5
\$\begingroup\$

I recently acquired 50 used li-ion cells (18650). I'd like to efficiently determine which cells are good matches (i.e. which cells have similar: capacity, charge times, & discharge times) so that I can put them into battery packs that will perform optimally (e.g. they don't punk-out early because one or more cells discharge too fast or over-charge or over-heat as slower charging cells lag near the end of charging).

What would you suggest would be the best way to test multiple cells to get such information?

\$\endgroup\$
  • 1
    \$\begingroup\$ This is a big investment of time for something that might not work very well. But, if the cells are the same make and model, the main thing is to match them by capacity. So you need to measure each cell's capacity, equalize them to the same state of charge, and put them in a pack. But if the cells are used, their capacities may diverge after a relatively short number of cycles. \$\endgroup\$ – mkeith May 14 '17 at 16:50
  • \$\begingroup\$ Putting used cells in a battery pack is similar to putting used bearings in a motor. A lot of work for something with an unknown amount of life left in it. \$\endgroup\$ – mkeith May 14 '17 at 16:51
  • \$\begingroup\$ @mkeith - Indeed, however, the price is hard beat (free) & the applications are not critical, so they will likely suffice. I'm just hoping to find a way to efficiently test them so I can match them well. Got any ideas? I'd hate to have to test them individually using an iMax B6... \$\endgroup\$ – zeffur May 14 '17 at 17:01
  • \$\begingroup\$ that's what Tesla module recyclers do on individual cells . Some just charge all then test Voc on each cell and record V to 3 decimal places and , label and bin. New they are matched 0.1%, old becomes 1% mismatch then 10% weakest becomes 1st one dead \$\endgroup\$ – Sunnyskyguy EE75 May 14 '17 at 17:03
  • \$\begingroup\$ use slow bulk charge to prevent overheat weak cells \$\endgroup\$ – Sunnyskyguy EE75 May 14 '17 at 17:06
2
\$\begingroup\$

Just like car batteries, use Cranking amp with a constant voltage sink pulse (7.5V for 12V car CCA and ( 3.0V from 3.7 for LiPo) and open cell voltage Voc after a xx seconds of fully charged cells.

This can be done with a pulse with sample and hold on current sense using a diode driven PNP darlington TO220 on a heat sink such that Ve is 3.0V at 10 A more or less for Isense using 0.01 Ohm non-inductive rated power resistor on collector to ground for 100mV sense at 10V. Pulse width = TBD

Then record I load and Voc on label and sort into bins of 1% or less .

Batteries and load transistor should not get hot in <<1/10 second with 30 Watts more or less being transferred.

But the iMax B6 is cheap and does similar thing.

more details

http://www.mpoweruk.com/chargers.htm

\$\endgroup\$
  • \$\begingroup\$ Can you provide me with a url to a circuit that does essentially what you've described? \$\endgroup\$ – zeffur May 14 '17 at 17:13
  • 1
    \$\begingroup\$ Well, if it is anomalously high, there is probably something wrong with the cell. I guess it is mostly an issue for rapid discharge applications. If you are not discharging the cells rapidly, it may not matter as much. It is definitely not true that resistance of LiIon cells remains flat throughout service life. It goes up as the cell ages. However, it may depend somewhat on history of the cell. How it is charged, what temperatures, etc. Rapid charging at low temperatures is very bad for a cell. \$\endgroup\$ – mkeith May 14 '17 at 17:26
  • 1
    \$\begingroup\$ It varies a lot by cell. Cells I am working with right now are high-discharge rate cells. The datasheet says 4A max charge rate (for a 2.5Ah cell). Max discharge is 25A (10C). These are used in vehicles and power tools. But laptop cells are not designed for such high charge and discharge rates. I think you will be safe for sure at 0.5C, not matter what type of cell you have. If you don't know what C is, just try charging at 1A for an 18650. V max is 4.2 for all Lithium Ion batteries I have ever seen. If you reduce Vmax, you will get longer life but less capacity. \$\endgroup\$ – mkeith May 14 '17 at 17:51
  • 1
    \$\begingroup\$ yes 4.1 gives almost 90% capacity but almost twice the life and using only 50% DoD gives almost the same life cycle improvement in Ah cumulative lifespan. Arhennius Law applies for battery MTBF too. Although ESR and thus CCA is not the best indicator, it is valid , except puncture flaws from microscopic insulation breakdown from contaminated conductive dust can cause infant mortality in the dielectric. \$\endgroup\$ – Sunnyskyguy EE75 May 14 '17 at 17:58
  • 1
    \$\begingroup\$ Pulse charging is like switched current pulse from SMPS. so it can be a BIC lighter at low currents or a "Blow torch" solution at lower duty cycle for deoxiding at risk of low percentile failure modes from excess H field force stress on conductor walls. The main avoidance should be cell junction temp rise but efficiency drops with higher current. overvoltage increases capacity but sharply reduces life. but main sales game is Ah per cycle while ignoring N cycles affected by C rates and T rise \$\endgroup\$ – Sunnyskyguy EE75 May 14 '17 at 18:25
4
\$\begingroup\$

This is obviously a lot to produce but it's how we once solved the problem you've described...

