Testing the big 15 and 20Ah LiFePO4 cells is tough! *Pics*

[cell_man]

If you're willing to DIY, you can do it A LOT CHEAPER.

It seems to me, cell_man, that the most expensive part of these systems are the load. For single cell testing, galvanized fencing cut to *just* the right size submerged in water can take a whole lot of watts due to the high surface area and the resistance can be just suitable for a single cell. Look for the tightly meshed type with a lot of criss-crossing tiny wires. For frequency testing, I have no idea if it's appropriate due to inductance, but for a constant DC load, it seems far more than adequate and doesn't cost more than $10.

I've already got all the resistors I need for a single cell but it cannot be integrated into a computerised system that will generate specs and a nice graph. If I can ever get the video I took just now hosted you will see a cell discharging virtually it's entire capacity at around 400A with no cooling and still maintaining a reasonable voltage. The file is around 20mb and if someone can put it on youtube for me you can all have a look. I live in the land of the not so free, so it's not so easy to post up videos from here.....

 

[swbluto]

If you're willing to DIY, you can do it A LOT CHEAPER.

It seems to me, cell_man, that the most expensive part of these systems are the load. For single cell testing, galvanized fencing cut to *just* the right size submerged in water can take a whole lot of watts due to the high surface area and the resistance can be just suitable for a single cell. Look for the tightly meshed type with a lot of criss-crossing tiny wires. For frequency testing, I have no idea if it's appropriate due to inductance, but for a constant DC load, it seems far more than adequate and doesn't cost more than $10.

I've already got all the resistors I need for a single cell but it cannot be integrated into a computerised system that will generate specs and a nice graph. If I can ever get the video I took just now hosted you will see a cell discharging virtually it's entire capacity at around 400A with no cooling and still maintaining a reasonable voltage. The file is around 20mb and if someone can put it on youtube for me you can all have a look. I live in the land of the not so free, so it's not so easy to post up videos from here.....

Depending on the value of your time (The average wage for your skills are... ??? In China?), it might be cheapest just to manually record values, graph it using something like Open Office's calc, and do a curve fit to generate the nice looking curves. If you plan to do volume testing, then this might not be most productive, but if it's just for testing the performance of a representative cell, it might be the most time-cost effective.

However, your English sounds native so I'd guess your skills aren't bound to China... or are they?

 

[cell_man]

Here's a video I took tonight which shows how a cell typically performs. Please do not right off these cells by what was shown in previous clips of a cell that has been damaged.

http://www.clipshack.com/Clip.aspx?key=A2031A06336A2EBA

That same cell took 17Ah of charge. So that cell was virtually fully discharged in that test. I had also run the cell a little before the test to get a little heat into it as it's a bit cold here, so the clip is not the full discharge.

I'll be getting some a123 26650s and will see how they compare. I don't expect the C rating to be quite the equal of the 26650s but we will see.

 

[CamLight]

Maybe you could ask John from Camlight for the CC-400 he is building!!
Probably that for 3000$ he could assemble enough modular unit to carry more than 2000W
I think it’s somewhere around 500 and 550$ for 500W so probably that for 6 unit for 3000$ you could get 3000W
These 500W units are very high quality and nice desing... and half size than the CBA amplifier for the same power.

http://www.rcgroups.com/forums/showthread.php?t=938266

<snip>
Doc

Thanks for the kind words Doc!

The CC-400's are $389 each, 400W continuous, 500W for the first 15 seconds. That's a conservative, high-reliability rating though. I have one client who asked me to remove the firmware's safety limits and he's operating them continuously up to 500W with bursts to 700W!
Of course, I don't recommend this though.

 

[Bikemad]

Notice the current clamp has exceeded it's 400A max rating. Initially the cell delivers 400A plus at 2.61V and then after some time it drops to 2.40V but still delivers 400A plus.

Just a suggestion, but have you considered using a pair of identical cables in parallel and just put the clamp on one of them?
You should be able to measure up to 800 Amps by simply doubling the reading on your current clamp.

Alan

 

[swbluto]

Notice the current clamp has exceeded it's 400A max rating. Initially the cell delivers 400A plus at 2.61V and then after some time it drops to 2.40V but still delivers 400A plus.

