

In this example the battery current is switched between 53A and 0A at the EUCs pwm frequency (some kHz). That's just 11,7A in average (with the duty cycle of this example), but switching this currents and supplying this over the wires from the batteries over the motherboard to the mosfets is not trivial. With modern wheels (this example was from 2017!) requirements and burdens greatly increased. Here a big capacitor getting "slowly" charged by the battery over the parasitic inductances of the wires and providing the short and high current bursts once the mosfets switches makes a big difference. Did not simulate this with such an capacitor in  LTspice back then...

With my second wheel (KS16B) the legs of this capacitor eroded/melted away - the wheel behaved like switched to soft mode (with some errant moves - ?maybe the pid control loop could not really handle this anymore or would have needed very different parameters for this situations?). At the next steeper incline the fuse burnt...

[Planemo]

That's great Chriull, makes sense. So the caps are primarily used as 'smoothers' to cater for the high speed pwm switching of the mosfet current, rather than general rider current demands.

In your opinion, do you feel that the capacitor rating on modern wheels is sufficient? I believe that it is always better to go too big with caps than smaller (especially given our EUC batts probably have no difficulty 'filling' them up), but how small is too small...

[Chriull]

3 minutes ago, Planemo said:

That's great Chriull, makes sense. So the caps are primarily used as 'smoothers' to cater for the high speed pwm switching of the mosfet current, rather than general rider current demands.

Exactly. At least this is how i understand this...

3 minutes ago, Planemo said:

In your opinion, do you feel that the capacitor rating on modern wheels is sufficient?

No idea - should be "easy" to calculate  - can try it lateron...

3 minutes ago, Planemo said:

I believe that it is always better to go too big with caps than smaller (especially given our EUC batts probably have no difficulty 'filling' them up)

They have no problem filling them way too fast, too. With bigger capacitors one could need some charge limiting. Already by now it's borderline sparking once one connects the battery to the mainboard...

[Chriull]

3 hours ago, Planemo said:

So the caps are primarily used as 'smoothers' to cater for the high speed pwm switching of the mosfet current, rather than general rider current demands.

I tried to resimulate the example linked above. Instead of the mosfet (did not find the model of this - it was an unsuitable type anyhow...) i used a switch.

No idea what to take as parasitic inductance for the wires (tried between 1uH and 100nH - did not make to much difference).

What was important ("more than the capacitance") was the ESR (internal resistance) of the capacitor. And it has to be capable of bearing high currents - look for specified ripple current in the datasheet!

Edited June 22, 2020 by Chriull

[Planemo]

44 minutes ago, Chriull said:

No idea what to take as parasitic inductance for the wires (tried between 1uH and 100nH - did not make to much difference).

What was important ("more than the capacitance") was the ESR (internal resistance) of the capacitor. And it has to be capable of bearing high currents - look for specified ripple current in the datasheet!

Interesting, thanks. I do know that I and many other ebikers were having issues with the caps blowing (overheating) in our 2A chargers. A reccomendation was made to swop them out for larger ones and everyones charger has been fine since. I appreciate that in the above scenario it wont be the same motor/mosfet switching issue but the solution worked great on the charger.

I now believe I have come across an MSX where the caps have blown for no obvious reason. I need to get more info but my line of thinking is around upgrading the caps before it happens to me. If I hadnt done the ebike charger I prob wouldnt have thought too much about it, but it got me thinking.

[Chriull]

18 minutes ago, Planemo said:

I do know that I and many other ebikers were having issues with the caps blowing (overheating) in our 2A chargers. A reccomendation was made to swop them out for larger ones and everyones charger has been fine since

I just looked a bit around digikeys offers - i did not see (by accident) any capacitor in this range with ripple current higher then 10A.

But the higher the capacity the higher their possible ripple current. So this could be a positive side effect by choosing a "bigger" one?

.... But this is about my first encounter with this details - so someone experienced with this should be asked for real advice.

22 minutes ago, Planemo said:

I now believe I have come across an MSX where the caps have blown for no obvious reason. I

I'd guess ("gut feeling") - too much ripple current and overheating. They are also hand soldered - so maybe in combination with a bad solder joint?

Or just low quality/bad charge/... capacitors?

Check all the copper traces from and to this caps!

Btw - you mean the real big ones and not the couple of "medium caps" for the switching voltage supplies?

27 minutes ago, Planemo said:

I need to get more info but my line of thinking is around upgrading the caps before it happens to me. If I hadnt done the ebike charger I prob wouldnt have thought too much about it, but it got me thinking.

