Mosfets and wheel heat generation

[esaj]

What I've begun to wonder is how would it affect the heat if the wheels were using top-of-the-line newest MOSFETs. For example, 75NF75 (used at least in older Ninebot mainboards) has an Rds(on) (resistance caused by the mosfet itself when it's fully conducting) of less than 11 milliohm. But then, just as an example, International Rectifiers' IRFP4368 has Rds(on) of 1.8 milliohm, giving a staggering 350A Imax (maximum current, probably theoretical number?) vs. 80A in 75NF75. What I don't know is whether it actually could be used, as it has larger gate charge (380nC vs. max. 160nC in 75NF75). But there are some power MOSFETs with less than 5 milliohm Rds(on) and gate charge around the same values as 75NF75, for example IRFP3077 is 75V max / 3.3 milliohm / 160nC and goes up to 200A.

I'm not that familiar with all the datasheet values in MOSFETs and how they affect the behavior (since it's used just as a switch in the half-bridges, probably the gate charge needs to be small enough so that it can turn on and off fast enough, while being capable of withstanding the voltages & amperage?). My understanding is that Rds(on) is measured in saturation (the MOSFET is fully conducting), so it actually (probably) passes through the non-saturated region (don't remember what that was called, linear region?) which could of course affect things, but the reason the MOSFETs heat up is the amount of resistance there is, causing a (relatively small) voltage drop over the MOSFET, which means power dissipation. While the voltage drop is small, with large current those watts start to add up and the heat caused by that can be substantial.

Just as an example, and these are purely theoretical numbers based on my understanding, shooting 20A through a 75NF75 and the IRFP-mosfets mentioned above, the power dissipation at full conduction (minimum resistance) is:

75NF75:  VDrop = 20A * 0.011ohm = 0.22V    Power = 0.22V * 20A = 4.4W

IRFP4368:  VDrop = 20A * 0.0018Ohm = 0.036V   Power = 0.036V * 20A = 0.72W

IRFP3077: VDrop = 20A * 0.0033Ohm = 0.066V  Power =  0.036V * 20A = 1,32

So both the IRFP's should warm up much less than 75NF75. But there are things these simple calculations don't take into account (passing through the linear-region, the resistance will be higher for a brief period, large gate charge might mean that larger dead time insertions need to be used to be sure the mosfet is fully closed before opening the another mosfet in the same half-bridge etc.), and I don't know what sort of other challenges they might have in relation to motor drive design  (gate voltages/charges etc).

Firewheel actually uses surface-mount IRF7759's, which (according to the datasheet) have similar Rds(on) = 1.8milliohm as the TO-packaged IRFP4368 above, but higher gate charge (200nC). I've never had my Firewheel overheat (at least to my knowledge ) or heard of it happening with anyone else, but there is one in the french forums which had one of its mosfets blown, don't know if that was overheating or a bridge shoot-through, or something else.

[US69]

@esaj

But it are not just the mosfet's that generate heat, or?

We are blowing up to 3000 Watt to the Motor through this Little control board...so i think with the over the last time more powerfull Motors, rising from 300Watt IPS to 1500Watt GotwayACM, heat is becoming more and more of a Problem. So trying to get this under control by a power blower fan like this is a good solution. As we can guess from the Photo the "blower fan"  is just additional to cool the heatpipe under the board.......

 

 

[esaj]

3 minutes ago, KingSong69 said:

@esaj

But it are not just the mosfet's that generate heat, or?

We are blowing up to 3000 Watt to the Motor through this Little control board...so i think with the over the last time more powerfull Motors, rising from 300Watt IPS to 1500Watt GotwayACM, heat is becoming more and more of a Problem. So trying to get this under control by a power blower fan like this is a good solution. As we can guess from the Photo the "blower fan"  is just additional to cool the heatpipe under the board.......

AFAIK, most of the heat is generated by the drive bridges (basically the mosfets) alone, then maybe the step down to bring the 50-67V battery voltage to 12V or 5V (depending how it's done?)... but I could be wrong, I'm not that good with electronics   If you look at the boards, usually nothing else except the mosfets are attached to the heatsink(s).

[SlowMo]

I could see tiny ventilation shafts in the above picture just to the left and right of the cooling fan or am I wrong?

[US69]

11 minutes ago, esaj said:

AFAIK, most of the heat is generated by the drive bridges (basically the mosfets) alone, then maybe the step down to bring the 50-67V battery voltage to 12V or 5V (depending how it's done?)... but I could be wrong, I'm not that good with electronics   If you look at the boards, usually nothing else except the mosfets are attached to the heatsink(s).

For sure you have a lot more knowledge as me

i barely know nothing about the boards...just agree with KS to put some more thoughts into heat reduction

7 minutes ago, SlowMo said:

I could see tiny ventilation shafts in the above picture just to the left and right of the cooling fan or am I wrong?

I think these "blowers" take the air out of the middle and push it to a big fat hole???

[SlowMo]

3 minutes ago, KingSong69 said:

I think this "blowers" take the air out of the middle and push it to a big fat hole???

I was using similar fans in my pentium pc a long time ago.  Really, there should be some ventilation shaft somewhere in there otherwise the inside area would become hot like an oven.

