Introduction

In a previous post, I described the nine modules that make up myDoorbell. These are shown in the following figure. In this post I’ll address two modules that are difficult to describe separately, the microcontroller near the center of the figure and the MP3 decoder to its right.

Basic Operation

In my previous post I described a circuit that will translate a doorbell button press to a signal appropriate for consumption by the microcontroller. When this signal is received by the microcontroller, it takes two actions:

• It raises a signal to the second microcontroller enabling it to raise external events. These events will be discussed in a future post.
• It reads a MP3 encoded audio file from the memory and writes it to the MP3 Decoder module.

The MP3 Decoder module receives the streamed data from the microcontroller and decodes it, resulting in a small signal audio output.

The Microcontroller and Software

The microcontroller chosen for this application is the Atmel Atmega 328P. This processor is simple to use, relatively inexpensive and more than meets the requirements for this portion of myDoorbell. The schematic below illustrates the circuit used for this application. The labelled input and output signals are grouped into several main functions:

• Input from the doorbell button detection circuit described in the previous post.
  • FBELL_BUTTON and RBELL_BUTTON
• Output signals to the second microcontroller.
  • FBELL_EVENT and RBELL_EVENT
• Serial Peripheral Interface (SPI) signals for communicating to the MP3 decoder and programming this microcontroller
  • BELL_SCK, BELL_MISO and BELL_MOSI
• Memory control and data signals
  • MEM_SEL, MEM_GRANT, MEM_REQ, MEM_RESET, MEM_RDY and MEM_SB
• MP3 decoder control and data signals
  • VS_XCS, VS_XDCS, VS_DREQ, VS_RESET and MUTE
• An external clock running at approximately 8 MHz.
  • EXT_CLK 

The LED and the associated current limiting resistor are available as a programmable output that is easily visible. This can be used to indicate state or for initial debugging. The push button enables the device to be manually reset. The capacitors are 0.1 µF devices that act as bypass capacitors, see Basics to Remember below. The six pin header, in the upper right corner of the schematic, is for programming the microcontroller.

The software that drives the signals identified above can be found in the myDoorbell repository on github.com. While there is considerable amount of detail, that is best addressed by viewing the source code, a brief overview is useful.

The “main” routine initializes the memory module and the MP3 Decoder module by communicating with them over the identified control and data lines. This communication takes place using the SPI protocol. This routine then loops checking the state of the front and rear doorbell buttons and requesting or relinquishing control of the memory device to / from the second microcontroller.

If the front or rear doorbell buttons are pressed, the microcontroller accesses its built in EEPROM to determine the volume at which the ringtone should be played. It then plays the appropriate ringtone for the front or rear door.

The ringtone is played by determining its length and repeatedly reading portions of it from the memory and streaming it to the MP3 decoder. When the data is consumed the device goes back to checking the state of the doorbell buttons.

The MP3 Decoder

The MP3 decoder is based on the VS1011 MP3 Audio Decoder. The following schematic is recreated from it datasheet.

The input and output signals, on the left of the schematic, are connected to the identically named signals from the microcontroller schematic already presented. The outputs on the right are the two audio output signals and the common between them. Lots of bypass capacitors and a few other components, as illustrated in the datasheet, are included.

The VS1011 is capable of consuming MP3 and WAV formatted audio files and generating low level audio output. The audio output is sufficient to drive a pair of headphones, but in our application will drive an audio amplifier and associated speaker.

A Failed First Attempt 

The first design had a serious flaw that only manifested itself when in use. This misstep led to the design described in this post. However, since this design was put into use, two others have been proposed and will be described later.

The first design had a single microcontroller that interacted with the WiFi module, the doorbell button circuitry, the memory and the MP3 decoder. There was a major issue with this approach. When the doorbell button was pressed, should the MP3 file be played before an event is sent to the user, such as a text message, or should the event be raised and then the ringtone played?

