Pics/Info: Inside the battery pack

[wk057]

Interesting. So it seems I have no idea where the voltage is measured during supercharging.

In your video on the 85kWh charge from dead you look away from the screen at the moment it would have had 0A flowing but a voltage reading. But, by the time you go back its only about 20A or something and is reading ~320V.

In the 60kWh video I see 273V as a low, so, 3.25V per cell... which is reasonable. Extrapolating to the 85, that would be 312V per cell.

Seems if I stick with say 3.3V to 4.1V for my SoC range I'll probably be fine.

 

[stopcrazypp]

To be clear, when you saw the voltage at 12 miles was current flowing into the pack? Because as soon as current flows the voltage is driven much higher, especially at a supercharger.

That's an important point, as during that point, it'll be the charging voltage, not the actual battery voltage. Things will also be different when measuring an open circuit voltage (OCV) as in the salvage battery versus a battery that is still connected (and running accessories and/or vampire drain).

I should add though that having a narrower range is always safer, although you get less capacity out of the pack.

 

[wk057]

Added a few pics to the first post of this thread.

 

[FredTMC]

So I figured out what the sharpie lines near some cells are. Looks like the cell level fuse was misplaced or something and manually repaired. Red for when the messed up fuse is found, blue for when it is repaired.

The pack I have only has 3 such occurences visible.

very interesting!

 

[JRP3]

Worth noting that during cycle life testing Panasonic takes the cell to 2.5V under load, for 2000 cycles, so anything above 3V should be fine. They also only charge to 4.1V in that test. Page 34 http://www.embedded-world.eu/fileadmin/user_upload/pdf/batterie2011/Sonnemann_Panasonic.pdf

 

[tom66]

The contactors are probably different parts because one will have to interrupt full rated current, whereas the other will never have to, if they design the software right to disconnect them in order.

I wonder if/where the inrush limiting is. Maybe in the inverter, but space is very cramped, for another contactor. Any sign of a pre-charge resistor in the pack?

 

[FredTMC]

Worth noting that during cycle life testing Panasonic takes the cell to 2.5V under load, for 2000 cycles, so anything above 3V should be fine. They also only charge to 4.1V in that test. Page 34 http://www.embedded-world.eu/fileadmin/user_upload/pdf/batterie2011/Sonnemann_Panasonic.pdf

Thanks for posting this! Model S has 3.1AH cells. Based on this published graph, the cells have amazing life even after 2500 cycles (> 70% capacity remaining).

I currently have 36k mi on my 60 kWh battery. That's ~200 cycles. I get 198 on a range charge. It was 208 mi when brand new. So, I've experienced ~5% degradation. I'm doing better than the graph. I'm sure that's because I don't do deep discharges and don't do many range charges.

This graph implies my battery will still have over 70% capacity after 300k miles. That's insane.

- - - Updated - - -

Panasonic graph

- - - Updated - - -

Graph.

 

[JRP3]

In fact your range "loss" may be just a balance or software calculating issue, since some people have "recovered" range by doing extended full charge/discharge cycles and leaving the car sitting plugged in for a while. Not that I'd want to do it with the minimal changes you've had.

 

[Dave EV]

This graph implies my battery will still have over 70% capacity after 300k miles. That's insane.

Just keep in mind that accelerated cycle life tests may have very little to do with actual battery life over 300k miles unless the conditions of your battery match those of the cycle test (which is highly unlikely). Capacity loss over time plays a large role in how much capacity one will lose over the life of the battery pack.

 

[scaesare]

Thanks for posting this! Model S has 3.1AH cells. Based on this published graph, the cells have amazing life even after 2500 cycles (> 70% capacity remaining).

Are you surmising this, or have some authoritative source?

I ask because it's widely held that Tesla's cell chemistry (which is not disclosed) differs from that of the standard Panasonic parts.

Earlier in this thread it was suggested that they are 3.3Ah cells, which, along with the observed voltages of the packs, makes the overall pack capacity numbers match more closely.

 

[FredTMC]

Are you surmising this, or have some authoritative source?

I ask because it's widely held that Tesla's cell chemistry (which is not disclosed) differs from that of the standard Panasonic parts.

