30Kw EV Motor Controller

Irving

Joined Jan 30, 2016
1,845
80W of heat??? :oops:
Not quite, I've corrected it. But yes, if you under-drive the MOSFETs the turn-on & turn-off times increase and so do the switching losses. It's easy to validate with a simulation. It sounds high but then those MOSFETs are switching 87.5A at 60v, 5250W, so ~1% isn't bad. A MOSFET with a lower gate charge would help, but at this level most have geometries that have that result.
 

shortbus

Joined Sep 30, 2009
8,706
It sounds high but then those MOSFETs are switching 87.5A at 60v, 5250W, so ~1% isn't bad.
That sounds like my original idea of how mosfets worked. But after posting on many sites and being told I was wrong I changed my thoughts, are you now saying this is how heat is figured in a mosfet?

What every one that knows way more than me about electronics were telling me is, the heating watts of a mosfet is based on the On resistance of the device times the voltage. Not the amperage of the load. I ask this with also saying, I don't really know but really do want to understand better, not a trolling question.
 

cmartinez

Joined Jan 17, 2007
7,368
That sounds like my original idea of how mosfets worked. But after posting on many sites and being told I was wrong I changed my thoughts, are you now saying this is how heat is figured in a mosfet?

What every one that knows way more than me about electronics were telling me is, the heating watts of a mosfet is based on the On resistance of the device times the voltage. Not the amperage of the load. I ask this with also saying, I don't really know but really do want to understand better, not a trolling question.
What you're saying is news to me... this thread has now my undivided attention...
 

Irving

Joined Jan 30, 2016
1,845
MOSFET losses are primarily switching losses + conduction losses. In the static case, conduction losses are primary, and they are Rds(on)*Id^2*D or Vds(on) * Id *D, where D is duty cycle, eg t(on)/(cycletime).
Which formula you use depends on what values you know or can reasonably guess at.

The switching losses are a different beast. It's a function of the switch over from low current/high volts to high current/low volts and the rise and fall times. Essentially the power loss is the average current through the transition * the average voltage, or Imax * Vmax * 1/2 the transition time (i.e tr or
tf, 2 transitions per cycle) integrated over the cycle time, so * switching frequency.

Switching losses can easily exceed conduction losses in a high frequency SMPS or motor driver.
 
Last edited:

shortbus

Joined Sep 30, 2009
8,706
Thanks for this. But the formulae and understanding them is the big reason I didn't become an engineer.

So what your saying is that the heat is coming from the fact this is a constant switching(PWM) circuit not a on-off type circuit? It's from the mosfet going through the different on and off states causing the biggest heating?

My use of mosfets was mostly as a switch not as a PWM .
 

Irving

Joined Jan 30, 2016
1,845
Thanks for this. But the formulae and understanding them is the big reason I didn't become an engineer.

So what your saying is that the heat is coming from the fact this is a constant switching(PWM) circuit not a on-off type circuit? It's from the mosfet going through the different on and off states causing the biggest heating?

My use of mosfets was mostly as a switch not as a PWM .
A picture (or several!) will make it clearer I hope...

Here's a simple simulation of a MOSFET M1 being switched on & off by a 50kHz square wave. In the upper panel you can see the gate drive current Ig(M1) in red is approx 2A peak and the gate voltage (green) is 15v. In the lower panel the drain current Id(M1) (red) goes to ~87.8A when M1 is on and drain voltage (green) at 60v when off.

1623163034031.png

Now lets zoom into the switch on transition...
Here we can see the drain current (lower pane, red) takes around 17.3nS delay t(on) before it starts to rise, and a further 38.1nS rise time (tr 10->90%). Drain voltage (green) takes similar timings to drop. During the transition M1 experiences a heating effect shown by the pink curve which peaks at 1.32kW, though only averaging 452W across the 100nS of that snapshot.

1623163967550.png

Looking further out in the cycle, we are in quasi-static conditions so the 'static' or, more accurately, 'ohmic' losses are in play. Drain current is 87.9A and drain voltage, Vds(on), is 223mV, giving Rds(on) as 2.54mOhm. Power dissipation is 19.6W (either V * I = 0.223V * 87.9A, or I^2 * R = 87.9A^2 * 0.00254Ω ). Remember this is only for (in this simulation) 50% of the time, so the overall 'ohmic' loss is actually 9.8W per cycle.

1623164285477.png

Finally, the switch off transition; a massive 97nS delay, then 88nS fall time, resulting in a peak dissipation again of 1.32kW but spread over a wider transition gives a higher average of 518W (over 200nS, >1000W per 100nS avg).
1623164737303.png

Taken across the whole 20uS cycle, power dissipation in M1 is 17.2W, of which 9.8W is static 'ohmic' losses and 7.4W switching losses.

