To profile a battery in a realistic way you have to choose a device activity profile that the battery will be discharged with. This profile consists of information about the current consumption during active and ideal (sleep) mode of the duty cycle. The acceleration is needed since nobody has the time to wait for a battery to drain for a couple of years to have the test results. With that said, the longer you choose to profile, i.e. less accelerated your activity profile is, the more accurate the results will be. 

This is how you can create an accelerated battery discharge profile:  

Step 1

profile your device, i.e., measure the current consumption profile of your device in active and sleep (idle) mode. 

Step 2

use the Otii Battery Life Estimator to get a first estimate of what battery size can fulfill the lifetime requirements for the given device current consumption profile.  
Pick a starting capacity, then select the active and sleep mode data of your device. Press ‘calculate estimates’. Iterate the capacity value until you have reached the battery life you envision for your device and use case. This estimate will be overly positive, the reality is that the battery life will be much lower. 

Step 3

profile the battery/batteries. A discharge of a battery according to your device’s current consumption profile will in most cases take months/years to perform, the calculated estimate tells you how long. You need to accelerate this discharge and create your own accelerated discharge profile.  

You can use the Battery Life Estimator in this step if you are discharging the battery with constant current (alternative is discharging with constant power but this is a separate topic).
Start by replacing your average current consumption with the highest current consumption seen in your device’s profile. Do the same with the active time, that this current level is on. Calculate the battery life estimate.  
The battery life will be most likely still unreasonably long for actually testing so the next step is to adjust the sleep time to a much lower value. Calculate the battery life estimates with the new settings. If the time is still too long for reasonable testing, then increase the lowest discharge current from the sleep mode. Iterate until you have a battery life estimate that is below 1 month, which is reasonable time for testing, yet keeping credible level of accuracy. 

Speaking of accuracy, here are a few things to consider when you accelerate your discharge profile to avoid introducing too many errors. The sad truth is that every step that you take away from your device’s original profile will introduce errors, smaller and bigger. 

• Coin cell batteries are very sensitive to high discharge, thus try to keep your high discharge the same as your device peak current. 
• Lithium Ion rechargeable batteries are very happy to give the energy to the device so here you can accelerate the discharge also by increasing the high discharge current higher than the device peak current . 
• Some batteries, like alkaline batteries, have a recovery period where the battery voltage and available energy recover during the low discharge. Analyze how much of a long sleep/low period you need so that the battery has time to recover. 
• Alkaline batteries are not as sensitive as the coin cells to increasing the high discharge, but also not as insensitive as the lithium ion cells. 
• Try to get the discharged energy in the accelerated profile to be as close as possible to the original profile, this is however, not always possible. 

Below we show this procedure with an example of a LoRaWAN device and a coin cell battery:  

We do not have time for 11.3 months or 14.1 years profiling, so we need to accelerate the discharge.

Since we are working with coin cell batteries, it is important to keep the high current discharge the same (or as same as possible) as the high peak of the device. In the LoRa case this is 40mA for roughly 1s (this is of course dependent on the spreading factor and the battery voltage). The reason for this is that coin cell batteries suffer from high discharge and the available energy from the battery is greatly dependent on how quickly you discharge the battery.

So, our starting point is then that our high or active discharge will be 40mA for 1s.

If we want to profile the battery with constant power instead of constant current (as the 40mA is) then a 105mW discharge is a good choice (this is actually a sensitive LoRa strip power consumption during the active phase)

We then have something like this:

It still takes 1,5 years to profile, this is not good (here the time has risen as 40mA for 1s is a lower average than 17,3mA for 3,8s as we originally started with – but now we have a higher peak).

Now we start playing with cycle time. We do not need to sleep  almost 15 minutes during the discharge, let’s assume 3 minutes instead.

Almost 4 months is still a very a time-consuming profiling task, so we increase the sleep current discharge also, and aim for 1 month discharge.

So, if we choose 40mA for 1s and 600uA for 179s we estimate that the discharge of the battery will take roughly 1 month. So, this is our accelerated LoRa discharge profile.

Since we do not expect the battery to be able to deliver its full capacity of 620mAh, the actual battery profiling will be shorter, but depending on the coin cell battery, this time will vary from a week to almost 4 weeks.