Help in making my BLDC controller

[Ludo91]

great, tnks

there is now a discussion between engineers about how to make the liquid cooling: Is it better to have the small heatsinks mounted on the right face of the mosfet and then have the liquid flowing on the heatsinks or having the whole mosfet submerged in the cooling liquid?

someone says that if we have the whole mosfet submerged in the liquid the flow must be pretty fast to take away the hot cooling fluid from the mosfet`s metal surface so the thermal conductivity of the luquid doesn`t even come into the game.

someone else says that using a heatsink will improve the contatct surface of the fluid and the mosfet but thinks that this is an advantage only with a not-so-high flow rate.

What is your opinion? Since I have a coolant flow of 8 liters/minute I would considet just sinking the whole mosfet body in a turbolent portion of the flow.

cheers!

 

[rhitee05]

Both. You need to have something attached to the MOSFET which will (greatly) increase the surface area exposed to the cooling liquid. The case of the MOSFET is totally worthless as a heat transfer surface. You need to have something attached to the back of the MOSFET where it will do some good. Turbulence and surface area are your friends here, so something like a pin-fin sink for BGA chips would be perfect. Covering the entire MOSFET in coolant will not improve heat transfer from the junction, as the case is a very effective insulator. It will however pull heat from the legs, which will increase the package limit to some degree (some, not a lot). Walk before you run, though. You have many more obstacles before the MOSFET legs will be your limiting factor, and immersing the whole thing raises additional issues.

 

[Ludo91]

I see. The point of the discussion was around this statement:
"if we have the back plate of the mosfet in contact with fast moving coolant we get the best dissipation possibile (possible from the backplate, not considering legs and so on) beacuse the plate is continuosly in contact with something cool, and adding a heatsink would improve the area of heat exchange with the fluid but since the heatsink has a "thermal resistance" itself we would end up having something hotter against the back plate than the fast and turbolent flowing coolant would be"

I actually kinda agree with this opinion but I find it to be a little unintuitive... what do you think?

 

[rhitee05]

False. It is no different than a conventional heat sink in air. Water has different properties than air, but it's still a fluid and the same fluid and thermo-dynamics apply.

Heat transfer occurs in the thin boundary layer of water (air) right at the surface. The transfer mechanism is convection, not conduction (water is actually a pretty poor heat conductor). The more surface area, the more boundary layer which heat can transfer into. Turbulence is good because it means the water (air) is being mixed up so the heat will be transferred effectively from the boundary layer into the full volume of water (air). Laminar flow is bad. Higher flow velocity causes turbulence (a simplification; look up "Reynolds Number"), so does discontinuity in the flow path. Something like this: http://www.aavidthermalloy.com/products/bga/371924b00032g would be ideal. The pin-fin design gives a lot of surface area and causes substantial turbulence. Turbulence means there will be more pressure drop and more pumping effort will be needed, but that's the price you have to pay. The same properties that make a good heat sink in air apply in water, with some caveats due to different density, viscosity, etc.

Just to give some ballpark idea, at 120 A using those 3077 FETs, the high- and low-side FETs will each be dissipating about 60W in resistive losses alone. Add in switching losses and you're probably looking at close to 200W of heat that needs to be removed in total (crude estimate).

 

[Ludo91]

I see. I will probably understand that better next year, when I`ll be studying "technical physics" and "fluid based machines" (both damn hard to pass but very intresting).
This means Im gonna make it 6 fets and give each fet its own dissipator. I`ll post updates here!

 

[Ludo91]

Hey! I cleaned the board, unsoldered the fets (without killing them!) and getting ready to reassemble it, ready to be tested with the current sensor working and also internal hall sensors

Here is the cooling system I made! mosfet, copper and a heatsink perfectly coupled and, when i`ll assemble the thing, a bit of good cooling paste in the middle.

http://imageshack.us/photo/my-images/594/mycooling.jpg/

Question: is it enought to pump a LOT of air in the plastic pipe to successfully cool the thing or I must make a better sealing with the heatsink and use oil? (no water because the heatsink is live)
I got the answer: Yes air is enough, I was running it at 60 amps and with a bit of air pumped thru it was almost ambient temperature.

Still, it is kinda "chunky" when spinning, do you think it is a timing issue??

 

[Ludo91]

it still spins super chunky.... do u have any idea of the reason??

 

[incememed]

Try and scope one of the motor phase leads. The curve form should be perfectly symmetric, at least during some point of the rpm interval. The rest of the interval it should be only slightly off.

If the curve form is assymetric with the heavy part to the left, the commutation is retarded. Move the halls slightly opposite the direction of rotation.
If the heavy part is to the right, it's advanced, and the halls should be moved in the other direction.

If the hall fixture is continuously moveable at least some 8-10 degrees, you should be able to find a sweetspot by sweeping through different positions (for all possible connections). If not, you have a different problem, such as a distorted magnetic field. Then try and move the halls to different positions along the length of the motor can, somewhere in the middle of the can the field should be ok to use, as tested by many on this forum.

(Can't tell you anything about internally mounted halls, have no experience of that).

 

[Ludo91]

It now spins fast and smooth!!!!!! Still, I have a problem: mosfet drivers are overheating!

 

[Arlo1]

It now spins fast and smooth!!!!!! Still, I have a problem: mosfet drivers are overheating!

Are they over clocked? What are they running for gate resistors and supply voltage?

 

[Ludo91]

The thing was running GREAT (smooth and with a 1A no load current at 14V, 40% trotthle), with fets going no higher than ambient temperature.
14V source was current limited at 5A.
Drivers are powered by and external source at 12 V.

Resistance between drivers and fets was 9.6 Ohm. the thing was being tested at 3.9 and 31.2 kHz.... well within the allowed range of both mosfets and mosfet drivers.

Now ( I am writing with the laptop on the workbench, or I`d better call it workbenchtop sinche on my laps I have the scope and the tester) I am checking what`s wrong and I see that also the mosfets are gone I am at least gutted, I am going to investigate the thing right now and report everything, any suggestion is Very welcome!

UPDATE: The Carnage was caused by my limited current power supply that, after each current limitation, when switching back to voltage mode, gave huge voltage spikes (up to 96 volts!). Well, it`s bad to destray things but at least I know why that happened.

 

[Ludo91]

THANKYOU AND SEE YOU SOON

University is about to start and, since I`m still (and I hope I`ll be there for long) in the part of the life where when you get truly interested by something you give almost all your time to that I had to put my controller in a box and put it in my room ready for next summer. I did that because when I started this amazing electronic journey I was still studying for final exams and I got very distracted by that, second college year is gonna be hell so I really can`t be distracted by something else, even if I love that
. I find the subject of BLDC & power switching extremely interesting and I thankyou very very very much for teaching and showing me the way to learn a LOT of things and build my controller. I cant believe I learn this much from a forum, you have are extremely generous and cool.
I started with almost zero knowledge and thanks to your advices I got to the point of building a (finally! ) my own working BLDC controller, understanding (well, beginning to understand) and learning a lot about this "form of art".
Today when it started spinning absolutely smooth and fast I was amazed as a few times before, looking at that lot of components actually working together like an orchestra.
In next days I`ll publish everithing in the google project page of the controller and then (sadly) forget about it till next summer, when I`ll finally put it on my ebike and have another summer of electric rolling (this summer I was running on a hobbyking esc).

Thankyou very much again, I hope this thread will be helpful to someone else in the meantime, See you soon!

Ludo