Would you like a very small Lebowski brain board?

liveforphysics said:
Gate drive and trace layout is maybe 90% of what makes a controller.

Hehe, I just updated the 18 FET power stage thread. During desat testing I saw just over 1000A at 64v with a 200uS pulse into a dead short and 750A into a 5uH coil. 750-800 is where I want my desat to trigger so I'll need to tweak my zener diodes a bit.

A kiloamp at 64v is impressive, but I was just playing around while testing. I knew I was within pulse spec of the parallel MOSFETs and I know they current share well due to how I matched them, but it's not something I recommend doing. I had a plastic shield over the power stage... just in case.

Sending a 5% DC 19khz pulse at 64v into the 5uH coil made from 10 guage produced a much stronger magnetic field than I expected (I was happy, magnets are awesome!). Hope I didn't wipe a credit card mag stripe.
 
usertogo said:
Sorry guys I hadn't noticed your replies!
The idea is to keep the current down, and I found affordable stock BMS up to 24s LFP, and want to build 2 banks at 77 Volt each with the option to put them parallel or in series. Kind of like electric gearing, for high speed and for mountain climbing.

This is not an efficient thing to do and adds unnecessary complexity. It also has nothing to do with "gearing". 144v puts in a sort of no mans land. The lower 77 volt option seems like a wiser decision, especially if this is your first design.

usertogo said:
The FET I found with low Rds and not too expensive is the IRFP4668 (irfp4668pbf). At peak operating current (48s) I would have a .3V drop and 7.5W loss at each individual FET (if its a simple 6FET bridge) I could if the .35% loss in FET is too much start paralleling transistors, which I may want to do

You can not look at MOSFET losses in this way. Losses = switching losses + conduction losses + body diode losses.

I just plugged through some estimated numbers based on that MOSFET and at 5% DC @ 18khz, 250nS switch time and a 1.3v body diode and 77 volt bus, you would have about 30W in switching losses, 3w from the conduction losses (RDSon) and 110Wfrom the diode losses.

This is about 143 watts of losses. As you approach 95% duty cycle the diode and conduction losses start to swap places and the conduction losses become higher and diode losses lower. You still end up with 100w of losses.

Do not forget that as temperature rises, so does the RDSon resistance. I tried to account for that in the above ballpark numbers.
 
Most dc friendly AC stuff runs fine on ~>80-90vdc.
 
@ zombiess
Let me know when you run your next batch, I will pay for an extra four (plus 4 of your 18FET power inverter).
Cheers
 
squeegee said:
@ zombiess
Let me know when you run your next batch, I will pay for an extra four (plus 4 of your 18FET power inverter).
Cheers

I have 2 more sets of boards if you want a set. I have not fully tested my current setup beyond the PWM test on the chip which means I don't know if everything works OK yet. I also need to edit the code that does the error handling in the 18F1320 PIC which I'm currently bypassing with the fault bypass juper. I'm planning to work on the controller this weekend so I hope to have more info.

Lebowski has made some changes to the controller and added hall sensors back in, so this setup isn't the most current as it's sensorless only. I want to redo it with the SMD chip he's now programming which will should make this setup even smaller. I'd really like to get down to a single board setup.

I can send you the files and schematics in KiCAD format if you want to edit or have the boards made yourself.

The most difficult part of building a controller is designing a robust gate driver and a good power stage. I've been going very slow on this since I picked it up again because I'm trying to figure out how good/bad my 18 FET TO-247 layout is, but I'm finding it hard to get good measurements due to EMF.
 
Can I get a set or two Zombiess? Looks like I'll be using the Lebowski on my next motorcycle.
 
I starting to work on my motor wind and still have ways to go... I can work on it in paralelle or wait for you to finish.
What would be most helpfull?
Note that for KiCad I'd need to learn it as have not done EE R&D work in ages and to my knowlegde KiCad did not exit then (can't remember what we used then :oops: ).
 
Futterama said:
zombiess said:
I think that's a good idea and eliminates the power supply issues since everything I am designing is using isolated supplies.
I don't understand the meaning of that sentence. What power supply issues are there when using isolated supplies?

Liniear Technology has a 150V input synchronous step-down regulator, but it's only rated 100mA:
http://www.linear.com/product/LTC3639

They also have a 400V input device, but it requires external switches and a whole bunch of other supporting components:
http://www.linear.com/product/LT3752


OK tried my hand at KiCad this week.
Found out there is a LTC7138 that has 400mA capacity, so made a small power board.
The unit fits, with screw type connector onto a 0.6"x1" board.
The 5V is the simple version, a couple more resistors for other output, 3 more for settable over and low voltage shut down, nice little chip.
Still got a long way to a "pro" board, but not a bad start.
 

Attachments

  • 5V_LTC7138.GIF
    5V_LTC7138.GIF
    66.6 KB · Views: 2,491
You need to have a bigger GND plane in connection with the exposed GND pad under the chip, it serves both as high current GND connection but also as a low thermal resistance connection. The few vias you have under it is not enough.
 
