[valerun]

Hi All,

wanted to get your opinions on the test I am about to do on a hybrid pack I am trying to build.

Basic idea: build a hybrid pack out of low-discharge-current LiFePo4 (sized for range) and high-current LiPo (sized for max current draw).

Objective: feed a 1000A controller such as Soliton1 for 15+ seconds of full power up to max RPM of Warp9/11 motor (i.e. ~170-180V on the motor). With a pack voltage of 250-300V this probably means ~10s of 750A draw on batteries (assuming other 5s spend in sub-top-RPMs where controller duty cycle is low enough not to worry about max battery current).

Requirements: Pack capacity: ~20kWh. Pack weight below 500lb (about as much as I can fit into my small donor car). And I do not want to kill the pack with excessive current draw.

The weight limit pretty much means that I can NOT just size my pack to satisfy 3C max that TS wants for anything beyond 3s duration (CroDriver posted elsewhere that TS can go for a few seconds of 10-20C but only 1-3 sec at a time per manufacturer). 3C in my application would mean a 300V pack of 200Ah cells - ~1,000lbs... Not going to work.

I believe this is the kind of dilemma faced by a lot of the builders here - how to get good acceleration up to the limits of the controller without having to have a truck to carry the batteries.

Now, I know about A123 and Headways but I am not super-comfortable and happy stringing 600+ little cylinders / pouches together. Also, this solution would likely be at least 2x more expensive.

Hence the attempt to marry the ease of install / safety of large lifepo4 prismatics and crazy power density of lipo (I do know the volatility potential of LiPo and do plan to take measures to monitor their state and contain any potential "events").

Setup: 2 battery strings in parallel:
1. "Slow" prismatics pack of X 3.2V nominal 100Ah CALB or TS cells rated for 3C continuous and 10C for 3s (per CroDriver). This would give up to 300A for 15+ seconds without impact on cell life. $0.35/Wh, $140/kW continuous
2. "Fast" Lipo pack of Y 3.7V nominal 10Ah RC LiPo cells rated for 45C continuous / 90C burst discharge (similar to http://www.hobbyking.com/hobbyking/store/uh_viewItem.asp?idProduct=14617). This would give 450A+ for 15+ seconds without impact on cell life. ~$0.8/Wh, ~$90/kW continuous.
The resulting 110Ah pack would have ~$0.4/Wh so just 14% more expensive than pure LiFePo4 pack but would give 150% more sustained power (if all the stars align, that is .

The X and Y (numbers of cells in the strings) will be selected to:
1. Have the "fast" pack provide majority of the current in high-current demand events
2. Have the "slow" pack do same in the low-current demand events
3. Have the "slow" pack to re-charge "fast" pack during the periods of the low-current draw (optional - if you size the 'fast' pack right, you would be running out of range before draining the fast pack - see below)

Obviously, all this depends on the 'fast' pack being 'stiffer' than the 'slow' pack on discharge - i.e. lower voltage sag at the same current. Then, since both packs are connected in parallel, there will be a cross-over in current-supplied vs. voltage curves which will make this idea possible.

To see if this concept even holds water, I am planning to do a number of tests. Some of these tests will be useful for people who just use large prismatics. I would love to get some feedback on the concept and the test design from the experts on this forum.

Test setup:
Equipment:
1. 20 LiPo cells for 74V nominal
2. 23-27 LiFePo4 cells for 74-86V nominal
3. Soliton1
4. Resistive load capable of dissipating up to 60-70kW. Designed to draw 1000A at ~50-55V (so that the battery draw would be at ~750A - max 15+ sec combined rating for the two strings). This would mean 50mOhms load. Will be a large DIY heating element submersed in a large reservoir of running water.
5. Hall effect DC amp meters on each string and a resistive load

Procedures:
I. large prismatic string alone
1. IR / voltage sag test. Apply 1/2/3/4/5/6C loads at 100/80/60/40/20%SOC and record voltage sag from nominal. Derive IR from the data at various points during the cycle
2. TBD (any ideas for what would be useful?)

