Power PCB for Lebowski Controller to be combined with Bobc's processor board

Lebowski

10 MW
Joined
Jun 27, 2011
Messages
3,412
Location
beautiful Zurich, Switzerland
This thread is for a power section that is designed to work with the processor board described in this thread:
https://endless-sphere.com/forums/viewtopic.php?f=30&t=87104

The basic idea is to have an easy to solder PCB with a connector for Bob's board. The power PCB will have a simple low inductance bus bar setup, 6 TO247 FETs, gate drivers (using IRS 2186) and current sensors (ACS 770 type). The goal is to have the board complete, so with no hand wiring necessary.


[pre]
Recommended gate resistors and deadtime for popular FET types:
FET R_off [Ohm] R_on [Ohm] deadtime [nsec]
--------------------------------------------------------
IRFB4468 4.7 15+4.7 600
IXTH180 15 47+15 500[/pre]


This is the database for the 'jan 2018' version:
View attachment Power_PCB_jan_2018.zip

For reference, measurement pictures for different R_on and R_off:
View attachment IRFB4468.zip
View attachment IXTH180.zip
 
The basic setup I have in mind:
DSC02006-800x800.jpg
On the top left the orange square is PCB copper on both sides of the PCB. Black circles / blue squares are xxx uF and yyy nF caps, one leg going through the PCB soldered to the bottom copper and the other leg bend flat soldered on the top copper. This is also where the main high voltage battery will attach. The idea is to leave the busbar copper exposed so it can be fattened up with solder or copper sheet/wire for high currents.
Below the busbar two rows of 3 FETs, sandwiching a thick (about 6mm or 1/4") aluminium piece attached to cooling. The idea is to follow my low inductance philosophy with the FET angled at 90 degrees:
DSC02007-800x800.jpg
Note that I use an aluminium L piece with threaded holes in the top for attaching to controller housing.

From the FETs wiring will go straigth down to the gate drivers (blue squares). Using twisted wires for the gate drive would be best, but I plan to use very close together short parallel wires to minimize inductance

Then finally on the right side are the 3 current sensors and below this (in purple) the connecting to Bob's processor board. Power will go from the busbar, probably via a small RC filter, into Bobs board. This one then generates the 15V for the gate drivers and 5V for the current sensor, so no low voltage supplies are necessary on the power PCB.

I will design the board and build a test version. I am thinking about using 4468's and 150A current sensors, as this works nicely on Izemans controller (built by Animalector). If you guys have different FETs in mind that make sense to me I will order a set of those to determine gate driver resistors (as the gate driver resistors are dependent on the FETs properties - gate charge)

Both Bob's board and the powerboard will have 100V components, though you can of course switch out the FETs for lower voltage / higher current ones. I think with 100V parts it should be usable from 36V upto around 80V (20s fully charged).

I hope to share progress here and to hear you guys suggestions...
 
3d mockup of lebowski proposed board2.png

This is a mockup made by Whereswally606 showing the general idea. Based on the size of the power stage and gate drivers I will need to think what to do about the processor board. I can lay it flat above or below the PCB, or stand it up straight. I'm thinking a few holes drilled in the PCB there will make it possible for a 3D printed plastic piece to act as a processor board holder. This may raise the processor board enough to 'float' above the current sensors...
 
Very cool!... I am looking forward to building one... maybe 2. What kind of volts/amps you think this will handle?

Thanks for the work.... and I guess this isn't so bad being all through hole (I still don't mind soldering some 1206/0805's.... but seeing how there's only a few components that would be different... would likely use the same FETs anyway... I suppose it wouldn't make much of a difference). Though guess we'll see, if/when whereswally makes the SMD version.

BTW, I love the fact that the boards under 100mm X & Y so far.... these will be quite cheap to get from pcbway (or similar)... I htink it's almost insane how cheap a set of 10 <100x100mm boards are... and my last order arrived at my door less than 96 hours after ordering (and 3 x 10 boards, 30 in total, plus DHL shipping was only $35... hard to beat that)

As soon as you got a design somewhat locked down, let us know... I already got a couple of SMD brains coming from you... hopefully can get a couple of bobc's boards ready to test with this.