[Chriull]

2 hours ago, esaj said:

What I've begun to wonder is how would it affect the heat if the wheels were using top-of-the-line newest MOSFETs. For example, 75NF75 (used at least in older Ninebot mainboards) has an Rds(on) (resistance caused by the mosfet itself when it's fully conducting) of less than 11 milliohm. But then, just as an example, International Rectifiers' IRFP4368 has Rds(on) of 1.8 milliohm, giving a staggering 350A Imax (maximum current, probably theoretical number?) vs. 80A in 75NF75. What I don't know is whether it actually could be used, as it has larger gate charge (380nC vs. max. 160nC in 75NF75). But there are some power MOSFETs with less than 5 milliohm Rds(on) and gate charge around the same values as 75NF75, for example IRFP3077 is 75V max / 3.3 milliohm / 160nC and goes up to 200A.

Should be a great idea! I would really prefer better Mosfets and maybe more sophisticated heat sink to a fan... Could be eventuelly also cheaper for the manufacturer?

If you look at the rise/fall times of the IRFP4368 ( 220/260 ns) against the 75NF75 (100/30 ns) it's not the big difference 

2 hours ago, esaj said:

I'm not that familiar with all the datasheet values in MOSFETs and how they affect the behavior (since it's used just as a switch in the half-bridges, probably the gate charge needs to be small enough so that it can turn on and off fast enough, while being capable of withstanding the voltages & amperage?). My understanding is that Rds(on) is measured in saturation (the MOSFET is fully conducting), so it actually (probably) passes through the non-saturated region (don't remember what that was called, linear region?) which could of course affect things,

The Mosfet changes between non conducting to conduction (saturation) through the "lineas region". So for the fall/rise time he goes from max voltage drop/no current to min voltage drop/max current.

So with every switching the Mosfet start with ~60V/0a and goes quite linear to the saturated phase(0,22/0,036/0,066V)/20A. So while switching there is in average (~60-0,0xx)/2V and (20-0)/2 A which leads to 60/2*20/2=300W - for something between 30ns and some 100 ns, depending on the circuit design and about 2*8000 times a second (for the new ninebot firmware for example with an PWM frequency of 8kHz).

So if one assumes a switching time of 400ns (which all of the three Mosfets can handle) with 16000 switching events per seconds, the 300W would be dissipated 16000*400ns=6,4ms every second. So in average a power disspation of ~2W. With the 75NF75 this could be theoretically reduced to ~0,5W by decreasing the switching time to their minimum value of ~100ns, for the IRF just to 1W for their ~200ns. Imho i have read somewhere that normally one does not use the minimal possible switching times because this would introduce too much high frequency electrical transients (for EMC). So this ~2W could be about realistic - dissipated by all 6 Mosfets of the bridge together.

2 hours ago, esaj said:

but the reason the MOSFETs heat up is the amount of resistance there is, causing a (relatively small) voltage drop over the MOSFET, which means power dissipation. While the voltage drop is small, with large current those watts start to add up and the heat caused by that can be substantial.

Just as an example, and these are purely theoretical numbers based on my understanding, shooting 20A through a 75NF75 and the IRFP-mosfets mentioned above, the power dissipation at full conduction (minimum resistance) is:

75NF75:  VDrop = 20A * 0.011ohm = 0.22V    Power = 0.22V * 20A = 4.4W

IRFP4368:  VDrop = 20A * 0.0018Ohm = 0.036V   Power = 0.036V * 20A = 0.72W

IRFP3077: VDrop = 20A * 0.0033Ohm = 0.066V  Power =  0.036V * 20A = 1,32

With an almost 100% duty cycle power dissipation for all six Mosfets would amount to two times this value (there would always be only one mosfet from the upper side and one from the lower side conducting). So this would make between 1,4 to 8,8W for the three half bridges.

This would sum up with the switching dissipation power to 3,4W (IRFP4368) to 10,8W (75NF75) for all six mosfets together.

An additional point is, that the Mosfets have a positive temperature cooefficient - so at 100°C junction temperature these mosfets have about 1.5 times the resistance and so 1.5 times the dissipated power from your example (but insignificant for the switching dissipated power). -> So this would lead to about 4,16 to 15,2 W

For 60V*20A=1200W continous power dissipation.

For 3000W peaks the 75NF75 would have to dissipate ~88W - which seems quite out of range for a passive cooling ;(, but for the IRFP4368 this would only mean ~18W.

2 hours ago, esaj said:

So both the IRFP's should warm up much less than 75NF75. But there are things these simple calculations don't take into account (passing through the linear-region, the resistance will be higher for a brief period, large gate charge might mean that larger dead time insertions need to be used to be sure the mosfet is fully closed before opening the another mosfet in the same half-bridge etc.),

The longer the Mosfets have to stay closed, the less power is dissipated -> less cooling needed 

2 hours ago, esaj said:

and I don't know what sort of other challenges they might have in relation to motor drive design  (gate voltages/charges etc).

 A major point could be how well the active freewheeling is implemented (you remember the video i found quite some time ago) where "switching peaks" and especially the reverse currents while regenerative breaking are not flowing through the body diode, but the mosfets get switched on by the controller. Flowing through the body diode the dissipated power is horrible (forward voltage about 1-1,4V) instead of the RDS on of the switched on mosfet.

