Choosing a battery for your next low power radio project

Just bung a 2032 coin cell in !

A large number of small electronic devices (from flickering faux candle night lights, to the keyfob that unlocks your car doors) are powered by the ubiquitous lithium “coin cell” and, while there are other sizes in reasonably common use, that probably means a “2032” cell.

So when you come to choose a battery for your next low power radio project (maybe a miniature transmitter, maybe an ultra low
power sensor node) it’s easy to select that same familiar power cell. It’s inexpensive, they are sold through a wide variety of outlets, and it is accompanied by a wide range of clips, holders and leaded versions. It even provides a conveniently compatible-with-modern circuits 3v output

It may be an easy choice, but please: think again

These common “lithium coin cells” are based around a lithium manganese dioxide chemistry. This yields a highly stable, long shelf life cell with a decent energy density, and a usefully high 3v (peaking initially at 3.3v) cell voltage.

The problem is that the one parameter where these cells do not shine is peak current supply capability Even a larger example like the 2032 is typically rated at a few milliamps Add in the cell’s considerable internal resistance, and you have a power source that is ideal for memory backup tasks, or powering ultra low power CMOS circuits (such as calculators, or bank card verifiers) … but
unfortunately this seriously falls down when powering typical low power wireless modules. Here the peak current will be (even for 10mW types) over 10mW, with many examples in the 20-30mA range … and these modules will likely require at least 2.7v, which we will see is a problem for the 2032

Now to be fair I am not condemning the lithium manganese dioxide cell family in it’s entirety: but I am flagging up significant failings when the device being powered is a low power radio module

This is a discharge curve for a Duracell 2032

Note that for the purposes of claiming it’s 265mA/Hr
capacity, the manufacturer defines the endpoint voltage is2.0v. If your circuitry will only function to 2.7v or so (as is not that uncommon) then this is really only a 200mA/Hr cell at 200uA, while at 3mA it is unlikely to deliver more than 150mA/Hr. Increase the current drain and that internal resistance will bite further into the usable capacity.

Energizer (a competitor) rate their 2032 at only 170mA/Hr to a voltage of 2.7v if the load is a 7mA 2s pulse every 2 hrs

Again, it can be argued that short duty cycle pulsed operation will still yield a much longer lifespan, but if we imagine an again not unreasonable 25mS pulse duration, this 5 minute test (to under 2.6v) only corresponds to 12000 discrete pulses. Possibly OK for a carefully designed “push-button” fob (which is almost always off), but not for something like a low duty cycle “node” such as an environmental monitor, or for an alarm, where the power drain is constant

A perfect 230mA/Hr cell ought to be able to power a 25mA load for just over nine hours (or over 1.3 million 25mS long pulses), but the 2032 is a long way from being that “perfect” cell

If we refer to the discharge characteristics of the silver oxide cell then we see a far “flatter” graph, with the terminal voltage holding up at around 1.55v all the way to almost 70% of the rated capacity (the graph shows a low current load, rather than our 30mA example, but is still pertinent)

For a low power wireless application, this effect alone means we will see a far longer battery life from a pair of 165mA/Hr silver oxide cells than a 265mA/hr rated 2032 lithium manganese dioxide part

The issue we are seeing here is that while the coin-cell is entirely capable of delivering it’s rated capacity into sub-milliamp loads, it falls down badly as the current is increased to something approaching that drawn by typical low power ISM band wireless modules, losing both operating voltage and usable capacity

(A further matter that is starting to raise it’s head here is that not all 2032 cells are the same. Even the two premium brands I’ve selected here have quite different data-sheets. Even the capacity varies (265mA/Hr for Duracell, but only 235mA/Hr for Energizer). If I was to throw “budget” brands into the mix I expect to see an even wider variance, while there are also relatively unusual lithium polycarbon monoflouride and lithium “organic” chemistry, cells, which are further optimised for even lower current operation)

So far we have just examined the 2032, and taken the SR44 as a hypothetical alternative in a similar size/capacity class

(In fact, if I have the space, I would be inclined to use three SR44 cells in series, and run them right down to their 1.2v discharge point to maximise capacity)

There are however a lot of battery chemistries in the market, and each of them have their own advantages and disadvantages:

