How To Build a Radio for a Robot
Written by Manas Menon, Lead Systems Engineer
At Shield AI, we develop artificially intelligent robots and systems for a safer and more secure world. Our first product, Hivemind Nova can operate autonomously, but to be most effective it requires a communications link to the operator. The operator uses this link to change robot plans, take manual control, and most importantly get a first person video stream from the robot.
I helped design this communications link, and this is the blog post I wish I had read before I started. This post will not dive into the details of building such a system - instead it’s meant to be a high level guide that shares the most important factors you should consider in building your own, along with a description of how they affect system performance.
Link budget determines if your transmitting radio is “loud” enough, and your receiving radio is sensitive enough to successfully pass a message. The following links are helpful in creating a link budget:
The key factors to consider (along with typical ranges) are:
Transmit power of transmitting radio (10 - 30 dBm)
Receive sensitivity of receiving radio at a given bitrate - the radio manufacturer should provide this information (-100 to -70 dBm)
Antenna gains. (0 - 8 dBi)
Free space loss - a function of frequency and distance. (40 - 120 dB)
Most radios operate on a single frequency. Software defined radios (SDRs) operate across a wide range of frequencies, but making an SDR work can be complicated. Most radios operate in unlicensed bands (free to use) such as 900 MHz, 2.4 GHz or 5.8 GHz. Higher frequencies let you push data at a higher rate, but lower frequencies can travel longer distances.
Many radios allow for software configuration of bandwidth (the width of the fourier transform of signal power around the carrier frequency). Note the distinction between bandwidth and frequency in this context. A larger bandwidth allows for more data throughput. A smaller bandwidth, assuming the same power, results in a higher signal to noise ratio because noise is typically constant across frequencies. This adds headroom to the link budget.
Size, weight and power (SWAP)
Because Hivemind Nova is an aerial platform, size and weight matter. An external amplifier can be used to boost transmit power, but these amplifiers require high linearity. This leads to high thermal losses, which could require a large heat sink.
A rough rule: a 1 watt radio (a radio transmitting 1 watt of RF power) can be expected to burn ~5 watts total.
Many radio manufacturers use a custom protocol and give little visibility into its workings, but they might describe scenarios in which it does and does not perform well. Spinning up a custom protocol is a significant effort. Some things to consider:
Frequency hopping - frequency hopping helps prevent jamming, but makes it harder to avoid frequencies that might have interference
Variable bit density - ideally the protocol allows for high bit density when signal strength is high (to push more data) and low bit density when signal strength is low (to extend range)
Mesh functionality - consider carefully if this is needed; mesh functionality can add significant cost.
MIMO - offers improved performance indoors (though quantifying this is difficult) and requires multiple antennas.
Encryption - not really a protocol consideration, but this seemed like a good place to put it. If critical for the application, many radios come with build in encryption.
Self interference / external interference - consider the case when you have multiple units of the same radio system operating simultaneously, or the case when there is external RF interference. Some protocols are better than others at handling each of these cases.
Interfaces are often an afterthought, but integration is far from trivial. Consider the following factors:
Communications protocol to the rest of the system
Mechanical mounting interface
Connectors / cables required
Physical location of the radio in the system to minimize electromagnetic compatibility issues
Antennas are passive elements - they do not increase the power coming from your radio, but reshape it. This means they increase power in some directions at the cost of reducing power in other directions. Antennas must be compatible with the frequency of operation, but they are agnostic to protocol. Some key metrics to consider include:
Efficiency: a measure of how much power is converted to useful radiated RF power versus wasted as heat
Directivity: a measure of how “directional” the antenna is. A satellite dish is very directional - high power in one direction while very little power in other directions
Gain: efficiency * directivity - this is the “bottom line” metric used in a link budget
Polarization: use the same polarization for a transmit and receive antenna or take a link budget penalty.
The first few tests you perform should verify the performance you expect out of the radio system.
Test throughput between a pair of radios at short range;
Test throughput between a pair of radios at the desired range using antennas similar to what will be on the final product;
Test the system against any additional challenges you expect (jamming, for example).
Antennas manufacturers will typically provide a data sheet with a radiation pattern that shows geometrically how power is transmitted from the antenna, but this spec is offered with the assumption that the antenna is in infinite free space. Mounting the antenna to hardware can change this pattern. It is important to measure this modified radiation pattern to make sure it’s not too badly affected by external hardware.
Measuring this requires expensive equipment. It is usually best to find a company that owns and understands this equipment and can perform this measurement. This is typically performed in an anechoic chamber.
The radio is usually plugged into a system that contains other components emitting electromagnetic radiation. It is important to test the effect these other components have on the performance of the radio. This is called desense testing. This testing usually involves setting up a pair of radios that can barely hear each other, and then turning on components in the system to see if they can jam the radio link. If an unacceptable level of jamming exists, changes to the radio system may be required.
This type of testing can be costly and take multiple weeks -- budget accordingly.
In closing …
Remember that each of these sections is only a high-level description and requires a deeper dive. It may be worth hiring a consultant to work through some portions of the design process. Often, vendors (start with the radio vendor) will suggest partners or answer your questions directly. At Shield AI, we talked constantly to vendors, consultants, and industry experts to make sure we had a thorough understanding of our state-of-the-art options. This resulted in a robust communications link upon which our customers can rely.