If the event is sent first it will delay the playing of the ringtone by at least hundreds of milliseconds, but often much more. This doesn’t sound like much, but when you push a doorbell button you expect to hear the bell. In practice it was very odd to push the button and not get immediate feedback.

If you play the ringtone first and then send an event it works great, unless someone picks a 30 second ringtone. In this case, by the time you receive the text message and arrive at the door, the person is long gone. I have personally found that this behavior is often a feature, but others disagree.

This issue led to the design described here where we have two processors that communicate via shared signals and a shared memory. This approach is obviously more complex and more expensive.

Physical Implementation

The schematic for the microcontroller shown above resulted in the following PCB layout.

This design required a few traces to be routed on the bottom layer. As with previous PCBs illustrated in this series of posts, this one also utilizes a 4-layer design with internal layers carrying ground and 3.3 V. The reset button is on the far left, the microcontroller in the center and its programming header on the right.

The schematic for the MP3 decoder, illustrated above, results in the following 4-layer PCB. The crystal that controls the clock for this unit is located on the left of the layout. Most of the input and output lines are located at the bottom of the board. The low level

audio output interface is located on the right. As in previous layouts, bypass capacitors and other components are placed as close to the device as possible. 

Two Alternative Designs

The first alternative design is based on the initial single CPU misstep. In the first design the problem was dealing with which came first, the raising of external events or the playing of the ringtone. A better approach may be to send the MP3 decoder the first 32 bytes of MP3 data and then start raising the external event. The MP3 decoder has a data request (DREQ) line that gets asserted whenever the decoder is capable of receiving another 32 bytes of MP3 data. If this were tied to the microcontrollers interrupt line it could be used to direct the microcontroller to send more audio data and then return to raising the event. The only question is whether the 8 MHz microcontroller is fast enough to switch between the two tasks and accomplish what needs to be done without introducing gaps in the ringtone.

Back of the envelope analysis to figure out if this would work goes something like this. Good quality stereo MP3 files are about 1 Mbyte per minute in size. Since we only need one channel we can reduce the file size to 500 Kbytes per minute. We can clock SPI data into the MP3 module at 4 MHz or 4 Mbits per second. With 8 bits per byte this translates to 500 Kbytes per second. With switching time between tasks and other overhead this won’t work, but it is likely that we can get by with far less quality than we expect from our digital stereo music. Reducing the quality will likely make this technique possible.

The second alternative design takes a radically different approach. For this approach we trade out the WiFi module, both microcontrollers, memory, and the MP3 decoder and replace it with a fast embedded ARM microcontroller that runs the Linux operating system. An example of this is the Onion, shown below.

The microcontroller runs at 400 MHz instead of 8 MHz, giving us plenty of horsepower to accomplish our task. The Linux operating system provides the task switching necessary to switch rapidly between raising external events and playing ringtones. Audio playback software can be used instead of the dedicated MP3 decoder hardware currently being used. Finally, this sort of board costs less than $20, which is less than the WiFi module used in the current approach. This approach is currently being pursued.

In my own defense let me state that these sorts of modules were not available when the first design was completed. In addition, education leads to doing things in a better way; I’ve learned a lot!

Outcomes and Artifacts

There are likely many ways to build this portion of myDoorbell and a few of them have been described. The Eagle schematic and PCB files are available from the myDoorbell repository at github.com.

Basics to Remember

Bypass capacitors are used in many digital systems. Each clock cycle a digital system has the potential to change from one state to another. When this occurs the device draws significantly more current for a very short period of time. Recall that every wire acts like an inductor and that inductors resist changes in current. If the PCB trace from the device to the power source is long or narrow it will restrict current sufficiently to starve the device from the energy that it needs during these transition periods. Bypass capacitors hold sufficient charge to supply the device with the needed energy during these short periods of time. They are recharged while the device is relatively stable. These capacitors should be located as close to the device as possible and located between the power source and the device. This close proximity will reduce the inductance of the connecting trace.

Introduction

In my previous post, I described myDoorbell’s power supply module. In this post I will describe the circuitry that translates a standard doorbell button press into something that can be easily consumed by a microcontroller.

