July 2015

Battery-2

In April of 2015 the Audi battery implementation process took sharp turn. I've got an offer to acquire brand new surplus battery packs manufactured here in the USA by EnerDel for Think City EV. The offer was too good to pass and since my Kokam packs were not ready (I had to give away part of my Kokam cells to a customer who managed to ruin cells in the pack I built for him, but this is whole another story), I evaluated suitability of these Think battery packs for the Audi. I already have so many lithium batteries - 100216216 40Ah Kokam cells, a ton of 15 Ah Headway cylindrical cells, complete A123 batteries from the Nissan Qashqai I recently took apart, and now - this offer for EnderDel batteries. One attractive reason to consider Think EV packs was integrated Battery Management System electronics and mechanically sound design. Well, as far as the BMS - half of it: all the slave boards are there, but there was no master controller offered, more on this - on the BMS page). That, however, was quite all right with me because I would not be able to take advantage of master BMS anyway - it relies on the communication and information exchange with the rest of the Think's electronics (namely its VCU), which obviously is not going to be part of Audi even if I were to possess it. So I would only need to build my own master BMS controller communicating with all the slave boards in both packs and come up with suitable BMS software for it. BMS systems are my specialty, so no problem there.

One Think's pack is nearly 400 VDC fully charged, so two in series would do just fine in the Audi voltage wise, but I had to consider max discharge rate and max power they can deliver since I use two inverters. Quick calculation revealed that the power wise EnerDel battery will be weaker than the one built with Kokam LiPo cells, and will become my power limitation if I choose inverters >100 kW each. However, unless I want to run the car on the race track, available power is still very adequate. Since two packs are connected in series, power delivery doubles as P=U*I and my voltage now is twice as much with the same discharge current. Of course I could connect both Think packs in parallel and retain same pack power, but as became evident from the simulation page, another advantage of high voltage system is very high motor RPM before its back EMF becomes large enough to approach battery voltage thus causing loss of motor torque. In other words, my constant torque (acceleration) region will now extend to much higher RPM than would be for just 400V battery. At 400V this limit is about 5000 RPM, and since I have no gear box, this is relevant for my setup.

Anyway, for the price and OEM assembly completeness I just could not resist to try this option, so few days later I gave my OK for the purchase. I got two new packs a week later:

Crates as delivered directly from the EnerDel factory in Greenfield Indiana.
Appearance of the battery pack out of the crate.
This is what you'd see after taking off top cover.

Here are major parameters of each Think battery, I will only provide data is available from public sources or something I could measure or calculate¹.

The official model of this battery pack is "Enerdel Lithium-ion pack PE700-394 Vigor+". The A306 model refers more to the design specific to the Think EV, that applies to hardware and software. Here are all major statistics*:

^*Some of these published numbers can be found slightly different, depending on whom you ask. Even EnerDel's own documentation contains discrepancies. But overall the data in this table is very close to reality.

Per EnerDel's naming convention the pack is constructed as follows:

• 2 cells in parallel = 1 element (3.65 VDC, 35 Ah)

• 12 elements in series = 1 module (44V VDC, 35 Ah)

• 2 modules in series = 1 subpack (88.2 VDC, 35 Ah) - this would be the assembly in sheet steel case with gray plastic cover, 410mm x 186.4mm x 295.3mm size.

• 4 subpacks in series = 1 string (352.8 VDC, 35 Ah)

• 2 strings in parallel  = 1 complete battery pack (352.8 VDC, 70 Ah).

The Power connector on a subpack level is Push-On Radsok® type, 6 mm diameter. Interconnects between subpacks are made with 4 AWG wire.

Simplified schematic of the original power connections of the pack. Two parallel strings have auxiliary contacts in the middle of each, breaking the pack into four subsections half the total voltage each. I'm not so sure if it was worth implementing for a sub-400 VDC pack, but certainly good idea for 800 VDC one.
 All the electrical stuff is mounted in the mid channel between subpacks. Auxiliary relays are tiny (rectangular gray on the photo), evidently rated connect or disconnect the pack under no load only (e.g. engage before main contactors and disengage after). There are also precharge relay, LEM current sensors and main fuse block, two fuses for each string.