We have already covered:

Silver Oxide: This chemistry has a terminal voltage of 1.55v and a very flat voltage/discharge curve. Peak current is better than a similar sized lithium, but energy capacity (especially mA/hr vs weight) is somewhat lower. Only available in small “button” sizes, and fairly costly. Self discharge is minimal and shelf life is long. Avoid confusion with alkaline cells sold in the same sizes (silver oxide SR44 vs alkaline LR44)

Lithium manganese: Nominal terminal voltage of 3v, but beware of high internal resistance, low peak current capacity and significant voltage/discharge fall off (to a 2.0v endpoint). Wide range of sizes (including our example 2032). Good shelf-life and wide operating temperature range. Relatively inexpensive Really suited only to low current tasks

However: there are other suitable types (as the reader is no doubt well aware)

Alkaline.    By this we refer to the generic class of ‘one and a half volt’ common usage primary cells. Historically, these are ‘zinc-carbon’ cells, although modern construction and chemistry has resulted in far higher performance units.

Found in a range of single cell sizes (AAA, AA, C and D, plus “button” cells like the LR44) and a few higher voltage multiple cell designs (the small 9v PP3, and various small, very low capacity 6v and 12v tubular batteries often seen in older remote control designs) the alkaline cell is inexpensive, and capable of producing large pulses of current. Holders, boxes and clips for these common forms are available in a bewildering variety. Shortcomings include a relatively short shelf life (2-3 years), capacity degradation at low temperatures and at high current, unimpressive energy density and a non-constant terminal voltage (an alkaline cell delivers over 1.5v when new, but falls to 0.9v at the end of it’s life)

Although cheap, the very cheapest examples ought to be avoided, as performance can be uncertain, and leakage of corrosive electrolyte materials far more likely. Most of the higher end manufacturer’s products are interchangeable, and have similar cell capacities (despite deliberately vague advertising claims by competing makers)

Lithium thionyl   As well as the LiMnO2 coin cells already described, there is also the potentially more useful, higher current capacity lithium thionyl chloride cells.

These cells have a high terminal voltage (3.5 to 3.7v, reasonably constant throughout the battery life) and(with careful choice of the specific type) moderate peak current output, with the typical maximum current for an AA cell being 55-100mA. Operating temperature range is excellent (to at least -30 degrees, with many type going to below -55) and shelf life is long. In addition to the standard AA and D alkaline sizes, lithium thionyl cells are made in 1/2 and 2/3AA, and the larger diameter 2/3A sizes. In addition to plain ‘holder’ type cells, versions are available with PCB mounting tags or axial axial wires, and occasionally custom sizes will be encountered, but oddly never an AAA size.
Primary shortcomings are high cost and limited availability (a good alkaline AA cell costs 40-60p, a lithium thionyl AA will be over £5), and a perceived chemical hazard, as although modern examples are not prone to exploding under overload like some earlier types were, they cannot be readily shipped by post.

Lithium iron disulphide. Marketed as a superior alkaline replacement, this chemistry has a terminal voltage of only 1.5v, but excellent peak current capability, high energy density (capacity), long shelf life and flat discharge curve. Only seen in AA, AAA and sometimes PP3 sizes. Expensive compared to the alkaline cells that it effectively supplants

Lithium ion (used as a primary cell). This is a recent development “borrowed” from the vape market. There are now small (theoretically rechargeable) lithium ion or lithium polymer which are marginally cheap enough to be used as disposable primary cells. They are invariably supplied in “non standard” packages, usually with wire connections and inbuilt protection circuitry, and (unless purchase volumes are very high) are costly.
Terminal voltage is 3.6v, peak current is high (especially in types intended for use in drones) and other characteristics resemble lithium thionyl. Be very careful in choosing the right part, as some have excellent self discharge rates, but not all do. Take care, as this is a new area.

There are a few other primary cells that are rarely useful for wireless applications, but which I should list: Zinc/air. Optimum for constant moderate current drain applications, these cells are rarely seen outside of hearing aid applications. They require oxygen from the air, and once the ‘seal’ is removed their self discharge is very high (unsealed shelf life is in months). Small “button” sizes only Mercuric oxide. No longer used, owing to the toxicity of the mercury, these cells were functionally equivalent to modern silver oxide cells but had a very constant 1.35v terminal voltage.