Just a suggestion, but have you considered using a pair of identical cables in parallel and just put the clamp on one of them?
You should be able to measure up to 800 Amps by simply doubling the reading on your current clamp.

Alan

Or if it's a pretty steady current, you can just add up the currents in each cable measured individually. This would skip over any problem caused by slight resistance unbalances.

 

[RoughRider]

hey guys...this cells have demonstrated their potential...

the 15Ah gave over 400A...so its a 30C cell...this is clear now...

there are no further tests necessary...

@Paul
keep your money...you dont need to make any more testing...
make some capacity-tests @10C charge and discharge and thats it...

 

[swbluto]

hey guys...this cells have demonstrated their potential...

the 15Ah gave over 400A...so its a 30C cell...this is clear now...

there are no further tests necessary...

:? - Is that sarcasm?

But, 10C discharge tests would give just about all the information we need. Independent tests would help corroborate the claims.

 

[RoughRider]

hey guys...this cells have demonstrated their potential...

the 15Ah gave over 400A...so its a 30C cell...this is clear now...

there are no further tests necessary...

:? - Is that sarcasm?

But, 10C discharge tests would give just about all the information we need. Independent tests would help corroborate the claims.

Why is that sarcasm?
15Ah * 30C = 450A...the test was over 400Amp...

what kind of tests do YOU need...???

what kind of projects do YOU have, that you need a 30C(450A) discharge?

All i say is: The cells are good for 30C...that more than enough...and there is no need to spend another 3000$ for testing...

 

[AussieJester]

Well said roughrider...

KiM

 

[Jeremy Harris]

I think it would be good to put a dose of realism in here, wouldn't it?

We only need insane discharge levels for short bursts, no matter what the application. Most will want at least 30 minutes endurance from a battery, which means a 2C continuous discharge rate should be fine. The ratio between average and peak power, even on a high powered motorcycle, doesn't exceed about 10 to 1 and is more often around 5 to 1. This means that cells that can deliver 20C peak, 2C continuous, without too much voltage sag, will be fine for even the toughest users. I mean, drawing 20C from a pack implies an endurance of just THREE MINUTES!


Luke,

I suspect that the problems you're having are very much to do with layout. Lead inductance at the currents and speeds that you're trying to switch will produce very high voltage spikes that will quickly take out devices, as you're finding out. You need to design this thing as if it were a microwave device, with very short, wide, connections, ideally stripline that's impedance matched to the loads if you build it on PCB. The gate connections need to be as short as possible, take a look at mounting the FETs in a doughnut configuration, so you can bring all the leads into the centre (a couple of annular rings cut from copper plate and stacked vertically with a spacer might be an idea for the high current connections). Cut the gate leads as short as you can and fit the gate resistors close to the package. Bring the other end of the gate resistors to the centre of the ring, keeping the leads as short as possible. Fit the gate driver in the centre, with it's output pin as close as possible to the junction of the gate drive resistors. Fit high frequency decoupling to the gate drive power supply as close as possible to the package. Use really good high frequency capacitors, backed with low ESR electrolytics in parallel, to maintain a stiff and transient free gate drive supply. Finally, connect the main power connections to the two rings so that they are diametrically opposed, this will help with balancing the current in the devices.

Hope this helps, good luck!

Jeremy

 

[cell_man]

hey guys...this cells have demonstrated their potential...

the 15Ah gave over 400A...so its a 30C cell...this is clear now...

there are no further tests necessary...

@Paul
keep your money...you dont need to make any more testing...
make some capacity-tests @10C charge and discharge and thats it...

Thanks roughrider I appreciate your support. However that cell was actually a 20Ah cell so it's wasn't at 30C, more like 20C. I'm not gonna try to pull the eyes over anyones eyes, what's the point, how long would it be before I was found out??

The 30C that I originally mentioned was a figure that I was given but had never proved and was never written in any documents. I was hoping that tests would have given the actual measured value, but now I will make best efforts to demonstrate the capabilities myself. They will not be continuous current tests due top the nature of my test equipment, but I have enough switches on my test kit to control the current to with say within 20A or so. I had left the tests on the back burner, but now I'll get some clear data for everyone to see.

