This is the fifth article in our series of articles realized in partnership with established experts from the IoT world. We asked them to share their knowledge and best practices, giving you a step-by-step guide on how to successfully lead your IoT project, from its conception, to its life on the field.

This week, we have invited Wilfried Dron, CEO & Co-founder at Wisebatt, a great simulation tool that allows IoT developers and electronic engineers to create a virtual prototype of their solution in just a few minutes and to simulate its battery life. So it was only normal that we asked Wilfried his expert tips on how to optimize power consumption.

If you’d like to read more inspiring articles, just click here.

Wilfried, in a few words, can you explain the mission of Wisebatt and how the simulator helps your customers optimize their devices’ power consumption?

Wisebatt is a spinoff start-up from Sorbonne University founded 4 years ago. Our solution helps electronic engineers design more reliable smart devices, faster. Our simulator allows them to create a virtual prototype of their device in just a few minutes and to simulate its battery life, among other things. In no time at all, they can access complex modeling results that would usually take weeks to generate. IoT developers and electronics engineers can then experiment with several parameters to make the optimal choice between cost, battery life and performance, very early in the design cycle.

Our simulator can allow them to save 3 months and up to 200 000€ in research and development. And that’s not including the direct and indirect savings made thanks to a higher battery life.

How does the solution work and what process do you apply when testing the power consumption?

We are the only extensive system simulation tool on the market. There exist partial simulation tools using theoretical approaches or empirical approaches to test smart devices but none are as reliable or as affordable as our solution. Today, to simulate the battery life of a smart device, there are several approaches:

• The theoretical approach involves making a simple rules of 3: we look at the consumption of the device, divide the theoretical capacity of the battery by its consumption, add a 10%-20% or even 30% factor that corresponds to the cell capacity that can’t be accessed and it will give us the expected autonomy. Often it stops there.
• The empirical method: we select a battery, design a partial prototype, simulate a mini discharge of the battery with the partial prototype in its normal use and then extrapolate the results after adding a security factor.   That second approach requires having already designed a partial prototype, hence having made some technical choices. You need measuring instruments to equip the partial prototype and batteries to discharge. It therefore represents time and money but moreover, you can only apply that approach once you have moved forward with the prototyping and engaged in making some technical choices. This method only allows you to test one or two designs, which means that it can confirm technical choices you made but it doesn’t efficiently help you to decide between multiple options.
• Another approach is to do some accelerated discharge. This is where you select the battery, and discharge 10 times the current that the prototype could potentially use and extrapolate the results. In this approach, you will also need to acquire measuring instruments and to repeat the tests, which will cost both time and money.

None of these approaches are completely accurate. They don’t really capture or reveal the behavior of the battery in real discharge conditions because the real battery’s behavior depends on the exact power consumption profile (maximum peak current value and duration, average current drawn, etc.). They don’t let us define the battery life of the system once the smart device is deployed in the field and as a result, they can mislead the engineer in his choice of architecture. Beyond that, the error factor is very important, as the battery life can be sometimes overestimated x10 to x12 times more than the actual one. This can lead to catastrophic results, especially for health, security and maintenance applications.

Note from Saft: Besides, accelerated discharge implies many bias, we will soon release a full blog post about this topic!

Wisebatt is different in the sense that it builds realistic models that can be directly applied in the selection and the characterization of the device’s architecture. We have compared our models to reality and have made experimental validations on more than 200 different architectures. We mastered the environmental parameters in discharge, both for primary and rechargeable batteries, and we are now able to address the charge.

There is mathematical proof behind our models that relies both on empirical data and on information and know-how directly transmitted by the manufacturers. We model batteries and components based on their datasheets and we generate simulations using a methodology that we developed ourselves. This allows us to get very close to reality. Indeed, the fact that we validated the approach and the methodology allows us to offer a certain guarantee in the extrapolation of the results for the systems we didn’t test.

In a simulation, the absolute truth doesn’t exist. The simulations’ results are only the expression of the manufacturer’s specifications and may not be in accordance to reality if the manufacturer specifications are not accurate. We only have a body of evidence that allows us to extrapolate a reality. But instead of doing it the empirical way – a way that is expensive and not very reliable - we do it through simulation. We validate the model’s performances and collaborate with the manufacturers to produce models that are as close as possible to reality.

When an engineer looks at battery choices, he may have already defined an architecture that he can then test and validate. Or he may not, in which case he can use the tool to select the correct battery for his device, do the simulation and validate the results.
 

Wisebatt Power analysis results: Battery life estimation, detailed view of your components' power consumption and instructions
 

What problems have your clients mainly encountered before they come to you?

We work with all kinds of actors, from the small start-up to the big engineering consultancy that works for large corporations. The problems vary significantly between users.  