Standard Operation

The standard doorbell circuit is shown in the following figure. Recall that when the doorbell button is not pressed (open), current flows from the transformer, through the lamp, and through the solenoid, completing the circuit. With the doorbell button open, my test setup has approximately 50 mA of current. When the button is pressed (closed) a larger current, approximately 1 A, flows through the circuit.

myDoorbell is complicated by the system requirement to use the existing doorbell button and retain the functionality of its lamp.

myDoorbell Button

The figure above can be represented by a schematic where the lamp is replaced by a 320 Ω resistor. We model the lamp with this value resistor because with 16 V_rms across it,  the current flow is ~50 mA.

myDoorbell does not need the solenoid in the above circuit. However, if we simply replace it with a wire we will create a short circuit when the button is closed. In this case, approximately 0 Ω will exist, resulting in nearly infinite current, a bad idea. So replacing the solenoid with a resistor, to limit current, is necessary, but what value still needs to be determined. A large value resistor will result in little current, which is attractive, but too little current and the lamp won’t be bright enough to be useful.

As previously discussed, resistors have resistance values measured in Ω’s, but they are also rated for the amount of power they can dissipate. The power dissipated by a resistor is equal to the product of the voltage across it and the current through it. Through manipulation of Ohms Law it can be shown that the power dissipated by a resistor is also equal to the product of its value and the square of the current through it. With this in mind, if we choose a low value resistor its power rating will need to  be increased.

For example, if we choose a 10 Ω resistor, for the top resistor in the above figure, then approximately 48 mA will flow through the lamp when the button is open. The 10 Ω resistor will have to dissipate 0.02 W, well below the 0.1 W rating typical for small surface mount resistors. However, when the button is closed 1.6 A will flow through the circuit and the resistor must dissipate 25.6 W, it’s going to catch fire!

Let’s choose a higher value resistor, say 10 kΩ. Now when the button is open approximately 1.6 mA of current flows through the circuit resulting in the 10 kΩ resistor dissipating just 0.02 W, which is great, but the lamp will be very dark, bad!

So let’s compromise. We’ll choose a 220 Ω resistor that results in a current flow of 29 mA and a lamp that is visible when the button is open. In this mode the 220 Ω resistor will dissipate 0.185 W. This will require a larger resistor, but manageable. When the button is closed the 220 Ω resistor results in a current of nearly 73 mA and a power dissipation of 1.17 W. This will require quite a large surface mount resistor, but is necessary to meet our design requirements.

Detecting the Press

When the button is open, the voltage across it is equal to 320 Ω * I_lamp. Where I_lamp is equal to 16 V_rms / 540 Ω, or 29 mA. So the voltage across the button when it is open is 9.48 V_rms. When the button is closed, the voltage across it is 0 V. What we need is a circuit that detects this difference and generates a DC voltage appropriate for an input to a 3.3 V microcontroller. The figure below depicts just such a circuit. 

To the left we see the portion of the circuit described earlier, consisting of the transformer, push button, the 320 Ω resistor representing the lamp, and the 220 Ω resistor added to restrict the current flow when the button is closed. The circuitry to the right needs explaining.

Let’s start with the component outlined in the rectangle, an optocoupler. This device consists of a LED that emits infrared light onto a light sensitive driver. When the input current is sufficiently high, infrared light is emitted by the LED, and the output of the driver is pulled to ground. When the input current is sufficiently low, the output is left in an “open collector” state where no current flows. With no current through the output resistor, there is no voltage drop across it. In this state, the output level is equal to the voltage attached to the other side of the output resistor, in our case 3.3 V. A representative device is the H11L1M. Its datasheet can be found here.

In our application, when the doorbell button is closed there is 0 V across the LED of the optocoupler and the output will be 3.3 V. When the button is open the 320 Ω resistor, representing the lamp, will have 9.48 V_rms across it. The peak voltages are -13.4 V and 13.4 V. According to the optocoupler’s datasheet, its LED cannot tolerate reverse voltages exceeding 6 V, so the external diode is added to protect it. The external diode conducts when the input voltage is negative, resulting in roughly 1 V across the LED. 