My feeling is that 20C or a little more continuous discharge is achievable for both cell types and for short bursts considerably more. I will back this up with some tests within the next couple of days and give proof of shunt caliubration by comparison with the 2 current clamps I now have. I think some temperature control is necessary to achieve the very best results but that's not so easy to achieve. Soom cooling is the best I can offer for now. I'm sure there'll be a few headaches getting all these tests done, but I've never been a quitter and will not let problems either big or small stand in my way.

I bought the following test equipment today:
- 2 channel oscilloscope, 1 channel for cell voltage, 1 for shunt resistor with USB connection for data logging.
- 1000Adc stand alone current clamp to confirm current measured on shunt and give proof. A shunt can be anything you say it is.
- DMM with temperature sensing and connection to PC, to log temperature during tests.

View attachment 1

View attachment CIMG0315.JPG

Also have 10pcs of A123 26650s arriving soon for direct comparison.

I'll speak with Cam Light about some test kit and try to get something arranged in the near future.

Also bought some materials for the pack assembly including cooling panels for high current applications. I'm a bit delayed with the pack termination but will be visiting a supplier on Friday which I use for Hub motors who have offered to do the necessary machining for me. The deign is almost there and will make a new post soon with some info on options and construction details.

Thanks for all your patience with this I'll do my best to get it done properly and ASAP

 

[cell_man]

I think it would be good to put a dose of realism in here, wouldn't it?

We only need insane discharge levels for short bursts, no matter what the application. Most will want at least 30 minutes endurance from a battery, which means a 2C continuous discharge rate should be fine. The ratio between average and peak power, even on a high powered motorcycle, doesn't exceed about 10 to 1 and is more often around 5 to 1. This means that cells that can deliver 20C peak, 2C continuous, without too much voltage sag, will be fine for even the toughest users. I mean, drawing 20C from a pack implies an endurance of just THREE MINUTES!

Jeremy

Hi Jeremy I do agree that for most applications a more modest discharge rate is adequate, but these cells are a bit more expensive than adequate cells and they need to perform. Some people have suggested using them in parallel with a much bigger bank of standard LiFePO4 to help with the peaks and reduce sag and in that case they will be pushed pretty hard. Someone else was looking to use these in a sports EV with 2000A. He was looking to do it with 3p cells. I suggested 5 would be be more reasonable.

I use a small 36V 6Ah pack with 5C constant cells for my 800W peak hub motor. I also use a 36V 10Ah pack with supposed 3C cells. The smaller, but higher C rated pack gives much better performance and delivers it's full capacity with little drop off right to the end. I can't wait to see how a pack of these cells will perform.

 

[bigmoose]

I made a copper heat-sink for the FET driver that was getting hot, insulated the bottom side of it, then ran it again, and this time it made a little crackle noise and let the magic smoke out... :x

WTF would cause this?? Any ideas?

Sorry I'm a bit late... the ol' guy was tired and retired at 9pm last night...

On to the question. Your observations are good. Jeremy's advice is spot on. What everyone is learning with you, is that at these currents and switching speeds, even a bussbar has inductance and hence generates flyback voltage according to V= L*dI/dt.

That said, I think that your ground is bouncing all over still, and that you might actually be reversing Vcc on the driver due to ground bounce. Where did you place the Vcc power return path for the driver TO220? I would wire the Vcc supply and return directly to pins 5 and 3 of the TO220, and bypass with a 100uf electrolytic parallel with 0.1 film poly right at that location with short leads. For perspective, you are likely driving the UC3710 with say 12 or 15 volts, if you can feel the inductive shock, that pulse is up around 40 to 100 volts or more... That reversal thru the UC3710 is the likely culprit. Edit: This is assuming the Vcc supply for the UC3710 is a floating supply and the return path is not connected to any other various and sundry grounds that really aren't ground!

Got any scope traces "sneaking up" on max current. These ground bounce artifacts will show up before max I and at a nondestructive level, so you can see them before they let out magic smoke.

As you can see high current, and as Jeremy said, hi frequency layout go hand in hand, and truly is an art form. That's why good RF guys are graybeards... and why they closet themselves up in the bowels of the lab and charge enormous fees for consultation... even more so for EMI/RFI guys... spent all day yesterday "searching for the truth" with a subsystem vendor's EMI guy, trying to determine who's subsystem was hosing the system, hence I was totally tuckered out and spent!