There are a few other things that should be different.
But I'll stop hyjacking this thread and if I get a multi-ouput unviversal version (jumper & resitor configurable) done, I'll post a new thread.
Cheers
 
zombiess said:

The above pic is 2 boards that will be separated and stacked after they are made. Made them all on one board for fabrication, it's easier to solder.

I would like to see more people trying out Lebowskis controller and helping him develop more features. The hard part about building a controller from scratch is you have to design EVERYTHING and it's a pretty daunting task. I believe with some simple kits we can get more people up and running on these controllers.

I have been working on a pretty small version of a Lebowski brain board for his latest chip version. The completed setup is 2 boards that stack with headers and is 3" long by 1.5" wide and I'm guessing about 0.75" thick. I incorporated my own error handling / detection system using a PIC 18F1320 if you so desire, it can also be bypassed by a jumper. The construction is a combination of SMD and through hole, but I made as many components SMD as possible. The board accepts the standard 40 pin DIP package he uses to allow for easy swapping of chips as he upgrades his code.

I have already designed and am using a larger version of this board, but this new one is about 1/5th the size of it. There are no current sensors on board for maximum flexibility and size.

If there is enough interest, I can have the PCBs pre fabricated with all the components and sockets for the chips which are flashed, but I don't know exactly how many people we would need to make this an affordable option since it would a low volume run. I know many are concerned about working with SMD components because I myself was until I tried about 6 months ago and not it's my preferred method of construction because I find it's easier than through hole in many cases, the opposite of what I expected!

Post up if you have interest. If we don't have much interest I can supply all the parts as a kit.

I'm also thinking about designing a small power stage for a complete controller setup that fits into a pretty small box. It would probably be 12 TO-247 100/150V MOSFETs and be good for 200A phase which is enough to make your bike pretty darn fun. You also get really cool features like variable regen!

Hi Zombie, I sent you a pm Monday maybe, if you still plan to make or have some of these I am interested by 2 ;) finally pulled the trigger on the bigger brother 36 Fet and I can't wait for it. Even more interested in this tiny board!

Seriously this forum has cool guys crafting cool stuff around, my wife will smash me :lol:
 
I am hesitant to give out this version because it is not fully compatible with Leboskis later version which added hall sensors back in. I also want to redo this so it works with the smd chip which will allow a smaller board. I'm not happy with my choice of connectors either as they do not lock into place. Once I have everything tested I use hot glue to secure them.

The good news is I should be testing soon. My power stage is working and I'm finishing up tuning the switch time on the bench now.
 
zombiess said:
I am hesitant to give out this version because it is not fully compatible with Leboskis later version which added hall sensors back in. I also want to redo this so it works with the smd chip which will allow a smaller board. I'm not happy with my choice of connectors either as they do not lock into place. Once I have everything tested I use hot glue to secure them.

The good news is I should be testing soon. My power stage is working and I'm finishing up tuning the switch time on the bench now.

I'm working on a next version which will be able to support up to 8 digital temperature sensors (DS18S20), to dial back the power when the object the sensors
are attached to get too hot. I opted for 8 as then in a big power stage you can attach 1 to each of the 6 power switches, and still have 1 or 2 in the motor.

Anyways, maybe it'll be a good idea to include a 3 pin connector for the temp sensors (only need 1 3-pin connector for all 8 sensors as they are all
connected in parallel):
WP_20150111_12_51_09_Pro.jpg
One pin is the ground, one is the 5V supply and one is the dataline. The dataline has a 4.7 kOhm pull-up resistor to the 5V, and is connected to the pin
called 'OC2/RD1' of the 30F4011. On the 40 pins DIL this is pin 18, for the 44 pin TQFP this is pin 37.

:D At the moment I have a glass of water in my freezer with inside frozen solid one of the temp sensors, later when I get home I'll see what reading this gives me :D
 
That's excellent thinking on the 8 sensors my friend. Thank you for the continued outstanding controller work guys!
 
Hehe, going to be fun to run serial data in a high current inverter.
Isolated supplies, twisted pairs, shielding and reasonably slow data rate are probably going to be required.

I believe that running serial lines near the silicon devices could cause issues for the MCU on some setups.

I hope the data rate is adjustable... hello EMI noise and EMF coupling :lol: I'm not complaining, it's still better to have the option of temp sensors as they have a multitude of uses. I'm just thinking of monitoring the MOSFET case temp in the above noise scenario.

By any chance do the serial pins you are using have A/D functionality as well? Might be handy to have an option to use them as settable with dual function so analog sensors may be used with a formula, I like K type probes (most likely buffered through an op amp) since they can be soldered to stuff, MOSFET leads, tabs, etc.
 
I wanted to use digital sensors as then there is no ambiguity... with PTC's and NTC's there are so many types all with different temperature curves etc etc.

The data rate for the 1-wire sensors is really low, about 10 kilobaud only (which helps with noise immunity). Plus it has a CRC check to see if the data was
received correctly. I myself am more worried about how the actual sensing part of the DS18S10 is going to be affected, especially when the sensor
is inside a motor. I know Arlo used this type of sensor once to measure the temperature inside an output stage...