II. LiPo string alone.
1. IR / voltage sag test. Apply 10/25/50/75/90C loads at 100/80/60/40/20%SOC and record voltage sag from nominal. Derive IR from the data at various points during the cycle
2. TBD (any ideas for what would be useful?)

III. Hybrid pack (2 strings in parallel)
1. Based on the IR data derived in I & II, and charge / discharge profiles, determine X & Y (# of cells in each string).
2. Build the strings, connect in parallel, connect resulting pack to a controller
3. Apply resistive loads of up to 1000A with 100A increment. Monitor currents at each string.
4. Understand the current distribution under different loads. Tune for desired behavior: I_slow>>I_fast at I_total<<300A; I_fast>I_slow at I_total>500A. Tuning done by adding / subtracting 'slow' cells.
5. Repeat #3 for different SOC of 'slow' cells. Confirm cell ratios still work.

What do you guys think? Has anybody tried hybrid packs like this before? Is this a crazy idea? What are the 'gotchas' I haven't thought about? (for example, longevity of LiPo pack can be a concern with such kind of usage. However, some prelim calcs show it's probably going to be fine. A typical 12-15s acceleration event (1/4 mile pull) would consume ~10-15% of the LiPo cell capacity. A typical 5-7s event (0 to 60/freeway) would consume ~5-7%. In normal 'spirited' street (not race) driving one could expect to have <10 events of the latter type per EV range hence we could assume 1 'fast' pack 70% DoD cycle per 1 'slow' pack cycle which is not too bad).

Please chime in. I will be doing these tests anyway within the next few weeks (as soon as I get my components) but want to make sure that I get all your input into the design, etc.

Thanks!

 

[Duncan]

Wonderful!!

I look forwards to seeing your results

It would probably be worth measuring the cell temperatures while testing as well

Just as a thought the 90Kw load may be difficult to arrange - could you do your testing at lower voltages using lower AH cells

You could test using multiple loads at lower voltages (not using your controller)

 

[major]

wanted to get your opinions on the test I am about to do on a hybrid pack I am trying to build.

Hi valer,

First off, I do not want to discourage you. I like it when people get down and really test stuff as opposed to just typing on a keyboard about it. Some of my favorite tests here are by jackbauer. He likes to really push limits

A few thoughts:

By the time you buy the test equipment and test cells (which you may not be able to reuse), you maybe could afford a pack of the good cells large enough for your vehicle desired range.

You will likely need two complete BMS and charging systems.

If you plan to use a motor controller for a resistive load, you will need inductance in the load circuit. I built a buck converter (basically a motor controller) for my battery discharge stand. It can do about 400V & 500A. I used 24 large (1.2 kW) braking resistors with fan cooling for the load. These are edgewound resistors which seem to provide more than enough inductance for the converter. I can configure the series parallel resistor connections to keep the controller duty cycle reasonable all the way down to 15 volts. Works pretty well, but I have lost () a few IGBTs over the years.

Years ago, before the Lithium age, I did experiment with a hybrid PbAcid/NiCad battery pack. If I remember correctly, it was 2 SLA and 21 NiCad 'D' cells. Was testing at 350A continuous and looking for same DOD on the 2 types. Actually worked pretty well in the lab. But never got sized up and installed into a vehicle.

Here's my test stand:

Regards,

major

 

[peggus]

I applaud you for testing it out of the car first, should be some interesting test results.

Don't forget to test the internal resistance at different temperatures too, it changes drastically from cold to hot.
The discharge resistance and charge resistance also changes with SOC, they tend to be about the same at 50% SOC but diverges from there. Charge resistance goes up with SOC and discharge resistance does the opposite. This will affect how your LiPo and LiFePo4 share current.

Don't skimp on the protection, I've seen a LiPo traction pack on fire in real life, it is not something you want to experience.

Monitor every cell V and T with a good BMS, monitor current, open up the contactors and completely isolate the LiPo pack if anything goes out of bounds.

Don't forget to fuse both string individually.

Personally I think this is more headaches than it is worth, and I have a NiMH + Li hybrid pack in my own car.