Your doing great Luke, hang in there, and thanks for sharing. We are all learning with you!

 

[swbluto]

In response to RoughRider,

Amps without voltage is meaningless. Just about any cell out there can be used for EV applications can do 30C bursts given a sufficiently short burst period, so labeling such as cell as "30C" is meaningless. Electrical equipment is often rated continuously and that means a long time without damage, otherwise, if it's defined as burst, it's often mentioned.

Sure, someone may not use their pack for 30C continuously typical applications (Given the implied 2 minute run time), but the output voltage will determine the motor current which affects things like torque. If I'm going to be drawing 30C for, oh, say a minute for climbing a hill, I want to rest assured that the cells aren't going to be damaged and that it provides the POWER one needs. (Power = voltage*amps)

I'm not trying to justify the expenditure of billions of dollars on testing, and it's true that some people may not care about 30C performance, but there are quite a few who do. We need accurate tests in this regard, and let's say all the test results so far haven't been as consistent as one would like.

(I don't mean to interrupt the excellent EMI and ground bounce discussion. Anyways, I will quit on this tangent since I see no reason to argue with the technically illiterate.)

 

[cell_man]

In response to RoughRider,

Amps without voltage is meaningless. Just about any cell out there can be used for EV applications can do 30C bursts given a sufficiently short burst period, so labeling such as cell as "30C" is meaningless. Electrical equipment is often rated continuously and that means a long time without damage, otherwise, if it's defined as burst, it's often mentioned.

Sure, someone may not use their pack for 30C continuously typical applications (Given the implied 2 minute run time), but the output voltage will determine the motor current which affects things like torque. If I'm going to be drawing 30C for, oh, say a minute for climbing a hill, I want to rest assured that the cells aren't going to be damaged and that it provides the POWER one needs. (Power = voltage*amps)

I'm not trying to justify the expenditure of billions of dollars on testing, and it's true that some people may not care about 30C performance, but there are quite a few who do. We need accurate tests in this regard, and let's say all the test results so far haven't been as consistent as one would like.

(I don't meant to interrupt the excellent EMI and ground bounce discussion. Anyways, I will quit on this tangent since I see no reason to argue with the technically illiterate.)

As far as I know for a LiFePO4 to be given a C rating it must be able to do that discharge level at a voltage greater than 2V. So on that basis I think you will find that there are not many cells that will actually do 30C no matter how short the burst as their internal resistance simply will not allow it. These cells have internal resistance in the range of 0.8mohm and 1.2mohm (pre checked and marked on every cell, who else does that??). My test that I showed previously on video shows approximately 20C continuous (more initially) and it was still well above the cutoff voltage of 2V.

The offer has been there for a few weeks now for anyone who has the means to test these cells accurately to take a sample and do so. Short of me spending another 3000$ on getting some equipment and waiting for it to be delivered I cannot do much more than I already have. I'll have some graphs done in the next few days once I've figured out all the new test equipment and got it all working properly. Even if I did have the equipment there would still be people doubting my data. I cannot do a constant current discharge but I can keep the discharge current to within reasonable limits and from data collected can easily calculate W/Hrs, Ah etc from data collected, regardless of whether the collected data is constant current or has some variation. As a matter of fact I would say that a constant load (resistance) is more like a real situation, than a constant current. Controllers do not ramp up the load as the battery depletes. Now I don't claim to be an expert in LiFePO4 cells and their testing but I can assure you that I will have sufficient expertise in the not too distant future. The only reason I haven't done more testing is due to limited time and the belief that independent tests would be forthcoming.

I'm not going to send out any more free cells. I've done that already and it has got me nowhere. For now, If someone wants a a sample for tests, they have the means to test it properly, will give me their results and intends to do get a pack sized order later, I'll do something to help a bit with shipping costs on the sample shipping when they come to place a bigger order. Or you could wait till a certain place in Spain supposedly gets some cells this month and pay them 138USD per 20Ah cell plus shipping.

 

[swbluto]

Got any scope traces "sneaking up" on max current. These ground bounce artifacts will show up before max I and at a nondestructive level, so you can see them before they let out magic smoke.