I helped another engineer design and develop a cell matcher, which he went on to sell. This was 25 years ago with NiCd cells used for remote-control vehicles. Our system profiled cells using a constant-current discharger and an ADC in a data logger (actually a home computer back then). Cells were selected and matched by examining the discharge graph and key parameters for each cell. The latter were figures such as time to Vout dropping to something like 80%, time to fall from 80% to 20% and a few others.

The discharge circuit was simple enough: a current regulator to ground with an op-amp controlling it.

schematic

simulate this circuit – Schematic created using CircuitLab

Our discharger actually had about 5 x TIP31A connected in parallel to dissipate the heat.

The whole system of computer interface and discharger worked very well, he sold enough of them. The customers were RC enthusiasts who already had their own chargers.

\$\endgroup\$
  • \$\begingroup\$ Nice! Couple your system with Tony Stewart's load tester idea (below) & a self-discharge monitoring/logger & you'd have quite a nice system! \$\endgroup\$ – zeffur May 14 '17 at 17:35
  • \$\begingroup\$ yes just need a regulated voltage pulse here but a constant voltage sink measuring Current , I believe is better to prevent overstress . But I agree, above works. But CC measures ESR directly from current via V drop ESR=delta V*Icc , while Pulse mode reduced Pd wasted but must allow for //RC time constant of absorption C different from main C of battery chemistry. (minor memory effects.) \$\endgroup\$ – Sunnyskyguy EE75 May 14 '17 at 18:10
  • \$\begingroup\$ @TonyStewart.EEsince'75 - re: "...different from main C of battery chemistry." Different in which way/s? \$\endgroup\$ – zeffur May 14 '17 at 18:12
  • 1
    \$\begingroup\$ Dielectric Absortion or the Double Layer charge effect in all Caps and batteries account for memory effects in ceramic, supercaps, lead acid, NiCad and to a small,extent LiPo's although common wisdom is LiPo's don't have memory but if you pulse charge there are two series ESR*C time constants in parallel and also the rest. NPO ceramics and plastic film do not have this. That's why X7R cap is worst choices for a high speed S&H cap. \$\endgroup\$ – Sunnyskyguy EE75 May 14 '17 at 18:19
  • 1
    \$\begingroup\$ hysteresis effect in Vbat is due to above effects. mpoweruk.com/chargers.htm \$\endgroup\$ – Sunnyskyguy EE75 May 14 '17 at 18:35
3
\$\begingroup\$

Analyzing the charge and discharge curve characteristics is the most reliable but also the most time consuming.

Measuring the internal resistance of an Li-ion offers little value as Li-ion keep a low and consistent resistance throughout their life cycle. It may identify the cells that have little or no life left.

Measuring the impedance curve between 1hz and 10hz is a reliable quick method but a bit complex.

A 3.7V, 18650 cell is an Li-manganese which is used in power tools because it can withstand short heavy discharge rates. Some cells can withstand up to a 5 second 30C discharge. This lends itself to testing its capacity by measuring the electrochemical dynamic response with the Pulse Discharge Test method. This measures the ion flow between the positive and negative plates.

The Pulse-Discharge method is the quick and somewhat simple method.

It is important the cell not be fully charged and ideally at about a 40% charge level.

A resistive load with between .1C and 2C discharge is applied for between 1 and 6 seconds, not to exceed 6 seconds.


A strong cell recovers quickly, weaker cells are slower getting back to the pre-pulse voltage.

pulse discharge waveforms


  1. Measure voltage before discharge pulse.
  2. Measure discharge voltage within one second after falling edge.
  3. Measure open circuit recovery voltage within one second after the rising edge.

This is a patented process Patent Number 7,622,929
The patent has much more detailed information.

\$\endgroup\$
  • \$\begingroup\$ Is this "A strong cell recovers quickly, weaker cells are slower getting back to the pre-pulse voltage." true for li-ion cells? I've noticed the exact opposite effect in lead-acid batteries (bad batts recover to near pre-test voltage quick & good batts recovery much more slowly). \$\endgroup\$ – zeffur May 14 '17 at 22:07
  • 1
    \$\begingroup\$ The pulse discharge is used only for capacity testing of Li-ion as it measures the ion flow between the positive and negative plates. Also it is not used on high capacity batteries like those to power vehicles. Works best with less than 1500mAh batteries. \$\endgroup\$ – Misunderstood May 14 '17 at 23:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.