Good luck

 

[Jozzer]

I think mixing LifeP04 and Lipo would be very risky.
Much easier would be to mix low rate TS cells with some high rate A123's. I often ran A123's in parallel with LifeBatt's to give packs a little extra kick, as long as the cells are tied together individually you should no problems or complications. It would be easy to make up 10 cell blocks to add to each TS..

 

[valerun]

I think mixing LifeP04 and Lipo would be very risky.
Much easier would be to mix low rate TS cells with some high rate A123's. I often ran A123's in parallel with LifeBatt's to give packs a little extra kick, as long as the cells are tied together individually you should no problems or complications. It would be easy to make up 10 cell blocks to add to each TS..

Thanks Jozzer - that might work in a lot of applications actually. However, I thought it's very hard to get ahold of these cells. And to make the hybrid packs worth the trouble, the 'fast' cells should have ability to provide at least 20-30C for at least 10 sec... Otherwise too much extra weight.

 

[valerun]

Hi valer,

By the time you buy the test equipment and test cells (which you may not be able to reuse), you maybe could afford a pack of the good cells large enough for your vehicle desired range.

You might be right ;-) Big part of this is satisfying my curiosity for why this hasn't been done before. Also, I'd love to work out something that others could use.

If you plan to use a motor controller for a resistive load, you will need inductance in the load circuit. I built a buck converter (basically a motor controller) for my battery discharge stand. It can do about 400V & 500A. I used 24 large (1.2 kW) braking resistors with fan cooling for the load. These are edgewound resistors which seem to provide more than enough inductance for the converter. I can configure the series parallel resistor connections to keep the controller duty cycle reasonable all the way down to 15 volts. Works pretty well, but I have lost () a few IGBTs over the years.

Here's my test stand:

That looks real serious, major! Good setup. I am hoping to cut down on the test load engineering by employing liquid cooling - one can get away with much crazier dissipation requirements that way.

On inductancies required - do you know what minimum inductance I would need to do it reasonably safely with controllers like Soliton1?

Thanks!
V

 

[Jozzer]

Thanks Jozzer - that might work in a lot of applications actually. However, I thought it's very hard to get ahold of these cells. And to make the hybrid packs worth the trouble, the 'fast' cells should have ability to provide at least 20-30C for at least 10 sec... Otherwise too much extra weight.

It is possible to get hold of 26650 A123 M1 cells, at the time the best source were dewalt drill packs. It is possible to cut and reuse the spot welded tabs so you dont need to weld them yourself, dewalts spotwelds are good for around 45A (the cell is capable of more, but the welds will heat the cell fast above this rate). So 1 packs worth of cells (10) can add 450A peak discharge to your otherwise 3C pack.
I do have 60-100 cells sitting on the shelf, but am in the UK, I would think you can get them cheaper in the USA (most of these came from the US market).


Steve

 

[valerun]

It is possible to get hold of 26650 A123 M1 cells, at the time the best source were dewalt drill packs.
Steve

Thanks Steve. I was trying to avoid dealing with individual cells. Look what I found yesterday, as well - http://www.hobbyking.com/hobbycity/store/uh_viewItem.asp?idProduct=10311. A LiFePo4 40C(!!) pack 19.8v nominal, 4.5Ah for <$60. That's ~$0.65/Wh vs. ~$0.4/Wh for CALB large prismatics. In terms of power per $, CALBs are ~$100/kW (assuming 4C max discharge - I know they are rated 12C burst but burst in their case means milliseconds I think...), these 40C cells are ~$10/kW.

What do you think of that?! Again, need to figure out the ratio of # of cells in each string so that they share the current as I intend...

 

[MN Driver]

Thanks Steve. I was trying to avoid dealing with individual cells. Look what I found yesterday, as well - http://www.hobbyking.com/hobbycity/store/uh_viewItem.asp?idProduct=10311. A LiFePo4 40C(!!) pack 19.8v nominal, 4.5Ah for <$60. That's ~$0.65/Wh vs. ~$0.4/Wh for CALB large prismatics. In terms of power per $, CALBs are ~$100/kW (assuming 4C max discharge - I know they are rated 12C burst but burst in their case means milliseconds I think...), these 40C cells are ~$10/kW.

