What a miserable saga. From a sanity challenging 1100 miles trip to Houston, to multiple equipment failures, a notable brain fart and 1 out of spec cell.
As described in the post Sourcing LiFePo4 Cells from China the cells came from Docan Power Tech, direct from China. I inspected each one at the warehouse in Houston. They came well protected in 2-cell packs, and none displayed any sign of swelling, prior use or improper handling. They all measured between 3.292V and 3.297V, and 0.165mΩ to 0.18mΩ. Within expected range, but that doesn't tell us much though.
Interestingly, there wasn't any safety or regulation handling mark on the boxes, contrary to shipping rules and regulations for Lithium batteries. That's apparently one of the contributors to the lower price from sellers direct from China.
In the end each cell required about 100AH to be topped off. So they were shipped at a high-ish ~65% SOC, instead of 30-50% per regulations or for shelf life.
So far so good. Having survived 4 days of a soul crushing succession of cotton / cow / corn fields during the trip to Houston, it was finally time to do a capacity test on these babies. The first 10 cells did not look good: 274, 264, 276, 278, 280, 277, 274, 277, 273, 276AH. What in tarnation ???
So started a wild goose chase...
First, the Zketech EBC-A40L cell tester / charger gave up the ghost while testing the 2nd cell after emitting a sparking sound. The whole unit was replaced, but after 12 cells the internal power supply of the new unit released magic smoke. Fortunately, it was promptly replaced by the TKingok seller.
Bottom lime: it is a very affordable and smart $280 40A charger + tester. Its data plotting & cycling software is great. Nothing comes close to it at that price. But its generic off-the-shelf internal power supply is not reliable. More notes on the EBC-A40L at the end(*).
Back to the underperforming cells, one hypothesis was that the tester was not reporting the correct current value. Indeed, it always displays 40.00A with no wiggling of the last digits. Clearly not a measured value. Its current data log file was no help either, as it also seemingly just reports the useless 40A target value:
So, a Bside ACM92 6.000 count DC current clamp meter was procured. It was checked with a lab power supply at 20A against an Extech multimeter, and at 10A against an Agilent U1242 10.000 count multimeter (+/-0.6% +5 accuracy @ 10A scale). A 4 loops test was also conducted to scale to 40A on the clamp meter while maxed out at 10A on the U1242A. On average it showed a ~3% overreporting, which is borderline compared to its +/-2.5% +10 @ 60A spec.
Time to check if the EBC-A40L indeed draws 40A or not. Well, most of time the clamp meter did read 39-41A. So, in the ballpark, but not good news. Indeed, if the latter truly overreports by ~3% then the cell capacities might actually be even lower than reported by the tester. I.e. 264Ah would in fact be 256AH in reality !
That's when I started drinking...
But, aha, a handful of times, during a 7 hours discharge test, the clamp meter would display 36A ! And, over 4 days, it also showed 44A a few times. Busted ! Now the EBC-A40L was suspected of misbehaving and erratically drawing unexpected current values.
Those spurious current events definitely had to be investigated... How long do they last ? What is their actual value ? Why is this happening to me ?
Being absolutely fed up with taking manual measurements every 10 minutes for days, it was time to stop drinking, enter full debug mode and properly plot the current. A 100A / 100mV shunt did the job, connected to the scope. Alas, the DC ground from the EBC-A40L is not connected to chassis ground so the scope was measuring more EMI noise than it was current...
Oh how I miss the professional multi-GHz Agilent & Tektronix scopes from work, and their superb active differential probes. But they cost as much as a Tesla Semi Truck, so not in my budget 😂
So, onward to procuring a battery powered datalogger, independent from the AC supply, to serve as a pseudo differential probe across the shunt. Wanting to measure down to ~0.25A steps via the 100A / 100mV shunt, meant a 0.25mV resolution. Assuming a 2.5V data logger input, that's 1 step in 10.000. Which is between 13 and 14-bit resolution.
Unfortunately, most affordable voltage data loggers (like EL-USB-3) can not deliver that. Enter the very versatile and well designed Onset Hobo UX120-006M 4 channel analog data logger. It claims 40µV resolution @ 2.5V scale, i.e. 16-bit resolution. And, in my experience, it actually delivers at least 14-bit. Got a great $100 deal, with 2 cables, on ebay:
Checking the logger + shunt against the multimeters, like described above, showed a ~0.85% underreporting. At this point, finally, a real discharge current plot could be captured, using a 1 second sampling rate:
Max current measured: 39.86, Min: 39.79, Avg: 39.83. Corrected +0.85% numbers: Max = 40.2A, Min = 39.82A, Avg = 40.17A.
So my EBC-A40L actually underreports the drawn current by 40.17 / 40 = 0.42%. Good news, it means that, ideally, its AH capacity measurements should be multiplied by 1.0042 to get the actual capacity.
And, tellingly, during some of these recorded tests the clamp meter did show a handful of 36A measurements...
Conclusion: the EBC-A40L is properly drawing current and calculating the AH, while the clamp meter is a piece of crap that sometimes lies by -10% or +10%,, so I spent 3 weeks and money on a red herring !
Baffled by this ghost goose chase, I then spent an hour obsessively turning on / off the clamp meter, zeroing it, changing scale, varying button push & release timings, only to find out that in about 1 take out of ~20 it shows ~36A, and ~44A in 1 out of ~60 takes. I.e. +/-10% erratic behavior. Darn frigging cheap stuff from Amazon, and penny pinching me... 😡
So, now certain that the cells were NOT delivering the promised 280AH, I contacted Jenny Wu, the seller of the cells from Docan Power Technology. Well, she spotted the low ambient temp during my tests right away. Being Winter season here in Colorado, my workbench room oscillates between 50°F and 60°F (10-16°C). Brain fart on my part, should have thought of it.
280Ah is rated at 25°C (77°F), and needs to be derated down several % at lower Temp. Generic example:
After holding the ambient Temp at 24°C (75°F), the 273AH cell and 2 new cells did reach 280AH ! So I am now confident that all the cells that tested above 273AH before will reach 280AH. Thank lithium heavens !
However, the outlier cell that was reading 264AH only reached 274AH at 24°C, which I am now discussing with Jenny Wu. Its fate will be covered in a coming post.
Update And, unfortunately, it still leaves a number of cells a smidge under 280AH, as described in detail in the next post: Are the 32 LiFePo4 Cells from China any good ?
And that, folks, is the end of this initial capacity testing saga. Time to move on to the remaining cells. What a massive waste of time and resources. So many lessons here... I hope that at least some of it will be of help for someone out there ?
Shame on me, avoid my mistakes !
Comments
Post a Comment