First off:
For the time being, I don't recommend discharging modules down to 0.0 volts/cell. Doing so will permanently render the module unusable, whereas we are still gathering information on how to proceed. If you feel the need to discharge modules before we have final guidance, my recommendation is to discharge to 3.4 V/cell, which is ~10% SoC. Storing modules at 10% SoC is much safer than fully charged modules, and is also recoverable (whereas discharging to 0.0 volts is not). An even easier option is to simply place the modules outside until we have further guidance.
...
Over the past few days I inspected QTY220 modules for visual and electrical failure signs.
A: Recall results (from the modules I've inspected over the past few days):
-QTY195 MP0217NS002A modules, which are on the recall list, and;
-QTY25 MP0214NS002B modules, which are not recalled.
As I've mentioned previously, I'm not convinced that whether or not a module is on the recall list is actually relevant. Given that Samsung doesn't understand the root cause, it's illogical to conclude that only those LX68 modules on the list could contain the (unknown) failure mechanism. I suspect Samsung's recall list only contains batch numbers where a failure has actually occurred.
Without a root cause, I suspect that all LX68 modules are equally as likely to contain the (unknown) failure mechanism. Therefore, I think we should treat all LX68 modules as suspect, and adjust our risk tolerance levels the same amount, whether or not your specific modules are recalled. Put more simply, I propose we treat ALL LX68 modules as if they could contain the (unknown) defect.
B: Voltage results¹ (from the modules I've inspected over the past few days):
-QTY215 modules have less than 8 mV delta between highest and lowest cell voltage, and;
-QTY5 modules have higher cell delta. Specifically: 11 mV, 14 mV, 19 mV, 24 mV, 25 mV.
These five modules have higher internal cell discharge rates than all others. The worst performing module – with 25 mV delta between cells – has lost ~0.5% SoC discharge over several months, which is ~235 mAh. Therefore, the individual cell that's 25 mV less than the other cells inside that module has ~100 uA internal cell leakage. That leakage is ~17x higher than the QTY2635 other cells I tested.
However, these QTY5 modules with 'higher' self discharge are still well within Samsung's specified 3%(max)/month
² self-discharge rate (versus the 0.5% worst case I observed after several months). Since we're comfortably within Samsung's specifications, I don't know whether these outlier voltages are noteworthy.
I've placed these QTY5 modules into my daily driver for further testing. LiBCM can certainly detect 100 uA leakage by logging per-cell BMS discharge time over several days... but again, I'm not certain the 100 uA internal cell leakage I've observed on the worst performing module actually indicates the separator contains the "micro stress" defects mentioned in the recall. In other words, it's unclear whether or not the voltage deviation on these five modules is an indicator that the separator is failing.
C: Deflection results (from the modules I've inspected over the past few days):
In general, most of the modules I inspected had visible deflection. Deflection wasn't appreciably different between recalled and non-recalled modules. None of the modules I inspected had more than 1 mm of deflection.
Here are two modules from my car that are each approaching 1 mm deflection (right click the image for a much larger version):
This picture is taken in such a way as to minimize optical distortions. Specifically, the camera:
-is placed far from the modules, and;
-is optically zoomed in as much as possible, and;
-optics have minimal curvature, and the modules are well within frame to further reduce distortion, and;
-has diffuse lighting entirely surrounding the sensor, and;
-is mounted orthogonal to the battery top surface, and;
-the (known straight and parallel edged) ruler is verified to actually be straight and parallel in each digitized image.
Samsung specifies the "maximum allowed elongation at end of life" is 5 mm
²:
Unfortunately, Samsung doesn't provide any guidance on what the 'elongation' measurement is. Therefore, the safest assumption is that each aluminum end plate can deflect up to 2.5 mm before the module no longer meets Samsung's specification. I can't conceive any other worst case interpretation, but am open to suggestions. In other words, if your modules are bowing less than 2.5 mm, it would appear they are within Samsung's specification.
Honestly, prior to obtaining this official Samsung specification, I was going to set the limit at 1.5 mm deflection... maybe even 1.0 mm. 2.5 mm
feels like too much, but that appears to be the specified limit (unless I'm interpreting the spec sheet incorrectly).
I am still performing a test to determine whether the deflection magnitude appreciably changes at different SoC percentages. (edit: updates below).
FYI: One additional note about the end plates:
Samsung's mechanical drawings
² clearly show why there appears to be a small protrusion behind the insulator directly abutting each aluminum end plate:
The two ~20 mm half-circular cutouts on the 'Insulation Membrane' strongly suggests that Samsung's design allows for the aluminum 'Plate End Support' to deflect. These cutouts are almost certainly designed to reduce localized pressure that would otherwise occur as the 'Plate End Support' flexes. Without these cutouts, localized pressure would occur due to the two slots at the top of the 'Plate End Support'. These cutout in the 'Plate End Support' exist to provide clearance for the plastic tabs that secure the plastic high voltage insulator (not shown).
...
Overall, my days-long analysis of QTY220 modules didn't yield any concrete results. The QTY5 deltaV(cell) outlier modules warrant additional analysis, but whether or not they are actually outliers remains unknown.
...
Next I'm going to theorize software algorithms to test for the defect. As I've noted previously, Samsung's first recall remedy is to install software that can detect defective cells prior to catastrophic failure. Right now I have two leading theories:
1: Test for abnormal internal cell leakage, using per-cell balancing timer, and;
2: Test for abnormal voltage recovery after heavy acceleration (as Ford is doing in their test).
...
Footnotes:
¹module test history prior to storage (this happened months ago):
-charged to ~85% SoC, then;
-all cells in each module balanced to within 1 mV, then;
-QTY4 or QTY5 modules (enough for one car) discharged into a 14 ohm load (~3300 watts), then;
-discharge continued until SoC reached ~27%, then;
-each pack was verified to have less than 8 mV cell delta, then;
-modules stored at ~27% SoC (~3.54 volts/cell).
²Don't ask me how I got this information. Samsung is notoriously secretive about their specifications. Fortunately, an anonymous source helped me out with this detail. To protect their privacy, I will not discuss how I procured this information. If moderators require proof, please send a PM and I will discuss with John Doe which information they are willing to share. Note that the graphics shown are slightly modified to obfuscate the confidential data sheet I retrieved them from.
www.insightcentral.net