Earlier in this thread it was suggested that they are 3.3Ah cells, which, along with the observed voltages of the packs, makes the overall pack capacity numbers match more closely.

Assumption due to discussions on TMC. It's at least 3.1. Yes, chemistry TM uses is tweaked from stock Panasonic. Since stock graph from Panasonic shows out to 2600 cycles, I believe TM tweaked batteries are at least as good as graph from Panasonic.

 

[TonyWilliams]

The NHTSA crash test of the 60kWh Model S (14 modules) lists a minimum operating voltage of 210V (2.5V*14 modules*6 series group per module) and max operating voltage of 350V (4.167V*14 modules*6 series group per module).
http://www.teslamotorsclub.com/show...ack-Pics/page2?p=449465&viewfull=1#post449465

Another point is that Tesla says the voltage window for the Roadster was 3.0V to 4.15V:
http://www.teslamotors.com/blog/bit-about-batterieso

Not a whole lot of reason to change from this.

So, Tesla quotes two different minimum cell voltages. Since I have physically ran down to 2.6v sag (2.8v OCV) per cell in the Rav4 EV, I'm going to suggest that the 3v threshold from the Roadster no longer applies. The prototype Rav4 EV used Roadster 2200ma cells, but the production Rav4 EV version did not.

Again, 2.5v minimum per cell appears to be the current threshold. Yes, that is well past the voltage knee, and the voltage drops fast past about 3.1v. I'm also not suggesting that's a good idea, or not. I'm merely following the data.

As was properly pointed out earlier, there is a big difference between OCV versus under charge / regen and / or discharge.

 

[yobigd20]

Thanks for posting this! Model S has 3.1AH cells. Based on this published graph, the cells have amazing life even after 2500 cycles (> 70% capacity remaining).

I currently have 36k mi on my 60 kWh battery. That's ~200 cycles. I get 198 on a range charge. It was 208 mi when brand new. So, I've experienced ~5% degradation. I'm doing better than the graph. I'm sure that's because I don't do deep discharges and don't do many range charges.

This graph implies my battery will still have over 70% capacity after 300k miles. That's insane.

FWIW I am at 54k miles now. outside of my first 4 or 5 months where range charge went from 265 to 259 over around -14k miles, the last 13 months doing another 40k miles I have seen exactly 0% degradation and my range charge is still 259.

Given this rate, at 300k miles I fully expect to have way more than 70% capacity.

 

[NigelM]

FWIW I am at 54k miles now. outside of my first 4 or 5 months where range charge went from 265 to 259 over around -14k miles, the last 13 months doing another 40k miles I have seen exactly 0% degradation and my range charge is still 259.

Although there's been a few tweaks of the SW calculations in that time.

 

[scaesare]

I'm trying to see if we have consensus on thinks like voltages within the pack for various states, cell capacity, balancing schemes, etc... However, there's a number of variables and assumptions thus far in the thread:

- Observed pack Vmax ~404V (= Cell vMax 4.21), however that was while supercharging
- Suggested cell Vmax of either 4.1, 4.15, or 4.2V
- Suggestion the Vmax value of 4.15 is at only 90% charge, whereas others have suggested that's at 100%
- Suggested nominal cell voltages of either 3.6 or 3.7 volts
- Suggested Vmin of 3.6, 3.2, 3.0, or 2.5 volts
- Suggested cell capacity of either 3.1, 3.3, or 3.4Ah
- Pack has active balancing via variable impedance vs. bleed-off resistors for charge-only balancing
- Some odd observations with the pack at 352V and trying reconcile rated miles reported vs. energy consumed during a charge

Most of the observations wk057 have posted are direct measurements and documented, however they are OUTSIDE of the car, so we aren't sure what the normal voltage extents would be.

Several observations folks have made are from the console when supercharging, so those were NOT at rest voltages.

A number of voltages tossed around are from the Roadster, which is understood to have different cell chemistry.

A number of capacity, voltage, and chemistry claims are made with no substantiation or source cited.


So I'm hoping to determine what we actually KNOW about the pack as it operates WITHIN the Model S. It seems like that really is only rated pack capacity (~85kWh), logical pack layout (74P/962), and max supercharging voltage (~404). There's some good evidence for bleed resistors, but that doesn't preclude some other balancing capability as well.