Again referring to the datasheet, we need 1.6 mA of forward current on the LED to get the device to activate, driving the output low. As the input voltage rises from 0 V to 9 V we’ll allow the output to remain high, but when it arrives at 9 V we’ll switch the output to low. This will result in a 60 Hz sequence of 3.3 V pulses while the button is open. To accomplish this we need the optocoupler input current to be less than 1.6 mA with input voltages less than 9 V, and greater than 1.6 mA when the input voltage are higher.

Imagine we don’t know the value of the 5600 Ω resistor in the above circuit. To meet the requirements described in the previous paragraph, we need to size this resistor appropriately. If there is 9 V on the left of the resistor and 1 V on the right side, due to the forward voltage drop V_F across the LED, there is 8 V across the resistor. This is explained by Kirchhoff’s Voltage Law that states that the sum of voltages around a loop must be zero. To get 1.6 mA at 8 V we need a 5 kΩ resistor, we’ll use the more common 5600 Ω device. At our peak voltage of 13.4 V, we’ll have 2.2 mA of current through the optocoupler’s LED, well below the 30 mA maximum value the device can handle. 

If the optocoupler’s output signal is tied to the input line of a microcontroller, a simple software missing pulse detector can be used to detect when the doorbell button is closed, see the sample below. In this example we check the value of the optocoupler’s output once each ms for one 60 Hz cycle, or one 17 ms period. If it is ever zero the button is open.

Physical Implementation

Our design is sensitive to the value of R3, the 5600 Ω resistor. For this reason we chose a device with a 1% tolerance. This costs a bit more, but in this case it was important. Also note that we chose a 3W resistor for R1 to ensure we don’t have a catastrophic failure, a fire!

The printed circuit board layout for the above design is shown below. The pads on the far left are for TB1 which is a connector where the two wires from the doorbell button are connected. Notice R2 in the lower right hand corner of the PCB. It is the 680 Ω output resistor and is a normal sized surface mount resistor. Compare this to the size of R1 which is the 220 Ω 3 W resistor, quite large in comparison.

This circuit results in a simple layout that doesn’t require a 4-layer PCB. However, to simplify connections to other myDoorbell modules, a 4-layer technology was used. No routing on the bottom of the board was required. The inner layers are used for ground and 3.3 V and are brought to the top layer using the three visible vias.

When integrated into the final system, small changes were made to optimize the overall layout. In the above photo of the myDoorbell button detection module, you’ll notice that the two smaller resistors and the diode have been move to slightly different locations. The final system contains two of these modules, one for the front door and the other for the rear door.

Outcomes and Artifacts

This post describes a simple circuit and associated software that meets the requirements of keeping the doorbell lamp functional and detecting when the button is pressed. There are likely many alternative techniques for meeting these requirements, but one worth considering is to use an AC line monitor chip such as the MID400 which is intended to raise a signal if the power line voltage is interrupted. In our scenario it would raise a signal if the AC voltage across the doorbell button was absent, or in other words, the button is pressed. This unit costs slightly more than the circuit described above, but would eliminate the need for the external diode and the missing pulse detector software. The output could be tied to a microcontroller interrupt line making the system more responsive, reliable, and perhaps simpler.

The Eagle schematic and PCB files are available from the myDoorbell repository at github.com.

Basics to Remember

The power dissipated by a resistor is equal to the product of the voltage across it and the current through it, P = VI. However, we often don’t know the voltage across a resistor, but rather know the resistance value and the current through it. In this case a simple manipulation of Ohm’s Law yields a solution, P = I²R.

The sum of voltages around a circuit loop must be zero. This fact is known as Kirchhoff’s Voltage Law. It is useful for circuit analysis and as used in this post to determine the proper circuit element values to meet requirements.