Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Tesla Model S 18650 Cell Test Data

This site may earn commission on affiliate links.
**mostly copy paste from my post on e-s**

For those who are not aware, I have acquired (http://endless-sphere.com/forums/viewtopic.php?f=9&t=67661) a full Model S battery pack, that's basically in new condition. I have disassembled the pack and removed individual cells.


For your pleasure, here are individual cell test results Tesla Model S cells, using a 4-wire programmable DC load; cell was charged to 4.20V with a 50mA cut and let to rest for 45 minutes.


I included a common e-bike cell, a new(from fast-tech) Panasonic NCR18650PF (which is rated for 10A cont.) as a comparison. The PF is a very similar cell with the same chemistry with probably only minor differences in electrolyte additives and separator structure.



Tesla Model S (1).png

(My annoying DC load stopped the 10A run at 2.9V for some reason, so I added some extrapolation [dotted line] based on prior tests to 2.5V.)


During the 10A run, I reduced current to 3A for 83 seconds to allow the cell to cool a bit. I did this to point out a couple of interesting, and quite useful performance characteristics of these cells:
-DCIR falls significantly at higher temperatures. This helps to self limit thermal build up of the... as the cell heats up, it tends to actually produce less heat!
-Panasonic NCA cells exhibit significant "surface discharge," that is, they perform well in bursts, where DCIR is lowered for a short period of time after a rest.


Here is a zoom of both effects:
Tesla Model S zoom.png



-These cells would run great if bursted to 15A for ~5 seconds, we're talking nearly 50W from this tiny, very high capacity cell.
-Because they are so dense, they have ~350% of the energy capacity per volume when compared to LiFe cells, yet they can still compete fairly well with LiFe cells on a power/volume basis.


I would rate these cells at up to 10A continuous, for up to 12 minutes at a time from full charge, then reduce current to prevent overheating.

- - - Updated - - -



Now, here is a torture test on the cell that I ran (by accident):


I set up another 3A discharge on the cell I had just run at 10A to see how it held up.
It actually picked up even more capacity! However, in addition, my DC load decided this time to not shut off. Not only did it over-discharge below 2.5V, it kept going to 0V, and kept pulling current at 0V for three more hours!!:
Overdischarge after 10A run.png



The voltage reading is a separate 4-wire voltage sense, so the cell voltage was really reading 0.010V or less. Now my DC load isn't good at measuring very low current, which it was pulling once it hit 0V, so it reality, it had pulled a little more then shown... maybe 3.35 - 3.4+ Ah.
Surely this cell is now trash? That's what they say about cells over-discharged this badly - dispose of them safely. Good thing it wasn't a HK Lipo in my living room...



Finally, after 3 hours it had shut off, and 10 hours after that, and the voltage of the cell had recovered on it's own to 1.8V
So, I decided to give it about 20mA for a while, bumped it up to 50mA, stopped a few times, and eventually charged all the way to 4.2V proper. The cell seemed ok. Held a charge, no self discharge, no heat, nothing:
Charging overdischarged cell.png


Now, surely the cell was damaged in some way... maybe the capacity was severely reduced, or DCIR would go way up...
Cycle 8 after overdischarge.png



Holy crap, I was right about the durability of the cells!! It lost only ~1% capacity, and after going to 2.55V, it recovered to 3.32V, telling me it had a little more in it, too.
That sure does make me feel a little better about taking them all the way to 2.5V under normal usage....


Stay tuned for a whole suite of torture tests.. :)
I also have an 8-channel automatic programmable cycle tester now that I can use to do cycle testing of the cells.
 
Last edited:
  • Informative
Reactions: hoang51
**mostly copy paste from my post on e-s**

For those who are not aware, I have acquired (http://endless-sphere.com/forums/viewtopic.php?f=9&t=67661) a full Model S battery pack, that's basically in new condition. I have disassembled the pack and removed individual cells.


For your pleasure, here are individual cell test results Tesla Model S cells, using a 4-wire programmable DC load; cell was charged to 4.20V with a 50mA cut and let to rest for 45 minutes.


I included a common e-bike cell, a new(from fast-tech) Panasonic NCR18650PF (which is rated for 10A cont.) as a comparison.



View attachment 76414
(My annoying DC load stopped the 10A run at 2.9V for some reason, so I added some extrapolation [dotted line] based on prior tests to 2.5V.)


