"De"-nginer

I hope to make this a running post for a one-off 2005 Prius PHEV conversion project utilizing an existing 4kwh Enginer. The target is a 10 kwh 35-Mile / 55+ mph pure EV; poor man's Volt!

The Project

For latest source and technical information see the PHEVCU git hub project

Battery
I'l be starting with the 4kw Enginer's, I need to add an additional 6kwh (10kw target) and end up at 240-260v pack. To do that I'll split the two Enginer packs into two again. That yields 100v/40Ah. I'll add an additional 48 cells of 40Ah LiFePO4, that gets me 250vnom / 40Ah, 9984Wh (lets call it 10kwh). I'll BMS the whole pack. ~$4k all said.

From my EV experience, I will be building a proper insulated, ventilated and heated battery box. This makes a big difference in real world.


Design:

PHEVCU - Plug-in Hybrid Electric Vehicle Control Unit

- Build a battery ECU (Dionmite w/dual Can-Bus') that can send the information the rest of the car (mostly HV ECU) wants.​

- Interface (CAN) with the El​

con 2500 Watt charger to set charge rates

0x346 / 0x34C: See other threads, but something will need to be done to spoof EV mode

I'm confident I know how to get all that done, but I'm concerned about fail safes, over-speed, over/under-temp, etc?​

BTW; I'm a big fan of simplicity, I don't need/want any kind of display. This is my wife's car, if she has to touch a button or it goes into error-state, this is a total failure. Ideally, I should program the ECUs to fail gracefully into whichever mode makes most sense without feedback. ​

 

<- it all makes more sense from this angle

yes, but I'm ignoring them. Worst case, I need to replace the entire pack.

I'm using miniBMS in my BMW EV. Calling it a BMS is a stretch, it's really a balancing and fail-safe system. That would be my preferred off the shelf solution. But... I have another idea, BMS cost is a direct function of voltage (ie, more volts, more cost, very linear). A BMS for 256v is $1250 just for the cell units. But here's the rub, LiFePO4 don't go wildly out of ballance, it lasts much longer with a gentle nudge in the right direction over time.

There are EV folk out there going entirely un-regulated or managed, but they manually check the battery periodically and balance by hand. I could do this as well, but I'd like to consider a scale-able solution. I have a profressional EE friend with a Twike (they run 380v, 16ah), he needs to replace his batteries. The trick is, the BMS costs more than the batteries! I've pitched him my MUX'g BMS

The X/Y gates are 1Amp dual mosfets configured like a SSR ($0.18/unit).

Yup, this is one of those things the amateur overlooks (I did) and it's really critical to year round performance. The battery MUST stay between 50-120F (10-48C). Too cold, they won't give any juice, running a little too hot lowers the cycle life and a way too hot (>150F); the cells swell and that's the end of your pack.

To keep the batteries happy, you need a closed insulated container. Above the insulation, but below the batteries, I put a thin (0.025) aluminum sheet (6061-T6) to which I zip-tied-and-glued an 42 watt thermostat driven soil heater. I hooked that to the charger line so if it gets too cold, it turns itself on while charging all night.

On the cooling side, I use a fan connected by 95F (35C) thermal switch to the car's 12v system. I perforated whole in the bottom of my battery box for draining and ventilation on the bottom (goes through the aluminum sheet). This draws cool air through the entire box.

Agree, fortunately, the duinmite mega has a VGA output, I will get a small 12v display and dump logs/data while I test drive. Once I'm done, I'll pull it.

I think I will expand EV mode slowly, I'll start with the stock range, then add 5mph. Many others have done the stealth EV thing, so I should find out their results. This is realy two initiatives in one project... First, make a Prius run with a huge EV battery; second, how can I run it faster on EV only?

Great feedback and keep me thinking!

 

It's just docs.google.com... I'm a lead software engineer / architect, so I do diagrams a lot.

 

A WORD OF CAUTION.... The #1 cause of EV fires is BMS failure at 4AM when nobody is paying attention and a small short and plastic fire turns into a house burning down. So don't do my wild idea out of hand, take basic fire prevention precautions (steal case/fire resistant materials), test it a lot and out of limits. That's my plan.

 

It doesn't have the first CAN bus exposed as well. But otherwise, yeah, it's better.

 

I could kick myself. I'm getting way too fancy again with the BMS. From my experience, once balanced, these things don't need hand holding, just a gentle top balance. I've been reading some old sites on balancing PBA setups with nothing more than a zener shunt. A 3.6v/5W shunt is good for 1320mA. $0.39 each! That's plenty. Put a 3.6v LED inline for feedback, since it's so cheap, why not a second set at 3.8V with a red LED.

 

Oh yeah, easily. The documentation says the first transceiver is on the board, I just can see where they pin'd it. The second I was just going to duplicate the circuit they provided (see the schematics). The pins to the C2TX/C2RX are still there, I think they're hijack'd for a serial port or something though. Worst case, you need a dremel. The basic language supports the first can for sure, the second has an I/O, but I'm not as confident.

 

In all honest, I'm feeling lackluster on the tools for the dionomite, it reminds me of my Apple IIGS, very cool, now what do I do with it? I do think it will do what we want, but the tools are not what I'm looking for. I've written plenty of BASIC (I grew up in the town of it's invention; Hanover, NH). It's an awful language for low level.

I will make a stab at writing on-chip, but I don't have high hopes it will do everything I need and I know it wont get written the way I'd prefer. I am very concerned that all the hardware functionality I'd like to use isn't available (interrupts, gate settings, etc) and I'll end up writing something that kinda works, but doesn't really.

I do think that if I invest $1000 in the compilers and IDE, I'd be sitting pretty. But out a grand on something I won't use much.

There are few platforms that have the power and I/O to handle bridging. The one's I like most (Atmel or Cygnal) just don't have it. They can do it, but not on a single chip and it would be messy; if I can't get the dionomite to work, I may try again with an two Arduions w/CAN drivers. it will be much more complicated hardware, but it can be done.

-Bruce

 

I'd love to be wrong here; but PIC is a very different instruction set, so, I think not. The other "mite" boards do have that capability, but they are atmel(?) based.

I should have stated the Arduino, but I said Atmel, which isn't fair. The Atmel line is very robust and a good choice for this solution as well.

I revere your work on the message decoding. You are saving me a ton of guess work. From what I glean, I'll still intercept 038.B0.b3 in EV mode, pass everything else both ways.

I've refined my "BMS" system as such... It's combination balancer and v-split detector

The zener will have to be tightly matched (tested and returned). It would be much easier to match if I could use pairs, but I can't find power zeners at 1.8v. 3.3 is the smallest in this series. But the cost for this system would be under $100.

For detection, I'm looking at Lee Hart's "Batt Bridge".

Watch this YouTube on the batt bridge

Any thoughts or critique?