Samsung INR18650-25S batteries, or what replaces them

About a year ago, information about a new battery model, the INR18650-25S from Samsung, started appearing more and more often. Naturally, it caught my interest, so I bought a couple to test. Today I'll tell you how they performed.


For quite a long time, Samsung had a very popular model, the INR18650-25R. I've already reviewed it, and I think many people who have worked on power tool conversions or used e-cigarettes know it well.
This model is considered a high-current cell, although I personally put it in the mid-range category, since its specified continuous current is 20 A, while there are 18650 cells rated for up to 25–30 A.
But Samsung decided to give the old-timer a little "upgrade" and released a new model, the INR18650-25S. As the name suggests, the only difference in the model number is the last letter. That's right: the capacity is unchanged at 2500 mAh, which is why the new model is positioned as a replacement for the 25R.

The link in the title goes to the Nkon store, where I originally planned to buy the new cells, but they sold out, so I ended up buying them offline. The situation is a little paradoxical: it turns out that right now it's easier to buy the 25S offline than from Nkon or AliExpress. That's partly why I'm not linking to any particular store. You can find them at Keeppower, batareyki.com.ua, and many others, and even the price is comparable to Nkon's.
Since this is a new model, at the moment that means:
1. The batteries are probably fairly fresh, since they only recently went on sale.
2. The chance of getting counterfeits is low, since they haven't had time to become popular yet. But that's a separate topic.


Since the package was traveling within the same city, the packaging was very simple: a couple of pieces of cardboard and some tape.


I found a datasheet online, which lists both the specified parameters and the conditions under which they were measured. I found two datasheets: the one linked above, from late 2017, and a slightly older one from the middle of the same year. The first revision dates back to late 2016, so this "new model" is actually already 4 years old.
The same datasheet shows that the battery has two upper current limits:
1. 25 A without temperature monitoring
2. 35 A, provided the battery is shut off if its temperature exceeds 80 degrees.

Sometimes seller pages describe these limits somewhat inaccurately:

and you can see that it heats up to 80 degrees at 35 A, but I’ll explain a possible reason for this description at the end of the review.


Interestingly, the first datasheet turned up on the... VapCell website. Google found it, since there are no direct links to it on the site itself.
While searching, I came across the VapCell H25 battery, with “25S” in parentheses. I assume these are repackaged Samsung 25S cells, although on AliExpress they list them as VTC5A.


The temperature limits are the same as for the 25R: it delivers its full capacity at 25 degrees and above. At lower temperatures, the percentage capacity reduction at a 10 A discharge current is shown below. If the batteries warm up during use, they’ll of course be able to deliver more than the values shown.


They look like ordinary, familiar Samsung cells. Interestingly, Samsung seems to be the only company that still doesn’t use a matrix code to verify authenticity. The markings themselves are also a little faint in places.
The positive terminal is flat and doesn’t protrude beyond the casing. Under the heat-shrink wrap, you can make out markings such as L5J4 44131 and F2EA, identical on both cells.

According to the SDI code marking 2J43, the batteries were manufactured:
J - 2019
4 - April
3 - third week, from the 15th to the 21st.
I wouldn’t call them exactly “fresh off the line.” I assume the sellers bought them as soon as they became available, but haven’t sold them yet.



The dimensions match those given in the datasheet: the length is just under 65 mm, and the diameter just over 35 mm.


A conditioning cycle at a 1 A discharge current yielded 2507 mAh for the first cell and 2544 mAh for the second. That’s pretty good, both in terms of capacity and how closely the cells match.
The batteries were almost discharged to begin with; the first charge took 2000–2050 mAh each, meaning they had about 20% charge left.


The remaining tests were carried out using two battery testers: the EBC-A20 (up to and including 20 A) and the EBC-A40L (above 20 A).

The cells support two charging modes: normal, at 1.25 A (0.5C) with cutoff at 125 mA, and fast, at 4 A with cutoff at 100 mA. I’m more used to high-current models being charged at 1C, but the manufacturer specifies 0.5C, so that’s what we’ll use.

At 0.5C, charging from fully discharged at 0.2C took 2 hours and 7 minutes—not bad at all.


