Skip to content

Development Boards to Go

I have a somewhat long trip coming up, and I won't be taking my workshop with me. Still, I want to be able to work on this project while I'm traveling. The decision I came to was to build a couple of development boards that are generic versions that I can use while I travel. I can slot these into the backplane, and pack just a single 12V power supply.

Goals

The project consists of two kinds of things: boards that slot into the chassis and contain some amount of UI, and endpoints that affect the world around them. I'm pretty sure I can build a single board and get most of what I need from that via some I2C magic. The hope is that this lets me have enough to work on most things while I'm away from home without having to carry a lot of extra things with me.

Universal Card

The card needs some basic components:

  • DIN 41612 backplane connector to get power and send CAN-FD signals.
  • WeAct STM32G474 Long MCU module
  • Some status LEDs for LED.

I'm planning to have the following I/O on the board:

For 1 each of the buttons, I'll use a physical debounce (discussed below). For the other, I'll do it in software. Let's see how that plays out.

In addition, because the card is fed with 5V from the backplane, we'll need a 3.3V LDO regulator, probably a 1117-style in fixed output.

Relay Card

The relay card is going to be a bit more "interesting" in a lot of ways. Because I want to be able to (potentially) handle mains voltage (110-120VAC in the US), I have to be careful and take quite a few precautions.

The board will have an AC relay (Omron G2RL-1A-E-ASI DC5) driven via a DRDC3105F. Power in will come via an C13 connector and provided out through a NEMA 5-15 receptacle.

Physical Debounce

While it's possible to do debounce in software, and quite common today to do that with the surfeit of CPU cycles most people have available, I feel like seeing what it would be like to use a physical debounce circuit. First, what the heck is switch bouncing and why do you need to debounce it? I'm gonna borrow some diagrams from this article to tell the story.

First, let's take a look at what it looks like without debouncing on a normal SPST switch:

Image showing both a switch with a pull-up resistor, and the behavior of it with both clean and dirty bounces

You can see how, due to the mechanical nature of the switch, there can be some bouncing of the output, even with a pull-up resistor. Now, let's take a look at it when you add an RC circuit:

Switch shown with an RC circuit on the output, and charts showing the discharge and charge cycle

Here we use an RC circuit where the resistor controls the charge and discharge of the capacitor, and a Schmitt trigger to ensure that we get a clean transition. The general guidance is that 20ms is plenty of time for the switch to settle, so if we aim for a 20ms time constant (τ) on the RC circuit, we get:

\[\begin{align} \tau & = R\times C \\ & = 100K \times 220nF \\ & = 0.022 \end{align}\]

Close enough for Seattle.

For the Schmitt trigger, I'm just going to use a 74LVC1G14 which are available in a small SOT package.

For the SPDT switch, however, I want to use an alternate strategy based on NAND gates in an SR latch configuration. The Cadillac of debounce, as it were.

Debounce circuit showing an SPDT switch connected to a dual NAND gate

For the NAND gate, I'm going to use the basic 74HC00, which unfortunately in the form I have is in an SOIC-14 package. But I paid less than $0.09 each, so who is complaining?