Best would be to look with an oscilloscope at the waveforms and deduct the currents!

Smoothing could be very ok - but too much ripple current flowing and by this prematurely destroying the caps. Alone if one looks at the legs of the caps one sees that they cannot be designed for this burdens?

[Planemo]

Sorry @Chriull, it must have looked like I ignored your last post way back in June! I must have missed it, apologies. I only picked it up after you linked to it in another thread.

Probing the caps via an oscilloscope is way beyond my technical abilities unfortunately.

Yes I did mean the 2 big caps.

We didn't find anything else of note on the failed board in question - just the endcaps themselves swollen/leaking/burst. Also this board had worked well for over 6k miles so if there was a solder/trace fault I think it would have showed up previously.

[Chriull]

BTW - as i looked recently at @EcoDrift's V12 teardown at https://ecodrift.ru/2021/05/11/inmotion-v12-vnutrennie-detali/ Inmotion imho seems to have a nice "solution" to this capacitor issue. They have the big fat capacitors a bit aside and along the high current rails with the mosfets 3 smaller capacitors:

Have seen in one of the teardowns afair a better picture showing this... ;(

So current is available decentralized where it's needed, overcoming the stray/parasitic inductances once the mosfets are switching and the big capacitors can be moved a bit away from the mosfets...

And the ESR is lowered by more capacitors in parallel...

[Chriull]

As in another, by now gone topic some intermediate capacitor bank was mentioned, here some more "refined" simulation. I have no real idea if the choosen values are very appropriate, but hopefully they are enough to show some tendency?!

So i have taken a 100V battery with an internal resistance of 0.15 Ohm (~24s4p 21700 cells). For the wires ~10nH and ~8mOhm (AWG 14) per meter. For the motor 0.3 Ohm and 300uH for two coils in series. With an PWM Frequency of 10kHz starting with an duty cycle of 10% and then jumping to 50%.

In the middle of the battery wires a capacitor bank is placed - 10mF, at the board before at the inverter two 560uF capacitors as used with begode wheels - the tan delta of ~0.2 could hopefully be around this 470mOhm internal resistance?

First simulation without the capacitor bank:

and then with the capacior bank:

Main differences i could see:

The battery current gets more smoothed with the capacitor bank - not as much pulse burden for the li ion cells.

The buffer capacitors C1 and C2 have lower current burden.

There is a little bit higher final motor current!  Did some additional Simulation for a step response. There the final max current is about the same, just a bit an additional current "overshoot" peaking after some 2-3 ms. So this "artefact" depends very much on the internal resistance of the capacitor bank and the exact placement ...

 

 

[RagingGrandpa]

Average pack current decreased, but average motor current increased!?
An efficiency gain of a very-impressive 15% (at least at your example motor speed)... Can that be right?

Also makes we wonder what more could be gained by adding capacitance directly to the control board, and not forcing the current pulses to endure L1 & R1.

"All models are wrong; some models are useful"... sounds enticing though, would be great to get some empirical data.

Edited December 22, 2021 by RagingGrandpa

[Chriull]

5 hours ago, RagingGrandpa said:

Average pack current decreased, but average motor current increased!?
An efficiency gain of a very-impressive 15% (at least at your example motor speed)... Can that be right?

It's the capacitor bank providing additional current for some milliseconds - i did some graphs which could have had some better timescale. Once i asked myself the sane question i came to

7 hours ago, Chriull said:

Did some additional Simulation for a step response. There the final max current is about the same, just a bit an additional current "overshoot" peaking after some 2-3 ms. So this "artefact" depends very much on the internal resistance of the capacitor bank and the exact placement ...

But i did not have time to redi the other graphs or make additional ones showing this.

5 hours ago, RagingGrandpa said:

Also makes we wonder what more could be gained by adding capacitance directly to the control board, and not forcing the current pulses to endure L1 & R1.

Should be imho much better. But needs more simulations.

To get some comparisons of the effects on battery and capacitor rms currents.

5 hours ago, RagingGrandpa said:

"All models are wrong; some models are useful"... sounds enticing though, would be great to get some empirical data.

Yes. Someone doing a coil resistance and inductivity measurement would be great!

[Chriull]

Here again the bigger timeframe until the system settled after the jump from 10% to 50% PWM. With cap bank an then without. (This time i inverted I(V1), I(C3) and I(C2) so positive currents are "going" out and for the coil "going in")

One sees nicely here the "additional" power for some milli seconds brought from the capacitor bank.