During the 10A run, I reduced current to 3A for 83 seconds to allow the cell to cool a bit. I did this to point out a couple of interesting, and quite useful performance characteristics of these cells:
-DCIR falls significantly at higher temperatures. This helps to self limit thermal build up of the... as the cell heats up, it tends to actually produce less heat!
-Panasonic NCA cells exhibit significant "surface discharge," that is, they perform well in bursts, where DCIR is lowered for a short period of time after a rest.


Here is a zoom of both effects:
View attachment 76415


-These cells would run great if bursted to 15A for ~5 seconds, we're talking nearly 50W from this tiny, very high capacity cell.
-Because they are so dense, they have ~350% of the energy capacity per volume when compared to LiFe cells, yet they can still compete fairly well with LiFe cells on a power/volume basis.


I would rate these cells at up to 10A continuous, for up to 12 minutes at a time from full charge, then reduce current to prevent overheating.

- - - Updated - - -



Now, here is a torture test on the cell that I ran (by accident):


I set up another 3A discharge on the cell I had just run at 10A to see how it held up.
It actually picked up even more capacity! However, in addition, my DC load decided this time to not shut off. Not only did it over-discharge below 2.5V, it kept going to 0V, and kept pulling current at 0V for three more hours!!:
View attachment 76422


The voltage reading is a separate 4-wire voltage sense, so the cell voltage was really reading 0.010V or less. Now my DC load isn't good at measuring very low current, which it was pulling once it hit 0V, so it reality, it had pulled a little more then shown... maybe 3.35 - 3.4+ Ah.
Surely this cell is now trash? That's what they say about cells over-discharged this badly - dispose of them safely. Good thing it wasn't a HK Lipo in my living room...



Finally, after 3 hours it had shut off, and 10 hours after that, and the voltage of the cell had recovered on it's own to 1.8V
So, I decided to give it about 20mA for a while, bumped it up to 50mA, stopped a few times, and eventually charged all the way to 4.2V proper. The cell seemed ok. Held a charge, no self discharge, no heat, nothing:
View attachment 76423

Now, surely the cell was damaged in some way... maybe the capacity was severely reduced, or DCIR would go way up...
View attachment 76424


Holy crap, I was right about the durability of the cells!! It lost only ~1% capacity, and after going to 2.55V, it recovered to 3.32V, telling me it had a little more in it, too.
That sure does make me feel a little better about taking them all the way to 2.5V under normal usage....


Stay tuned for a whole suite of torture tests.. :)
I also have an 8-channel automatic programmable cycle tester now that I can use to do cycle testing of the cells.

excellent testing. Thanks for posting.

Yes, very interested in cycle testing. Would love to see how cells hold up after hundreds and thousands of cycles.
 
I wonder about cycle life @ 10A

If it's decent enough the batteries don't seem to be the real power constraint in the P85D, but rather the cooling system.

My tester can do max 10A so I will be running that.

But 10A in the Model S is only 255kW battery power at full charge. or ~320 mechanical horsepower.
The P85D is absolutely limited by the batteries to no more then 500 mechanical horsepower (~16A per cell for a few seconds)
Cooling of the motor and inverter might be an issue, not so much the batteries. The Model S "coolant" that's piped through the modules is really only for environmental temperature control to maximize calendar life and improve performance in the cold. Not so much for cooling the batteries due to load.
I'd wager it's used for heating the batteries much more often then cooling them.
 
"no more than 500 mechanical horsepower"

Well, this may be why the S85D is doing quite well in some timed-tests for 0-60 mph and doing nicely with its two 188hp motors. So, don't tell P85D owners about limiting to well below 692hp - or is there some computation of the torque which takes it up to 692hp for part of the early mph area of the acceleration?
 
Wow that was surely impressive that the cell was able to recover after being, for practical purposes, completely discharged for several hours! Impressive. Wonder how it would have held up if it had a 2, 4 or 6 year calendar life?

Great post and great testing!

What are you doing with the rest of the pack?

Edit: read your post on Endless-sphere - you're selling them ;)
 
Wow that was surely impressive that the cell was able to recover after being, for practical purposes, completely discharged for several hours! Impressive. Wonder how it would have held up if it had a 2, 4 or 6 year calendar life?

Great post and great testing!

What are you doing with the rest of the pack?

