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

Joined
Oct 28, 2008
Messages
14,539
Location
Manhattan Beach, CA, USA
So, I've been doing cell testing on the cells sent to me by Cell_man. They are 15Ah and 20Ah cells rated for 30C discharge.
There has been 2 threads about them:

http://endless-sphere.com/forums/viewtopic.php?f=14&t=14719

http://www.endless-sphere.com/forums/viewtopic.php?f=9&t=14832

I badly underestimated the amount of work it would be to test these cells.

I first made 4awg leads, and used 1/16"x1/4" buss bars for terminals. The problem is, when you're trying to pull ~500amps at just a couple of volts, keeping the voltage drop happening at the loadbank (where you want the heat to go) is tough! Pulling 500amps for a minute continous also takes professional crimped connections on 4awg, and turns them into a smoking mess that melts into the carpet. lol So, I thought, hmm... lemme double up on all the cableing... NOPE! Still a fail! lol

For anyone looking to do cell testing at these current levels, I will save you the time and expense of doing it wrong twice so you can just start out doing it right. Use 000, or 0000awg cable. Use cable ends that say they are rated for 2,000amps continous, the little "500amp" ends have too much V-drop, and get hot, and V-drop is a big problem when trying to pull absurd currents at tiny voltages.

For bus bars, use something absurdly oversized. You will be glad you did, and this was the only way I finally got my conductors and connections to stay cool.

batteryleads.jpg


battterminal.jpg


When doing a FET stage for 450-600amps at ~1.8-2.0v, you're basically SOL. I designed for 800amps, this worked for about 4 seconds before I got fireworks. I designed the second one for ~2,000amps, and it just died today. The FETs do NOT have a good Rds at 2v, and they do not behave like they do at higher voltages, even just 5v they perform fine, drop them to 2v, and all bets are off. From looking at the scope, they seem to switch slowly, which I'm thinking adds the boatload of heat to the silicon, which in turn jumps up the Rds, and they quickly fail. Also, I can't stiffen this with caps and things, because that would spoil the clean cell voltage ripple I need to read on the scope to read Ri on the cells. Because of no caps and crazy currents, could the FETs be doing inductive current loops or other wacky things that are killing them?

I'm determined to get some 100hz, 1khz, and 10khz Ri measurements from these cells at full 30C discharge, so the new plan is to go with 20 of this fet:
http://www.irf.com/product-info/datasheets/data/irfp4004pbf.pdf

I'm going to mount them directly with solder paste onto a 1.5"x1/4" x 20" long copper buss bar using the solder paste/hot-plate method, then snip the center tabs off each package, tie the gates together with a 10ohm resistor in series with each gate, then lay a 1"x 1/8" copper buss bar across all the source legs, and drill the ends of the bus bars and bolt the 0000awg terminals onto them. I've got 7amp FET driver chips to drive the gates.

Does anyone with good experience at dealing with high currents at extremely low voltages have any special insight to help? Does my new FET bank idea seem like it would be tough enough? On paper it should good to switch about 3,800amps, but I'm not sure if it will handle 500-600amps at 2v, because the last bank I built should have been good to 2,000amps, and it made a sound like firecrackers and blew up testing at 1khz and ~500amps, yet the sink never really got warm. It was like the FETs just exploded inside rather than overheating.


I'm struggling with this, and I'm long past-due on wanting to get this all done. Any advise from you pro-cell testing guys would be great!

Thanks!

