Micro Lebowski Controller - DIY 3kw

Animalector

10 kW
Joined
Mar 17, 2012
Messages
976
Location
Mareeba - Queensland
Hi all,

Update - BUILD PICS START ON PAGE 4

I've been toying with this for a little while in my spare time. I think it's time to put it up to see what the experts around here think of the layout and concept.

The plan is for around 3kW, up to 80V max due to component availability. The boards will sit one on top of the other, with vertical interconnects. Currently I am considering my options for the BAT+ and BAT- interconnects. I am thinking standoff posts, but am concerned about their current carry capacity so may just hardware them.

The CPU board sits on top, the driver board and battery buss, with caps is in the middle with FETs and Current sensors on the bottom.

comments welcome. I started on this a while ago, so may have missed some updates to Lebowski's original thread. I don't want to spend money on boards and parts until I have some opinions on the layout. then maybe I'll buy a few boards, and sell them off to anyone keen so I can afford to buy parts.
UPDATED 2015-02-06 This is the file used to produce the PCB's for production / prototype
View attachment Lebowski_Controller_Schematic_r1.0.pdf

**edit**
This one shows the high current path which is excluded from the mask
View attachment Controller_Lebowski_topPaste_layer.pdf

View attachment Controller_Lebowski_v0.2.pdf
View attachment 4
View attachment Controller_Lebowski_top_layer_v0.2.pdf
View attachment 2

All of the parts I have chosen from Element 14 Australia. The part numbers are interchangable with Farnell / Newark I think for other countries.
View attachment Lebowski_Parts_List.pdf

Andy
Cairns Australia.

_______________________________________________________________
moderator edit: I have received permission from Animalector to post this cross-links:

"Lebowski's motor controller IC, schematic and setup manual" (by Lebowski, 48 pages)
http://endless-sphere.com/forums/viewtopic.php?f=30&t=36602

"Build your very own Lebowski controller" (by Lebowski, 8-pages)
http://endless-sphere.com/forums/viewtopic.php?f=30&t=57877

"The big Lebowski controller" (by Lebowski, 2-pages)
http://endless-sphere.com/forums/viewtopic.php?f=30&t=42655

"Micro Lebowski Controller - DIY 3kw" (by Animalector, 5 pages)
http://endless-sphere.com/forums/viewtopic.php?f=30&t=65297

"ZombieSS's power stage for Lebowski's controller" (by zombiess, 19 pages)
http://endless-sphere.com/forums/viewtopic.php?f=30&t=51342

"My new 18 FET TO-247 layout for Lebowski Controller" (by zombiess, 8 pages)
http://endless-sphere.com/forums/viewtopic.php?f=30&t=58341

"Would you like a very small Lebowski brain board?", (by zombiess, 4 pages)
http://endless-sphere.com/forums/viewtopic.php?f=30&t=60699
 

Attachments

  • Controller_Lebowski_schematic.pdf
    55.1 KB · Views: 446
Thanks. see what people think and maybe get some built. I've done heaps of SMT soldering before and designed a few other low power digital boards, but this is the first high power (noisy) design... I've just tried to keep things relatively simple..

Andy
 
Hi,
as far as I know, Lebowski ships the chip in DIP package only, isn't it?

After my 2 controllers I built I don't consider myself a power circuit expert, but I can tell my thoughts :)
- Pull-down resistors between each mosfet gate and source would be useful to prevent the gates from floating in case of driver circuit failure.
- For the output signals did you take into account the current capability of the PCB wires? Or do you solder copper bars on the PCB?
- I don't see how the mosfet board and the gate driver board can be connected. The connector placement of the sensors do not match?
- About the power caps I shared my opinion in this thread: http://endless-sphere.com/forums/viewtopic.php?f=30&t=61937&start=100
Based on this I would suggest more capacitors, but it depends on several factors, like the phase current, battery impedance, ripple current rating of the capacitors, etc. Too small electrolytic caps can burst due to the high ripple current.
Briefly, the role of the big capacitors is to filter the PWM frequency from the power supply current and to keep the ripple voltage of the power line low, and the role of the film or ceramic capacitors is to supply the mosfets during switching with low inductance. So the ceramic caps should be close to the mosfets on the mosfet board, and a few uF of them is needed for being able to charge the internal capacitances of the mosfets with low voltage drop.
 