Footnote : DC-DC converters
Sooner or later it’s going to occur to you that (given the availability of modern parts) one approach would be to adapt the power requirement to the battery, by means of a DC-DC converter circuit, to efficiently step supply voltages up or down. Then a low battery mean or end voltage can be made to match with the needed supply voltage into the module. A couple of alkaline cells (3v maximum, but 1.8v endpoint) could then supply a radio module and/or other circuits needing a solid 3.3v rail Simple ?
Well, yes, but (as usual) there are ‘down-sides’. Effectively three of them:

1. The obvious (but occasionally overlooked) issue is that the DC-DC converter based power supply will require additional circuitry to be added to what may already be a physically small design. The PCB footprint of a usable converter is likely to be at least equal to adding another LR44 cell, although that said, if the unit is slightly bigger, and the chosen power source was (for example) an AAA alkaline cell, then the extra circuit real estate needed to convert up to 3v is unlikely to be crippling. The BOM cost could be a different matter.
2. This isn’t a cure for high internal resistance or low peak current capability. As the input voltage falls, a DC DC converter will draw a proportionally higher input current to maintain it’s output level, and if the battery cannot supply it, then the power supply system will collapse. A 2032 coin cell at it’s 2v end of life point isn't going to power a 25mA output 3.3v switch-mode (although an alkaline at 0.9v might well be able to)
3. Noise. At a simplistic level, a DC-DC converter is inherently an oscillating circuit, and all AC signals have the potential to cause interference. At the core of any such supply design is a switch device, driven by an alternating waveform, carrying a significant amount of power. The switch waveform itself, harmonics of it, and secondary effects are potential interferers, and can affect the radio module by direct radiation, through the power supply rails or even by magnetic induction. This does not mean that a DC-DC converter is out of the question (the environment is full of periodic waveforms, from logic signals to the domestic mains, and untold numbers of radio links continue to provide excellent performance) but it does need to be designed in carefully, with significant EMC and RF testing

The issue with any design that integrates a switch mode supply with a sensitive radio element is therefore down to identifying the sometimes subtle ways that radio performance can be degraded and by minimizing or eliminating the possible interference issues. Receivers are inherently more at risk than transmitters (owing to the minuscule signals they handle, and the large amounts of circuit gain), but even if they remain apparently functional, transmitters can suffer sufficient interference to become non-compliant with relevant legal regulations.

If implementing a DC-DC converter based power supply:

• Keep the power levels low. The less power it handles, the less interference it is generating. Use the DC-DC converter to supply only those elements of the circuitry that actually need it.
• Minimise radiated effects: do not locate the switch-mode close to the radio. Within the physical limits of the design, place it as far away as possible. Include shielding
• Make as much use of decoupling as possible. Don’t run the wireless device from ‘raw’ switch-modeoutput rails: include passive decoupling, or intermediate linear regulators, to ensure the supply is as 'clean’ as possible. If you can see the noise on a simple ‘scope, then it’s too much
• Optimize the switching regulator itself. Topologies are divided between constant frequency types and constant on-time (where the switching frequency is varied with the load). The constant frequency type is preferred, as the actual interfering signal can (within limits) be placed where it willdo least harm. Frequencies in the baseband (AF) range of the link should be avoided, as should frequencies around the channel offset and IF frequencies, and obviously (for high frequency switchers) the channel frequency and sub-harmonics of it are also forbidden.
• Layout techniques should follow good RF practice, and all tracks carrying high power alternatingwaveforms (including the ground returns) need to be short, and low impedance (wide). Low ESRdecoupling capacitors will be needed, and both low and high frequency decoupling must be provided. Lastly, the power inductor should be chosen for lowest external magnetic field (toroid's or other closed magnetic circuit types are superior to drum or spindle types)
• Be prepared to try several different designs. Effects can be unpredictable, and committing to a single approach too early in the design/testing process can be fatal. You will need to test (or get tested) the RF performance of the modules with the DC-DC converter running, to absolutely ensure that no degradation is occurring (and factor this into the project cost).

Myk Dormer for Radiometrix Ltd, August 2025