At a 4 A charging current, charging took 51 minutes. The CV phase lasted noticeably longer, and the battery temperature was barely above room temperature.


The tests were run at nine rates: 0.2, 0.5, 1, 2, 4, 6, 8, 10 and 12C, where C = 2.5 A. It took quite a while, but I wanted to gather as much data as possible in one go.
The cells deliver their rated 2500 mAh only just—though still within the datasheet specifications.


Temperature at 10, 15 and 25 A compared with the 25R (below). You can see that:
1. The new cell really does run cooler, so it can handle a higher load current.
2. The temperature had already exceeded 80 degrees at 20 A, even though the manufacturer specifies that as the maximum.

Yes, the maximum temperature for the 25S is 80 degrees; for the 25R it’s slightly lower, at 75 degrees—another advantage for the new cell.

In reality, the difference is smaller, since the air temperature during the 25S tests was 23 degrees, while during the 25R tests it was around 27–28 degrees—in other words, 4–5 degrees higher. That's why the difference becomes more noticeable as the current increases. At 10 and 15 A, the temperature is almost the same.



At 25 and 30 A, though, things were already worse: essentially, every additional 5 A of load current added about 10 degrees, bringing the temperature to over 90 and 100 degrees, respectively.

I thought about testing at 35 A too, but didn't want to risk the batteries. Still, I can say that if you don't stop the discharge, the temperature will reach around 112–114 degrees.


The second sample behaved identically, so I won't include a comparison here. If you're interested, though, you can just go to the corresponding review and read it.


Temperature at 10, 15 and 20 A


And at 25 and 30 A


The internal resistance is significantly lower than that of the 25R (12.5 mΩ).
As always, here are the results as received, after “conditioning,” and the final results.


While looking for a cover photo, I happened to come across a statistical sample of internal resistance measurements based on 100 cells. I assume they were from the same batch, so the sample isn't entirely representative, but still.


Naturally, I really wanted to compare them with other batteries—you could say that was the whole point of this exercise.
Below is a comparison chart at 1C and 20 A for these batteries:
Samsung INR18650-25S
Samsung INR18650-25R
Basen Y25 18650 2500 mAh 22 A
Murata US18650VTC5A
Vapcell K25 18650 2500mah 20A

As you can see, the batteries chosen for comparison had similar capacities and comparable specifications.

Overall, I can say that the Basen Y25 came out on top in terms of capacity in Ah, in both test modes.
But in terms of capacity in Wh, the 25R came first at 1C, while at 20 A it was the same Y25, with the new 25S close behind.

Although, looking closely, the actual difference is so small that only the Murata VTC5A really stands out—and only for lagging behind the others.



When preparing reviews, I almost never read other reviews of the same product, so another author's opinion won't influence my own. I did the same this time, but after I'd finished almost all the tests, I came across a video review of these same batteries on YouTube. I watched it and wanted to add a few comments.
But I recommend not watching the video right away. First read what I've written below, then come back and watch it.

I also want to point out that the 25S is compared with the VTC5A, not the 25R, which would have made more sense—especially since, as the comparison chart above shows, the VTC5A clearly lagged behind the others in power delivery.


Let me say right away that this is in no way a criticism of the author. It just so happened that this one video brought together several mistakes I've seen in many places, and I wanted to explain how to avoid them—or at least reduce their impact—and what they can affect.

First, the production date. The video says the batteries were made on April 3, 2019, and the code is exactly the same as on the ones I show, so they are either from the same batch or at least from around the same time.
At first I thought the author had simply misspoken, but then I realized the date-check page had misled him; it also showed April 3.


As a reminder, for most batteries with SDI markings, the production date is determined by three characters. But even simple logic tells us that the marking for April 3 would be 2J403, not 2J43, since the date takes two characters and the week takes one.
Other companies (LG, for example) specify the day of the year, while Molicel uses the month and day.

But in my opinion, this really doesn't matter, since there's no significant difference, and knowing the month and year is quite enough.


There are exceptions, though, depending, if I'm not mistaken, on the place of manufacture and the cell type. The Samsung 35E is one example: it uses a different order and a different month/week code. I actually failed my own test of the Samsung 35E, though.