Edit: read your post on Endless-sphere - you're selling them ;)

Selling the rest that I'm not going to use for a few projects. :-D

- - - Updated - - -

"no more than 500 mechanical horsepower"

Well, this may be why the S85D is doing quite well in some timed-tests for 0-60 mph and doing nicely with its two 188hp motors. So, don't tell P85D owners about limiting to well below 692hp - or is there some computation of the torque which takes it up to 692hp for part of the early mph area of the acceleration?

Not really. They just rate for peak power of both motors and inverters. Commonly done with hybrids.
http://endless-sphere.com/forums/viewtopic.php?f=1&t=66514&start=75#p1007033

The batteries are def the limiting factor. 700 hp would require like 25A due to voltage sag and would really over-stress the cells.
Unless the P85D has different cells, which I think is unlikely.
 
Selling the rest that I'm not going to use for a few projects. :-D

- - - Updated - - -



Not really. They just rate for peak power of both motors and inverters. Commonly done with hybrids.
Endless-sphere.com View topic - Tesla P85D Insane Mode

The batteries are def the limiting factor. 700 hp would require like 25A due to voltage sag and would really over-stress the cells.
Unless the P85D has different cells, which I think is unlikely.
@okashira
Just for completeness...

Can you calculate how much capacity the battery would need to have -- with the same chemistry -- for the battery to be able to fully support both motors fully and continuously? Are we talking 120 kWh? 200 kWh? 500 kWh?

I'm curious because it might give us a feel for the (perhaps aftermarket) upgradability opportunity for the P85D as battery technology matures.
 
My tester can do max 10A so I will be running that.

But 10A in the Model S is only 255kW battery power at full charge. or ~320 mechanical horsepower.
The P85D is absolutely limited by the batteries to no more then 500 mechanical horsepower (~16A per cell for a few seconds)
Cooling of the motor and inverter might be an issue, not so much the batteries. The Model S "coolant" that's piped through the modules is really only for environmental temperature control to maximize calendar life and improve performance in the cold. Not so much for cooling the batteries due to load.
I'd wager it's used for heating the batteries much more often then cooling them.

Terrific testing! I'd love to see some cycle testing using Tesla voltage parameters, but that's probably asking too much. I thought about setting up a test bench to do that.

I'm thinking these NCR18650BE's are designed for pulse discharge at 20A, which is 6C or 510 kW.
 
@okashira
Just for completeness...

Can you calculate how much capacity the battery would need to have -- with the same chemistry -- for the battery to be able to fully support both motors fully and continuously? Are we talking 120 kWh? 200 kWh? 500 kWh?

I'm curious because it might give us a feel for the (perhaps aftermarket) upgradability opportunity for the P85D as battery technology matures.

You mean the same cell? There are cells with the same chemistry that could handle 691 HP in bursts no problem, such as the Samsung 25r, but capacity would be much less.

For the same cell.. completely continuous, which is no normal automotive use(say a Model S re-geared for top speed run from full charge to dead, or towing 15,000 lbs up a mountain at 70mph)
Id keep it to 10A max, that's 36W at full charge, and ~28W - 30W where you'd want to shut down.
If we use 30W, you'd need 691hp*0.745699872kw/hp*0.95(assumed eff)/(30 W/cell) = 16,317 cells or a ~184kWh pack.
Account for volatge drop of all the fuse wires, connections and wires... make it a cool 96s175p pack for 16800 cells.

More realistically, 9000-10000 cells would do it in bursts like the car currently allows you to hold peak power for.

- - - Updated - - -

Terrific testing! I'd love to see some cycle testing using Tesla voltage parameters, but that's probably asking too much. I thought about setting up a test bench to do that.

I'm thinking these NCR18650BE's are designed for pulse discharge at 20A, which is 6C or 510 kW.

I think the cell would be OK pulsed at 20A, but you'd sag to ~3.3V, so that would be 468kW for the whole pack, peak.

The cells love the heat. DCIR drops by almost 60% when they are nice and toasty (>50°C) so you can extract a-lot more power when they're really warm. Perhaps you P85D guys will see better times on a super hot day letting the car sit in the sun.
The cells would be fine. You could store them at 60C for a month and barely be able to measure the capacity lost.
 
You mean the same cell? There are cells with the same chemistry that could handle 691 HP in bursts no problem, such as the Samsung 25r, but capacity would be much less.