 

[esaj]

1 hour ago, Chriull said:

Should be a great idea! I would really prefer better Mosfets and maybe more sophisticated heat sink to a fan... Could be eventuelly also cheaper for the manufacturer?

Probably the very newest technology can be a bit pricey, but then again, Firewheel mainboards do use the 1.8 milliohm SMD-versions, so it shouldn't be that expensive. And probably the mainboard manufacturing price isn't that high compared to the batteries and motors. But I'm just guessing, I haven't got much idea what it costs, but still I'd expect it to be relatively cheap (when buying components & manufacturing in bulk). No idea on design & prototyping costs, though.

 

Quote

If you look at the rise/fall times of the IRFP4368 ( 220/260 ns) against the 75NF75 (100/30 ns) it's not the big difference

Although it's over double, it doesn't sound much as the bridges aren't really being driven in high frequencies (like megahertzes ). If I calculated it right, 520ns (2 * 260ns, rise + fall in succession) should get closer to 2Mhz, so (up to) a few tens of kilohertz for motor drive shouldn't really be a problem?

 

Quote

The Mosfet changes between non conducting to conduction (saturation) through the "lineas region". So for the fall/rise time he goes from max voltage drop/no current to min voltage drop/max current.

So with every switching the Mosfet start with ~60V/0a and goes quite linear to the saturated phase(0,22/0,036/0,066V)/20A. So while switching there is in average (~60-0,0xx)/2V and (20-0)/2 A which leads to 60/2*20/2=300W - for something between 30ns and some 100 ns, depending on the circuit design and about 2*8000 times a second (for the new ninebot firmware for example with an PWM frequency of 8kHz).

So if one assumes a switching time of 400ns (which all of the three Mosfets can handle) with 16000 switching events per seconds, the 300W would be dissipated 16000*400ns=6,4ms every second. So in average a power disspation of ~2W. With the 75NF75 this could be theoretically reduced to ~0,5W by decreasing the switching time to their minimum value of ~100ns, for the IRF just to 1W for their ~200ns. Imho i have read somewhere that normally one does not use the minimal possible switching times because this would introduce too much high frequency electrical transients (for EMC). So this ~2W could be about realistic - dissipated by all 6 Mosfets of the bridge together.

With an almost 100% duty cycle power dissipation for all six Mosfets would amount to two times this value (there would always be only one mosfet from the upper side and one from the lower side conducting). So this would make between 1,4 to 8,8W for the three half bridges.

This would sum up with the switching dissipation power to 3,4W (IRFP4368) to 10,8W (75NF75) for all six mosfets together.

An additional point is, that the Mosfets have a positive temperature cooefficient - so at 100°C junction temperature these mosfets have about 1.5 times the resistance and so 1.5 times the dissipated power from your example (but insignificant for the switching dissipated power). -> So this would lead to about 4,16 to 15,2 W

For 60V*20A=1200W continous power dissipation.

For 3000W peaks the 75NF75 would have to dissipate ~88W - which seems quite out of range for a passive cooling ;(, but for the IRFP4368 this would only mean ~18W.

I had to read through this a couple of times to get it  But yeah, that's a huge difference in the end with high peak power, 88W (in total) vs. 18W dissipated by six components the size of your fingernails. 3kW could probably be reached (momentarily) during a very strong braking... It does make me think that the manufacturers should upgrade their mosfets instead of (or in addition to) trying to make the cooling better. For comparison, my computer has an older generation 4-core Xeon with 130W TDP and it needs a huge heatsink + active cooling fans that get really loud during high loads, like running complex mutltithreaded LTSpice -simulations or running server software with hundreds of threads 

 

Quote

The longer the Mosfets have to stay closed, the less power is dissipated -> less cooling needed 

Didn't think of that, but true   Something like a few microseconds of DTI is probably peanuts, considering for example a 10kHz PWM, where the period length is 100µs...

 

Quote

 A major point could be how well the active freewheeling is implemented (you remember the video i found quite some time ago) where "switching peaks" and especially the reverse currents while regenerative breaking are not flowing through the body diode, but the mosfets get switched on by the controller. Flowing through the body diode the dissipated power is horrible (forward voltage about 1-1,4V) instead of the RDS on of the switched on mosfet.

Excuse my ignorance, but that does raise a question... since the high-side mosfets sources are connected to the motor, and in the reverse current flow the motor back-EMF gives for example 70V during braking, doesn't that mean that the gate voltage must be the back-EMF + 10V (or 5V for "logic-level mosfets" or whatever is needed to enter the "saturation")? If so, then the follow-up questions are:

-Can the gate withstand higher voltages than drain-source -connection? Those IRFs (and 75NF75) I've used as an example are rated for 75V max drain-source -voltage, but can the gate be even higher? In the above example, if the gate needed to be source-voltage + 10V, it would already be 80V
-If the gate-voltage isn't high enough to make the mosfet fully conductive, does at least some part of the current run through the body-diode then?