Finally, lets look at the effect of increasing R2 from 5 to 15ohm, reducing gate current from 2A peak to 900mA peak....
This increase turn on time to 41nS (from 17nS) and rise time to 91.5nS (from 38nS). This results in a massive increase in turn-on switching losses from 45uJ to 107uJ, equivalent to an increase from 457W to 1070W, and overall for the 20uS cycle from 17.2W to 27.4W (switching losses from 7.4W to 19.4W v 9.8W static losses.)***

1623167720242.png

***Note these don't specifically match the calculations in my post #79 as those are based on worst case figures and the simulation is using 'average' figures and some data-sheet parameters don't translate completely into the simulation (and vice-versa). Both are valid for assessing viability of design, sizing heatsinks, etc. but at the end of the day only the building and extensive (indeed possibly destructive) testing of several prototypes can fully validate a design.
 

shortbus

Joined Sep 30, 2009
8,706
@Irving, thank you for that! I never thought before of the heat generated when switching on and off as a problem with mosfets, but on second thought and being shown I see where it is! This kind of thing is why this is my favorite electronics forum, people like you willing to take the time to explain things to a old guy that is mostly self learning this stuff. Again thank you for showing an old dog something new to think about.

Hope you don't mind another dumb question. Why the 50KHz frequency for this? I have been under the understanding that for motor control 18 to 20KHz is fast enough, just out side the audible range.
 
Last edited:

Irving

Joined Jan 30, 2016
1,845
hope you don't mind another dumb question. Why the 50KHz frequency for this? I have been under the understanding that for motor control 18 to 20KHz is fast enough, just out side the audible range.
From one old guy to another, not dumb at all. How else do we learn?

I used 50kHz here because that was what @Dragonoid referenced, however it depends on the motor construction. The more poles & slots the motor has the smoother the running and the less cogging torque (the effect of the rotor pulling back or jumping forward to the last/next magnet position), but the higher the number of poles/slots the faster the magnetic field has to rotate. To quote Microchip's AN889 application note (attached) "To vary the speed, these signals should be Pulse Width Modulated (PWM) at a much higher frequency than the motor frequency. As a rule of thumb, the PWM frequency should be at least 10 times that of the maximum frequency of the motor. When the duty cycle of PWM is varied within the sequences, the average voltage supplied to the stator reduces, thus reducing the speed. Another advantage of having PWM is that, if the DC bus voltage is much higher than the motor rated voltage, the motor can be controlled by limiting the percentage of PWM duty cycle corresponding to that of the motor rated voltage."
 

Attachments

Last edited:

shortbus

Joined Sep 30, 2009
8,706
From one old guy to another, not dumb at all. How else do we learn?
You would be surprised at those who think either that they shouldn't be questioned, or that the knowledge they have gathered shouldn't be shared! And asking is how I learn. Just so glad you are willing to share and take the time to explain! THANK YOU.

I understand the cogging but thought it was only a problem when the motor was operating in slow speeds. One of my bucket list things is to make a SRM switched reluctance motor. The one big disadvantages to them is cogging.

Thank you too for the PDF. they show a different type of BLDC in it than what I'm used to, ones that the stator looks more like a hobby style BLDC. The one pictured looks more like an induction motor stator. Will have to take the time to read that PDF.
 
Last edited:

Irving

Joined Jan 30, 2016
1,845
I understand the cogging but thought it was only a problem when the motor was operating in slow speeds. One of my bucket list things is to make a SRM switched reluctance motor. The one big disadvantages to them is cogging.
It is mostly, but many slots & poles resolves it.

I've been designing/simulating a BLDC motor for wheelchair usage for about 5y now. Ideally I want a gearless, brushless, hub motor inside a 6" hub which is kind of tricky. Cogging is a big issue at low to moderate speeds, but so is lack of torque. Only 1 wheelchair manufacturer has ever fitted brushless motors in a mainstream chair and they are lacking, despite being a 48-slot/16 pole motor in an 8" hub, which compromises the amuont of tyre left... (& I'm not including those cheap n nasty 12v 8" hard-rubber-tyred chairs from China).
 

MaxHeadRoom

Joined Jul 18, 2013
23,348
An aside:
I recall when I got into CNC retro fitting, mostly DC brushed motors at that time, we had a salesman for a motor company visit and he extolled the virtue of BLDC versions of their motors, but I recall him saying they needed gearing due to cogging at low RPM's.
After experimenting with them and a PC based CNC motion card (Galil) I found that this was only true when used open loop.
When used with the closed loop Galil, they were as smooth as any high quality DC brushed. :cool:
 

Irving

Joined Jan 30, 2016
1,845
An aside:
I recall when I got into CNC retro fitting, mostly DC brushed motors at that time, we had a salesman for a motor company visit and he extolled the virtue of BLDC versions of their motors, but I recall him saying they needed gearing due to cogging at low RPM's.
After experimenting with them and a PC based CNC motion card (Galil) I found that this was only true when used open loop.
When used with the closed loop Galil, they were as smooth as any high quality DC brushed. :cool:
Quite so, in fact the direct drive BLDC servo has much in common with a bipolar stepper motor in that respect - I've built several CNC machines as well as retro-fitting my mill and lathe.