The temp sensor data will be requestable over the RS232 from the controller IC, so it'll be quite easy to use a separate u-controller to poll this and show
it on a display.
 
I too am concerned about noise in my application. At this point I'm envisioning a motorcycle with a hub motor capable of 15 or 20 kW continuous and about 100V of battery. My experience at A123 as well as in automotive audio and data control in power applications generally tells me that I would favor using thermistors and analog signals to monitor power devices. Slow 1-wire data buses can be OK if the slow data is acceptable and you route wires carefully, but if you need any real speed a differential bus is, IMO, the only way to go. This is where CAN comes from. Most automotive audio systems use balanced or partly balanced audio signal chains for the same reason--and this is in the analog stages. Erratic behavior from a motor controller is a serious annoyance, or worse--but of course you know that.

I was showing your board layout at the beginning of the thread to a colleague who is a really good PCB designer. He was impressed with your work (which is saying something) but did raise one concern I haven't noticed being addressed here. In the lower left corner of the lower board, the drive outputs (that's what they appear to be, anyway) have two traces (shown in red) that run down the middle of the pads for other drive channels. He worried this could be a clearance/shorting risk, and suggested moving those traces to the opposite side of the board to avoid that possibility. Just wanted to put that out there for you. Looking forward to the final version!
 
wb9k said:
I was showing your board layout at the beginning of the thread to a colleague who is a really good PCB designer. He was impressed with your work (which is saying something) but did raise one concern I haven't noticed being addressed here. In the lower left corner of the lower board, the drive outputs (that's what they appear to be, anyway) have two traces (shown in red) that run down the middle of the pads for other drive channels. He worried this could be a clearance/shorting risk, and suggested moving those traces to the opposite side of the board to avoid that possibility. Just wanted to put that out there for you. Looking forward to the final version!

I'll take the compliment. I think this is my 9th or 10th PCB layout now, but it's only the 3rd and my first double decker I've done since I started learning power electronics, layout is everything in gate driver and power stage design.

Those traces in the bottom left are only for the status LEDs which is why I felt OK with running them down the middle (something I dislike doing. The clearance is set to 0.010" and the OSHPark in clearance is 0.006". I would normally place ground plane under those LEDs to help minimize the chance of noise creeping in, but directly above them are connections directly to the motor and I was concerned about proximity. If you could ask your colleague what the best practice is in this case I'd appreciate it. Is it better to install a ground plane really close to a noise source and reduce loop area or better to omit it? I couldn't decide and decided to route the traces down the middle to keep them further away from the motor phase inputs.
 
zombiess said:
wb9k said:
I was showing your board layout at the beginning of the thread to a colleague who is a really good PCB designer. He was impressed with your work (which is saying something) but did raise one concern I haven't noticed being addressed here. In the lower left corner of the lower board, the drive outputs (that's what they appear to be, anyway) have two traces (shown in red) that run down the middle of the pads for other drive channels. He worried this could be a clearance/shorting risk, and suggested moving those traces to the opposite side of the board to avoid that possibility. Just wanted to put that out there for you. Looking forward to the final version!

I'll take the compliment. I think this is my 9th or 10th PCB layout now, but it's only the 3rd and my first double decker I've done since I started learning power electronics, layout is everything in gate driver and power stage design.

Those traces in the bottom left are only for the status LEDs which is why I felt OK with running them down the middle (something I dislike doing. The clearance is set to 0.010" and the OSHPark in clearance is 0.006". I would normally place ground plane under those LEDs to help minimize the chance of noise creeping in, but directly above them are connections directly to the motor and I was concerned about proximity. If you could ask your colleague what the best practice is in this case I'd appreciate it. Is it better to install a ground plane really close to a noise source and reduce loop area or better to omit it? I couldn't decide and decided to route the traces down the middle to keep them further away from the motor phase inputs.

Ground plane is good, but only if there is a straight shot back to gnd source. Get any series impedance to the ground plane area, and it becomes an antenna that sets loops singing. Ignoring this option, these possibilities were suggested:

Assuming the LEDs you're laid out for here are 1206 size, you could narrow the width of the two traces that go under the LEDs and put them closer together, further away from the LED pads than they are now. You could also downsize the LED's to 0805 size, rearrange their placement on the pcb, route one trace above the LED's (where the other two already are) and have only one trace right down the middle of the LED pads. It may also be possible to go to 0603 size LEDs, rearrange the order, and get all four traces up above the LED's.

Not sure how much help that is....My friend says if you're willing to share the Gerbers he could help optimize things if you like...hard to offer precise advice without those files and time to mull them over.
 
zombiess said:

I incorporated my own error handling / detection system using a PIC 18F1320 if you so desire, it can also be bypassed by a jumper. The construction is a combination of SMD and through hole, but I made as many components SMD as possible. The board accepts the standard 40 pin DIP package he uses to allow for easy swapping of chips as he upgrades his code.

Trying to put together this one. Is the error handling/detection system chip on the postition U2?
 
Back
Top