Although I did build a simple step-up converter to drive around 45V 15W led-lamp from 6V batteries, I must admit that I don't fully understand how the voltage of "inductive kickback" acts... Does the induction (collapsing magnetic field of the coil) only generate voltage that's "sufficient" to start current flowing somewhere, or otherwise the voltage keeps ramping up (of course there's some upper limit)? Probably I should try it in a simulation and hit the books again, at one point I felt that I start to understand electronics, but was quickly returned down to earth after starting reading "Intuitive Analog Circuit Design", which throws differential equations at your face in pretty much every page...   Not that intuitive for me, especially since I haven't done differential equations since my math-classes maybe in 2007 or 2008

[Chriull]

 

4 minutes ago, esaj said:

Although it's over double, it doesn't sound much as the bridges aren't really being driven in high frequencies (like megahertzes ). If I calculated it right, 520ns (2 * 260ns, rise + fall in succession) should get closer to 2Mhz, so (up to) a few tens of kilohertz for motor drive shouldn't really be a problem?

The "switching" frequency for the "?slope/edge?" itself is in the MHz area, also the PWM frequency itself (the rectangular pulses) generated are in the tens of kHz range...

If the switching itself gets too slow the dissipated power will again get way too high... ;(

4 minutes ago, esaj said:

Excuse my ignorance, but that does raise a question... since the high-side mosfets sources are connected to the motor, and in the reverse current flow the motor back-EMF gives for example 70V during braking, doesn't that mean that the gate voltage must be the back-EMF + 10V (or 5V for "logic-level mosfets" or whatever is needed to enter the "saturation")?

Ups. Yes - do much writing and not enough thinking. The Mosfet would have to magicaly change to a p-channel Mosfet (or jump on the pcb and change the drain and source connection)... ;). So for regenerative braking only the body diodes can be used with their quite high forward voltage. There only special diodes (like schottky with very low forward voltages) could help to minimize power dissipation...

But to shorten the "inductive kickback" it is usable - so maybe if one switches the Mosfet fast enough so no negtive DS Voltage can not build up it should work. Maybe i try to find this youtube video about active freewheeling again - was, as far as i remember quite well made...

4 minutes ago, esaj said:

Although I did build a simple step-up converter to drive around 45V 15W led-lamp from 6V batteries, I must admit that I don't fully understand how the voltage of "inductive kickback" acts... Does the induction (collapsing magnetic field of the coil) only generate voltage that's "sufficient" to start current flowing somewhere, or otherwise the voltage keeps ramping up (of course there's some upper limit)?

The induction tries to keep the current flowing - to achieve this the voltage will rise until the magnetic field can (fully) release its energy. That's a good way to create sparks or destroy semiconductors...;) So for this one puts the freewheeling diodes antiparallel to bipolar transistors switching inductive loads - mosfets can handle this with the body diode, but there exist much better diodes to handle these situations...

 

[esaj]

35 minutes ago, Chriull said:

Ups. Yes - do much writing and not enough thinking. The Mosfet would have to magicaly change to a p-channel Mosfet (or jump on the pcb and change the drain and source connection)... ;). So for regenerative braking only the body diodes can be used with their quite high forward voltage. There only special diodes (like schottky with very low forward voltages) could help to minimize power dissipation...

Ah, ok, so the mosfet cannot conduct in "reverse" (except for the body diode). I have some MBR30100's on the way (http://www.secosgmbh.com/datasheet/products/Schottky_Rectifier/TO-220J/MBR30100.pdf  100V / 30A power-Schottky), so it should be possible to bypass the the mosfet "in reverse" without overheating the mosfet itself due to body diode voltage drop? 100V should be enough to prevent the actual battery voltage running towards the motor... At least "large enough" Schottkys seem to be available (there's also MBR40100, 50100, 60100... ). Do you know if the wheels actually have separate bypass diodes or just rely on body diodes during regenerative braking (at least they don't seem to have anything as big as those TO-220 -packaged power-Schottkys).

I just ordered them without thinking too much what I'd use them for (probably for a high current voltage source or such), but now that you mentioned it, I could place them in parallel with the mosfets for the 3-phase driver (although I probably wouldn't need that high amperage, as I'll likely drive much, much smaller motors with it ).

 

Quote

But to shorten the "inductive kickback" it is usable - so maybe if one switches the Mosfet fast enough so no negtive DS Voltage can not build up it should work. Maybe i try to find this youtube video about active freewheeling again - was, as far as i remember quite well made...

The induction tries to keep the current flowing - to achieve this the voltage will rise until the magnetic field can (fully) release its energy. That's a good way to create sparks or destroy semiconductors...;) So for this one puts the freewheeling diodes antiparallel to bipolar transistors switching inductive loads - mosfets can handle this with the body diode, but there exist much better diodes to handle these situations...

I did build that step-up converter without really knowing what I was doing... it was at first meant to run 9V / 700mA power LEDs, but after I saw huge voltage spikes with the oscilloscope (I don't remember what I was using as load at that point), I decided to test if it could actually drive a bigger 15 * 1W LEDs in series -lamp   So far it has worked fine (it was meant to replace my desklamp, but currently I've used it for soldering, really helps with seeing what I'm doing: http://imgur.com/vukqzJf), but it's using a BJT-power transistor instead of mosfet (and the transistor & inductor both do get pretty hot after a while). I've been writing some monologues about my (pretty basic) electronics projects in the off-topic forum:

EDIT: Sorry for the thread-hijack (once again ), I probably should split the mosfet-talk somewhere else... General discussion?

