Jump to content
Marty Backe

The Gotway Gods Destroyed My ACM

Recommended Posts

13 hours ago, DaveThomasPilot said:

Maybe another whole section on brushless motor power and efficiency analysis, but briefly---

Some of high current paths don't conduct the current 100% of the time.  The speed controller is probably running around 100 Khz, or at least 10's of Khz.

I don't know if anyone has actually measured it from the gate-drivers, but it seems the wheels use lower PWM-frequency, <10kHz... this is mostly based on the sound emitted and someone did run it through a spectral analyzer to see the peaks, the fundamental for KS16 was somewhere around 5.5-6kHz, if I remember correctly, with some spikes at higher harmonics.

If I had to guess, I'd say they use lower frequencies to minimize switching losses (mosfet passing through the linear region at turn on / turn off) and maybe to minimize the skin-effect on the longer wiring going to the motor? But I haven't got a clue how the parasitic inductance of such cables should actually be calculated, it's multi-stranded but not Litz-wire, so I don't know if you should just take the outer diameter or model the shape of the strands pushed together, as there's more surface-area then... probably doesn't even matter that much in the wheels, but I have my own reasons for being interested in the parasitic inductance in longer thick wiring  (pushing up to around 1000A / 1ms pulses through a 16mm2 / about 5AWG -cabling, leading to very high di / dt ...  :whistling:)

Share this post


Link to post
Share on other sites
13 minutes ago, esaj said:

I don't know if anyone has actually measured it from the gate-drivers, but it seems the wheels use lower PWM-frequency, <10kHz... this is mostly based on the sound emitted and someone did run it through a spectral analyzer to see the peaks, the fundamental for KS16 was somewhere around 5.5-6kHz, if I remember correctly, with some spikes at higher harmonics.

Imho it was 8 kHz - but i have no screenshots of my "measurements" on my harddrive. Would have to search in the old posts....

13 minutes ago, esaj said:

If I had to guess, I'd say they use lower frequencies to minimize switching losses (mosfet passing through the linear region at turn on / turn off) and maybe to minimize the skin-effect on the longer wiring going to the motor? But I haven't got a clue how the parasitic inductance of such cables should actually be calculated, it's multi-stranded but not Litz-wire, so I don't know if you should just take the outer diameter or model the shape of the strands pushed together, as there's more surface-area then... probably doesn't even matter that much in the wheels, but I have my own reasons for being interested in the parasitic inductance in longer thick wiring  (pushing up to around 1000A / 1ms pulses through a 16mm2 / about 5AWG -cabling, leading to very high di / dt ...  :whistling:)

pffff... nice parameters...

On the english wiki

https://en.wikipedia.org/wiki/Inductance :ph34r:

are the formulas for the self-inductance of two parallel wires. Maybe you should use the formula for two parallel wires (high frequency), since high currents behaviour is sometimes quite similar to high frequency situations.

Or there is also the inductance of a wire parallel to a conducting wall, if this better meets your requirements?

Share this post


Link to post
Share on other sites
5 minutes ago, Chriull said:

Imho it was 8 kHz - but i have no screenshots of my "measurements" on my harddrive. Would have to search in the old posts....

pffff... nice parameters...

On the english wiki

https://en.wikipedia.org/wiki/Inductance :ph34r:

are the formulas for the self-inductance of two parallel wires. Maybe you should use the formula for two parallel wires (high frequency), since high currents behaviour is sometimes quite similar to high frequency situations.

Or there is also the inductance of a wire parallel to a conducting wall, if this better meets your requirements?

It's for my capacitor bank spot-welder, a project that has been dragging for quite long, as for example I had to wait 3 months for the caps (that should have been originally in stock within a week from order), but also because I've had lots of other stuff to do recently (like riding ;))...

Even a single wire has stray inductance, and I used some wire inductance calculator (using just the diameter/circumference of the cable, which probably isn't exactly correct) to get a value of 525nH per cable (two 50cm cables in total, welding prongs at the ends). The current changes fast with the spikes (basically shorting a 470000uF bank at 12V max over the cables, pieces to weld and paralleled mosfets), which in simulations caused over 100V spike at turn-off, probably that would fry the low-side mosfet-switch bank (made from 5 x 40V, 190A/1080A IRFP2804 mosfets in parallel), so I had to order some high-amperage fast recovery diodes that can freewheel and keep the voltage from rising too much and am changing the switch-bank to use actual gate drivers instead of just a push-pull array to speed up turn-on / off and looking into using ferrites and gate resistors to "correctly" handle the parallel gates. But I'm still not sure if the stray inductance is correct, or if the diodes are enough (3 x TO-247 RURG3060), or if I need place them just in reverse parallel next to the cabling or only/also at the switch-bank. I'd rather play it safe than end up blowing the switches (or the caps)... Further discussion of that probably belongs to off-topic. :P

Share this post


Link to post
Share on other sites
1 hour ago, esaj said:

I don't know if anyone has actually measured it from the gate-drivers, but it seems the wheels use lower PWM-frequency, <10kHz... this is mostly based on the sound emitted and someone did run it through a spectral analyzer to see the peaks, the fundamental for KS16 was somewhere around 5.5-6kHz, if I remember correctly, with some spikes at higher harmonics.