BTW, those 4468's look pretty damn good for the job... tho if you weren't expecting to go over 100A continious... the 4768's appear to be a pretty solid choice as well (given the 250V is likely unnecessary, but seems to have better RDS among a few other specs... and can even be slightly cheaper, tho pretty close in price to comparison). Though maybe I am missing something... would expect them to be very similar chips... and would think the 4768 could likely handle more amps than it's spec if your running less than half the max Voltage. But I could be wrong.
 
Maybe you could benefit from SMD gate resistors. They are big enough so that soldering them is not a problem, and they are inherently very very low inductance, whereas regular TH resistor are inductive unless you specify a non inductive and more expensive part.
 
I plan to use through hole resistors... I got a gazillion of those from when ST-Ericsson closed down here, makes it easy to experiment and find the correct values for the type of FETs. Another reason is that I do not have too much faith (for high currents) in PCB via's, I always try to switch PCB layer through a component wire. Also, never seen problems from the inductance of normal resistors.

Plus, I gather that if people really want to use SMD they can just lay it flat on the little solder pads that every through hole connection has.
 
Although that one doesn't appear to be available in TO247 package; maybe the legs could be bent out enough to still mount in the holes (I think I've done that before with other FETs).
 
I would have expected teh TO247 package to have better heat dissipation/conductivity, and higher leg-amp ratings. And allow for larger die size for more current handling/heat dissipation.

I'd assume from what you said that it doesn't?
 
One thing I also like about the 247's is that you can use normal screws to attach them to the heatsink as the hole though the package is non-conducting. You only need some Kapton tape or heat conducting pad or something for drain isolation, no plastic screws or fiddly plastic washers.

Took a screenshot from Digikey, just searched for TO247 transistors, NMOS, between 75 and 100V, over 100A, NMOS, sorted by price:
FETs.png

The IXTH180 looks nice for a 150A controller...
 
At least some of the TO247's have completely plastic cases, requiring no isolation pads. I don't know about modern ones, but I have some old 100A parts around here that are built that way. Can't recall the p/n.
 
Pats are in :D so I've started the PCB design. This is how I image the PCB busbar and the FETs (still thinking about where to place the caps, each FET pair will have 2 caps as shown). Notice how the source of the low side FET has its leg clipped off and is soldered to the top of the PCB. This allows the + battery to go underneath it (continuing the busbar idea) to the high side drain.
The caps only have a drill hole for the + terminal to reach the + busbar at the underside of the PCB. The - terminal is bent flat and soldered straight to the - busbar (the top side copper). About 1 cm of room is kept between the caps and the FETs for beefing up the PCB and to be able to remove any FETs in case repairs are necessary.
pcb1-800x800.png
pcb2-800x800.png
 
busy with the gate driver layout... I've decided on a 2.54mm grid and a 2mm track width, giving 0.54 mm spacing. This small spacing should limit the parasitic inductance, as twisted wiring to the gate is not possible on a PCB... The 0.54 mm spacing in the gate lines limits the by the gate wiring enclosed area (and thus the inductance).
Also, because of the high peak current I am changing between front and back wiring only via component pins. This is why R29 is close to the FET and not close to the rest of the gate driver components.
The output wiring is 4.5 mm wiring on the front layer, later it should be possible to add lots of solder or even a thick copper wire here.
pic2.png
pic1.png
 
Very interesting to see how this evolves including the comments on your decisions. Are all the components identical to your original design, except that there are 6 FETs now (2 caps per FET-pair (TO247)) instead of 2 caps for the full bridge (4110, 4115)? It would be very helpful to stay below 80mm for the total width for the installation in a bicycle frame... I hope this is will be possible.
 