The rest appears to be conjecture.

It would be nice to know at some point what the reported SoC-to-voltage correlation is (altho firmware revision dependent it seems), and if it's linear.

For example: It's been suggested that 90%=4.15V and 100%=4.20V... if that's true, there's either some significant non-linearity or the V Vmin is something like only 3.7V (which doesn't seem likely).

Unfortunately, it seems like the answer to most of these questions will require either making pack measurements while connected to the car somehow, or access to some other authoritative data (or something like the BMS service screens).

How did the roadster guys get some of this data?

 

[tom66]

Just a note: There won't be a direct SOC-to-voltage translation. SOC will be calculated from an energy balance equation, measuring total pack energy output, and decreasing a counter as the pack energy is decreased. Possibly the below zero mile shut down will occur at a specific cell voltage, but this will not be used to determine capacity.

 

[djp]

How did the roadster guys get some of this data?

Good summary.

Back in the Roadster days, Tesla wrote a technical blog that covered behind-the-scenes details on the car's design and operation (I miss those!). The Roadster also dumps a set of log files every time a properly formatted USB stick is inserted in the port, which members on this forum have reverse engineered. Parsers and graphing tools are available on several threads. A few members also have access to the service screens, which have real-time brick voltages and SOCs.

Model S is a lot more locked down by comparison, but it's just a matter of time before the collective intelligence of this forum figures it out.

 

[scaesare]

Just a note: There won't be a direct SOC-to-voltage translation. SOC will be calculated from an energy balance equation, measuring total pack energy output, and decreasing a counter as the pack energy is decreased. Possibly the below zero mile shut down will occur at a specific cell voltage, but this will not be used to determine capacity.

Ah, good points. I'd also assume that local variables such as temperature might be included.

Nonetheless, I'd expect that some things, such as a full 10% difference in reported charge from 90 to 100% correlating to only a 50mV voltage differential might set off some alarms as being unlikely.

Nonetheless, it would be interesting to see if we can at least document the extents as within the car's normal operational envelope.

- - - Updated - - -

Good summary.

Back in the Roadster days, Tesla wrote a technical blog that covered behind-the-scenes details on the car's design and operation (I miss those!). The Roadster also dumps a set of log files every time a properly formatted USB stick is inserted in the port, which members on this forum have reverse engineered. Parsers and graphing tools are available on several threads. A few members also have access to the service screens, which have real-time brick voltages and SOCs.

Model S is a lot more locked down by comparison, but it's just a matter of time before the collective intelligence of this forum figures it out.

I should have said: "How did YOU Roadster guys get this data!"

Seriously, a lot of trailblazing was done there, and even if not all the specs area directly transferable, a number of the principles and methods no doubt are.

I agree that it's threads like this one that will help document what the S and it's pack does. My post was not intended to throw any cold water on this effort, quite the contrary: I applaud this thread and the work wk057 is doing.

What I'm really hoping for is clarification from folks so we can determine what is known as fact, either via direct observation/measurement, or from some authoritative source (manual, part # spec, etc..). For the rest of the input, I think it would be helpful to qualify it as a informed speculation (and upon which basis), and/or extrapolation from elsewhere (Roadster data, similar cell specs, etc...)

Cool stuff...

 

[Banahogg]

For AC charging, the REST API reports the battery current as well as the charger power. If we assume the battery current is measured rather than computed internally and ignoring losses (a less reasonable assumption), then we can get the pack voltage as (charger_voltage * charger_actual_current) / battery_current. Running that through my database and throwing out derived voltages outside [300V,400V], I get what looks like a wide straight line on the graph. The spreadsheet puts the fit line at about 313.6V + 0.685*battery_level (SOC percent from REST) and from eyeballing it, the band looks to be a pretty constant 30V wide at a given SOC.

Pretty indirect, but seems to give plausible values and is easy for anyone logging the REST values from their cars to replicate.

 

[scaesare]

It appears that the SOC% is dependent on 100% being 4.15v and balanced, as you suggest. Here a shot of 99.0% with the highest and lowest cell sub-group voltages of 4130mv and 4134mv. The three zeros are net power used in kWh at startup:

Tony, this is the best evidence yet I've see of ~4.15 being Vmax for 100%. What's that a screenshot of?