PS: Here is a video of this poor 15Ah cell taking some abuse. :) I've kept one 15Ah cell freshly cycled, and not abused, and this other poor guy has been my little test dummy. He has been dead shorted a few times by mistake, he vaporized a wrench for me, he has been stepped on and folded a bit (oops!), and has taken 500amp continous discharge from 3.7v to 2.0v about 10 times. Amazingly, his capacity has actually increased just a whisker, and he still performs great, though I think I've managed to increase his Ri a little (I got him too hot a few times). This idea was to determine of these things were delicate cells or not, because sh*t happens in EV applications, and it's important to know your pack isn't going to be killed if a controller shorts inside or something. I think we can rest assured these cells, and the tabs on the cells are very much up to the task handling the full 30C. I've hit this 15Ah cell with 700amps (47C) and the tabs stayed fine, voltage dropped quite a bit though. Here is a quick little video to see the sort of abuse I'm putting the cells through, and they take it like champs. This is my whipping-boy abused/damaged/overheated cell, and he still performs. Note that voltage gauge is a 0-200vdc gauge that has crap accuracy at these voltage ranges. The current gauge is a 0-800amp shunt type, and I calibrated it to match my fluke, I wouldn't trust it right down to the amp, but I would definately trust it to be +-10amps. I use a scope when I do Ri measurements of course, but it doesn't show up on video worth a damn unless you're focused on on it, so the meters are just to be visible in the video and give an idea of the current these things can pump.

[youtube]i4CefGEnryc[/youtube]
 
Here is another little video of the same whipping-boy cell. :)

[youtube]QvmrH-CFfqc[/youtube]
 
These cells have been on the bench at the ev shop for a couple months and they have been fast charge and discharged with success. I remember thinking "wouldn't it be great if the E-S guys had access to these" and lo, there you are doin the damn thing. I'm doing my taxes e-file style with direct deposit as soon as they'll let me and you can bet your ass that money is going to be in my account for less than a day before the cell man has it. Keep up the good work bro.

Now make me a cordless tesla coil dammit!
 
They’re beautiful!
 
Looking Good Luke :mrgreen: You're the man :mrgreen:

Yeah it's not easy is it.... I was only messing with resistors and even that was a real PITA. I can't see any of the pics or the vids (must be blocked by the Chinese firewall) but it sounds very very good. I've pulled out random cells from the next batch I've received and they all perform just the same :wink:

I've made some progress on a nice simple, no solder and fairly cheap method (need to get some prices for the bits this week) to link the cells. Will start a new thread maybe tomorrow with my progress so far and to get a bit of feedback. It's gonna be PCB based with terminals made from sections of aluminium. Will get some bits put into small production this week so we'll have complete packs within a few days. Some BMS and Chargers are on order (12S and 16S initially).
 
We lost a few things from yesterday, but this is where I'm at with the cell testing.


20100102100934.jpg


fetresistors.jpg


fetdriverchip.jpg


This time, you can see the gates have zeners (6.8v), and 100ohm gate drive resistors rather than 10ohm. I also added 2.2k pull down resistors between the source and gate. The purpose of those was to get the gate switched off in a timely manor (minimize time in the linear region) if I want to manually trigger the gates high to use the FET bank like a manual On/Off switch.

This is where I'm stopped at the moment, because I want you folks with way better electrical design skills to give it a look over and let me know if anything looks like a potential failure brewing. :) If I can avoid the inductive problems and gate bouncing problems this time, on paper it should be a 0.25mOhm switching array with 2,400amp capability, and only 0.15v of drop at 600amps. If the FETs can be controlled correctly, at 600amps, it should only be needing to dissipate 90w of heat, which it would be easily able to handle. It really seems like it should stay together just fine IF we can get these FETs to switch and behave as they should.
 
Luke, it's worth a shot. Try, if you have suitable loads, to sneak up on it if you can. Try a few switching cycles at say 100 amps with the scope. Ch 1 gate to source; Ch2 drain to source; 100 nS/div say to start and see what it's doing.
 
bigmoose said:
Luke, it's worth a shot. Try, if you have suitable loads, to sneak up on it if you can. Try a few switching cycles at say 100 amps with the scope. Ch 1 gate to source; Ch2 drain to source; 100 nS/div say to start and see what it's doing.


Good idea :) I will see how it goes, and see how the waveform looks.
 
I see we have lost some recent comments due to the last evening 'outage' so ill say it again...

"Luke...you fuckin rock!"

KiM

p.s if you ever do get to OZ that accent of yours will ensure MUCH female attention...if you brings
a cap from one of the US0fA Navy boats your a dead certain to be sexin anything single in a
skirt with very little effort LoL ...Just pop on the 'cap' and tell the wiminez you steer the boat or point the big guns
in the bad guys direction and your set...
 
