Ah my bad, its 4 devices for the phase current, so 8 per half-bridge, 4 top, 4 bottom x 3 phases = 24 devices @ 60W per device = 1.5kW switching losses approx (5% isn't bad) so you'll need a heat-sink of better than 0.07K/W @ 1440W if all devices are on one sink. You'd be better off using 3 sinks, one per phase, which only need to be 0.21K/W @ 480W, a much easier call.

Thank you for clarifying.
So I came up with this layout.


8 Mosfet Per Phase so a Total of 24 Mosfet will be used.
Bus bar used in the PCB are 40x9mm for VCC and 40x9mm for GND
40x9mm = 431 Amp + PCB Trace with 35um Thickness will give a current conductivity of 50 amp @35c so Current coming from the Source will be properly managed.

 

I'll refer you back to AN11599.pdf.... read it...

Start with your heatsink => its going to be a large flat surface as a thin rail just won't have the thermal capacity to get the heat away. So your MOSFETs per heatsink are going to be in 2 groups of 4 arranged in a cross formation. Pins bent 90deg into the PCB which is parallel to the heatsink face...

 

[/QUOTE]

I'll refer you back to AN11599.pdf.... read it...

Start with your heatsink => is going to be a large flat surface as a thin rail just won't have the thermal capacity to get the heat away. So your MOSFETs per heatsink are going to be in 2 groups of 4 arranged in a cross formation. Pins bent 90deg into the PCB which is parallel to the heatsink face...

The Copper is basically just the bus bar, Not the heatsink. Is it possible if you could show me a picture of the setup

or similar to this?

 

The second approach is much better...

I'll try and find some examples...

 

For the kind of power that's being discussed here, I'd give serious consideration to adding either large cooling fans, or a liquid cooling system.

 

For the kind of power that's being discussed here, I'd give serious consideration to adding either large cooling fans, or a liquid cooling system.

The heat-sinks I referenced in post #38 from coolinnovations are just that, from 60 x 60 x 10mm with one fan to 250 x 250 x 65mm with 4 x 120mm fans.

They are actually better & more cost effective than generic liquid cooling systems assuming you can get the required overall air flow. For EV use this shouldn't be too difficult unlike battery-cooling systems which often require liquid-cooling because of the form factor.

The smallest that might do the job, in terms of thermal resistance, as per post #40 is their model 3-343407RFA which is 86 x 86 x 18mm + a single 80mm fan @ 84cfm. You'll need 3 - 1 per phase, with all 8 MOSFETs mounted on it. They retail for around CAN$ 100 each excluding shipping, fan and taxes. Suitable 12 or 24v fans are around $15 - $20.


So, for the board layout I'd probably do something like this.... the MOSFETs are bolted to the heatsink and the leads formed with a 1mm radius bend vertically away from the heatsink. The PCB is placed over the pins with spacers to maintain a fixed height off the heatsink of approx 5mm and soldered to the pins. The holes in the PCB are large enough to remove the fixings in the event MOSFET(s) needs replacing. The PCB is 86 x 86mm, has 2 zones on the rear for V+ and GND mirrored with zones on the front for attachment to busbars (stitched together with copious vias - only a few shown as an example) and a final zone on the front for the phase output busbar, with a gap for the gate drive resistors. This isn't final, there's more work to do on the mechanicals, like how the busbars are attached and further thermal assessment as, on reflection, it does feel a little tight. If it were my project I'd probably do some CFD modelling to check this heatsink has enough thermal mass to channel 480W for any reasonable period of time; I might opt to go up to 100x100 (92mm fan) or even 120x120mm (120mm fan), but that's a more expensive option so the modelling is important to avoid spending 100's of $ on the wrong heatsinks! Copper is at least 2oz 70um and preferably 4oz 140um. There is still an issue with skin effect on the busbars to be resolved...

Front_______________________________________________________Back

 

The heat-sinks I referenced in post #38 from coolinnovations are just that, from 60 x 60 x 10mm with one fan to 250 x 250 x 65mm with 4 x 120mm fans.

They are actually better & more cost effective than generic liquid cooling systems assuming you can get the required overall air flow. For EV use this shouldn't be too difficult unlike battery-cooling systems which often require liquid-cooling because of the form factor.

The smallest that might do the job, in terms of thermal resistance, as per post #40 is their model 3-343407RFA which is 86 x 86 x 18mm + a single 80mm fan @ 84cfm. You'll need 3 - 1 per phase, with all 8 MOSFETs mounted on it. They retail for around CAN$ 100 each excluding shipping, fan and taxes. Suitable 12 or 24v fans are around $15 - $20.
View attachment 238736View attachment 238737

So, for the board layout I'd probably do something like this.... the MOSFETs are bolted to the heatsink and the leads formed with a 1mm radius bend vertically away from the heatsink. The PCB is placed over the pins with spacers to maintain a fixed height off the heatsink of approx 5mm and soldered to the pins. The holes in the PCB are large enough to remove the fixings in the event MOSFET(s) needs replacing. The PCB is 86 x 86mm, has 2 zones on the rear for V+ and GND mirrored with zones on the front for attachment to busbars (stitched together with copious vias - only a few shown as an example) and a final zone on the front for the phase output busbar, with a gap for the gate drive resistors. This isn't final, there's more work to do on the mechanicals, like how the busbars are attached and further thermal assessment as, on reflection, it does feel a little tight. If it were my project I'd probably do some CFD modelling to check this heatsink has enough thermal mass to channel 480W for any reasonable period of time; I might opt to go up to 100x100 (92mm fan) or even 120x120mm (120mm fan), but that's a more expensive option so the modelling is important to avoid spending 100's of $ on the wrong heatsinks! Copper is at least 2oz 70um and preferably 4oz 140um. There is still an issue with skin effect on the busbars to be resolved...