This would be good spot to insert a side note: Interesting "feature" of Think's setup is that precharge resistor is mounted on the main controller PCB while precharge relay (most right on the photo) is mounted remotely, in the mid channel. Beats me, I just don't understand the rationale behind such design. I have not seen Think's main controller PCB (they call it MLEC) but I've heard that this tiny (just about 3W !) precharge resistor tend to burn out and take PCB traces with it if user attempts to turn on the car while any DC load is on. This is due to the fact that as entire HV load current will passes through this poor precharge resistor which then will glow red hot in a fraction of a second. In contrast, A123's precharge circuit for Nissan Qashqai's very similar pack has wire wound 100W precharge resistor (actually consisting of two 50W ones in parallel) mounted on a heat sink right next to the powerful precharge contactor. That is more like it. The way it should work is that the VCU logic should prevent turning on any DC bus loads before MC is engaged. That's how it works in Qashqai. If such attempt happens, the DC bus voltage just will not reach battery voltage within allowed timeout and precharge contactor will release with the VCU throwing an error, but the precharge resistor will stay cold. I dare to say that the way this circuit and function was implemented in Think was inexcusable engineering oversight, or considering mandatory OEM vehicle testing before release, could have been deliberate cutting corners to reduce cost. We will never know for sure.

The very first thing after taking delivery of the packs was to make sure all the cells are sufficiently charged, but not to 100% SOC. Initial voltage measurement revealed very uniform condition - all the modules were at 39.4 VDC +/- 50mV. Very uniform. The power supply was set to 45 VDC which corresponds to about 80% SOC and it took roughly 2 hours for each module to get charged. I timed the CC->CV transition to judge if the modules are balanced and how consistent they are. The bulk charge was completed within 1 hour where current dropped from initial 13.6 A  to 1.6 A...1.7 A and it took another hour for the current to taper to 0.1 A. I did not bother to connect an Ah counter, but based on the supply current starting to decline in about 35 min, amount of charge put in was initially 13.6 A * 0.35 h = 4.8 Ah and eye-balling integration of the current in CV mode I'd say 2A average for another 90 min put in additional 2 * 1.5 h = 3 Ah. So each 35 Ah module received about 7.8 Ah total (22% of its capacity). That means the modules were shipped to me charged to 80% - 22% = 58 % SOC which is about what EnerDel evidently maintained. Very good value for a long term storage.

Here is what a subpack pulled out of the battery metal case looks like.
Top cover removed from the subpack. Each module has got two slave BMS controllers EnerDel calls RLEC (Remote Lithium Energy Controller), model A306. Each RLEC serves its 24 cells (2P12S), which is half of the subpack module The cells and embedded thermistors connection to RLEC PCB is implemented via flat flex polyimide ribbon cable (brown on the photo).  
Close up photo of RLEC PCBs. I'll be constructing a test jig with small 18650 lithium cells to one of RLECs to work out access to it over CAN, read voltages and temperatures and turn on/off shunting resistors.
Schematic of the original Think pack
Proposed simplified schematic of the Audi pack. Its topology is partially dictated by how Think's packs were configured. In short I will have two strings of about 750 VDC worth of lithium cells in parallel. Each cell is actually two cells in parallel, so the real pack arrangement becomes 2P192S2P. Each series string will be broken in two halves for decent charging ability and safety. One BRUSA water cooled NLG513 charger per half of single string (per quarter of the battery), four chargers total for 10kW charging power. Same idea as I've implemented in my Leaf, except that DC outputs of chargers in the Audi are not going to be in parallel. The BMVCU will handle J1772 comm if I ever will need to charge from an EVSE. Again, experience with J1772 interfacing gained from upgrading the Leaf will be directly applicable here.

Initially for the software development access to RLECs will be done from a PC environment. This capture of CAN database symbol editor screen shows CAN IDs for all the registers for each RLECs in a pack. Having graphical representation of how each byte of voltage, temperature and other information is packed within CAN messages and bits position within each byte makes working out access to this info via the software quite straightforward, albeit tedious just because of number of RLECs and registers in each. More on this - on the RLEC specific page which will get updated as I progress in testing and designing BMVCU. Considering the way subpacks fit in Audi I will use one Think battery as is (in original enclosure) and mount it behind rear seats. I'll take another battery apart and mount six subpacks in pre-welded aluminum enclosure under hood while two more will go on the back.

