Working in IoT we sometimes need to handle large data streams of information, that might or might not be totally accurate. Streams might contain noise, inaccurate/unreal readings and other unwanted data.

Switch debouncing

Debouncing can be done on the hardware itself, or in software. Hardware debouncing can be done either using an S-R circuit or an R-C circuit. Two famous algorithms to do software debouncing is vertical counter and shift registers. Despite being well-known, in literature, these methods are typically presented as a code dump with little or no explanation. In this article, I will touch upon these circuits, methods and other algorithms and their use in IoT debouncing.

Understanding Switch Bounce

When the contacts of mechanical switches toggle from one position to another, these contacts bounce (or “chatter”) for a brief moment. During the first millisecond, the bounces are closely spaced and irregular, and although all of it happens in the course of milliseconds, high-speed logic will detect these bounces as genuine presses and releases.

A button release produces bounces too, but it is common for a switch release to produce less bounce than for a switch press.

Switches usually become stable after 5-20ms depending on the quality, size and electronics of the hardware.

Hardware Debouncing

Debouncing using S-R circuits

Switch debouncing using S-R circuit is one of the earliest hardware debouncing methods. In this circuit, S-R latch avoids bounces in the circuit along with the pull-up resistor. It is still the most effective debouncing approach.

The figure below depicts a simple digital debouncing circuit which is used quite often.

The circuit utilizes two cross-coupled NAND gates which aim to create an S-R latch, A SPDT (Single Pole Double Throw) switch and two pull up resistors. Then the resistor produces and generates a logic ‘one’ for the gates and the Switch pulls one of the inputs to ground.

If the switch is kept in a position as  seen in the figure, the output of the upper gate is ‘1’ regardless of the input of the other gate and the one created by the bottom pull up resistor which stimulates the lower NAND gate to zero, rapidly in turn hustles back to the other gate. If the switch moves back and forth like a pendulum between the contacts and is suspended or halted for a while in neither one of the regions amidst the terminals, the latch preserves its’ state because ‘0’ from the bottom NAND gate is fed back. The switch may move between the contacts but the latch’s output assures that not in any way it would bang back and therefore, the switch is bounce free.

R-C Debouncing

Although S-R is still common, it’s bulkiness cause problems when you try to use it frequently. You can see that it uses many hardware pieces. Another drawback to using S-R circuits is SPDT switches are more expensive than SPST switches. Thus, a new approach of debouncing emerged using an R-C circuit. The basic principle behind it is to use a capacitor to filter out swift adjustments or changes in the switch signal.

The following image demonstrates a basic R-C circuit which is used for debouncing.

It is a simple circuit which uses two Resistors, a Capacitor, a Schmidt trigger hex inverter and an SPST switch.

• In the event where the switch opens, the voltage across the capacitor which is initially zero begins to charge to Vcc through R1 & R2. The voltage at Vin is higher and hence, the output of the inverting Schmitt trigger is low (logic 0)
• When the switch is closed, the capacitor discharges to zero and subsequently, the voltage at Vin is ‘0’ and output of the inverting Schmidt trigger is high (logic 1)

At the time of the bouncing condition, the capacitor will halt the voltage at Vin when it comes to either Vcc or Gnd.

You may wonder why a standard inverter is not used. There is a problem for using the standard inverter gate here. TTL defines a zero input when the applied voltage is between 0 and 0.8 and the output in certain circumstances or situations is very unpredictable or unforeseeable. Thus, we must use a Scmitt trigger hex inverter. Thereby, the output remains constant even if the inputs vary or dither and it also ensures to prevent the output from switching due to its’ hysteresis trait.

Software Debouncing

We can debounce switches using the software as well. The basic principle is still to switch signals and filter out glitches if any. The most used algorithms used for that are counters and shift registers.

Counter Method

The first approach uses a counter to time how long the switch signal has been low. If the signal has been low continuously for a set amount of time, then it is considered pressed and stable. 

Let’s see the steps in the Counter method. 

First, we set up the count value to Zero. Then set up a sampling event with a certain period, say 1 ms. You can use a timer for that. On the sample event, Do the following things.

If the switch signal is high, reset the counter variable to 0 and set the internal switch state to ‘released’. If the switch signal is low, increment the counter variable by 1 until it reaches 10. once the counter reached 10, set the internal switch state to ‘pressed’. 

Shift Register Method

Similar to that of the counter method. The only difference is that it uses a shift register. The algorithm assumes unsigned 8-bit reg value usually found in microcontrollers

First, set up the shift register variable to xFF. Set up a sampling event of period 1 ms with the help of a timer. On the sample event, Do the following things.

First, shift the variable towards MSB, the most significant bit. Set LSB, the least significant bit to the current switch value. if the shift register value is equal to 0, set internal switch state to ‘pressed’. otherwise, set internal switch state to ‘released’. 

IoT Sensor Bounce

Recently my team has been working on telemetry involving OCR decoding of License Plates. I consider data from an OCR routine, a temperature sensor or a push button the same thing and debouncing the telemetry can be done very much in the same way.

First of all, we needed to clean up the data stream by filtering out incorrect values. Since there are not control digits on license plates, we chose to trust the result if the camera would return three similar plates within five iterations.

If you want to know more about how to debounce data streams or if you have any questions, please reach out to me: bjorn.nostdahl@gunnebo.com