Front_______________________________________________________Back
View attachment 238734View attachment 238735

PIN Heatsinks are definitely superior in terms of active cooling solution and reliability, Considering the cost I guess I'll opt for these soon, But in parallel, I will use a radiator and water pump liquid cooling system as well.
Definately a CFD analysis is important before moving on with the thermal design let me set up the environment for that soon.
The Topology that I opted is similar to what you have specified, I would like to thank you for your time in drawing it for me. I have used 90um Copper with 20x15mm Coper to carry a current of about 520 Amp with + 5 Degree C increment.
Have a look at the designs now, Capacitor placement is remaining as well as the controller board that will be stack above the power stage.

 

Interesting!

How are you going to fix the MOSFETs to the heatsink? Or do you bolt them down then solder them to the board? A right pain if you need to replace one, you'll have to unsolder all 24! That why I went for 3 separate heatsinks.

Personally I'd split the busbars and have a separate 60v feed to each phase, the single (dual?) cables from the battery to the busbar will have significant voltage drop - another justification for separate phase modules.

That busbar arrangement means a custom heatsink, and that narrow piece from the mounting face to the bulk of the main heatsink body is going to have a detrimental effect on its performance.

 

Interesting!

How are you going to fix the MOSFETs to the heatsink? Or do you bolt them down then solder them to the board? A right pain if you need to replace one, you'll have to unsolder all 24! That why I went for 3 separate heatsinks.

Personally I'd split the busbars and have a separate 60v feed to each phase, the single (dual?) cables from the battery to the busbar will have significant voltage drop - another justification for separate phase modules.

That busbar arrangement means a custom heatsink, and that narrow piece from the mounting face to the bulk of the main heatsink body is going to have a detrimental effect on its performance.

 

Again just my simple minded thinking on this. How are you going to keep the recommended ~1"/25mm distance from the gate driver to multiple mosfets? I don't, I guess, also understand the resistance to use a single or as it's industry number, SOT-227 mosfet or IGBT. That package eliminates having to use an insulator with the heat sink, has a better thermal profile, is a proven device made for power electronics, ect.

Then it can also use a heavy gauge cable for connection instead of a buss bar. And with only one there is less chance of one of the multiples failing due to power hogging. To my way of thinking an all round better way.

 

Again just my simple minded thinking on this. How are you going to keep the recommended ~1"/25mm distance from the gate driver to multiple mosfets? I don't, I guess, also understand the resistance to use a single or as it's industry number, SOT-227 mosfet or IGBT. That package eliminates having to use an insulator with the heat sink, has a better thermal profile, is a proven device made for power electronics, ect.

Then it can also use a heavy gauge cable for connection instead of a buss bar. And with only one there is less chance of one of the multiples failing due to power hogging. To my way of thinking an all round better way.

TBH you're right, 6 big IGBTs are probably a better bet, apart from the price!

 

 

That's looking good.

How big is that central busbar? What is the expected maximum fundamental frequency of the output voltage to the phase?

 

TBH you're right, 6 big IGBTs are probably a better bet, apart from the price!

I wasn't talking those half bridge IGBT modules you showed earlier but just good industrial strength ISOTOP/TO-227 mosfets. Ones that probably cost less than the pin heat sinks you suggested.

 

I wasn't talking those half bridge IGBT modules you showed earlier but just good industrial strength ISOTOP/TO-227 mosfets. Ones that probably cost less than the pin heat sinks you suggested.

I had a look.. the best I could find (at Mouser) was an IXFN400N15X3 which is limited to 200A - a limit of the ISOTOP/TO-227 package - so you'd still need 4 per phase @ £30/each plus they have 90W loss per device so 360W rather than 480W (8 x 60W), but essentially they'd need the same heatsink... If you can find a better one....

I did find this, which is good for 1200v and 800A and is a complete phase (2 MOSFETs), but at £950 its a tad overkill & expensive!

 

Hi,
I have designed this module with a Gate driver to reduce inductance and reduce parasitic oscillation on gate pins. For Current sensing I am using HAS-300A It Outputs +-4V .

 

That's looking good.

How big is that central busbar? What is the expected maximum fundamental frequency of the output voltage to the phase?

PCB Copper Thickness is 50 Micron and the copper bus bar is 10x15mm thick.
The frequency will be around 200Hz to 800 Hz depending upon the motor design. (Multiple Iteration of Motor might be carried out)

 

OK, at 800Hz the skin depth on the phase bus bar is 2.3mm - this is where the bulk of the current will flow. The effect of that is illustrated in the diagram...

 

OK, at 800Hz the skin depth on the phase bus bar is 2.3mm - this is where the bulk of the current will flow. The effect of that is illustrated in the diagram...

View attachment 239095

Thank You for figuring that out, I have changed the bus bar accordingly and this has to help to reduce the gate driver track length to a great extent too. Have a look.

 

What would be the best method to measure current?