 

The details at the final stage:

With capacitor bank: (avg and rms value obtained for one period between 29ms and 29.1ms)

Battery current: AVG 78.5A, RMS 78.2A

Coil Current: AVG&RMS 145.7A

I(C3) capacitor bank: AVG 0.1A, RMS 34.4A

I(C1) buffer capacitor: AVG 0A, RMS 12.4A

 

Without capacitor bank: (avg and rms value obtained for one period between 29ms and 29.1ms)

Battery current: AVG 76.1A, RMS 85.9A

Coil Current: AVG&RMS 141.6A

I(C3) capacitor bank: AVG -, RMS -

I(C1) buffer capacitor: AVG-0.1A, RMS 17.4A

21 hours ago, RagingGrandpa said:

Average pack current decreased, but average motor current increased!?

So here in the final stage, once the capacitor bank has perfomed it's about equal again... Just a bit less "performance" in average without the bank - but this again seems to be just a "shift" which can be regulated by the PID algorithm. The final maximum power at 99% (following graphs) is about equal again.

But RMS battery current decreased by the capacitor bank - takes some burden from the cells.

And maybe as interesting maximum case - a jump from the 10% PWM to 99% PWM:

The final coil current is about identical (213A vs 213.8A) but with the capacitor bank the coil current peaks up to 237A after ~2.5ms. Maybe the little bit of extra torque needed to overcome some obstacle without faceplanting?

(As within 2.5ms at 30 km/h one travels ~20mm)

The nicest thing seems to be the relief for the two buffer capacitors C1 and C2, but with these numbers they still would be overburdened... The drop from 17A to ~12A is better than nothing but still not enough - specified Ripple current @120Hz 1.91A and @100kHz 4.29A for the 560uF/160V KXJ from United Chemi-Con.

And this would be just 50% PWM ... Of course all this under the premise, that my choosen numbers are not too way off reality. But underlines the report of the EUC service technician that one of the main weakesses of (Begode) motherboards are the buffer capacitors - they often are atomized on returned burned boards...

21 hours ago, RagingGrandpa said:

Also makes we wonder what more could be gained by adding capacitance directly to the control board, and not forcing the current pulses to endure L1 & R1.

Does not change too much - just more relief for the buffer capacitors. For 50% PWM quite similar to the other 50% PWM version

Edited December 23, 2021 by Chriull

[RagingGrandpa]

Personally speaking, I don't feel limited by torque when riding my MSupers... so the 'transient response' headroom offered by adding capacitance doesn't sound enticing to me.

However, your simulations suggest the possibility of a steady-state efficiency gain (and less stress to components). I hadn't considered it before.

[Mark13i]

On 12/22/2021 at 4:09 PM, Chriull said:

As in another, by now gone topic some intermediate capacitor bank was mentioned, here some more "refined" simulation. I have no real idea if the choosen values are very appropriate, but hopefully they are enough to show some tendency?!

...

 

Can you modify your simulation by replacing 10mF capacitors with LiPo cells with a total resistance of 50 mOhm (there are LiPo cells with discharge currents above 60C and resistance <2mOhm).  

It would also be interesting to simulate the flow of currents between LiIon and LiPo cells as a result of their uneven discharge, and the successive cycles of charging / discharging the LiPo buffer

[RagingGrandpa]

3 hours ago, Mark13i said:

replacing 10mF capacitors with LiPo cells

Like... a whole 24 cells?!

[Chriull]

On 12/24/2021 at 1:37 AM, Mark13i said:

Can you modify your simulation by replacing 10mF capacitors with LiPo cells with a total resistance of 50 mOhm (there are LiPo cells with discharge currents above 60C and resistance <2mOhm).  

Unfortionately not - i have no lipo/li ion modell available for ltspice.

Quote

It would also be interesting to simulate the flow of currents between LiIon and LiPo cells as a result of their uneven discharge, and the successive cycles of charging / discharging the LiPo buffer

But as it seems, low impedance, very fast chargeable lipos will relieve the li ions from pulse burdens and the buffer capacitors.

But, imvho such a project would *not*) make the EUC safer. No idea how many small partial high current charge/discharge cycles a lipo could take before going up in flames?

Edit: And for "real" flow of currents one would need accurate values for the equivalent curcuit diagram for a specific wheel. One should also refine the diagram - this shows the maximum burden, a blocked wheel without any firmware current limitation. Chances are high that it would not even make this shown 30ms without going up in smoke...

This gives, with the choosen/ guessed numbers just a very very rough impression of possible magnitudes and eventually possible (dis)advantages! Nothing more to be gathered from this.