Incremental testing and design, gotta love it. I find it's much quicker to develop software when one incrementally builds and tests, and it seems that in real life, not doing so not only costs time but actual equipment and money.

 

[GGoodrum]

Anyways, I will quit on this tangent...

Good idea.

I've probably got as much experience with a123 cells cells as anybody, and I can say that I personally think enough has been demonstrated already, to convince me that these are the real deal, and seem to have similar performance to what I've seen with the many a123-based packs I have used since the first DeWalt packs came out about four years ago.

That said, I love what Luke is doing, as we all will learn something new, I think. I also appreciate the insight and suggestions coming from the "big guns", like BigMoose and Jeremy.

-- Gary

 

[cell_man]

I'd just like to take this opportunity to make a public apology to Luke as I've been a little rude and unappreciative of his efforts. Between 1 thing and another I was a little concerned when people took some videos that Luke had taken of a damaged cell and taken the values shown to be a valid indication of these cells performance, even though Luke had stated the cell was damaged. The timing was also not so perfect, with the cells arriving at Luke's place just before Christmas, so putting delays onto getting things underway. I forget sometimes that people have more to life than this, as I've pretty much spent my entire time for the last few weeks pretty much entirely occupied by my business.

Luke has put a lot of work, effort and money into his tests and for no benefit to himself. I hope Luke can accept that I am in a very tough situation at the moment and although it is no excuse, I hope he can understand that I'm not such a bad guy and hope we can be friends. I think without Luke's words I maybe wouldn't have taken the massive plunge into getting these cells in the first place, so I do owe him a lot.

Thanks Lucas,

All the best
Paul

 

[liveforphysics]

I'd just like to take this opportunity to make a public apology to Luke as I've been a little rude and unappreciative of his efforts. Between 1 thing and another I was a little concerned when people took some videos that Luke had taken of a damaged cell and taken the values shown to be a valid indication of these cells performance, even though Luke had stated the cell was damaged. The timing was also not so perfect, with the cells arriving at Luke's place just before Christmas, so putting delays onto getting things underway. I forget sometimes that people have more to life than this, as I've pretty much spent my entire time for the last few weeks pretty much entirely occupied by my business.

Luke has put a lot of work, effort and money into his tests and for no benefit to himself. I hope Luke can accept that I am in a very tough situation at the moment and although it is no excuse, I hope he can understand that I'm not such a bad guy and hope we can be friends. I think without Luke's words I maybe wouldn't have taken the massive plunge into getting these cells in the first place, so I do owe him a lot.

Thanks Lucas,

All the best
Paul

Thank you a lot for that Paul. It takes a big man and a strong character to make a public aplogy. Thank you, and I forgive you.

I want to aplogize for putting up a video using a damaged cell, I wanted to show how the setup was able to vary the current by changing the tap position on the buss, but I shouldn't have put of anything that had damaged cells in it, because people would naturally view it as an indicator of cell performance. I just didn't want to risk a good cell until I had the test setup refined, but it was foolish of me to make any videos until I had a refined setup and testing good cells. I'm sorry for that Paul, and I'm sorry about getting so pissed at you for not appreciating the work I was doing to help you.

On a brighter note, BigMoose and I had a very helpful talk while I was picking up more parts at the electronics store. I had thought he was a bit nuts for thinking that the buss bar was having enough of an inducted voltage spike when the FETs turned off to be raising the ground voltage tap point to above VCC of the FET driver. He convinced me though, and I re-designed everything to tap from a single ground, and guess what? No more FET driver heating. Using my sucky-scope, I was able to capture a little hair-line spike of about 70v across a buss bar that you would never dream was possible to have a spike of anything over a few mV. Two big lessons learned, one of which I've been re-learning a few times a year for about 7-8 years now.

#1. Listen to BigMoose. Always. If he tells you that you've got more than VCC voltage reversing across a buss that your brain can't possibily imagine getting more than a mV or two of voltage difference across, don't doubt him, just accept it. If he tells you if you paint yourself purple, stand on a dishwasher holding a rake and jump, and you will land on the moon, you better invest in a space suit, because no matter how impossible it seems, this guy has never once given me wrong advise on anything for any topic, no matter how utterly unlikely it seemed to me.

