
by Joe Madden, founder and chief analyst, Mobile Experts, Inc.
Setting up a private wireless network for basic connectivity is becoming routine. In the U.S., a CBRS network can now be set up in either “general” or “priority” access mode. In many enterprises, CBRS will be the perfect solution, providing IoT connectivity and a simple way to connect push-to-talk handsets, broadband access, and advanced devices like automated ground vehicles.
But what happens when the enterprise needs more capacity? Priority licenses in the CBRS band use between 10 MHz and 40 MHz of bandwidth. Assuming a small cell uses the full 40 MHz channel, each one can support about 104 Mb/s average (more than 160 Mb/s peak) in everyday operations. For voice communications and most basic IoT applications, this is more than enough.
However, for enterprises with heavy video analytics, a single location can drive demand to a few gigabits per second, so scaling up in CBRS would be painful with the deployment of large numbers of small cells. With this in mind, let’s look at alternatives.
Cost Comparisons
To illustrate the costs involved, we examine two case study scenarios. In both cases, we want to equip a 40,000 ft.2 factory space with a private wireless network. In Scenario 1 we have 30 robotic welding machines, each using two 8K-resolution video cameras to control the welding tool for high quality in real-time. This application calls for about 40 Mb/s per camera or 2.4 Gb/s of total capacity. In Scenario 2, we have the same factory with only three robots instead of 30. Everything else is the same so we can understand how capacity influences the cost picture.
Our choices include:
- A CBRS-based private network based on small cells. In this case, a private network can be deployed by a mobile operator or by an independent system integrator. Up to 40 MHz of spectrum can be available on a priority basis or about 80 MHz on a ‘general availability’ mode that is shared with other users.
- A Distributed Antenna System (DAS). These networks have been used in stadiums for the past 20 years and use a signal source provided by the mobile operator. They have the advantage of using multiple bands or even multiple operators to provide more capacity.
- A millimeterWave network. In this case, a private network can be implemented by a mobile operator, or in some countries, using a privately licensed millimeterWave band. Because each millimeterWave gNodeB has very high capacity, repeaters can be used to fill in coverage gaps.
The 40,000 ft.2 facility and the required network hardware for each contender are shown in Table 1.

We also considered Wi-Fi networks, as almost every enterprise has a Wi-Fi network, and the wide bandwidth possible in the 5 to 6 GHz bands can be useful for some applications. However, for critical industrial operations where any hiccup will cause quality problems in the product, we now see a strong preference for licensed-band radios. Wi-Fi will be used in the factory for human broadband use and non-critical IoT devices like asset tracking. But the critical machines will run on either wires or licensed wireless. In the end, we did not include Wi-Fi in our calculations.
The High-Capacity Scenario
In this scenario, the CBRS private network requires 20 small cells (Figure 1a) to reach the required capacity and may require some fine-tuning of power levels, antenna locations, and channel parameters as all the small cells share the same 40 MHz of bandwidth.

The DAS network has the advantage of using multiple bands and using a single operator, so a DAS system could incorporate about 120 MHz of spectrum. In this case, about eight DAS sectors would be required throughout the facility, as each sector would support about 310 Mb/s (Figure 1b).
A millimeterWave network would more easily meet the capacity challenge, with a single gNodeB unit providing about 3.6 Gb/s of capacity. In this case, deployment would focus on coverage, spreading the signal through the area instead of trying to minimize power and frequency reuse in separate zones. One gNodeB radio would cover roughly 8,000 ft.2 so we calculated the cost using a single gNB and four repeaters (Figure 1c).