Next, differences in measurement methods and equipment can affect the test results.

First, the holder. The author uses a clamp to press metal plates against the battery terminals. This method is generally acceptable, but for low load currents or tests that don't require much accuracy, I've seen some other reviewers use the same connection method.


For more accurate measurements, holders with a four-wire connection are used. Each contact has two connections, one for power and one for measurement. I've mentioned this before.
At high currents, ordinary two-wire setups start to give inaccurate capacity readings in Wh. Also, voltage drop across the contacts can cause the discharge to stop prematurely.


But even if the author had a holder like that, it wouldn't have helped much—or, to be precise, it wouldn't have helped at all, since he uses an iCharger 406DUO charger/discharger.
It's a great device: it can do a lot, displays a ton of information, has good specs, a balancing function, and up to 1400 W of discharge power (1000 W per channel) in regeneration mode. The problem is that it's not a battery tester, just an excellent hobby charger, if only because it uses a two-wire load connection.
It also looks really cool, and as for the price...


By comparison, ZKEtech devices look as if they were assembled in a hobby club, and cost much less. But they are battery testers, partly because of their four-wire connection and partly because of their specialized functionality.

But as I said, the iCharger 406DUO is a good device (as are others like it); it's just intended for somewhat different uses, particularly balancing, fast charging, two channels, and so on.


What the lack of a four-wire connection to the battery means.
Here's a screenshot from a video of the iCharger 406DUO testing a Samsung INR18650-25S battery. You can see a current of 30 A, an input voltage of 3.119 V, power of 96.9 W, and 542 mAh discharged.


And here's a screenshot of the EBC-A40L running at the same 30 A with the same battery, and even at almost the same point in time, since 538 mAh had been discharged. You can see that the actual battery voltage at that moment was 3.357 V and the power was 100.7 W.
That's a difference of 0.238 V, which means at the very least that the load won't cut off when the battery reaches the set 2.5 V, but at 2.738 V instead. And at high load currents, the discharge curve is usually flatter toward the end because of the battery's internal resistance, so the measurement results will differ accordingly.


Can you work around this? Of course. A four-wire holder is preferable here too. First, load the battery at the test current and measure the difference between the load's reading and the actual voltage at the battery terminals. Pause the test, adjust the cutoff threshold to account for the voltage difference, then resume the test.
For example, here I had to set the threshold to 1.91 V to get it to cut off at an actual 2.5 V.


Unfortunately, even that won't help much. If you look closely at the display, among all the other information, one of the key measured parameters is missing: energy delivered in Wh. I was actually a little surprised.


Why this matters.
Below is a comparison graph of two batteries: a Samsung 20S with a capacity of 2000 mAh and a Murata VTC5A rated at 2500 mAh. It all seems clear and logical, but I'll quote the conclusion from the 20S review.

You can also see how different their graphs are.


The author now uses a more sophisticated load. It has a four-wire connection, measures Ah and Wh, and comes with a bunch of useful features. What can I say—it really is an excellent electronic load. But it runs into the same old problem: the battery holder is two-wire, and this is just a load, not a battery tester, although it comes very, very close. I wouldn't mind having one myself, but I can't afford it yet :(


Another error is temperature measurement. This is what I was talking about above when I mentioned 80 degrees at a current of 35 A.

During the tests, the video author recorded the following temperatures:
15 A - 54.0
20 A - 61.6
25 A - 66.0
30 A - 71.4
35 A - 77.0

As a reminder, I got somewhat different readings. At 30 A, the temperature actually exceeded 100 degrees. How is that possible?


The reason is that a contact method is used to measure the temperature. Not only does it introduce error because a flat or round sensor is usually strapped to a round battery, but the sensor itself is also partially cooled from the opposite side.


The problem is easy to explain.
If we measure temperature using a sensor with a flat-sided housing, like the one in the photo above, it can only work reasonably accurately with flat surfaces.
The problem is this: heat from the object being measured warms the sensor, but the sensor also releases heat into the surrounding environment. The result is an intermediate reading that depends on the air temperature, the position of the temperature-sensitive element inside the sensor housing, and, of course, the thermal contact.
The situation is even worse when the object being measured isn't flat: the thermal contact point becomes much smaller, while the heat-transfer area increases.

