"In my opinion, a main driver of IoT is low cost. I think the large-scale rollout of IoT will be enabled by low-cost devices, low-cost platforms, and low-cost connectivity solutions," says Danish IoT expert Anders P. Mynster.
| Share |
|
We have invited Danish IoT expert Anders P. Mynster to a Q&A about the practical challenges to be considered when choosing and implementing a wireless sensor network solution. Here are his thoughts on power consumption, radio smog, overall cost of ownership, and much more.
Anders P. Mynster is an engineer specialized in wireless engineering and working at hi-tech consultancy company FORCE Technology as a senior consultant within wireless and EMC. Anders advises companies that consider stepping into the world of IoT. Also, he is chairman of the IEEE European public policy working group on ICT and co-manager of the Nordic IoT Centre, based in Copenhagen.
New emerging technologies often use existing infrastructure. That goes for wireless sensor networks as well. They utilize wifi, as wifi infrastructure is already in place in many companies and city spaces. What are the upsides and downsides of that?
– The obvious advantage of wifi is, that you avoid rolling out your own infrastructure. But wifi has some significant downsides as well. Wifi is power-hungry. Many developers would like their network to operate on one or two triple-A batteries. But the power consumption of wifi technologies give a battery lifetime of days – at best – and nothing near the weeks or years we would like to achieve.
– In my opinion long battery lifetime is a must. IoT solutions should last for preferable 5 to 10 or even 15 to 20 years on a set of batteries. So, I think the power consumption of WiFi is one of the key challenges in rolling out IoT utilizing that technology.
– Of course, another important aspect of the existing wifi infrastructure is that you have to deal with the administrators of the networks. In companies it is often the IT department that administrates the wifi. The administrators are unsure of the cyber security of the new devices connected to the wifi, and in general they are unsure of what these devices are. They are used to laptops etc. connected to the wifi, but now all these IoT devices are being connected, and what is that all about? So, we see a lot of companies that already have wifi in place as infrastructure, but the usage in relation to IoT are hindered by the IT administrator policies for the networks.
The future will bring vast amounts of sensors connected wirelessly. Will that lead to difficulties regarding congestion, interference, radio smog etc.?
– One of the challenges of Low Power wide area network technologies that are based in the 868 MHz band, is that we start to see more interference. Some researchers from Aalborg University published a very nice article about the interference situation, and I’d like to recommend it to people interested in the details of this.
– The obvious challenge is that the low power wide area networks have a very long range from a single base station, and they rely upon a very sensitive radio. With all the Radio Smog that is emerging in the 868 MHz band we will soon see some of these devices being challenged by interference. A lot of research is going on at the moment to clarify what impact this will have.
– In regards to mesh networking, one of the interesting things about mesh is that the devices are connected to each other. Mesh is a multi-hop solution, so you can get large area coverage without having long-range links. That is an obvious benefit of mesh.
– Also, together with shorter radio links, the sensitivity of the receiver does not need to be very low, and therefore the smog level has to be much higher to interfere with the mesh network radio links.
Do you see specific use cases for 100 per cent battery powered sensor networks? And in the future, how many sensor networks will be mains powered, and how many will be battery powered?
– That is a very good question. The answer depends on how you understand IoT. In my opinion, a main driver of IoT is low cost. I think the large-scale rollout of IoT will be enabled by low-cost devices, low-cost platforms, and low-cost connectivity solutions.
– Having a mains powered device requires installation, and if you consider an average salary in Europe, it is about 20 euros an hour. That amount quickly becomes comparable to the device cost itself. Furthermore, you have to roll out cables, install the mains power in various situations, and also plug in all of these devices. So your cost will increase rapidly, also because the electronics supplying the device itself is costly as well. In my opinion, the whole low-cost paradigm of IoT is supporting wireless connectivity.
– I am sure that a lot of IoT devices will become battery powered, not least due to the mobility aspect of IoT. A lot of IoT devices are being moved around, either in a large-scale area like tracking of goods within a factory, or tracking the employees that are moving around in vehicles in a municipality area. For this you need battery-powered solutions. So, this aspect also supports the battery-powered operation.
– All in all, I see non-mains powered solutions as the main trend here. Of course there will also be many mains powered devices, but this will typically be more costly systems and therefore I don’t expect we will see the same amount as the non-mains powered. Power for these could come from batteries, but also from solar cells, vibrational energy harvesting, or other kinds of energy harvesting.
What are the downsides of mesh networking?