What do you think of that?! Again, need to figure out the ratio of # of cells in each string so that they share the current as I intend...

Careful. Do a search on Google to find places where people have tried these, other forums with people who have tried them have been disappointed. Their LiFePO4 cells haven't met their specs in the past, bad voltage sag. Their LiPo cells however are great but I'd be careful with their LiFePO4 and if you really want to try them, try ordering a sample first and testing them.

 

[galeson]

valerun, check these out.

www.endless-sphere.com/forums/viewtopic.php?f=14&t=23029

 

[valerun]

valerun, check these out.

www.endless-sphere.com/forums/viewtopic.php?f=14&t=23029

Thanks Galeson - very good find but only 100-150 cycles quoted. For all others reading this: the above thread is about a LiPo 20Ah 6s (22.2v) pack for $109 (~$0.25/Wh).

I still think that for LiPo one of the 90C burst packs from Turnigy would work best (37v 5Ah for $160 or $0.85/Wh, 1000 cycles quoted). For LifePo4 same brand makes 40C 10sec pack (19.8v 4.5Ah for $60 or $0.7/Wh, 1000+ cycles). I intend to test both.

 

[valerun]

from http://www.rcgroups.com/forums/showthread.php?t=1206770&page=7:
"The question as to whether or not the Turnigy LiFePO4 battery packs are any good for use at primary power packs in RC applications has been answered as far as I am concerned. And the answer is a clear and resounding, No!"

see post #90 on that thread for details if interested... they recommended A123 instead. duh.

so looks like its back to just LiPo 'stiffener'. Unless I start sourcing DeWalt packs for A123s, that is...

 

[Jozzer]

Running a Lipo and LifePo4 pack in parallel will give rise to all sorts of problems. THe discharge curve is completely different, as is the voltage range. I strongly urge that you experiment carefully before trying this on a larger scale.

Steve

 

[GizmoEV]

You might want to look at the graphs on this page: http://ebikes.ca/index.shtml. Go down roughly 1/3 of the way to the section called Battery Discharge Curves. It is just above the April 2010 date heading. There are discharge curves for several types of batteries including LiPO and LiFePO4. You can really see the difference between them.

 

[valerun]

You might want to look at the graphs on this page: http://ebikes.ca/index.shtml. Go down roughly 1/3 of the way to the section called Battery Discharge Curves. It is just above the April 2010 date heading. There are discharge curves for several types of batteries including LiPO and LiFePO4. You can really see the difference between them.

Thanks GizmoEV - good site! Yes, LiFePo4 has flatter discharge profiles which is exactly why this hybrid battery might work. If one engineers the two string such that LiPo discharge curve intersects a flat LiFePo4 discharge curve at above 30-40% SOC of LiPo pack, LiPo will never get deep-discharged and will effectively serve as a high-power buffer. What is also needed is lower IR from LiPo string than LiFePo4. Then, if the strings are matched at nominal 100% SOC, the load will draw the more current from LiPo4 the higher the load is.

 

[valerun]

ok, ordered a bunch of high-discharge LiPo and LiFePo4 cells (including A123) with chargers from HobbyKing - will see what we get there. Some CALB cells are coming in a couple of weeks, as well. stay tuned ;-)

 

[todayican]

I wonder about using a "watered down" version of this plan. Heaways and Thunderskys, anyone heading down that path?

 

[valerun]

I wonder about using a "watered down" version of this plan. Heaways and Thunderskys, anyone heading down that path?

I thought about this but even the highest power density Headways (8Ah) are 20C so to get a meaningful boost, you would have to get 3 cells in parallel. Or total of 180 cells for my 60-cell CALB string. Could probably work but gets into that high cell count I did not want to get into.

Although, if my high-discharge LiFePo4 packs don't deliver promised 50C 10sec rating, I might go back to Headway+CALB/TS plan

 

[valerun]

ok got all the stuff (A123 25C continuous, Turnigy LiFePo4 high discharge 30C continuous-40C 10sec, LiPo Turnigy High discharge 45C continuous-90C 10 sec), chargers, etc.