In practice, you can think of the temperature sensor in this setup as a small heatsink that absorbs heat from the object and releases it into the surrounding environment. You can improve measurement accuracy by applying thermal paste and/or using strip-style temperature sensors.



Temperature is also sometimes measured using the thermocouples included with multimeters. I think many of you know the little bead on a wire.
This can be better or worse. It's worse because the thermal contact point on this type of sensor is usually tiny and barely changes depending on the shape of the object being measured. It's better because the sensor usually has little mass and is therefore less sluggish. But the small contact area also means greater measurement error, and sometimes the sensor is strapped on with something that also acts as a heatsink, which again introduces considerable error.

When measuring battery temperature, though, it has an advantage: if pressed firmly, it partially pushes through the battery's insulation, increasing the contact area. The tape used to attach it for convenience also creates a greenhouse effect.

These sensors work well for immersion measurements, where the sensor is submerged in the medium being measured, whether water or molten solder. But when measuring surface temperature by contact, it will still give you a bogus reading—for example, when you try to measure a transistor's case temperature with it.

The best, though still imperfect, method is non-contact measurement, for example with a pyrometer, but that has its drawbacks too. A pyrometer averages the temperature across the measured spot, so it works well for relatively large areas. In addition, the surface's emissivity has a strong effect on the measurement. For example, measuring an aluminum heatsink and a piece of wood will give you different readings even if their actual temperatures are the same.
The first problem is easy to get around with a thermal imager, which can “see” the hottest or coldest spots. The second can be addressed by applying a correction.


Non-contact measurements with a thermal imager and a pyrometer give roughly the same readings: 100–102 degrees at 30 A and an ambient temperature of 23–24 degrees. That's a substantial difference of 30 degrees, wouldn't you agree?

This difference depends heavily on both the temperature itself and how quickly it rises. For example, at 15 A the difference was 13 degrees. Part of the reason is that the temperature can even be calculated fairly accurately, since it's directly related to the current and the battery's internal resistance.
For example, at 10 A and 10 mΩ, the power dissipated is 1 W. At 20 A it will be 4 W, but since the test takes half as long, the battery will dissipate only twice as much energy, not four times as much.

So, with an ambient temperature of 23 degrees, the battery heated up to 55 degrees at 10 A and to 80 degrees at 20 A. The temperature rise was 27 degrees in the first case and 57 in the second, i.e. The relative temperature change is proportional to the heat generated in the battery.
Accordingly, if we increase the current by a factor of 3 (to 30 A) and reduce the test time by the same factor, the temperature should be:
23 (ambient temperature) + 27 (temperature rise at 10 A) × 3 (current relative to 10 A) = 104 degrees. I measured 102 degrees with a thermal camera, which once again agrees with the calculations.


And finally, a demonstration of the temperature sensor's lag. The test is over and the temperature is 77 degrees. The video was sped up a little after the test, and you can see that the temperature rose to 81.6 degrees. The author explained this as follows:


But friends, a battery isn't a nuclear reactor. Its temperature doesn't rise after the load is removed; it drops, and very quickly, since the battery weighs only 45–50 grams.
No, there is a little thermal lag: the heat does move from the center toward the edges, but the effect is so small that it can be ignored.


For example, when discharging the battery at 30 A, I measured its temperature with a thermal camera immediately after the discharge ended and got 102 degrees. I set it aside and measured it with a pyrometer: 101 degrees. I took a couple of photos, measured the temperature again, and got 99.8 degrees. These measurements were taken about 0, 10, and 30 seconds after the discharge was stopped, respectively.



And now for the conclusions.
Overall, I liked these batteries. They perform noticeably better than the 25R at currents of 15–20 A, and can also handle 25–35 A. Their capacity is roughly the same, although the 25R comes out ahead at low currents.
At 30 A and above, the voltage sag becomes noticeable, as the graphs show.

If the 25S is sold at the same price, it's a good buy, and it's also a good choice when you need to draw a lot of current. But if your current is up to 20 A and the 25R costs less, I'd go with that one. I don't think you'll notice the difference.

That's all from me. I hope you found this useful.