For the same cell.. completely continuous, which is no normal automotive use(say a Model S re-geared for top speed run from full charge to dead, or towing 15,000 lbs up a mountain at 70mph)
Id keep it to 10A max, that's 36W at full charge, and ~28W - 30W where you'd want to shut down.
If we use 30W, you'd need 691hp*0.745699872kw/hp*0.95(assumed eff)/(30 W/cell) = 16,317 cells or a ~184kWh pack.
Account for volatge drop of all the fuse wires, connections and wires... make it a cool 96s175p pack for 16800 cells.

More realistically, 9000-10000 cells would do it in bursts like the car currently allows you to hold peak power for.
Thanks for running the numbers. I asked for two reasons: [1] general curiosity interest and [2] data for some of the, um, "hotter" threads on TMC regarding the current performance of "stock" Model S vehicles.
 
Thanks for running the numbers. I asked for two reasons: [1] general curiosity interest and [2] data for some of the, um, "hotter" threads on TMC regarding the current performance of "stock" Model S vehicles.

Well you could straight up replace the Model S cells with Samsung 25r's

But you would only get a small boost in power, because the Model S (P85D) would still limit to ~16A per cell. Voltage sag would just be less.
You'd need to hack the programming to increase draw to 25A (or more)
At that point you'd want to upgrade all the wiring as well!

25R could do 30A in bursts easily or a cool 700kW! Such a pack would be "only" ~63kWh, but it would be more efficient because the cells have less resistance, so range would still be great.
But they wouldn't last as long, probably.

I think the best way would just be find a way to heat the pack to 50C or even 60C. This would be easy, just install your own coolant heater or hack the firmware to heat it more then usual. Hack the firmware to allow 20A or 25A (1850A) from the pack. You could easily get 40% or more power out of the pack for drag runs.
Oh and upgrade the wiring, bus bars, connections, contactors, etc! :-D
 
Awesome testing! Could you perhaps test how the cells handles deep freezing and charging after ajusting to room-temp? Do you know what V is at 100% charge and 0% is in the car? Does your testing imply that Tesla could have used higher and lower voltage and the batteries still would have survived with minimal degradation? It`s comforting 2 know that our cars probably has some of the best battery chemistries among EVs :)
 
Awesome testing! Could you perhaps test how the cells handles deep freezing and charging after ajusting to room-temp? Do you know what V is at 100% charge and 0% is in the car? Does your testing imply that Tesla could have used higher and lower voltage and the batteries still would have survived with minimal degradation? It`s comforting 2 know that our cars probably has some of the best battery chemistries among EVs :)
No, from what I can tell, Tesla uses ~96% DOD. They go all the way to 100% with a range charge (4.2v 100ma or 50ma cut)
And from the voltages I've seen, they go to ~3-5% SOC when the car stops. The trick is that no-one actually drives like that. Most do 50% - 90% or something like that.

- - - Updated - - -

Great thread. Out of interest, do you know what revision of pack these cells came from?
Was a late B pack. About to get a Module from a guy in Cali from an early A pack.

- - - Updated - - -

Interesting capacity tests. They look like barely 3.1 ah cells, x 3.7V nominal = 11.47Wh, x 7,104 cells = 81,482 Whs. Have these cells lost some capacity, or is 7,104 not the correct cell count?
Capacity depends on how you test and rate them. Nominal voltage depends on discharge rate. I'll run one at ~100mA load to see the actual energy capacity and update the chart to show watt - hours instead of amp-hours on the x-axis.
The pack is 7104, and the cells I have have lost ~1%-2% capacity from new.
 
Capacity depends on how you test and rate them. Nominal voltage depends on discharge rate. I'll run one at ~100mA load to see the actual energy capacity and update the chart to show watt - hours instead of amp-hours on the x-axis.

Wait, do you mean that the battery Tesla rates at 85kWh could be rated differently with another type of usage pattern (i.e. another average discharge rate)?
 
All batteries are like that. Lead-Acid battery are WAY worst than that. Under load, you can't even get 30% of the capacity. It all depends on internal resistance and battery chemistry.

OK. So in the Gigafactory, if they only build the one NCA cell type and make say 74 million of those and put them in 1) Model S packs 2) Home storage packs (Different use patterns discharge wise) then they would report two different numbers when they were going to say how many GWh they manufactured that year?