*) Edit: have just seen that i forgot an important not above - should be obvious by the following argumentation, but could be misleading/irritating!

Edited December 25, 2021 by Chriull

[RagingGrandpa]

On 12/22/2021 at 5:44 PM, Chriull said:

a coil resistance and inductivity measurement would be great!

Ok why not

 

On 12/22/2021 at 10:09 AM, Chriull said:

motor 0.3 Ohm and 300uH

Looks high for both then?

[Chriull]

1 hour ago, RagingGrandpa said:

Ok why not 

Thanks!

1 hour ago, RagingGrandpa said:

Looks high for both then?

Resistance should be the 10kHz ESR? So ~2*0.9 Ohm! I'd assume not the DC resistance? Have to look which comes nearer to your measured Z at 10kHz?

Inductivity 2*120μH was not too far off

Will try it once with these values - i'd assume no real change? Let's see...

 

[RagingGrandpa]

5 minutes ago, Chriull said:

should be the 10kHz?

Probably...
The commutation rate is slow; phase current control is fast. If we're modeling the current control, yeah I vote use the 10kHz numbers.
(100Hz also shown if you click thru, hope you found that)
 

11 minutes ago, Chriull said:

ESR?

Probably...
Your ltspice model put it as a plain series resistor. I don't know if it's better characterized in DC or at 10kHz (ESR). Maybe a motor expert can comment further.
 

8 minutes ago, Chriull said:

Inductivity 2*120μH

The magnetic circuit is very different with the magnets and rotor present, so I think it's better to use the 2nd set of numbers (w/ rotor). They do depend on the rotor position, so the phases won't measure equally... pick the highest one?

And with this WYE-connected 3-phase stator, my measurements are across 2 motor terminals (third terminal left open), so I think inductance is just as I measured, not 2x.

So, 90uH @ 10kHz. Not 300.

[Chriull]

11 hours ago, RagingGrandpa said:

So, 90uH @ 10kHz. Not 300.

Yep. and 0.9 Ohm@10kHz. (1)

Also the test leads only numbers are interesting. The leads were just going "distributed" from the LCR meter to the motor? Or more or less near at each other? Or both cables are "fixed" together?

And which length? As i assumed 0.5uH and 4 mOhm per 0.5m.

Your cable measurement says 20 mOhm and 0.2uH. The higher resistance should be the connector/clamp?!

Imho i should maybe also reduce a bit the ESR of the 560uF capacitor - the 470 mOhm came from the tan delta < 0.2 as (very) worst case. As your 220uF/100V cap measurements shows imho more relistic numbers around 150-200mOhm?!

Edit (1) - only for the ripple. "Most" of the current is DC in the coil - there is just a small pwm ripple on it...

Edited January 5, 2022 by Chriull

[RagingGrandpa]

9 hours ago, Chriull said:

The leads were just going "distributed" from the LCR meter to the motor?

The leads-only measurement had the leads simply clipped to eachother (shorted) with no connection to the motor.
You could subtract the lead resistance values from the motor measurements, although since the values are small I think it won't change the simulation conclusion.

And then to measure motor properties, it was as shown in the picture below.
Funny that the photo its showing a slightly lower ESR than in my notes... I must have bumped something and shed 3% resistance. Fussy.
 

9 hours ago, Chriull said:

And which length? As i assumed 0.5uH and 4 mOhm per 0.5m.

Very short... about 25cm circuit length in the leads.
I could tell that contact resistance was quite important: wiggling the clamps, or swiveling the jack in the meter receptacle, increased the lead-only DCR as much as 15mOhm. So I chose to attach the leads and wiggle them until a minimum DCR was shown, and then left them undisturbed for each motor measurement.

Perhaps better to treat the leads-only data as a meter zeropoint offset; and not use the numbers to infer resistance of the leads themselves.

Edited January 5, 2022 by RagingGrandpa

[Chriull]

1 hour ago, RagingGrandpa said:

photo its showing a

It seems to have wrong permissions - it's not to be seen.

Was just the question if both leads covered a big area (higher inductance) or beside one another (lower ind).

[RagingGrandpa]

beside one another

(img fixed now)

Edited January 5, 2022 by RagingGrandpa

[RagingGrandpa]

On 1/4/2022 at 3:20 PM, RagingGrandpa said:

C30 84V MSX:

Also, I noticed my C38 is wound slightly differently.
Still 3 turns but only 16 conductors. (Less copper?!) 
And no idea what +12 might mean...

(C38 left; C30 right)