Recently, I have to do some intercom equipment and test, each time I speel the microphone, I can always hear myself from the speakers in front of them. If you think about occasionally call, you will have a similar situation. Just don't know why your phone has your own speech, I just felt the phone problem, or the signal skewers, I didn't think about it. This is a face-to-face encountered this phenomenon. After understanding this is one of the most difficult problems in the design of the system, echo. After a period of fighting and echo, there are many experiences in echo elimination, and this experience is shared to everyone today. It is mainly used to use the tone test board.
This time in order to share the complete design, the tool I use is Easyeda. This website is very good for hardware design open source, and GitHub is a bit similar, but focus on open source hardware design, and individual feels more than Github's people, GitHub only has English version, and the born is tailorned for the gods. GitHub focuses on the software's open source, with strong version management functions, fully records the modification history of each text format, and compares the difference between the different versions of the file, and supports online reading code. However, this powerful management function is limited to files in text format. For files in non-text formats, GitHub believes that it is a file, only logs its version, if you modify this file is not detailed, It is very troublesome when you need a rollback version, and GitHub is online browsing that does not support this file. Easyeda supports online hardware design and has good project management and version management. Picking two pictures Obvious explanation Easyeda because the following is mainly on it.
I spent a dotted water and describing a design tool. Workers must have a good thing, so I take some time to popularize some basic knowledge to the partners who have just met the echo.
Solicitor's echo principle
Its basic principle is that, as shown below, the sound source (user speech) is collected from the microphone of the local device, sent to the remote device through the network, the remote device plays the sound of the sound source from the speaker (usually enlarged), if The remote device does not make a cancellation of the sound process, then the sound of this remote device is played by the microphone of the remote device to be collected by the microphone of the remote device, and is transferred back to the local device via the network and enlarged, from the local speaker playback process .
The simulation intercom system has fewer intercourse questions, there are two reasons: 1. The transmission delay is small, even if there is a sound, it is usually hidden in his own speech, and the user is not easy to detect. 2. Due to the first reason, the unique sound processing of the simulation system is also relatively simple (with partners who know the echo cancellation algorithm should know the key parameters of this echo return, this is not discussed here, so don't repeat it) Directly subtract the output and input to subtract, and the cancellation effect is basically guaranteed. We often use the simulation intercom system with traditional fixed-line systems, while IP phones are (numbers) network intercom systems.
Talking about echo and howling difference
The figure below simply describes the formation of howling, the basic reason is that the audio collected by the microphone is broadcasted from the speaker after zooming. If the speaker is relatively close to the microphone, then the sound wave passes back the microphone, the microphone will enlarge the sound Re-acquisition and send it to power amplifier, then broadcast from the speaker, with this, the sound is getting bigger and bigger, from the perspective of the observer, the more speakers, listening to the harsh loud noise. This and the tsunami is a bit like, the earthquake wave pushes the waves, and later waves will be superimposed on the waves in front, and finally the higher the higher, form a giant waves, causing an amazing destructive power.
The echo can also cause a whistle, but this back-off path is not only including the space (the acoustic echo propagation path), but the reason is the same because the reason is the same, there is an audio loop, and there is a ring Enlarge.
Remove the sound common means
As early morning, the algorithm is not perfect. In order to handle the echo, whispering this kind of audio problem, the network intercom has to use a single-time intercom in order to bypass them, that is, when I talk, I can't listen to you, think about it, Taking a drop in the hospital, if the needle water is used, you need to call the nurse on the edge, you need to press the button to get the permissions of the speech, and the nurse has heard the call, and you need to press the button to recover. Half-duplex intercom system is such a tangled system.
Currently, the counting of the audio algorithm has made great progress. It is not necessary to use half-duplex's footwife to avoid the problem of opening and echo. It is understood that there are three processing methods, 1, using the intermittent dedicated cancellation These libraries are like WEBRTC or FreeSwitch to count the sound source version, and there is also a feewitch's resurrection of the Library business version; 2. Use Fudi's cancellation sound chip, this chip is actually a dedicated DSP plus curing The cancellation of the tone program (firmware) and opens the parameter adjustment interface; 3. Use the DSP to run independently developed canceled algorithm, because such algorithms involved more floating point operations, usually require high-performance DSP chips to support. The first way to extract the sound effect, but the usual adaptive algorithm needs to converge time, which converges relatively long. The second way is simple and easy, and the algorithm converges fast, and the effect can meet the requirements of most usage. In the third way, whether it is the highest effect or convergence time, but the technical threshold is high, the cost is also high.