Looking at the details (Table 2), we see that the millimeterWave network saves money because the number of gNodeB units is low, resulting in a smaller number of fiber runs and simplifying the installation. That said, one bit of complexity costs some money in the millimeterWave case: We must include a low-band 5G anchor network, as today’s state of the art demands this for high-reliability operation. But in our factory example, this simply means that we add a CBRS radio to two of our five millimeterWave radio nodes and the added cost is only about $6,000.
The number of gNodeBs or sectors required for the three network types is shown in Table 1.
The Low-Capacity Scenario
In our second scenario, only three robots are required on the same manufacturing floor, so there is no need to deploy large numbers of small cells and the 5G network cost drops dramatically. In fact, the CBRS network can be roughly half of the cost of the millimeterWave network because of superior coverage of the CBRS small cells. DAS is the most expensive alternative in all cases that we studied because of costs for signal sources, head-end equipment, and multiple fiber runs.
In these two scenarios, we consider only the upfront cost of installing the equipment. Several business models are possible for the in-building network, ranging from a managed service from a mobile operator to a privately installed and managed network in which the enterprise takes care of everything. For simplicity, our comparison simply looks at the network equipment itself and demonstrates the raw cost of the equipment and its installation.
In comparing our two scenarios, we see that the best choice of technology can depend directly on the level of capacity needed. In the low-capacity case, a simple CBRS network is best. But in the high-capacity factory case, a millimeterWave network can be much less expensive.
How can this be when it’s logical to assume that millimeterWave networks would always be more expensive, as their small cells cost more than CBRS small cells? In our case study and many other real-world scenarios that require high capacity, millimeterWave networks can be less expensive for a simple reason: The high-capacity fiber deployment can be implemented in one place and the rest of the deployment can use very simple repeaters to spread the capacity throughout the building. Using only the narrow radio band in the sub-6 GHz range leads to the deployment of far more small cells.
We’ve analyzed our two scenarios in terms of satisfying the capacity requirements of specific machines, but most projects are designed to include future capacity as well. If we have 30 robots today, we can implement a private wireless network to connect them, but we must be sure that the network is scalable.
Considering the CBRS and DAS options, scaling up from 2.4 to 5 Gb/s of capacity would result in huge numbers of radios in a small space. In fact, such an extreme density of small cells would be unlikely to work at all. In contrast, to double the capacity in the factory using a millimeterWave network, just one repeater node would be replaced by a gNodeB. This makes future capacity enhancement simple and straightforward.
In enterprise markets, we expect several scenarios to result in growing traffic demand over time:
- Factories using video analytics with resolution increasing over time
- Broadband usage by employees
- Videoconferencing that is now standard in many factories
- Medical/healthcare facilities should experience extensive growth in bandwidth requirements
- Automated ground vehicles and other second-tier IoT devices will be added for greater automation
The Benefits of Bad Propagation
MillimeterWave networks are known for their difficulties in terms of signal attenuation, penetration through walls, and other propagation problems. For indoor enterprise use, these factors are not drawbacks and are actually very positive attributes.
Lack of penetration through walls means that millimeterWave bands can be used indoors without interfering with the mobile network outdoors. Many mobile operators are reluctant to implement private networks on their licensed sub-6 GHz bands for fear of causing interference on the street outside. This concern is greatly reduced for millimeterWave signals, which can fill the indoor space without significant leakage into the outdoor environment.
Attenuation of the high-frequency signals can be an advantage in setting up multiple sectors inside a building. Our case study calculated that the enterprise would use one millimeterWave gNodeB and four repeaters to spread the radio channel throughout the building. In an industrial building, interior walls can become an excellent boundary between the coverage zones of each repeater, helping to avoid feedback loops or other distortions. Low-band radios require much more attention to detail in setting power levels and antenna direction because of overlapping coverage areas.
Time to market can be a significant advantage as well. In our study, the millimeterWave network required only a single fiber location, and the repeaters could be implemented on walls of the building using only AC power. That’s much simpler than a low-band approach with as many as 23 fiber runs, and we expect private millimeterWave networks will be implemented in hours rather than days or weeks.
Finally, the wide bandwidth of the radio channel means that each robot can use different resource blocks. In cellular systems, high capacity depends on frequency reuse and in a high-density environment such as our factory example, the radio frequency-and-time slots must be very carefully managed so that multiple small cells can reuse the same frequencies at the same time.
However, the millimeterWave channel is wide enough that each robot can use different frequency-and-time slots, reducing the chances of interference. We expect this simple aspect to result in higher reliability for millimeterWave networks in high-density applications, as well as a simpler deployment process.
In short, many people believe millimeterWave networks must be inherently more expensive than sub-6 GHz networks, and in low-capacity applications, they would be right. But automation of industrial processes is moving quickly toward very high resolution video to enable AI and machine learning to improve quality and productivity. This means that many enterprise applications will emerge in which millimeterWave networks will be the best approach. They can offer the lowest cost, the fastest deployment, and the simplest RF planning in complex high-density scenarios.
(231)