Checked the tracking number for china post for the new wson designed bobc brain board. They were in the country as of yesterday. Wont be long now I hope, just hoping no other unforeseen show stoppers
 
emmgee said:
Very interesting to see how this evolves including the comments on your decisions. Are all the components identical to your original design, except that there are 6 FETs now (2 caps per FET-pair (TO247)) instead of 2 caps for the full bridge (4110, 4115)? It would be very helpful to stay below 80mm for the total width for the installation in a bicycle frame... I hope this is will be possible.

The components are not all identical to the original design. Like before, the gate drive resistors depend on the FETs you want to use, and this time it is a 6 FET, so each gate driver only drives 2 FETs (meaning I can use the full 4A shutdown current for only one FET, halving the minimum gate off resistor).
About the caps, I think the main thing here is to make sure the caps don't heat up too much. From my current test controller I know one set of 220uF/470nF caps works for 50A phase current without heating. So for 150 to 200 A phase current 3 sets of caps should do it, though less may also be good enough....
Will keep the 80 mm in mind...
 
some more squeezing...
pic1.png
pic2.png
note how each of the gate drivers has its own vss line to the main bus bar. This to keep the low inductance. The vss lines at the gate drivers are also connected together, it is this line that will be extended and used as the vss for the processor board.
 
pic1-800x800.jpg

note the little to92 transistor like component at the left side of the power stage. This is a temperature sensor, it is placed here so it can easily be clipped or glued to the same (~1/4" or 6 mm) aluminium cooling strip as the FETs...

pic2-800x800.jpg

What do you guys think would be the best orientation for the processor board ? And would you place it on the component side of the power board or on the other side of the power board (where it can possibly be shielded by a big slab of ground plane) ?
 
Lebowski said:
pic1-800x800.jpg

note the little to92 transistor like component at the left side of the power stage. This is a temperature sensor, it is placed here so it can easily be clipped or glued to the same (~1/4" or 6 mm) aluminium cooling strip as the FETs...

pic2-800x800.jpg

What do you guys think would be the best orientation for the processor board ? And would you place it on the component side of the power board or on the other side of the power board (where it can possibly be shielded by a big slab of ground plane) ?

My 10 cents on the layout: I am sure at this power level this will work well but I would put the gate drivers where you have the snubber/DC link caps, move the caps to the other side of the fets and move the current sensors to where you currently have the gate drives. This would free up the area to the right for the brain board and keep it well away from the power pass components. I don't like the idea of having the brain anywhere near the fets or current sensors. If the layout is to remain the same, under the ground plane seems like the best place for the brain to minimize interference.
 
I would make the phase track leave the middle of the top/bottom track instead of from one of the FETs.
 
Nice idea with the temp sensor! Maybe three additional holes for another sensor glued directly on a FET could be nice. Is the brainboard placement over or under the drivers critical? Maybe 'fat' connections on one side, brain connections form the other?
Up to now <80mm width seems possible. What do others think about the dimensions?Layout_TO247_Lebowski_800.jpg
 
I hear your comments... Swapping the place of the gate drivers and current sensors:

pic1.png
pic3.png
pic2-800x800.png

I like the straight shot of the output to the current sensors. If I only use the top layer for the high current path, then I can route the gate signals underneath the current sensors.

What I don't really like is the long gate driver wires. An online inductance calculator tells me that the inductance fo two flat 2mm wires spaced at 2.54mm is 16nH per cm (so 160nH for 10 cm). The Ciss of the Fets is 7 nF (for ITXH180) or 20nF (for 4468's). The turn-off gate resistor is between 4.7 and 10 Ohm.

Damping is 0.5 * R * sqrt(C/L).

So the 4468 with 4.7 Ohm gives a damping of 0.83
The ITXH180 with 10 Ohm gives a damping of 1.05

Both numbers do not impress me, I would like to have had 2 or higher... the cause is of course the high wiring L.

P.S. no clue why some pictures show up bigger than others... even when I make all 800 pixels wide they still show up differently ?
 
Gate drivers on separate PCB, on top (or underneath, depending on the POV) of FET PCB?
 
Back
Top