What EIS actually tells you that a DCIR measurement can't
The single number that hides the full story
Direct-current internal resistance is the measurement most labs still lean on. Apply a current pulse, watch the voltage drop, divide one by the other, and you have a number in milliohms. It's fast, it's cheap, and for decades it was good enough to screen bad cells from good ones.
The problem is that a lithium-ion cell is not a resistor. It is an electrochemical system with at least three distinct loss mechanisms — ohmic resistance in the electrolyte and current collectors, charge-transfer kinetics at the electrode interfaces, and mass-transport limitations deep inside the active material. DCIR lumps all three into one scalar and throws the structure away.
If the single number is within spec, you call the cell good. If it drifts, you don't know why. That matters more every year, as cells get denser, warranties get longer, and second-life applications start demanding real-time state-of-health reporting that DCIR was never designed to deliver.
Where electrochemical impedance spectroscopy comes in
Electrochemical impedance spectroscopy (EIS) does the same measurement — voltage divided by current — but across a spectrum of frequencies instead of a single DC pulse. Typically from around 10 kHz at the fast end down to 10 mHz or even 1 mHz at the slow end.
At high frequency, you see the ohmic resistance of the electrolyte, separator, and contacts. In the mid-frequency band, usually a few hundred hertz down to a few hertz, you see the charge-transfer kinetics as a semicircle on a Nyquist plot — its diameter tracks the state of the SEI layer and the catalytic activity of the electrode. Down at the low-frequency tail, below a hertz or so, a 45-degree line appears that reports the Warburg diffusion impedance — how easily lithium ions move through the bulk of the active material.
Three different loss mechanisms, three different regions of the spectrum, one test.
What this tells you in practice
A manufacturing defect in the current collector or a dry joint on a tab shows up almost immediately as raised ohmic resistance — a shift to the right on the Nyquist plot, visible at the highest frequencies.
A cell that has been cycled hard or stored at high state of charge will show a growing mid-frequency semicircle, because the SEI is thickening and charge transfer is getting harder. DCIR will show a modest rise, but EIS shows exactly which part of the cell is ageing. That is the difference between screening and diagnosis.
A cell with lithium plating damage, or one that has lost mass-transport capability because of electrolyte decomposition, will show the 45-degree Warburg tail steepening and moving. DCIR cannot see this at all, because the measurement is over before the slow processes have had a chance to respond.
For anyone doing incoming inspection on cells, running accelerated ageing tests, validating a second-life application, or trying to correlate manufacturing variation to performance, this frequency-resolved view is not a luxury. It is the minimum data you need to make a decision.
The old objection: EIS is too slow
The reason EIS has historically lived in R&D rather than production QA is simple. Traditional frequency-sweep EIS measures one frequency at a time. A proper sweep from 10 kHz to 10 mHz takes several minutes per cell, and at the low-frequency end a single point can take a full minute on its own. Multiply that by a production line or a battery pack with hundreds of cells and the maths does not work.
This is why most labs settled for DCIR. Not because it was better — because it fit the clock.
Multi-sine stimulus changes the maths
The EA-BIM 20005 — the 20-channel battery impedance meter we stock from Elektro-Automatik, part of the Tektronix group — uses a multi-sine stimulus rather than a single-frequency sweep. It excites the cell at many frequencies simultaneously and separates the response in software.
The practical effect is that a full spectrum from 1 mHz to 10 kHz comes back in seconds, not minutes. Across 20 channels at once. Each channel gets a paired 4-wire PT100 temperature measurement so you can cross-correlate impedance with thermal behaviour on the same timebase. The integrated power stage can charge or discharge at ±1 A, so you can take measurements at controlled states of charge without moving the cell to a separate cycler.
Throughput at that level is what moves EIS out of the R&D bench and into cell QA, battery module validation, and state-of-health monitoring for second-life applications.
What a good 21700 spectrum looks like
On a healthy cylindrical 21700 lithium-ion cell at roughly 50% state of charge and 25 °C, you typically see:
- A high-frequency ohmic intercept in the low single-digit milliohm range
- A single, tight mid-frequency semicircle with a diameter that reflects the charge-transfer resistance, usually under 10 mΩ for a fresh cell
- A clean Warburg tail at 45 degrees starting somewhere below 1 Hz
On a cell with 300 cycles of hard fast-charge abuse, the ohmic intercept creeps up, the semicircle roughly doubles, and the Warburg region becomes noisier and moves away from the ideal 45 degrees. On a cell with incipient lithium plating, you often see a second, smaller semicircle appearing — a new loss mechanism that was not present in the healthy cell.
None of that is visible from a DCIR number. All of it changes how you would treat that cell, that batch, or that supplier.
Come and see it on your own cell
If you want to see what your own 21700 cell looks like on a Nyquist plot, we are at Battery Tech Expo on Thursday 23 April at The Wing, Silverstone. Stand A35.
Bring a cell. We will connect it to the EA-BIM 20005, run a full spectrum across 1 mHz to 10 kHz, and walk through what the curve is telling you. You will leave with the plot for your own cell and a concrete sense of whether EIS belongs in your own test regime.
Slots are limited and free. Book one at https://www.tek.com/en/lp/event/em-ea-auto-battery-tech-expo-with-partner-caltest-lp and we will have the instrument ready when you arrive.
