February 2019

[an error occurred while processing this directive]
(Click Message to Learn More)

The Anatomy of an Edge Controller

The reality is, none of these devices contain any parts that are truly proprietary or unavailable to the public.

Part 4 of 4 - The Software

Part 3 of 4 - The Hardware

Part 2 of 4 - The Processor

Part 1 of 4 - Introduction

Calvin Slater Calvin Slater
Member of
Project Haystack

Contributing Editor

The Hardware

The Hardware

New Products
[an error occurred while processing this directive]
Site Search
[an error occurred while processing this directive]
Past Issues
[an error occurred while processing this directive]
[an error occurred while processing this directive]

When you show someone a Beaglebone or Raspberry Pi and tell them that it’s possible to use the thing as a serious controller the reaction is usually total disbelief. I know this because that’s what I do all the time at work. The type of response I get can be divided into these two basic categories:

  1. I Don’t know what you are talking about. What is a pie?
  2. OK fine, but who cares? You can’t use that thing; it doesn’t have a plastic case.

I just returned from AHR for our Open-Hardware Open-Software session where we discussed this exact subject. It was great to see a large amount of interest in this topic. The room was packed with industry professionals as well as those who were just curious. There was a major shortage of seating for an already large room, with people standing or simply sitting on the floor. The amount of engagement and interaction from the audience was also surprising and greatly welcomed. Their reaction was more or less evenly divided between:

  1. I know what you are talking about, and no you are totally wrong.
  2. I know what you are talking about, and yes you are totally right.

The truth is both opinions have validity. The difference between success and failure are what steps you take to make a particular device suitable for real use. You can’t just take one of these development boards, add a couple small pieces to it, toss it on a VAV box and call it a day. There’s a lot more that goes into a production device that is not apparent when looking at it, such as months and months of development and redesign, followed by reliability and compliance testing. What separates a serious device that you can trust to run your central plant from something that is just a toy with a sophisticated processor?

Controller Baseboard

A lot of it has to do with the form-factor; if it looks like a controller, then it is a controller. What the Single Board Computer or developer board is missing is a ruggedized baseboard or carrier board with all necessary peripheral components as well as heavy duty connectors that make it suitable for industrial use. Half of the controller is essentially missing. These hobbyist SBCs have a strong commonality with the half of a controller often referred to as a System on Module (SOM). The SOM board is usually smaller than the baseboard and contains the applications processor, clock crystals, external RAM, flash memory, power management, and supporting passive devices. These smaller PCB boards are always higher in cost per board-unit-area due to the fact they are usually more than four layers, feature finer trace width, to accommodate a greater number of high-speed signals. The small board, which is essentially the brain of the device, must have some way to connect signals to its body. On most SOMs, signals are escaped to the baseboard by routing to edge, mezzanine, or pin-type, header connectors. Each style of connector comes with its own size, reliability, and cost impact. Take a look at two board hardware concept used in Google’s Android Things platform project. Google provides a selection of certified SOMs, and then the entrepreneur-developer provides their own baseboard hardware.

Figure 1

I/O and Connectors

The most important part of an embedded edge controller is the physical inputs and outputs. At the end of the day, having a complete personal computer on a chip is totally useless unless it’s connected to interact with the actual equipment we care about. Controllers attached to equipment are typically equipped with at least five wired sensor inputs.

To receive these signals, we attach simple passive circuity to an Analog to Digital Converter (ADC). Many embedded processors have on-chip ADC modules with multiple input channels.  A factory controller uses ADCs as the basis for what is commonly referred to as the Universal Input. If a greater number of inputs are needed, additional external low-cost ADC chips can be added. Adding lots of inputs to a controller is actually very easy. We will have to scale, filter, and protect the incoming input signal before it is applied to the ADC. This signal conditioning circuitry before the ADC input is often referred to as an Analog Front End. When you set the input jumpers of a typical building controller, you are connecting your signal to the appropriate AFE circuit. Each selectable jumper position connects the input to a voltage divider circuit that will give a suitable usable scaled range.

For example, a 4-20mA current transmitter signal passing through a 250-ohm precision resistor will scale to an input voltage range of 1V - 5V on a five volt ADC device. A pull up resistor connected to a voltage reference can be used for the Thermistor and Dry Relay Contact type inputs. Simple filtering circuits using resistors and capacitors are normally placed to eliminate electrical noise from reaching the ADC. For example, a 10K ohm resistor and the 1uF capacitor will form a simple low pass filter that will reject any noise from frequencies above 16Hz. The ADC module then returns a Count to the controller software application which is a numerical representation of the fraction of its maximum detectable voltage. The mapping of data from ADC counts to physical values is a task accomplished in software. If a particular sensor requires non-linear mapping such as a thermistor, then curve fit functions, or lookup tables must be implemented in the program.

It was interesting to see at AHR in the controls area exhibits, the increasing presence of board-component-level vendors showcasing the exact items you find on production controllers. Companies who sell circuit board components such as Sensirion and Metz-Connect were right there alongside vendors with controls products that use their parts.