#2. This is largely the lessen Moose taught me. Ground isn't "ground". There is no such thing as ground if you're dealing with conductors with high currents flowing. There are only various places you can tap on inductors that appear to be grounds anytime the current is at a continous state (or off). This lesson has been learned with all 3 of the true learning tools. Money, time/work, and skin (my damn foot). This means the lesson is solidly learned not to be forgotten, and it's going to help me a ton when I design controllers and things in the future.

 

[cell_man]

Thanks Luke

We all make mistakes and no matter how hard I try to do the right thing, I still continue to xxxx up occasionally/a bit too often.

Glad you're getting it all sorted and your flesh wound is healing :wink: I constantly underestimate what's involved in getting things done, I'm doing it every day, but I suppose if we all realised how hard everything was gonna be from the outset, we'd be less inclined to even try....

All the best
Paul

 

[liveforphysics]

[youtube]0YDy6Y1Ucyc[/youtube]

Datalogger is set to 0.75V/division. The next step down is 0.375V/division, which puts the resting voltage just above the display window.
The current meter has a slow sample rate, so it get's out of sync and displays an in-between current value if it collects the samples at the switching point. The current is always greater than 352amps and less than 356amps for every switching cycle.

20Ah cell:
3.259v resting
2.970v loaded @ 352amps.
Internal resistance 0.821mOhm.

I can switch them at 1hz, 10hz, 100hz, 1khz, 10khz perfectly now with no FET issues. I'm creeping up on current right now, and I want to get some good datalogs and graphs before I risk blowing up the FETs again (maybe it's not an issue now at all I hope?).

At 1hz, it makes an audible "click" sound when it switches, which must be from magnetic fields actually moving parts in a circuit that has no moving parts. At 10-1khz it buzzes, and at 10khz it actually sings! The FETs bank stays cool/warm, the battery tabs stay cool, the cell gets luke-warm, the wires get luke-warm, and the loadbank gets hot enough to blue the stainless!

I've got 4 channels I can record on the datalogger, so as soon as I get setup for recording current and voltage in the datalogger, I will submerge the loadbank in my 200gal water tub and make some full 100% DOD cell performance graphs for you guys. Enjoy the vid, and thank you to all the folks who helped me to problem-solve enough to make this video possible.

-Luke

 

[bigmoose]

Luke, Whooppeee it's working! Kudos for sticking with it!

...what Luke was referring to was my Dad's treasured advice to me, which I shared with Luke years ago... "If you want to learn something it is going to cost you; time, money or a piece of skin. No matter what it will cost you." I think it's modern equivalent is "The road of excess leads to the palace of wisdom." :wink:

Oh and keep in mind it's not the magnitude of the I that is causing the delta V's; it is the dI/dt!

Finally, remember a broken clock even get's the time right twice a day... sometimes you get lucky with a diagnosis, sometimes you just let more magic smoke out...

Back in the ol' days when disk drives were 2 feet in diameter and powered by 440V 3 phase, a technician came into the plant to tweak an ailing drive. He was working on it live when he dropped his screwdriver across the buss bars feeding power to the drive. A nice snap occurred with the gallery saying, "lightning is not supposed to reside in the computer room!" The tech just grunted and said, "This is going to take extra parts from the truck to fix!"

Kudos to cell-man too, we know what it is like to invest all in a dream... and it looks like your cells are good performers. I think we all knew that Luke was showing his tortured cell. At least he didn't beat it with a hammer first...

 

[oatnet]

Nice setup Luke!

Cell_man you will be hearing from me as soon as I figure out what to build around these cells.

 

[swbluto]

20Ah cell:
3.259v resting
2.970v loaded @ 352amps.
Internal resistance 0.821mOhm.

At 352/20 = 17.6C and comparing it to the graphs at http://www.fmadirect.com/support_docs/item_1229.pdf, it looks like a slightly better performer than an a123 cell. This is great confirmation that a123's original advantage is now "out in the open" and subject to less monopolistic pricing, which is pretty awesome. I'm starting to think it might be time to stock up on these before the price inevitably rises with the huge demand soon to follow.

Btw, did you measure the temperature?