If I had to guess, I'd say they use lower frequencies to minimize switching losses (mosfet passing through the linear region at turn on / turn off) and maybe to minimize the skin-effect on the longer wiring going to the motor? But I haven't got a clue how the parasitic inductance of such cables should actually be calculated, it's multi-stranded but not Litz-wire, so I don't know if you should just take the outer diameter or model the shape of the strands pushed together, as there's more surface-area then... probably doesn't even matter that much in the wheels, but I have my own reasons for being interested in the parasitic inductance in longer thick wiring  (pushing up to around 1000A / 1ms pulses through a 16mm2 / about 5AWG -cabling, leading to very high di / dt ...  :whistling:)

Usually motor switching frequency is chosen above the audible range, to eliminate unwanted noise.  That's why I assumed "10's of Khz".  But, I don't have anything specific info for EUC wheels.

I didn't pick up which wires had melted insulation on Marty's wheel?  Were they the wires from the battery to the control board, or wires from it to the motor?

I would think parasitic inductance of any wires driving the gate of the FETS would be important.  They gate drive net is probably entirely contained on the pcb?  But I'd think for any wire in series with a motor winding, it shouldn't matter--the windings themself are much more inductive than any wire parasitic inductance.  If there's a snubber network of some type on the motor end of the wire, the wire would NOT be strictly in series.  The current in the wire would NOT always equal current in the associated windings.  But, if the current path in the wire is 100% common (in series) with a motor winding, it shouldn't matter.

 

Edited by DaveThomasPilot
added NOTs!

Share this post


Link to post
Share on other sites
14 minutes ago, DaveThomasPilot said:

Usually motor switching frequency is chosen above the audible range, to eliminate unwanted noise.  That's why I assumed "10's of Khz".  But, I don't have anything specific info for EUC wheels.

Yeah, I can't claim to know that they use lower frequency for the actual PWM either, it could be also just something (windings) resonating in the motor at lower frequency.

 

Quote

I didn't pick up which wires had melted insulation on Marty's wheel?  Were they the wires from the battery to the control board, or wires from it to the motor?

The wires going from PCB to motor, it has happened in many Gotways. Earlier problem was that the connectors heated so much that the solder melted and the wires could come off during riding, now that they've switched over to heavier duty connectors (and crimping + soldering, apparently), the wire insulation has melted and shorted the phases on at least a few riders.

 

Quote

I would think parasitic inductance of any wires driving the gate of the FETS would be important.  They gate drive net is probably entirely contained on the pcb? 

Yeah, it's on the PCB. Only board using ferrite beads in series to mitigate ringing on the gates I've seen so far has been the Firewheel / SBU -board, which doesn't even have paralleled mosfets, Rehab1 sent me his dead ACM-board (the MCU's fried), there they seem to use just resistor with diode in parallel (probably to speed up turn off?) for parallel mosfets (the gates are tied straight together, no resistors in between), and no ferrites. But the gate drive is completely built on the board, just PCB traces and SMDs in the path.

 

Quote

But I'd think for any wire in series with a motor winding, it shouldn't matter--the windings themself are much more inductive than any wire parasitic inductance.  If there's a snubber network of some type on the motor end of the wire, the wire would be strictly in series--current in the wire would always equal current in the associated windings.  But, if the current path in the wire is 100% common (in series) with a motor winding, it shouldn't matter.

You're probably right, the stray inductance of the wiring is likely very low in comparison to the actual motor winding. But still, with high current spikes (mosfets turning on and off, direction of current changing), could the skin-effect come into play if the frequency is high enough? As in, would the wires start to "act" like a thinner wire, as the current flows closer to the surface and not through the entire cross-section of the wire? Or maybe I've misunderstood the whole concept of the skin-effect :P  According to the Powerstream AWG-table, the 100% skin depth for solid 14AWG is 6.7kHz and 11kHz for 16AWG, of course the actual wires use multiple conductors, and I don't know how the surface-area / cross-section carrying the current should be modelled then...

Edited by esaj

Share this post


Link to post
Share on other sites
8 minutes ago, esaj said:

Yeah, I can't claim to know that they use lower frequency for the actual PWM either, it could be also just something (windings) resonating in the motor at lower frequency.

 

The wires going from PCB to motor, it has happened in many Gotways. Earlier problem was that the connectors heated so much that the solder melted and the wires could come off during riding, now that they've switched over to heavier duty connectors (and crimping + soldering, apparently), the wire insulation has melted and shorted the phases on at least a few riders.