Engineering consultants will be looking for a super quick evaluation of the performance, the cost, the architecture and the autonomy to be able to respond to calls for bids.

Start-ups will search to compensate for a lack of expertise or knowledge internally. In general, debugging a system that doesn’t have the expected autonomy is very complicated. A battery’s behavior is not as predictable as other components’ operation and performances vary from one battery brand to the other. It is a very complex idea to grasp. You have to go through a thorough process and have enough experience to model its behavior, which not many people can do.

That is therefore the main reason why people use our system: to demystify battery technology and to count on more reliable results than a rule of 3.
And then of course there are all the advantages of a virtual prototype where you can test many different combinations without having to read a datasheet.
 

What are the components in an IoT device that usually use the most power and/or can be optimized?  And how can IoT designers optimize their power consumption?

We note that on average, between 25 to 35%, if not 40% of the power budget of an application is consumed by powering functions remaining in standby mode, that don’t have any function other than to stabilize the system.

To optimize it, we have the ability to do load switching which involves “disconnecting” some of the parts of the system when they are not used.

Then we can have optimizations linked to the use case. Typically, it’s about rationalizing the communications which are often the most power consuming in a device. Some micro-controllers or applicative processor architectures have a consumption that is greater or equivalent to that of a radio. We can get them to wake up only when communications are truly necessary.

Using a modular architecture can be interesting, but sometimes it is done to the detriment of battery life and cost. We need to rationalize this modularity. Sometimes it is more interesting in terms of performance to do two distinct versions of a device for two specific use cases, rather than do a generic one.

We also see a tendency to put more features in all the components to justify their premium positioning. Microprocessors or microcontrollers are equipped with artificial intelligence to process data for instance. If these features are seamless and useful, it is a clear advantage, there’s no question about it. But if it adds development complexity or if it’s just an additional peripheral/feature that will consume power, it is of little interest. So before overequipping your object with lots of power consuming features, ask yourself if the computing capacity will be actually used? Will it be useful? How will it impact the cost, the performance and more importantly, the power consumption? It will help you optimize the choice of components and optimize your power consumption.
 

If you had to give some advice to IoT developers trying to optimize their consumption, what would it be?

My first piece of advice that I give all the time would be to start from the use case and evaluate the need for battery life.  

The use case is the most important element when designing a product. Considering the battery life as the primary specification of the system is key. Leaving it until the end of the development process could compromise its chances of success.

I remember a client who designed a tracker to follow the GPS position of kids. He had developed a great technology. But we looked at the power consumption of the device and the tracker only had 12 hours of autonomy and was meant to follow the position of the child over one or two days, which meant that it had missed its use case.

Another example was a client who designed satellites but couldn’t get them started because the need for initial current was too high for the battery capability.

My second advice is to use a modeling method, like our simulator, to have a mathematical proof of the battery life and optimize it.  

Be mindful of the applications’ lifetime vs. the battery life. We used to design applications for a specific purpose, for a specific lifetime, for which no updates were possible. Today, the new paradigm is to conceive connected devices that are updated on a very regular basis, sometimes even weekly, in the hope that the device will last 10 years or more, and increase its capacity over time. There is no need to oversize a device in the hope that it will perform in 10 years. Because it will impact the battery life and in 10 years, it will underperform or might even be obsolete.
 

In your opinion, what trends will shape the IoT technology in a near future?

I see many projects based on a power mix between energy harvesting and “stored energy”, like for instance adding a solar panel to a rechargeable battery to extend the battery life of an asset tracker or a condition monitoring sensor. I believe that an autonomous device without a stored power source cannot exist, at least not for all applications. It’s a fantasy! Yes we already see for instance devices powered by vibrating energy that can be placed in tires to control the pressure, but then the operations complexity is fairly limited and the environment fairly controlled (i.e. a tire will always vibrate enough to produce enough energy for the sensor to measure and transmit the tire pressure).

The battery is the most reliable element to ensure a minimal service, so it will stay essential for critical applications. I believe that batteries and microbatteries will always be necessary as a main energy source or as a backup, because harvested energy depends on external parameters that are not reliably predictable in every single use case.

Again, it all comes down to the use case. We need to think about how the application will be used and the energy needed to make it work.
 

Thank you Wilfried!

If you'd like to go further, read Wisebatt's free White Paper: The digital revolution of the electronic components industry.

The digital revolution is increasingly shifting the industry’s attention on customers’ expectations. Behaviors analysis, needs comprehension and offers customization are now key to keep up with the competition. Based on market research, studies and articles, this report introduces how digital solutions will allow marketing and sales teams to adopt effective customer-centric strategies.

In this white paper you will find:

• New challenges in the electronic components industry
• Customers' changing habits and behaviors
• Practical approaches to adapt sales and marketing strategies