Hey bigmoose, with your post that was lost due to a server failure, your link to the data sheet that recommended higher resistance and higher inductances for a ferrite bead on the mosfet's gate was really valuable(and also mentioned a clamping zener diode didn't really solve the high-frequency oscillation problem). Do you have a link to it?
 
I kept a page open with that paper on my netbook. I'm on my phone right nowm but when I get home from my date I will post it up. Great info in that paper! Stuff I never knew about!
 
liveforphysics said:
when I get home from my date I will post it up.

Got PiCs? Tell her if things dont workout and shes after someone with a lil more "class" and "sophistication"i'll be here for her Luke well...as long as shes 'up to scratch'
(seeing your dating her I'm asuming shes a stunner...)

:mrgreen: :mrgreen: :mrgreen:

KiM
 
AussieJester said:
liveforphysics said:
when I get home from my date I will post it up.

Got PiCs? Tell her if things dont workout and shes after someone with a lil more "class" and "sophistication"i'll be here for her Luke well...as long as shes 'up to scratch'
(seeing your dating her I'm asuming shes a stunner...)

:mrgreen: :mrgreen: :mrgreen:

KiM


That is one awesome picture on the link :p
 
Luke, in my loads I'm easily getting over 150A down to 1.6V Vds with no problems using four PCB-mounted FETs (no bus bars) so I think your 20-FET setup can work just fine. That's an awesome bus bar setup you've got!! I just parallel five loads with 4AWG wiring to a terminal block to get 750A max. out to 000AWG cable.

The discrete wiring you have for the resistors and zeners won't be a problem (I think) for occasional on/off switching of the FETs but might be a problem at the higher frequencies you're looking to do the IR testing at. If you see a lot of ringing and overshoot at turnoff during testing, I recommend a very low inductance PCB setup to the gates of the FETs. Several good hi-amp FET drivers are probably needed too. Or, at least, have the board set up for them so you can just drop them in if needed.
 
These got lost in the shutdown also,

Cell_man here are the videos hosted here in MP4:

Obtained via http://googlesystem.blogspot.com/2008/04/download-youtube-videos-as-mp4-files.html
 

Attachments

  • get_video2.mp4
    1.9 MB
magudaman said:
These got lost in the shutdown also,

Cell_man here are the videos hosted here in MP4:

Obtained via http://googlesystem.blogspot.com/2008/04/download-youtube-videos-as-mp4-files.html

Thanks :wink:
 
Thanks for the great test Luke.

I wrote down the figures I could catch from the videos.


10C - 147A - 2,3V
12C - 181A - 2,5V
15C - 233A - 2,2V
20C - 296A - 1,8V
36C - 538A - 1,2V

I have to say that I'm a little disappointed.

My ThunderSky cells sag to 2,1V at 10C. This cells seem to be not that much better
 
CroDriver said:
Thanks for the great test Luke.

I wrote down the figures I could catch from the videos.


10C - 147A - 2,3V
12C - 181A - 2,5V
15C - 233A - 2,2V
20C - 296A - 1,8V
36C - 538A - 1,2V

I have to say that I'm a little disappointed.

My ThunderSky cells sag to 2,1V at 10C. This cells seem to be not that much better

Comparing to traditional a123, from http://www.rcgroups.com/forums/showthread.php?t=676587

A123 at rest voltage=13.5 volts-- loaded voltage=12.7. amps

It appears at ~5.5C(12.7A/2.3Ah), a123 droops to about 3.175 volts (12.7/4). This suggests the cells don't have the same performance as traditional a123 and if you consider a123 cells to be 30C cells, it appears these would be rated somewhere between 5 to 10C (I would like a 5.5C test just to make the calculations simple. I could use the SDT formula though.... hmmm.).

EDIT:

I just found a good graph of the a123 cell's performance at this link: http://www.fmadirect.com/support_docs/item_1229.pdf

It appears ~10C has about 2.88 Volts and 20C has about 2.61 volts.