FM1188 Dissuction Test Paper Design
A large section of Science has finally entered the theme, how to design the FM1188 to count the tone test board.
In fact, the programs of Fu Di in the unus of tone are still useful, and there are also small modules that have already burned the parameters on Taobao. If the following figure, it can be directly embedded in the circuit. However, this module is not convenient to debug parameters, and today's debug board fully considers the various situations of debugging.
As shown in Figure 2, each of the schematic diagrams placing a functional module of the board, and then tells the functions and design considerations of these functional modules. (The whole picture may look a relatively vague, there is a need for the partners to log in to the project link to discuss)
1, ALC5640 audio decoding circuit
The Alc5640 is a high-performance multi-channel codec chip introduced by Realtek. There are two groups of I2S interfaces that can be used as AUDIO HUBs, which can truly implement a full-duplex operation, which is encoded while coding. Moreover, small volume is usually used on mobile devices.
In order to ensure the sound quality, it is necessary to reduce the power supply noise, and the codec chip of the debug board is powered by a separate linear power supply, as follows. Each module on the board is important to keep low noise. When debugging, it can be better evaluated, designing the inhibitory effect on the true noise (spatial acoustic noise).
The ALC5640 built-in dual channel D power is used, the maximum support 5W power output, the output must be added filter circuit, as follows:
All control signals, audio input and output are introduced, and the package is packaged with a two-row pin of the 2.54mm pin, which is convenient to use DuPre line as needed as needed, as follows:
2, Amplifier power amplifier circuit
The amplifier module design contains two different output power amplifier chips, where the LM4950 is 7W, and the LM4871 is 5W, and the two chips have a mute control function, which is better than the subsequent debugging. In the case of the negative cancellation effect.
Like the audio codec module, the power supply is powered independently, reducing power supply noise; all control signals, audio input and output are introduced, and the package is used with a 2.54mm pin spacing, which is convenient to use DuPont line according to The interconnection circuit is required, as follows:
The LM4950 and LM4871 internal principles are similar, all of which are bridged to push the speaker work, as follows, without external connection, large capacitance, this is an OCL (Output Capacitor Less) power amplifier circuit.
3, FM1188 Qiji Touch Module
FM1188 is a language processing chip introduced by Fortemedia, which reduces environmental noise, or collecting sound, eliminating tone, etc. Its internal cancellation calculation algorithm or the cancellation logic is as follows:
Because it is a debug board, the configuration is unfained, increasing the jumper interface, and configures as needed when it is debugged. The configuration includes two types, one is the configuration of the parameter download mode, and the FM1188 supports 4 parameters download mode: 1, from 256 Byte EEPROM read parameters; 2, read parameters from 1 kbyte EEPROM; 3, read parameters from the SHI interface ; 4, read parameters from the UART interface. The modules sold on Taobao usually burn the parameters into the EEPROM, and the parameters are usually not much, and then configured to download the first parameter download mode, convenient for customers. There is also a configuration of the burn parameter location, which can burn the program to the different addresses of the EEPROM depending on the configuration.
The so-called configuration refers to setting the level when the function pin is powered on, reserved a two-row pin of the distance between the 2.54mm pin spacing, and then using the jumper cap can be jumped, as follows.
The volume adjustment pin is introduced, and the volume can be changed at any time when it is easily commissioned, so as to observe the cancellation effect in different loudness.
The schematic module is these, and then draws the PCB. The PCB map design is also implemented on Easyeda (the same thing is similar to the work, the partners interested in the tools Baidyeda, here is not detailed), herein indicates the attention of the design matter. First, the overall renderings are as follows:
After the model is designed, the effect of the production of the board is also able to preview with Easyeda, as follows, is it really like the model of the sample?
The final sample is as follows, because of the proof of the company channel, some company information, so the picture has played mosaic.
The microphone flux FM1188 performs sound acquisition and takes into account the echo, and noise; FM1188 processes the audio signal to the audio code decoding chip ALC5640 LINEIN for collecting and encoding; audio codec chip ALC5640 output audio Lineout on one aspect On the one hand, it provides a reference for FM1188 (no reference source is unable to eliminate echo, otherwise how FM1188 knows what speaker broadcast?)
Search for "Love Bo.com" to pay attention, daily update development board, intelligent hardware, open source hardware, activities, etc., you can make you master. Recommended attention!
[WeChat scanning picture can be directly paid]
Related Reading:
This kind of cheap USB device is perhaps the best way to deal with account hijacking
Our other product:
