Steaming audio over winsock
Posted on 2004-03-30
I have worked with winsock in VB many times before, but never for a purpose as complex as this.
I have 2 programs, serverside and clientside (running on seperate machines.) There is already a TCP winsock link between the two, sending simple text strings between each other. I want to setup up a second winsock link and with this i want to stream all the audio from the client to the server. This doesn't just mean Wave audio, but also Mic and Line-In audio too (just them three will do, but if its easier to stream all the audio then thats ok too.) I understand how to setup up the winsock link and connect, but the audio is a problem.
I assume that sound is handled at 44.1Khz 16bit (could be wrong,) if it is handled this way, then there is no need to downsample it (although downsampling it to 22.05Khz 8bit would be ideal if its not too difficult to code and not too CPU hungry when running, as this will cut down on LAN traffic.) However, the audio should preferably be converted to mono before sending (this isn't absolutly essential, so if its much easier left as stereo, then it can be left as stereo.) The audio then needs to be sent over winsock. When the server receives the data, it should output the audio to the default sound device (and so play the clients audio out through the servers speakers.)
The audio needs to be constantly streaming with as little delay as possible.
I am only assuming winsock is the way to go with this, I cant think of any other methods of communicating the data, although if there is a better way that i dont know about then im open to suggestions- It'll be another lil bit of VB knowledge for me :)
Really needing help with this please. Ive been trying to figure this out for ages but i don't even know where to start with the audio bits n pieces. I figured the way to do it was for the clientside program to somehow capture the audio stream and convert each sample to an ASCII charactor that represents the sample, then send this over winsock (or wait for a certain number of samples to be processed and compile a string representing all these samples, and then send it.) On data arrival, the server would convert the ASCII string back to samples, then somehow output each sample one at a time to the default sound device at the same frequency the client is processing at.
I don't know if thats at all close to how its really done, but it was just how I imagined it.
Another method at the server side would be to write the incoming data to a wav file and play the file (I came up with this method as i didnt know how to directly output to the sound device, but i knew how to play wav files in my programs.) You'd have to use two files, one to write to as the other plays, and keep switching between the two. The problem with this method is that with the constant file reading and writing, there may be slight delays which would not only cause hiccups in the audio, but may accumilate and cause a forever-increasing audio delay (and thus hogging memory as the line of queued samples gets longer and longer.) Plus, there may be slight delays between playing the first file and the next file. I suppose you could just cause the file playback to be asyncronus and play each file on time whether the other file had finished or not, although you still have the delay caused by reading and writing to harddisk and now you have slight overlaps before each file finishes, so consequently you end up with poor audio quality as it hiccups thousands of times a second.
Anyways, back to the problem in hand....
Thanks for your help in advance,