A friend of mine has replaced the brushed motors on his wheelchair with 90mm 8-pole BLDC units, using the original 32:1 rightangle gearboxes. That effectively kills any cogging or reduces it to near imperception... but those motors were custom built at nearly £3k/pair!
 

Thread Starter

Dragonoid

Joined Nov 27, 2017
53
@Irving
Thank you for taking the time to show the calculations on those MOSFET gate driver selection, They are definitely a refined approach for deciding the gate drivers.
As specified by the calculations required Mosfet gate current peaks 2 Amp so the total setup requires 8 Amp Gate Current.
Now We have IRS2186 Which is a 4 A /4A Driver, Which will definitely require parallelling of the 2 drivers. But I feel that dedicating2 Date drivers might be a risk factor when talking about system stability.
Is it safe to use 2 Mosfet Gate drivers in this System or We Can Use Low Side Gate drivers with an additional +15v on the High Side Gate Driver? Since the Low Side Gate drivers are available cheaply and with higher current conduction capabilities.
 

Thread Starter

Dragonoid

Joined Nov 27, 2017
53
The International Rectifier gate drivers aren't the only ones being made. There are other companies make half bridge drivers that have more amperage than the IR ones do. Like the ones from Fairchild/ On Semiconductor. Here's one with 8Amp output and input -
https://www.mouser.com/pdfDocs/NCV57080A-D.pdf
I found Something here, it's an IXYS Based SIC/IGBT Gate driver, I guess this will work for this application, Since its Cost effective too. And Provides 9A Current Capability.
These are Low side Drivers,So I wanted to know how to charge the gate of high side Mosfet with these Gate Drivers that are Specifically Designed for IGBT (Without Charge Pumps)
 
Last edited:

shortbus

Joined Sep 30, 2009
8,706
Drivers that are Specifically Designed for IGBT (Without Charge Pumps)
There is no difference that I know of between a mosfet driver and a IGBT driver, since they are both used to turn on a capacitive load, the gate.[edit] Should have said today's IGBTs, the early days of IGBT development there was a problem with something called tail currents, that need a reverse or negative spike to turn them off fast. They have since found a way to eliminate that problem in modern IGBTs. The notation, SIC is just a newer type of mosfet. https://toshiba.semicon-storage.com/us/semiconductor/knowledge/faq/mosfet_igbt/igbt-010.html

The part you linked to is only for low side, i can see no way to make it work for a high side switch. What to look for when looking for your project, is called a "half bridge driver". They give a way for both the high and low side switches to work.
 
Last edited:

Irving

Joined Jan 30, 2016
1,845
As @shortbus has said, you can't use a low-side driver for the high-side (well, not without isolated power rails and isolated logic drive).

A proper 1/2 bridge driver is your best bet, not least because it regulates the dead time between upper & lower MOSFETs and minimises the likelihood of shoot-through.
 

shortbus

Joined Sep 30, 2009
8,706
Not really my place to say but then again I will say it. Not only this project but many other peoples projects they come here for help, but when the correct device is told they seem to be worried about a few more dollars to spend. The idea that a sort of solution costing $ and the correct solution costing $$ seems self defeating to me. Kind of like the old Fram oil filter advertisement, pay me now or pay me later. Rant over.
 

Irving

Joined Jan 30, 2016
1,845
Not really my place to say but then again I will say it. Not only this project but many other peoples projects they come here for help, but when the correct device is told they seem to be worried about a few more dollars to spend. The idea that a sort of solution costing $ and the correct solution costing $$ seems self defeating to me. Kind of like the old Fram oil filter advertisement, pay me now or pay me later. Rant over.
Sadly I have to agree with you. Also those individuals who think they can change the laws of physics and many years of industrial knowledge by willpower alone...
 

cmartinez

Joined Jan 17, 2007
7,368
Sadly I have to agree with you. Also those individuals who think they can change the laws of physics and many years of industrial knowledge by willpower alone...
Some of us prefer to learn through our own hard experience, instead of relying 100% on that of others who're genuinely trying to help... learning things that way is more expensive, more time consuming, and most of the time also more frustrating... but trust me, lessons learned that way are never, ever forgotten ... just ask @shortbus about my little h-bridge driver project... ;) :p
 
Top