It's so awesome to see these next generation DIY controllers sprouting up. :)

A controller is a worthy project and I expect you will learn a lot, and hopefully help many others on ES learn with you as well.

A compact 3kW Lebowski controller would suit the needs of most all practical use ebikes very nicely.

Once you've got a design tested and refined, I bet you would find plenty of interest in folks who would like to buy at least the PCB's from you to build there own.
 
The circuit is (should be) exactly as per Lebowski's Build thread schematic. It's possible to get the SMT IC pre-programmed, but just in case I added the PinKIT programming port if I have to send the boards to Lebowski

Refer to the following picture. The HI/LO signals from the driver board to the FET board, will be twisted pair jumper wire. The phase_feedback is just a single wire, and the current sensor signals will be shielded twin wire, so 6 jumpered links. I can probably make this a bit simpler, by aligning the HI/LO and (maybe) the phase_feedback. but they still should probably be jumper wired rather than a solid structured link... maybe I can use standoffs for these as well? hmm.. now you have me thinking.

The CAPs have specs similar to those listed in The Lebowski build thread, but Surface mount. of course I can just solder the legs of a thru-hole CAP in place of these components. so I was not particularly worried about the CAP selection. what matters is the size of the CAP which will define the vertical separation between the boards. I can have one cap above, and one cap below the Driver board, so I think there will be plenty of space for rapid release energy storage. :lol:

Wiring.JPG

I'll fiddle with the signal interconnects some more, and look into PCB pricing.
 
i'd really like to fit all the pcb and parts into an infineon's 12fet case. if this could be done this controller would be an instant build/buy!!

what's the internal dimensions of the 12FET Case. I think the resulting profile will be higher than a 12FET infineon case. but extrusions are cheap. and people with 3d printers (SamD) could possibly help out with end-caps for any extrusion...

Andy
 
peters said:
as far as I know, Lebowski ships the chip in DIP package only, isn't it?
Lebowski got himself a TQFP socket for programming the 44 pin TQFP package of the dsPIC30F4011. I don't know whether he has the TQFP chips in stock, I sent him my own for programming since shipping from me to him was cheap.
 
Animalector said:
Refer to the following picture. The HI/LO signals from the driver board to the FET board, will be twisted pair jumper wire. The phase_feedback is just a single wire, and the current sensor signals will be shielded twin wire, so 6 jumpered links. I can probably make this a bit simpler, by aligning the HI/LO and (maybe) the phase_feedback. but they still should probably be jumper wired rather than a solid structured link... maybe I can use standoffs for these as well? hmm.. now you have me thinking.
The twisted pair is not fully clear for me on the picture. Normally the gate and source wires should be twisted, that is 6 pair of wires, but on the picture it looks like only the high and low side gates are twisted.
But when the sources are also connected with these wires, then it turns out to be a ground loop problem to resolve, because the GND is connected at several points... it is not easy.
There are also single row wrap connectors to consider, placed close to the GND rail and to the mosfets.

Generally, as I see, with these integrated non-isolated driver chips there are 2 options to reduce the potential noise problems: the first one is the Lebowski-type long twisted pairs which allows the bridge GND to deviate from the board GND at high frequency without generating noise on the driver board, and the second is a strong GND plane all over the boards which minimizes the GND "wandering". From noise point of view the first option seems better to me, but the GND loop is a question.
 
Normally the gate and source wires should be twisted, that is 6 pair of wires, but on the picture it looks like only the high and low side gates are twisted.
Yep you're right. I had noted this earlier, but neglected to follow it through.

Also, I'm going to use a 4-layerboard, with middle +5 and GND planes. On the controller board this makes routing much easier. and gives a large ground plane. I will Add GND fill on the driver board around the GND referenced parts.

I'll update the PCB thanks. that's why it's good for someone else to look at it. no matter how many times I look at it again myself there's always something I miss.

When I get some time I'l model it in solid works so we can get an idea about physical location of things in space

*** edit ***
like this?.
The high-side source is the phase_out
the low side source is GND

I figure I don't need special solder points on the FET board, since I can solder them to the transistor legs.

layout (Custom).JPG

Now that I look at it I might massage it a bit more so I don't have the phase signal plonked in the middle like that..

Andy
 
PCB files updated. with new connections for Hi_Gate + Phase, and Lo_Gate + GND for each phase. plus added ground plan bottom layer of driver board. Also isolated the +5V from the inner layer of the FET board to reduce noise induced into the CPU board via the 5V inner plane.