[Chriull]

3 hours ago, esaj said:

Ah, ok, so the mosfet cannot conduct in "reverse" (except for the body diode).

I assume so - but all the theory i knew a long long time ago...;)

3 hours ago, esaj said:

I have some MBR30100's on the way (http://www.secosgmbh.com/datasheet/products/Schottky_Rectifier/TO-220J/MBR30100.pdf  100V / 30A power-Schottky), so it should be possible to bypass the the mosfet "in reverse" without overheating the mosfet itself due to body diode voltage drop? 100V should be enough to prevent the actual battery voltage running towards the motor... At least "large enough" Schottkys seem to be available (there's also MBR40100, 50100, 60100... ). Do you know if the wheels actually have separate bypass diodes or just rely on body diodes during regenerative braking (at least they don't seem to have anything as big as those TO-220 -packaged power-Schottkys).

I just ordered them without thinking too much what I'd use them for (probably for a high current voltage source or such), but now that you mentioned it, I could place them in parallel with the mosfets for the 3-phase driver (although I probably wouldn't need that high amperage, as I'll likely drive much, much smaller motors with it ).

Should help the mosfets. Will be interesting if you make before/afterwards measurements. In case there is really some power to be dissipated with your smaller motors do not forget to give them a heatsink too

3 hours ago, esaj said:

 

I did build that step-up converter without really knowing what I was doing... it was at first meant to run 9V / 700mA power LEDs, but after I saw huge voltage spikes with the oscilloscope (I don't remember what I was using as load at that point), I decided to test if it could actually drive a bigger 15 * 1W LEDs in series -lamp   So far it has worked fine (it was meant to replace my desklamp, but currently I've used it for soldering, really helps with seeing what I'm doing: http://imgur.com/vukqzJf), but it's using a BJT-power transistor instead of mosfet (and the transistor & inductor both do get pretty hot after a while). I've been writing some monologues about my (pretty basic) electronics projects in the off-topic forum:

EDIT: Sorry for the thread-hijack (once again ), I probably should split the mosfet-talk somewhere else... General discussion?

For step up converters if found this pdf http://cdn.intechopen.com/pdfs-wm/13353.pdf where maybe section 3.2 and 3.3 could be interesting for you (if you use a similar circuit). Do put a snubber circuit to the coil to take the burden of your bjt.

[Hunka Hunka Burning Love]

9 hours ago, KingSong69 said:

@esaj

But it are not just the mosfet's that generate heat, or?

We are blowing up to 3000 Watt to the Motor through this Little control board...so i think with the over the last time more powerfull Motors, rising from 300Watt IPS to 1500Watt GotwayACM, heat is becoming more and more of a Problem. So trying to get this under control by a power blower fan like this is a good solution. As we can guess from the Photo the "blower fan"  is just additional to cool the heatpipe under the board.......

 

 

I wonder whether a future design could implement a heat sink sandwiched in between the clamshells so the cooling fins would be exposed to the big fan also known as the hub wheel that is constantly spinning underneath.  There must be a lot of cooling air being circulated as the wheel spins.  Why not use that instead of using a separate fan that eats up battery power?  

The heatsink could be housed with a large rubber gromet around the edges to help keep things waterproof or be permanently mounted and sealed.  One other idea would be to make the top 1/8 of the EUC out of aluminum and mount the mosfets to it so you have part of the casing be the heat sink with the wheel itself cooling fins underneath and the side parts dissapating heat as well.  I guess my main point is to expose some of the heatsink to the outside elements rather than being entirely enclosed in the plastic casing which just keeps the heat in.

[esaj]

28 minutes ago, HunkaHunkaBurningLove said:

I wonder whether a future design could implement a heat sink sandwiched in between the clamshells so the cooling fins would be exposed to the big fan also known as the hub wheel that is constantly spinning underneath.  There must be a lot of cooling air being circulated as the wheel spins.  

It's not a bad idea, and has come up from time to time. Probably the heat sink part inside the wheel casing shouldn't have any fins, though, as they'd get clogged up by mud & crap flying from the tire over time and drying in place. Firewheel actually has a large metal plate to which the mainboard & heatsink are attached to, that is next to the spinning tire (on the side, not on the top), that probably helps at least a little on heat removal (then again, there's nothing between the heatsink & plate, so I don't know how well it actually conducts heat to the plate).

Quote

Why not use that instead of using a separate fan that eats up battery power?  

Compared to driving the motor, the energy used by the fan is nothing... for example, I have a 80x80mm case-fan I intend to use for a constant current source cooling, and it's rated at 12V & 0.14A = 1.68W, and in comparison, the motor can pull over 1000W on peaks and usually hundreds of watts on average during riding. Even if the fan used in the KS was larger wattage, I doubt you'd see much (if any) difference in range or power whether it was running or not.

 

Quote

The heatsink could be housed with a large rubber gromet around the edges to help keep things waterproof or be permanently mounted and sealed.  One other idea would be to make the top 1/8 of the EUC out of aluminum and mount the mosfets to it so you have part of the casing be the heat sink with the wheel itself cooling fins underneath and the side parts dissapating heat as well.  I guess my main point is to expose some of the heatsink to the outside elements rather than being entirely enclosed in the plastic casing which just keeps the heat in.