Completed an initial set of tests, ramping from 1C all the way to 50C. Key learnings:

1. As expected, LiPo battery is way ahead of the LiFePo4, A123 or not. 15-20mOhms per AH at 53C (so 5Ah would be 3-4mOhms = 5 of 1AH in parallel). This amounts to a sag of just 0.7V per cell - at 53C! The IR tends to decrease with increasing temperature and (weakly) increasing current. So I expect ~1-1.1V sag per cell at the max rated 90C draw (10sec rating). With my 5AH pack, this means 450Amps at (3.7-1)V = 1.2kW per cell.

2. Turnigy high-discharge LiFePo4 are actually not that bad (for LiFePo4), showing just 0.7V sag at 25C. The internal resistance, as for LiPo, gets lower with temperature and current. The current dependence is MUCH greater than LiPo, though - 3-4x reduction from 5C to 25C. Which likely means that at the rated 10sec max of 40C, we will see less than 1V sag. For a 5AH cell, this equals (3.2-1)V*200A=0.45kW per cell. Obviously, quite a bit lower than LiPo. Also, they are ~20-25% bigger than LiPo for the same capacity, so you have ~30% of the POWER density of the LiPo...

3. You better not discharge Turnigy LiFePo4 below 40% SOC or IR goes through the roof and you (at least) don't get the expected bang out of it or (at worst) overheat the cell and damage it.

4. A123 LiFePo4 sags about the same as (or a bit more than) Turnigy at 25-30C, which was a surprise, considering that (1) it's at least 2x the price and (2) everyone seems to be obsessed with A123 at the moment... It does have a more flat IR profile with respect to SOC, so I guess there's some good news.

5. Both LiFePo4 batteries exhibited quite a bit lower capacity at extreme loads. LiPo - not so much.


All in all, my exec summary is:

1. If you want max performance, LiPo is the answer. With my setup (60 100AH CALBs, with claimed <0.9mOhm IR, and 4C continuous), I can get 72kW boost power out of 60 5AH LiPo cells - at ~$900 total cost (still need to figure out commutation to the main pack). HK has 10-cell packs in stock so you just need 6 - makes wiring a breeze.

2. There is that small downside of your car going up in flames with LiPo, though. If you think that no safety features are enough to counteract that, that means Turnigy LiFePo4. To get the same boost power levels, though, you'd have to build a 13Ah pack, at ~$2000-2500 total cost. IMHO, this makes little sense as a booster. Perhaps one can build a complete pack out of these puppies and then it might make sense. You can get a 6S 4.5AH at $70 from HK - or a 60-cell 45AH pack at $7K. It will support 30C (or 1,350Amps) continuous, 40C (or 1,800Amps) burst. Should be quite sufficient for any kind of high-power EV ;-) It's only about 8kWhr pack, though, so don't plan on more than 25 mile range on that...


Since all of range, power and cost are kind of important to me, I will likely go with LiPo as a 'turbo-booster' pack. I will plan to enclose each of the 10-cell packs into a LiPo sack, then slot into a 1/8" steel pockets, then enclose the whole thing into a 1/4" steel box. I will run water cooling tubing around the whole thing, and will get my Arduino car computer to check the pack temperature before and while the pack is connected. IMHO, this should be sufficient to prevent any blow-out and, if that actually happens, contain the impact.

After looking at the IR numbers I have obtained, I am convinced that direct connection of the main pack and this power boost pack are not going to work. So the next incarnation of the approach is a separate pack that is connected to the main pack only at WOT.

The LiPo pack cell count will be determined so that the sagged voltage of LiPo string at 90C matches voltage of the main pack at the main pack's continuous discharge current (400A).

This will give me 10sec burst Amp rating of ~850A. At this Amperage, sagged voltage will be ~170V. Total burst power ~140kW. With 80% motor efficiency, I will maybe get 160 wheel hp.

...maybe I should upgrade to a 300V pack... ;-)

V.

PS. Attached are the summary charts from my tests. These are generally averages of 2-3 tests for each battery type. I know it's not super-statistically sounds testing methodology but works for me ;-). Left column of charts is a cell voltage, right - Internal resistance (normalized to 1AH). X axis is % SOC (of nominal capacity). Enjoy