 

Yeah, it's on the PCB. Only board using ferrite beads in series to mitigate ringing on the gates I've seen so far has been the Firewheel / SBU -board, which doesn't even have paralleled mosfets, Rehab1 sent me his dead ACM-board (the MCU's fried), there they seem to use just resistor with diode in parallel (probably to speed up turn off?) for parallel mosfets (the gates are tied straight together, no resistors in between), and no ferrites. But the gate drive is completely built on the board, just PCB traces and SMDs in the path.

 

You're probably right, the stray inductance of the wiring is likely very low in comparison to the actual motor winding. But still, with high current spikes (mosfets turning on and off, direction of current changing), could the skin-effect come into play if the frequency is high enough? As in, would the wires start to "act" like a thinner wire, as the current flows closer to the surface and not through the entire cross-section of the wire? Or maybe I've misunderstood the whole concept of the skin-effect :P  According to the Powerstream AWG-table, the 100% skin depth for solid 14AWG is 6.7kHz and 11kHz for 16AWG, of course the actual wires use multiple conductors, and I don't know how the surface-area / cross-section carrying the current should be modelled then...

Wow, hadn't thought about skin effect---but, most of the current isn't switching due to the high winding inductance.  So, (I think) the skin effect would only impact power dissipation from the ripple current.

So, I guess you could calculate the increased power due to skin effect by using ripple current, not total winding current.

 

Share this post


Link to post
Share on other sites

Just another way to see the cable prob:

AWG 16 has an resistance of 13,6 Ohm per km, so 13,6 mOhm per m so something about 5,5mOhm for the cabling from the PCB to the motor (~40cm?)

AWG 14 has 8,55 mOhm per Meter, so ~3,4mOhm

AWG 8 has ~0,85 - so the first below 1mOhm

Considering that the Mosfets have ~5mOhm Rdson and get a heatsink...

Cables can withstand a higher temperature than a mosfet, but that just "buys" some time...

Share this post


Link to post
Share on other sites
23 hours ago, Chriull said:

Just another way to see the cable prob:

AWG 16 has an resistance of 13,6 Ohm per km, so 13,6 mOhm per m so something about 5,5mOhm for the cabling from the PCB to the motor (~40cm?)

AWG 14 has 8,55 mOhm per Meter, so ~3,4mOhm

AWG 8 has ~0,85 - so the first below 1mOhm

Considering that the Mosfets have ~5mOhm Rdson and get a heatsink...

Cables can withstand a higher temperature than a mosfet, but that just "buys" some time...

@zlymexposted measurements from his MSuper3s+ 

showing an about ~50A average going up a big hill.

As it seems Mosfets are under controll - with overheating warnings/tiltbacks once it gets too much.

Cable losses (6-3 mOhm with 50A are around 7,5-15W) seem to be enough to get them overheated over time - or maybe it was alway somewhere a not so good connection/"offended" cable (bendings or whatsoever... anyhow just too high resistance) that generated the heat to get the cables/insulation down?

If now the cables get thicker, connectors better and they are under controll - what's the next part to fail?! One weak link eliminated - next one dies?

Are the mosfets back again - or will the PCB traces go up in smoke?

Or we get something new? The motor has an internal coil resistance of something about 0,1 Ohm - so with this 50A it dissipates 250W. Should be cooled enough by all the metal inside - i assume this should work out?!

Then we have the batteries - now they use in these wheels 20s6p? If they have comparable cells like the LG MJ1 (37mOhm) that would result in an internal resistance of 0,123 Ohm - dissipating ~300W! They are in a plastic sleeve - no cooling? The BMS should have temperature surveillance -> hard shutoff once they overheat?

Edit: As found out in http://forum.electricunicycle.org/topic/7549-current-demand-versus-battery-voltage/ the average battery current is much lower than the motor current in many cases, so in such a case there should/could be just ~15A battery current leading to just 30W battery losses!

 

Edited by Chriull

Share this post


Link to post
Share on other sites
On 2017/6/22 at 7:06 PM, esaj said:

the low-side mosfet-switch bank (made from 5 x 40V, 190A/1080A IRFP2804 mosfets in parallel),

Do you happen to have the datasheet of this IRFP2804? Cannot find it.

Share this post


Link to post
Share on other sites
35 minutes ago, esaj said:

Thanks. I got some IRLB3034 and IRFB7430 which have the same voltage rating(40V) but larger current rating. I bought these for the same reason to DIY a capacitor welder, although I bought some SCR as well for alternative switch to avoid the turn-off spike, but I haven't started yet. These two MOSFET btw are primarily for motor drives according to IR.

Share this post


Link to post
Share on other sites
On 7/1/2017 at 3:14 PM, Leonte Mihai said:

Yes man, endurance test, and Gotway failed again, and think these toys cost a fortune (at least for me, I earn 700 euros a month).

I know what you mean.  You don't really "invest" in the fun you get from a candy bar, because they are trivially cheap.  If you get one that's stale or something, or drop it in some unholy muck on the side of the road, the loss is no big deal.  But for an EUC, we need that "investment" in fun to pay off  If its performance comes up short, you're going to feel it and it's going to hurt.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...