That was my thought too, outside it could also be finned to increase surface area, as you could clean it up without dismantling the wheel. Care should of course be taken that either the sink can't get too hot (large enough) or if it does, it should be placed so that you don't accidentally burn your hand when grabbing the handle or such. 

 

 

2 hours ago, Chriull said:

Should help the mosfets. Will be interesting if you make before/afterwards measurements. In case there is really some power to be dissipated with your smaller motors do not forget to give them a heatsink too

I did a simple(ish) simulation of three-phase BLDC with LTSpice (but it was probably too simple to actually give realistic results) and then tried it with & without diodes in parallel with the high-side mosfets. It would seem that the diodes help with the "reverse" flow, ie. much less current runs through the mosfets themselves. Of course I'll have to test it in reality to see the actual results, trying to simulate the back-EMF with PWM-drives and battery voltage etc. in LT-spice isn't that easy   It might be a while before I get back on the 3-phase BLDC project, as I've been sidetracked by building my own tools (currently waiting for assembly: constant current source & overvoltage-protector, a simple function & noise -generators would be nice, maybe an electronic load... ). If I wasn't so damn cheap, I could of course just buy real (much more precise & higher quality) tools, but I figured I'd learn a lot on the way too

Quote

For step up converters if found this pdf http://cdn.intechopen.com/pdfs-wm/13353.pdf where maybe section 3.2 and 3.3 could be interesting for you (if you use a similar circuit). Do put a snubber circuit to the coil to take the burden of your bjt.

I did read up on that, but it seems to use two coil-setup, mine only has one... but thanks for the tip, I'll probably read up more on snubbers & build a better step up at some point (actually, I'd need another step-up if I want to build a flashlight that uses 9V / 3W power leds... ). 555 would probably give a better PWM-source than my multivibrator-setups.

[Hunka Hunka Burning Love]

Are there vents in the Kingsong for the hot air to exit and cool air to enter efficiently or is it more a convection oven with those small venting tubes on the side?  I know it's tricky to find a balance between keeping things waterproof while having good breathability on a device like this.  

One issue with fans is their buzzing noise, and the fact that they eventually die sooner or later.  I see your point regarding power consumption.  It would be nice to find a simpler, more elegant solution like those water cooled CPU setups you see.  I wonder if moving the mosfets to a daughter board attached to a heatsink would allow more creative positioning to gain access to better cooling solutions than there are now.

[esaj]

2 minutes ago, HunkaHunkaBurningLove said:

Are there vents in the Kingsong for the hot air to exit and cool air to enter efficiently or is it more a convection oven with those small venting tubes on the side?  I know it's tricky to find a balance between keeping things waterproof while having good breathability on a device like this.  

From the pictures I've seen so far, I really cannot tell.

2 minutes ago, HunkaHunkaBurningLove said:

One issue with fans is their buzzing noise, and the fact that they eventually die sooner or later.  I see your point regarding power consumption.  It would be nice to find a simpler, more elegant solution like those water cooled CPU setups you see.

Probably the noise is quiet enough that you can't hear it (much) from inside the case, but the lifetime of the fan might be an issue. Water-cooling, while cool (pun intended ),  would probably be more trouble than it's worth; if the pump dies and the system cannot keep the temperature in check without it, you end up frying your board. If the pipes break or leak for some reason (shock from a fall or such), you could end up frying your board. Passive systems, being maintenance & care free would seem to me the way to go, maybe passive heatpipes could be used (but again, there's the danger of breakage).

2 minutes ago, HunkaHunkaBurningLove said:

 I wonder if moving the mosfets to a daughter board attached to a heatsink would allow more creative positioning to gain access to better cooling solutions than there are now.

That could be a good idea, although you need some wiring running between the other parts of the mainboard and the half-bridges (at least PWM-signals, hall-sensor signals and probably back-EMF / current measurement). Maybe running them in wires instead of board traces could induce too much interference (you certainly don't want the PWM- or hall-signals to get interference, that would probably lead to a face-plant really fast ).

Another thing that I've wondered about has been why they use such flimsy looking wires between the half-bridges and the motor itself? Usually it seems that the power to the mainboard/bridges is brought with something like 16AWG or 14AWG wires, and then you have these "strings" running to the actual motor   Not that anyone has reported that their wires have melted due to overheating, but still...

[Jason McNeil]

Yes, found it! Great thread, I remember reading in the "Electric Motors & Drives- Fundamentals, Types & Applications" 3rd Edition that states that the most problematic source of heat build up in MOSFETs is not when during the saturation phase, but rather during the transition.  

"...that some heat is generated when the transistor is on, and at low switching rates this is the main source of unwanted heat. But at high switching rates, ‘switching loss’ can also be very important.
Switching loss is the heat generated in the finite time it takes for the transistor to go from on to off or vice versa. The base-drive circuitry will be arranged so that the switching takes place as fast as possible, but in practice it will seldom take less than a few microseconds. During the switch ‘on’ period, for example, the current will be building up, while the collector–emitter voltage will be falling towards zero. The peak power reached can therefore be large, before falling to the relatively low on state value. Of course the total energy released as heat each time the device switches is modest because the whole process happens so quickly."

