I created a two differrent signals on my android phone, one to act as the clock signal (lower signal), and the other to be the data signal (lower signal), but the problem is that the audio the signal always drifts towards the "ground"
Why does it do this? Note that my signal isn't really a square wave it's more of a PWM.
The signal from the headphone jack is most likely AC-coupled.
The easy way to fix this is to add a diode from each signal line to ground: anode of both diodes to ground, cathode of each diode to your signal line.
The diodes act as a clamp and allow the signal to NOT go lower than about -0.6V. The remainder of the signal amplitude is above ground (positive).
Note that this works ONLY if the output is AC-coupled.
The reason I mention this is that the headphone signal on your phone might be AC-coupled but phones of different models or from different manufacturers might NOT be.
There is a way to fix that, though.
That'll be because you're creating an audio signal - an audio signal is AC coupled, and will always want to drift towards ground since ground is what it is centred around.
An android phone's headphone socket cannot create reliable digital waveforms since it's an analog output. You should instead choose something that is suited to what you want, instead of trying to use something that is completely wrong and getting terrible results.
To solve this problem you can instead modify the encoding so the sum of 1 and 0 bits actually sent over the wire converges to equal and the number of consecutive 1s and 0s is minimal. This means sending some extra bits over the wire but it will remove the DC component that the audio hardware can't handle. This means using a constant weight encoding or Paired disparity encoding.
Constant weight in it's simplest form that means sending a 01 for 0 and a 10 for a 1. This scheme has the obvious downside that desynchronization during decoding/capture means scrambling the entire message.
Going more advanced with paired disparity means creating 2 packets per symbol you want to send, one with more 0s and one with more 1 and then dynamically selecting based on the difference between the number of 1s and 0s already sent.
You can compensate for this in software!
As you can see from the shape of the signal (the curve that is slowly pulling your signal to ground), there is a time component involved. You can offset it in software by emulating the time constant and compensating for it.
This ability will be slightly limited:
If you lower the base amplitude of the signal, the available headroom can be used for a longer time, giving you the ability to compensate for longer pulses. But you have to keep in mind that your time constant compensator will have to recenter itself on 0V eventually, otherwise the DC bias will eventually creep to maximum positive or negative amplitude and then clip and distort.
So the idea is to come up with a filter that will boost what the headphone amplifier is limiting.
Here is an example in which I superimposed a triangle wave on top of the square wave. Note that this example is a simplification because the voltage of the triangle wave normally depends on the voltage of the square wave. This circuit will only work with that specific square wave.
The values were obtained empirically. The square wave at the left represents your signal. The two identical clusters on the right represent the capacitive coupling of your headphone amplifier. In the upper part of the circuit, the signal is sent through unaltered, resulting in the waveform in the first graph.
The lower part of the circuit adds a triangle wave on top of the signal. The phase is matched with the square wave. The op-amp part is simply a follower amplifier that attenuates the signal a bit so that the output is not too loud for the headphone amplifier.
By superimposing the triangle wave on the original signal, the output will look more like a square wave, like in the second graph.
The third graph shows the output before it is filtered by the capacitive coupling.
So what you should try to do is reimplement the triangle wave in software. Beware of the pitfalls!
Edit: and as Dwayne Reid pointed out, this capacitive coupling is specific to your phone. Other devices might have different characteristics, or not have coupling at all. So this is a solution for your device only.