– Traditionally, mesh networks have been very power-hungry, because you have to route the packet through the network, and you have to do multiple transmissions. But in the way that 2nd generation mesh networks like NeoCortec has designed their algorithm they seem to have overcome that by the scheduling of how the package is transmitted. That of course has a downside concerning the latency of the packet arrival. But in many applications, you do not need real-time communications, so I think some degree of latency will be acceptable for most applications.
You have some thoughts regarding the term “Real-Time” in relation to IoT sensor networks. Please explain.
– I think we need to define the meaning of “real-time” more precisely when it comes to IoT. Currently I participate in one of the study groups of the international standardisation of the real-time concept of IoT. There is an effort to try to define: What is real-time? Is it 1 millisecond or a tenth of a millisecond or is it 100 milliseconds? It all depends on the use case of the user.
– To many users it could be much more than 100 milliseconds. If you for instance look into the finances of a company, the CFO would like to know exactly what has been the daily turnover. If he gets that information the next day, that may be “real-time” for him. Whereas if you look into the industrial automation systems of a brewery, where you sometimes have ten thousand bottles passing through by the hour, then even 1 millisecond of delay can be non-real-time.
– I think it is very important to have a discussion with the end-users of the applications to figure out “what is real-time for you”. As an example could be smart farming with humidity sensing of the soil. This enables reporting the soil moistures, so the farmers know when to irrigate the fields. For them real-time does not mean within seconds. It can easily be within half a day and it is still real-time, so the farmer can turn on the irrigation system to water the fields in that area. The understanding of real-time differs a lot from application to application.
– A lot of applications actually have to be real-time. But it is a matter of being very specific about the concept. I think there are very few that have a clear definition of what real-time is to them. So, here we as engineers need to understand the focus of the users a lot better and to have these discussions – this can for instance be clarified by design thinking and user centric development.
It seems there are a lot of companies out there looking for the right IoT connectivity solution. Many of them find it hard to find the right solution and to compare the different technologies on offer. Why is that?
– The devil is in the detail. A lot of the sales people do not always have the detailed information that we as engineers need. They may not know all the specifications of the technology, and therefore they can be unaware of specific limitations. But this is maturing, in my opinion. Also, we need to develop a common vocabulary to talk about the different technologies. Similarly to what we just talked about regarding the concept of Real-Time we need to develop a common language when we talk about reliability or low power consumption.
– We all need to make an effort, sales people and engineers as well as consumers and purchasers. We need a common terminology to describe these technologies. Take for instance Narrowband IoT: There are a lot of parameters that you can tweak, when you implement it. If you do it poorly, then you’ll have high power consumption and if you do it really well, then you’ll have a low power consumption. So it depends. You need to know what you are doing when you implement the technologies.
– Concerning this, I think small sized companies have a strong selling point here. As a customer, you will be dealing directly with the designer of the technology. So he’ll be aware of all the small parameters you need to tweak when designing your product.
At FORCE Technology, a lot of clients come to us for an individual and independent 3rd party evaluation and comparison of different technologies. We can do that, of course. But quite often we need to go back to the client and tell him “We can see that this implementation is not performing power efficiently. So maybe you need to optimize that before we do the real testing and the A-B comparing of the technologies.”
Let us take a look at the cost of running a network. There are many different cost models out there. How can a customer get an overview over the connectivity cost to avoid any nasty surprises?
– The basic evaluation you need to do is to differentiate between a subscription model, where you subscribe to an existing network, and an establishment model, where you build your own infrastructure. Which one to choose depends a lot on the number of devices you have. The prices have dropped significantly for some cellular technologies. Some companies offer 10 years of subscription data with up to 500 Mb for 10 euros. That is, if you look into the value creation, a very, very small amount for the total cost of ownership, and in general, the cost of connectivity is going down.
– But obviously, you need to consider the total cost of ownership, so it is about predicting the subscription fee for the entire lifetime of the service, compared to rolling out the infrastructure yourself. To predict that you also need to know what happens if you want to scale up the network at some point. If you use the subscription model you are in the hands of the network operator. Is he willing to scale the network to your needs? If you have a client in a remote area where there is no coverage, would your network operator be willing to set up another base station to create better connectivity, just because you have one customer in that area?
– On the other hand, if you decide to roll out your own infrastructure, you either need the skills to do that in-house or you need an external partner, who can do it for you. This kind of resource costs money too.
– So all in all, the cost of connectivity, especially if you look a few years ahead, is quite difficult to calculate. But, to conclude, I see a strong case for non-mains powered network solutions with ultra-low power consumption, because, as I said earlier, the large-scale rollout of IoT will require low-cost devices, low-cost platforms, and low-cost connectivity solutions.
Anders Mynster, thank you for your time and for your willingness to share your knowledge with us.