The difference in magnitude appears to be about
Code:
(3.3-2.3)/(3.3-2.88) = 2.38
(3.3-1.8)/(3.3-2.61) = 2.17
So about 2.25 times worse. If you consider a123 cells to be "30C" cells, you'd expect these cells to be rated at ~13.5C. Still, that's not too bad. By the way, we need discharge graphs of ThunderSky batteries. That seems pretty good for "2C" cells.
 
CroDriver said:
Thanks for the great test Luke.

I wrote down the figures I could catch from the videos.


10C - 147A - 2,3V
12C - 181A - 2,5V
15C - 233A - 2,2V
20C - 296A - 1,8V
36C - 538A - 1,2V

I have to say that I'm a little disappointed.

My ThunderSky cells sag to 2,1V at 10C. This cells seem to be not that much better


That is my abused, overheated, dead shorted multiples times test cell. It's voltage didn't used to sag that badly before damaging it. Once I get a proven cell switching setup complete, which I've almost got finished (for the 3rd time...) then I will use the freshly cycled virgin cell to take test readings, and I will do it with my scope and fluke 189 DMM so we can get some accurate and precise figures.

I've got a temp-probe setup with a little digital display to stick on the cells as well during testing so we can be logging lots of things.


Once I get a rock-solid test setup finished, I'm going to set the cell on a block of ice, and get some Ri readings at 0-2degC, then set it on a heating pad, and get some Ri readings at 50degC, and of course the normal Ri readings will be at 25degC.

I'm also going to take some 100% SOC Ri readings, and 20% SOC Ri readings.

I'm going to do 10hz, 100hz, 1khz, and 10khz Ri measurements at the different temps as well.

This should be enough data to let us put together some graphs of temp effects on Ri, freq effects on Ri, and SOC.

This will be really valuable predictive performance data for folks building packs. Come to think of it, I think it will be the useful data for designing a pack that I've ever seen given for any battery. :)
 
That Nigerian bomber really has me pissed off now, because I'm sure that now there's a 0% chance the airlines will let you on the plane carrying 54 of those beauties. 108 is what I really need (for now), but I can get by on 54 or maybe even 27 for the short term. :mrgreen:
 
John in CR said:
That Nigerian bomber really has me pissed off now, because I'm sure that now there's a 0% chance the airlines will let you on the plane carrying 54 of those beauties. 108 is what I really need (for now), but I can get by on 54 or maybe even 27 for the short term. :mrgreen:

I could put them in my checked luggage. I'm not afraid to try at least. I will make a professional label for each cell that says NiCd battery, 5Ah 1.2v or whatever they would allow. ;)
 
LFP,
I agree with Camlight's suggestions for circuit reconfiguration, although I think he understates the potential problems from switching the gates directly through those inductive, leaded resistors. I would bet it's not the switching frequency, but the edge rates (high harmonic content) and are likely causing your problems.

Definitely use FET driver IC's. They have some switching hysteresis (better switching signal noise immunity) rather than bounce like directly driving the FET gates will do. One gate driver should be able to switch several FET's, but the really important point is to have a physically short, low impedance path (including a signal ground/reference plane) for your gate drive signal to avoid ringing while switching. I've only needed a few ohms (<10) on the gate to damp any bouncing with short, low impedance traces for much lower current FET driver circuits.

This layout/implementation technique should apply from your signal generator all the way to the FET's. Try using coax if a making a PCB is not convenient.

Thank you for all your time and effort on testing these cells!

Eric S.
 
swbluto said:
CroDriver said:
Thanks for the great test Luke.

I wrote down the figures I could catch from the videos.


10C - 147A - 2,3V
12C - 181A - 2,5V
15C - 233A - 2,2V
20C - 296A - 1,8V
36C - 538A - 1,2V

I have to say that I'm a little disappointed.