There's a very good paper written about calculating these losses, but have to confess that's outside my current level of understanding.

http://electronicdesign.com/boards/calculate-dissipation-mosfets-high-power-supplies

[Chriull]

12 hours ago, esaj said:

... Do you know if the wheels actually have separate bypass diodes or just rely on body diodes during regenerative braking (at least they don't seem to have anything as big as those TO-220 -packaged power-Schottkys).

Could be - I already got this idea while looking at the ninebot control board. There are 12 To-220 sitting on the heatsink. 6 are "named" and for sure Mosfets, the other six have nothing written on it and could be diodes and not paralleled mosfets...

6 hours ago, HunkaHunkaBurningLove said:

Are there vents in the Kingsong for the hot air to exit and cool air to enter efficiently or is it more a convection oven with those small venting tubes on the side?  I know it's tricky to find a balance between keeping things waterproof while having good breathability on a device like this.  

There has to be an airflow (fresh air in - hot air out). Just mixing the hot air in the small compartment would be a cosmetic measure - i do not think/hope Kingsong just produced a marketing gag...

6 hours ago, esaj said:

From the pictures I've seen so far, I really cannot tell.

Probably the noise is quiet enough that you can't hear it (much) from inside the case, but the lifetime of the fan might be an issue.

Normaly noise from fans is not the real issue - as long as the device is new. But about all of my stuff with fans gets louder and louder over the time...

The worst thing is if the fan starts "to resonate" with the chassis - then it get really annouing ;(

.. I already see in mods&repair threads starting to mod/change/improve the fan...

6 hours ago, esaj said:

Another thing that I've wondered about has been why they use such flimsy looking wires between the half-bridges and the motor itself? Usually it seems that the power to the mainboard/bridges is brought with something like 16AWG or 14AWG wires, and then you have these "strings" running to the actual motor   Not that anyone has reported that their wires have melted due to overheating, but still...

Also in the motor are just "normal" wires - thicker wires to the motor would maybe reduce a little bit the losses, but so they are maybe used as fuses to secure the motor

6 hours ago, Jason McNeil said:

...
Switching loss is the heat generated in the finite time it takes for the transistor to go from on to off or vice versa....

There's a very good paper written about calculating these losses, but have to confess that's outside my current level of understanding.

http://electronicdesign.com/boards/calculate-dissipation-mosfets-high-power-supplies

 

There they state something about switching times in the micro seconds range... Could be that the minimum values for t on and t off from the datasheet cannot be reached safely when driving especially inductive loads. That could easily bring the switching power dissipation in higher ranges than estimated above....

[esaj]

15 minutes ago, Chriull said:

Could be - I already got this idea while looking at the ninebot control board. There are 12 To-220 sitting on the heatsink. 6 are "named" and for sure Mosfets, the other six have nothing written on it and could be diodes and not paralleled mosfets...

Forgot about the "unknown" TO-220's in Ninebots. Diodes could make sense.

15 minutes ago, Chriull said:

There they state something about switching times in the micro seconds range... Could be that the minimum values for t on and t off from the datasheet cannot be reached safely when driving especially inductive loads. That could easily bring the switching power dissipation in higher ranges than estimated above....

It seems that the source was also talking about BJTs, not mosfets:

 in practice it will seldom take less than a few microseconds. During the switch ‘on’ period, for example, the current will be building up, while the collector–emitter voltage will be falling towards zero.

But I don't remember if (power-)BJTs have slower rise and fall times than mosfets, maybe? Should compare some datasheets with mosfets... At least their hFEs are usually quite low (unless it's a Darlington-pair)?

[electric_vehicle_lover]

On the OpenSource firmware I am using 3us dead time (you can see here: https://github.com/generic-electric-unicycle/firmware/blob/master/src/pwm.c):

TIM_BDTRInitStructure.TIM_DeadTime = 165; // 3us dead time

I got to that value looking at oscilloscope and I shared here: https://github.com/generic-electric-unicycle/documentation/wiki/MicroWorks-18km-h-controller-board

When we have the final firmware working, you will be able to change the mosfets as you like and tune to optimize the DeadTime :-)

[esaj]

Referring back to the question whether a mosfet can conduct in "reverse" (except through the body-diode): yes, it can (through the normal channel):

http://electronics.stackexchange.com/questions/105559/does-mosfet-let-current-flow-through-source-to-drain-as-it-allows-it-from-drain

Due to the body diode, most discrete MOSFETs cannot block in the reverse direction, but the channel will conduct in either direction when the gate is biased "on"

It still leaves me with the question whether the gate can be biased above the drain-source maximum voltage (75V for example) and by how much:

On 4.4.2016 at 6:28 PM, esaj said:

Excuse my ignorance, but that does raise a question... since the high-side mosfets sources are connected to the motor, and in the reverse current flow the motor back-EMF gives for example 70V during braking, doesn't that mean that the gate voltage must be the back-EMF + 10V (or 5V for "logic-level mosfets" or whatever is needed to enter the "saturation")? If so, then the follow-up questions are:

-Can the gate withstand higher voltages than drain-source -connection? Those IRFs (and 75NF75) I've used as an example are rated for 75V max drain-source -voltage, but can the gate be even higher? In the above example, if the gate needed to be source-voltage + 10V, it would already be 80V
-If the gate-voltage isn't high enough to make the mosfet fully conductive, does at least some part of the current run through the body-diode then?