I think it might be ready for a test? anyone else sanity check? did I miss anything else?

Thanks
Andy
 
Futterama said:
peters said:
as far as I know, Lebowski ships the chip in DIP package only, isn't it?
Lebowski got himself a TQFP socket for programming the 44 pin TQFP package of the dsPIC30F4011. I don't know whether he has the TQFP chips in stock, I sent him my own for programming since shipping from me to him was cheap.

Yep I have a TQFP socket for programming, and have 44pin TQFP chips in stock.
 
Yep the idea is to mount it to the base of an extrusion. using the bottom surface as the heatsink.

Here's the CPU board modelled. These are very rough just to give indication of assembly and general sizing. basically the only components that are true physical dimensions are the SMT CAPS, the FETs and PCB's. the SMT Caps are 16mm high round thru hole caps might be better and offer a lower profile if mounted in their side... one above one below the Driver board.
completeLebowskiModel.jpg

Andy
 
i measured the infineon 12fet case as i think that those are easy to find (and i already got that size installed on my bike *ggg*). the internal dimensions are 145x75x35.
looking at your drawing it seems those 9/10pin headers are not pushed together all the way?! so it seems the overall height might be smaller than shown on the picture?
laying the big caps down is a good idea for sure. this will make the whole thing lower as cap's height is normally bigger than their diameter.
overall the whole thing looks pretty small
btw: i bought a lot of smd tools - can wait to start building 8)
 
Laying the caps down will increase inductance and lower the thermal transfer from the caps to the board, both are due to the exposed leads of the caps. Generally this is not recommended, as far as I have read here on the forum.
 
those 9/10pin headers are not pushed together all the way
the board spacing is defined by the height of the Caps. 10 - 12mm would be sufficient for the 10-pin headers ( only show 9 but there are ten) but for that we'd have to use thru hole caps and lay them on their side and they would need to be < 10mm diameter. find some suitable caps on au.element14.com and give me a part number.

Laying the caps down will increase inductance and lower the thermal transfer
because of the vertical bus bars, the horizontal caps would probably butt right up against the bus bars so leads would be very short... This raises the issue of the vertical bus bars... is this design doomed from the start because of this design decision??

that was the point of this thread.. to get peoples opinions about whether having the layout stacked like this is good or bad...Or for 3kw 6-FET maybe it doesn't matter???
 
Animalector said:
those 9/10pin headers are not pushed together all the way
the board spacing is defined by the height of the Caps. 10 - 12mm would be sufficient for the 10-pin headers ( only show 9 but there are ten) but for that we'd have to use thru hole caps and lay them on their side and they would need to be < 10mm diameter. find some suitable caps on au.element14.com and give me a part number.
ok. so how i thought. in your picture the upper board is not (near) touching the caps. so i assumed that the boards will be closer together. i will see if i can find some caps.
btw: if you want to connect the fets to the case they should be turned around with the metal back plate touching the case instead of the pcb?!
 
It's the perspective view. the CPU board is smaller than the other boards, and in fact they are touching the top of the Phase B and Phase A caps.. so yeah..

if you want to connect the fets to the case they should be turned around with the metal back plate touching the case instead of the pcb
Also, this is just how they are modelled. they are actually designed to have the metal plate facing the heatsink. this is how the pin orientation is done on the PCB. it just looks like they are mounted face up because of how I drew them. I actually modelled all these parts from scratch in Solid works so I didn't have time to add features like the metal pad.

Andy
 
nice :) just wanted to be sure that this has been thought of. i sometimes overlook the most trivial things ;)
and unfortunately i can't contribute anything else but the trivial thoughts. no idea how these circuit really work and how a good layout is drawn. i can solder the parts to the board and program it. that's it ...
 
UPDATED V0.2 - I found a couple of unrouted nets. I plan separate the schematic into three main sheets to populate the three boards so I can do a proper un-routed net / design rule check.

but not tonight.

Andy
 
Animalector said:
UPDATED V0.2 - I found a couple of unrouted nets. I plan separate the schematic into three main sheets to populate the three boards so I can do a proper un-routed net / design rule check.

but not tonight.
Andy
not tonight?! pah! that's no excuse. come over here. it's daytime in europe right know. just follow the sun and you can work 24h a day. you could still sleep at night. ;)
 
How's it coming animal? I know many folks are interested in lower power Lebowski's.
 
Back
Top