Network Communications Ports

The EIA-485/RS-485 standard forms the Physical Layer for many building automation communication protocol schemes. RS-485 provides an economical method to bring wired network communication to multiple low-end devices that are only equipped with a simple UART peripheral.  Any device that has a UART can easily implement an RS-485 network node. All that is needed is the addition of an appropriate transceiver and supporting circuitry to create a communicating network node. One of the main media types we have relied on all of these years to connect most of the devices is BACnet MSTP. The reason we have been stuck with MSTP for so long is device cost.  Most microcontrollers have plenty of UART modules.

Ethernet and IP enabled controllers, however, require a much more sophisticated peripheral know as Media Access Controller (MAC) and also a physical layer chip often referred to as a Phy. In the past, the MAC was usually a separate chip that had to be added to the board and greatly added to the complexity and cost of a controller. Therefore, IP connectivity was only found on gateways, routers, and large plant-type controllers. These days there are least one or two MAC modules on most SOCs, and they are fully integrated into the applications processor itself. Not only that, many higher end microcontrollers now commonly have built-in MACs. This particular trend is one of the reasons we must start saying goodbye to our dear friend MSTP. Walking the exhibit hall at AHR, I saw all of the new edge controllers from different manufacturers being primarily IP based. Some controllers had serial ports, but those were put there only to support the addition of small secondary networks. The future controls network topology will most likely be IP based. With this drastic change, however, we now have the possibility of simplifying power installs by using PoE.

Ruggedized Power

This leads to another common criticism against using a single board computer for building automation; that they do not accept 24VAC power using the typical terminal block connectors. This is actually a simple thing to fix. Designing a five-volt powered board to run on 24VAC power is an extremely easy thing to do. 24VAC power is a standard voltage for powering building automation system controllers.

Higher DC voltages can be converted to lower ones through regulators. Switching regulators are excellent for their power efficiency and are useful in cases where the voltage delta between the regulated output and unregulated input is large. Entire PCBs featuring adjustable switching regulators can be found online for only a few dollars each with all of the necessary components fully populated. These handy boards are extremely useful for powering your embedded projects. The only possible drawback to these types of regulators is the potential to introduce switching noise. Switching regulators are not well suited for directly supplying noise sensitive analog circuitry.

Figure 2 

Complete low-cost LM2596 switching regulator board for DIY projects

Linear regulators are less noisy are the most efficient when the output voltage is very near the unregulated input voltage. Of course, we must ensure that the unregulated input does not go below the regulator’s dropout voltage. Since these devices are usually half-wave rectified, a reasonably large filtering capacitor will be required. Any filter capacitor will have to charge quickly and provide the required current during the negative cycle without dipping below the regulator’s dropout voltage.  Another important consideration is the lifespan of these devices. Some types of capacitors lose capacitance over time. Therefore, when considering the lifespan of the controller, a filter capacitor should be slightly oversized.

Configuration Switches

The most important consideration in deploying large amounts of these edge type devices is installation workflow. When commissioning one or two devices, it is easy enough for the operator to manually enter the initial settings such as a static IP address and other equipment data. When this same task must be done for hundreds of devices at once, the chore of initial setup becomes a lot more cumbersome and time-consuming. At startup, each controller will be in a default low-security initial unconfigured state. Each device needs some type of unique identifier so that it can be found on the network automatically and then receive the configuration data that was meant for it. An important concept to keep in mind regarding the deployment of these devices is that the person who ends up programming and configuring the device is usually not the same person who physically installs it. Not only that, these two activities are usually performed at distinctly separate times. The addition of a few rotary encoders or dip switches can be used to address and identify hundreds of devices. This should be an adequate number for any particular network segment of a building. 

Figure 3 

Rotary decimal based encoders.

Isolation and Circuit Protection

Another issue with using an SBC as a control device is the lack of protection of the board edge I/O. In industrial environments, damage to the device can occur from multiple sources. Damage to inputs can occur from things as simple as the terminal blocks being incorrectly wired by the user to surges caused by electrical disturbances and electrical fields. A common way to protect these board edge terminal inputs is the use of Varistors, Transient Diodes, and Fuses.

[an error occurred while processing this directive]We can reduce the risk of damage to other components by adding the following devices to edge connections:

Metal Oxide Varistors (MOV) are good choices for protecting Universal Inputs. If a voltage above say 30V is presented to an input, the current is quickly directed to chassis ground.  A good discussion on MOV selection can be found here. In general through hole varistors are capable of absorbing much more energy than surface mount devices.

Transient Voltage Suppression (TVS) diodes. These are good for protecting serial communication ports from short duration voltage overages that are caused by electrical disturbances.

Fuses are included where overcurrent is a possibility such as power inputs. Fuses can also be added to the RS-485 port. Large induced currents are a possibility from various electrical disturbances due to the long transmission line lengths associated with the port.

Compliance testing

Once all of these items have been added to the board, you almost have your own home-made controller. But don’t forget about compliance testing. In the last part of this series next month we will discuss the final piece; The Software. In the meantime, consider the cautionary tale from the design development of this PCB.


[an error occurred while processing this directive]
[Click Banner To Learn More]

[Home Page]  [The Automator]  [About]  [Subscribe ]  [Contact Us]


Want Ads

Our Sponsors