Current solutions to these kinds of “announcing system” are limited by the OTP (One-Time Programmable) voice chip with small capacity (normally use EPROM as storage media), not to mention the relatively expensive price (the price of OTP voice IC is determined by it’s capacity – voice recording time, normally 10s~200s). For instance, the OTP chip AP89010 (manufactured by APlus) is $0.48 (10 seconds OTP), and AP89341 is $2.43 (150 seconds OTP), and one announcing system may use multiple voice chips (2-10pcs) which also results in the complexity of hardware design. On the contrary, the price of standard SD card is going cheaper with the rapid development of storage technology, 512MB SD card (Kingston) only needs $3.65. Compared to OTP chip, 512MB storage can hold as many as 128 songs with MP3 format, can contain 10 lossless songs (WAV format), and the key point it that SD card can be easily formatted, songs or recorded files can be modified at will – as long as you have a PC supports FAT filesystem.
As mentioned above, the announcing system using OTP chip solution has a complicate circuit – different audio announcements need to be played from different chips, the control logic is quite a burden which definitely increase the circuit complexity including the PCB area. Quite the opposite, our system only need 6 main wires (SPI bus: 4 wires, PWM output channel: 2 wires) to play the songs or any recorded wave files. AVR chip can be purchased as cheap as $4 (ATmega328). So the whole system can be very cost-effective.
High Level Design
There are many MCU-Based MP3 player, however, these kinds of MP3 player solution not only needs extra hardware decoder, but also needs DAC chip, and the circuit is complicate. How to design a chip music player while reserve the audio quality? To substitute the MP3 decoder, we can simply use WAVE file which format is simpler than MP3, and it can be easily decoded by MCU. To substitute the DAC chip, PWM can be used to replace it with simple R-C filter, such as the Cricket Call Generator. Based on this kind of idea, WAVE player based on AVR wais born.
Since SD card is used as the storage media, except SDIO, the only way to commuinicate with it is SPI. Generally, SPI has four working modes.
To operate SD card using SPI, the working mode must be specified. In “SD Specifications Part 1 Physical Layer Simplified Specification“, Section 7.2 mentions part of the SPI protocol, but doesn’t specify any details. After several experiments, the suitable SPI working modes for SD card communication is MODE0 and MODE1.
Meanwhile, instead of reading RAW data from SD card, FAT16/32 file system is implented, so that WAVE files can be easily stored into SD card via any PC with SD card reader socket. The FAT file system is open source “Petit FatFs” which is a simpler version of popular “FatFs” developed by ChaN.
With knowledges above, I still needs to get familiar with the format of WAVE file which is actually a subset of “RIFF” format developed by Microsoft. RIFF audio format will be introduced in detail in software section.
At a high level, the logic structure is much simpler. It contains “Storage Module” and “PWM-DAC Module. With embedded hardware SPI interface and PetitFatFs module, WAVE files stored in SD card can be retrieved. Then ATmega1284 will parse the WAVE file and get the pure audio data which is used as the PWM output (4 channels). Four 8bit PWMs are divided into two groups (corresponding to audio left channel and right channel), and combined separately to two 16bit “PWM-DAC” outputs. These two output channels are finally connected to the loudspeaker.
User Experience (UE), Hardware (HW) and Software (SW) Trade-offs
There always exists trade-offs in UE, HW and SW. This project actually shows how to appropriately handle these trade-offs appropriately.
UE and SW
There exists two methods to retrieve audio data from SD card. We can write RAW audio data into SD card with many 3rd-party software tools. In this way, the software design is pretty easy: all MCU needs to do is just read audio data via SPI interface, and forward the audio data stream to PWM channels. However, it will be very annoying to replace audio data in SD card.
The other method is to use standard file system so that SD card can be operated normally in any PC with FAT support. With this method, it is very convenient to replace / update audio files, the difficulty is that a lightweight File System must be implemented so that MCU is able to retrieve data via standard FAT16/32.
HW and SW
MP3 is a popular audio format with smaller file size and good audio quality. If we use MP3 files as the audio source, then a hardware decoder must be implemented. ATmega1284 or any other low-cost 8-bit MCU is not able to guarantee the playback quality since it needs to parse the complicate compressed MP3 file. On the contrary, the format of WAVE file is easier: it is composed by a simple file header followed by RAW audio data which can be directly output to PWM channels (16-bit audio data needs simple subtraction operation).
Another advantage to use WAVE file is that it is lossless! As long as we can guarantee the playback quality (output audio data in exactly the same frequency as sample rate), theoretically speaking, we can achieve very high audio quality. Even though WAVE file is much bigger than MP3 file, but it would not be a problem: SD card is so cheap these days.
- SD Specifications Part 1 Physical Layer Simplified Specification (version 4.10);
- Microsoft Extensible Firmware Initiative FAT32 File System Specification (version 1.03);
- Microsoft New Multimedia Data Types and Data Techniques (version 3.0);
- ANSI-C Specification (ISO / IEC 9899).
- GCC Manual (version 4.9.2)
Petit FatFs system is a lightweight Fat File System which supports FAT12/16/32, and it is developed by ChaN. Petit FatFs is specially designed for those MCUs with low RAM capacity.