For example the datasheets for ST 75NF75 state:

VDS    Drain-source voltage (VGS = 0)    75V

VDGR   Drain-gate voltage (RGS = 20KΩ)   75V

VGS    Gate-source voltage    ±20 V

under maximum ratings, so the gate-voltage should be within 20 volts (either direction) from the source voltage? I'd interpret that this means that the gate-voltage can go above maximum Vds, but not sure. Also now I began wondering can you drop the gate to 0V, if there's large voltage on the source, as it could be much bigger difference than -20V between them

[Hunka Hunka Burning Love]

On 4/4/2016 at 4:41 PM, esaj said:

Another thing that I've wondered about has been why they use such flimsy looking wires between the half-bridges and the motor itself? Usually it seems that the power to the mainboard/bridges is brought with something like 16AWG or 14AWG wires, and then you have these "strings" running to the actual motor   Not that anyone has reported that their wires have melted due to overheating, but still...

I've wondered the same thing, but I guess there's not much current going through those wires due to the passing magnet so they don't need to be heavy gauge?  You would still think that they would use some beefier wiring just due to the vibration ad friction related to a moving device.  Repeated movement and vibration can cause wear on wire insulation resulting in a short and failure.  I've seen incidences with wiring harnesses in $5000 pieces of equipment developing frictional wear ending up in a short damaging it.  On a EUC going at speed, the last thing anyone would want is an electronics failure.  Considering some of these wheels are produced with the cheapest methods/components available, it shouldn't be that surprising to see on a generic EUC like mine. 

[Chriull]

 

8 hours ago, esaj said:

Referring back to the question whether a mosfet can conduct in "reverse" (except through the body-diode): yes, it can (through the normal channel):

http://electronics.stackexchange.com/questions/105559/does-mosfet-let-current-flow-through-source-to-drain-as-it-allows-it-from-drain

Due to the body diode, most discrete MOSFETs cannot block in the reverse direction, but the channel will conduct in either direction when the gate is biased "on"...

Interesting finding! So my first intuiton was right  - but i did not find anything to it shortly searching the web... Seems to be not the 08/15 usage - or so trivial that noone writes about it...

Another thing - somewhere here sometimes i read about a new gotway getting driven by a 100V battery? Was this just a guess, a rumor or a real anouncement?

Higher Battery Voltages should also help to keep the mosfets cooler - less current is needed for the same power and with the same RDSon less power is dissipated by the mosfets!

[esaj]

15 minutes ago, Chriull said:

 

Interesting finding! So my first intuiton was right  - but i did not find anything to it shortly searching the web... Seems to be not the 08/15 usage - or so trivial that noone writes about it...

Could be too trivial for most people, I tend to stick (too much?) on the trivial details...   I read about mosfets from a couple of books today, and now I think I realize the gate-voltage part (plus checking back on the motor driver schematics I have, the high side-gate bias/voltage pump is always connected to the source, so it stays within the limits). So basically it isn't related really to the drain-source-voltage limit, as it's (of course) in relation to the source voltage. So you could very well have +85V (compared to ground) on the gate, with the source at 70V (as long as the voltage difference between drain and source doesn't exceed the Vds(max) and Vgs stays within the maximum +- voltage of source). As the gate in a mosfet is actually more like a capacitor, too high charge  voltage difference will break the very thin insulation between the gate and the whatwasitcalled... (checks from a book) substrate.

 

Quote

Another thing - somewhere here sometimes i read about a new gotway getting driven by a 100V battery? Was this just a guess, a rumor or a real anouncement?

Higher Battery Voltages should also help to keep the mosfets cooler - less current is needed for the same power and with the same RDSon less power is dissipated by the mosfets!

It's a Rockwheel:

Not sure on the 100V -part, might be just a rumor

[MaxLinux]

On ‎4‎/‎4‎/‎2016 at 6:32 PM, HunkaHunkaBurningLove said:

One issue with fans is their buzzing noise, and the fact that they eventually die sooner or later.

I wouldn't mind fan noise very much, but I have a serious worry about fan failure. I've lost count of how many cooling fans I've seen go bad in computer equipment and power supplies. When the fan fails, in the best case scenario, the customer has major inconvenience when operation of the wheel is blocked or curtailed by the overheating protection system. In the worst case scenario, components are destroyed by excessive heat. I hope KS has subjected these fans to massive durability testing.

[Hunka Hunka Burning Love]

This is a classic thread from the past.  I enjoyed re-reading it and not understanding most of it a second time around.  

It sounds like the fan comes on rarely so that should extend it's lifetime even more.   Taking into account the running time of EUCs versus computers which can be on for hours or days at a time the likelihood of fan failure is likely small.  It would be nice if the control board was smart enough to issue a beep warning if the fan RPMs don't meet spec, but it's likely just a simple unmonitored fan.  My bigger concern would be where is the fan blowing out the hot air?  It sounds like it works well enough though judging from people's reports.