My ThunderSky cells sag to 2,1V at 10C. This cells seem to be not that much better

Comparing to traditional a123, from http://www.rcgroups.com/forums/showthread.php?t=676587

A123 at rest voltage=13.5 volts-- loaded voltage=12.7. amps

It appears at ~5.5C(12.7A/2.3Ah), a123 droops to about 3.175 volts (12.7/4). This suggests the cells don't have the same performance as traditional a123 and if you consider a123 cells to be 30C cells, it appears these would be rated somewhere between 5 to 10C (I would like a 5.5C test just to make the calculations simple. I could use the SDT formula though.... hmmm.).

EDIT:

I just found a good graph of the a123 cell's performance at this link: http://www.fmadirect.com/support_docs/item_1229.pdf

It appears ~10C has about 2.88 Volts and 20C has about 2.61 volts.

The difference in magnitude appears to be about
Code:
(3.3-2.3)/(3.3-2.88) = 2.38
(3.3-1.8)/(3.3-2.61) = 2.17
So about 2.25 times worse. If you consider a123 cells to be "30C" cells, you'd expect these cells to be rated at ~13.5C. Still, that's not too bad. By the way, we need discharge graphs of ThunderSky batteries. That seems pretty good for "2C" cells.

I've tested many of these cells and they all have performance which is far above the above figures when measured on my resistive load. I've already posted lots of pics of the cells under various load conditions. Also I have been instructed that these cells should never been loaded to under 2V so I would say that the cell under test has already been seriously damaged as stated by Luke and is therefore no longer performing properly.

Now lets say that even if a 26650 A123 cell has a slightly better C ratings than these prismatics, that does not take into account the much higher energy density of the prismatics and the form factor of the prismatics makes the volume much smaller than a pack built from 26650s. Not to mention the ease of pack building. A 16S pack of 20Ah cells is 110mm high, 165 wide and 227 long (not including terminations). How big is 16S, 9p (144 cells) of 26650s?

I would say a sensible approach to sizing a pack is to base it around 10C. This would give you plenty of headroom for higher peaks and also give you a more reasonable range. 10C will after all only give you 6 minutes usage. I will try to do some more tests on the resistive load and you can see the typical voltage sag at more reasoable loads. 30C is great for dragsters, but come on how useful is it in real world situations. A very high C rating just means that at more reasonable loads, the sag is is much reduced as is the heating.

There are plenty more cells if anyone has test equipment that can adequately check a 20Ah cell. They are welcome to contact me to obtain a couple of cells for test.
 
I'm going to get 10pcs of A123 26650s and do some comparisons myself. I'll happily give specifics and take a video and forward to anybody to be posted on youtube or whatever. I thought my images already posted using a resistive load would give some idea of the cells capability but I'm now thinking that I'm best to dump another 3000USD into a test system that can test to 2000W.

I've got some normal 26650s around here (but they're 3Ah cells so already 25-30% higher energy density than A123 26650s :) ) see the pics below to give you an idea of the size. The plastic former in the pic will accept 16 * 20Ah cell stack, 110mm high.

A block of 6 cells measures 69 * 55 * 82.5mm. So 2 by 4 of these blocks would be 69 * 165 * 220mm. So this would be total 48 cells. If A123 26650 that would be a total capacity of 110Ah if all in parallel. This is also the exact same size of 10 * 20Ah prismatics which would have a capacity of 200Ah :!: Even high density Lifepo4 26650 typically have about 3Ah per cell and they will give you max 5C. So that would be a total of 144Ah of high density standard LiFePO4 compared to 200Ah. The prismatics will have a termination method which can be DIY and will not require almost 100 tabs to be spot welded.

View attachment CIMG0310.JPG

View attachment CIMG0311.JPG

I've personally seen well in excess of 400A from both these type of cells and the voltage sag is still well above acceptable limits. I calculated over 550A from a single 20A cell and this was at just over 2.2V. The cells do not like continuous discharge at that kind of load but they will sustain it for short burst of a few seconds. At more reasonable 10C-15C, voltage sag is much more acceptable and heat is not so bad.

I've got Skype, I can fix a webcam and will happily demonstrate the cells capabilities live to anyone with an a genuine interest in buying some cells. I'd be very interested to know of any better cells currently available than these, when all things are considered, because I don't know of anything that comes close.
 
Back
Top