After seeing internal construction of the pack I've decided to have one complete enclosure to fit right behind rear seats and take apart another pack to try to locate up front as many modules as will fit. Since the original engine was heavier than all the front modules will be, this will also help the weight distribution which is less than ideal in stock vehicle. Since the modules are large, there is not that much room to maneuver under the hood - if the module is longer by 1-2 cm than the place it could fit, I'm out of luck, I'm not going to disassemble individual modules to reshape each one. As always, carton mock-up boxes helped to try different scenarios and work out best fit. Fortunately six modules were installed with two positioned vertically. Once layout got settled and drawings created, an aluminum box containing all 6 modules was welded. It has two supporting legs that rest on the front suspension beam and isolated from it by rubber bumpers - this helps to support the middle of the box and preventing bottom of the box from caving in. For the rear battery I placed original Think pack cover to represent the pack position and fit remaining two modules from the second Think pack right behind. Another aluminum box was welded just for those two modules, so I ended up with six up front and ten between rear seat and the hatch door. No internal space was compromised (as far as moving the seat or cutting into the floor).

Electrically 8 modules in both strings are still connected in series as in original Think batteries, but physically these 16 modules are split in 6 upfront and 10 on the back. Since RLECs addresses can be set in the range from 0h through Fh (16 modules), the BMS still accesses both packs as if there are 8 modules each. This is clearer from the proposed schematic of the Audi pack. Unlike stock harness controlling each contactor over separate wire to its coil, all the contactors and sensors are accessed over CAN bus, so the physical interface to the BMVCU is very simple - just power and CAN wiring.

The rest of the work was to re-design all the auxiliary components - precharge circuit, contactors and current sensors. Similarly to the Think pack each string in my battery is broken in half with middle contactors separating it in safer <400 VDC sections. I used five identical Tyco Szonka contactors for all internal string connections, except that pre-charge contactor does not have monitoring contacts. So five contactors (mid-left, mid-right, pack positive, pack negative, pre-charge) were positioned in the middle between rows of modules where stock ones use to be. The current sensors and fuse box were also repositioned to make room for the I/O board controlling the contactors and reading the sensors. This board also contains pre-charge resistor (consisting of four 50W resistors in series) and chargers' fuses.

More details about this I/O board design can be found on I/O board page.

At the end everything fit very nicely. Below are the photos of the major installation steps.

Mock up of the front modules fit.
Mock up of the rear battery and separate box fit
This is entire front end where the 6 modules will fit right above the differential and front motor.
Front box positioned and ready to be bolted down.
First two modules installed.
Retaining brackets had to be sawed off from the modules to allow them to fit. Of course they will be retained in place by other means (the lid).
All 6 modules are installed in the front box.
The plastic radiator and headlights assembly is on to resemble final look.
Spare wheel well is used for the chargers. The main battery goes on top of them.
Middle tunnel of the stock pack had to be reworked completely. All the components are removed and...
...repositioned to make room for the I/O board.
Assembling and wiring the battery components in progress.

Now a bit more about I/O board installation and finalizing wiring.

The I/O board, ready to be installed.
I used clamp cage type connectors for all the periphery in the pack. It is very quick to connect, never gets loose, no screws to be lost, over tighten or stripped threads. Love'em.
Here is where the I/O board sits.
Closer look at the I/O board.
This may appear messy job, but actually very simple and quick. Each contactor has four wires (coil and monitoring contacts) and respective connector is labeled as such. All I have to do is push the lever of each connector and insert bare wires from contactors and LEM sensors.
It tool less than 30 min. for all the periphery to get hooked up.
All the connections are made, not the loose wires had to be tied to the top edge of the I/O board where holes for wire ties are located specifically for this purpose.
Periphery wiring completed.

Completing rear battery installation.

A special fixture was built to lift and roll in the complete rear battery assembly.
Think tray with eight modules is showed in. Barely fits side to side, there are about 10 mm gaps between walls and the pack edges! It is also positioned to be against rear seat cushions. This makes fit very snug and I was pleased with the outcome.
After trial positioning the main tray, remaining two modules got their own aluminum box.
The EVision shunt board got home in the cavity meant for Think EV BMS' MLEC board. Very handy. At this point all the holes around both boxes were marked on the floor for studs tying them to the body. Both battery trays got removed for the last time before final installation. M10 rivnuts were installed around both trays and interface wiring connected to the 8 module electronics.
Here all the external interface wiring is installed and connected to the I/O board and EVision.
Cables from the chargers are being connected to the main contactors according to the schematic.
This is what completed installation looks like (no top lid shown).

¹ Yes, I know quite a bit more details about these packs and their BMS than I'm allowed to share in public domain since much of specifics as well as sources of such info is still considered EnerDel proprietary and confidential. The Think EV business is unfortunately bankrupt for a few years now, but EnerDel is alive and well, thus as with any major OEM I have to respect confidentiality of proprietary information.

Next - BMS (BMVCU)