2 AC signals on one line

it Is possible but it depends on a few things, what sort of power levels are you thinking of, and what frequencies are being transmit.

if you have quite different frequencies and they`re high enough, you can use a bandpass filter for each signal and then make the equiv of a superhet receiver, of course if the power levels are too high, it gets harder as these signals could then overwhelm the front end of the reciever.

when you think about it, an ordinary radio antenna is picking up Many different stations all at once in the same piece of wire, the radio then selects what station to listen to and ignores the others, the principal is the same

and Do make sure that the frequency difference of the 2 signals isn`t the same as the I.F used in the recievers.

You don't say enough about your application to be sure of the best way to do it but traditional telephone circuits do this. You don't need any frequency conversion as long as the line charactersistics are reasonably predictable at the signalling frequencies.

Look at: http://en.wikipedia.org/wiki/Hybrid_coil

Michael Kellett

Hi All,

Thank you for the feedback.

The scenario is I am wondering if it is possible to pick up the signal that is induced in a loudspeaker by sound originating where the loudspeaker is located and split this signal from the audio signal being sent to the loudspeaker (this is not for nefarious purposes! :-)).

The signal coming from the loudspeaker (if any, I am still testing this) will be much weaker and contain similar frequencies to the signal being sent to the loudspeaker. Therefore I can't use high-pass or low-pass filters. My understanding of the telephone (old fashioned multiplex system) is that it works by splitting one line into different frequencies and transmitting at different frequencies so therefore the signal can be demultiplexed down the line, which also isn't releavant here.

I was thinking of a high-speed bilateral switch which would cut out very small slices of the sound being sent to the speaker so much so that this would not be noticed by people listening to the speaker and in those time slices pick up the faint signal being induced in the speaker and then round out that signal similar to a CD / DVD music track.

The only thing is that I doubt that the physical characteristics of the loudspeaker would respond quickly enough to this switching, so I'm stuck :-(

I'm sure this should be possible though, maybe using DSP somehow...

Clueless

ec :-)

The classic telephone system does not split frequencies (and is entirely passive) but does rely on the signals input at each end being of similar amplitude - it will not solve your problem.

If you use a loudspeaker as a microphone the signal will be very small (mV or less) but when you drive it you will using signals of the order of a volt or so. This means that the return signal is likely to be between 60 and 80 dB lower than the drive signal. To make things more dificult it is likely that most of the return signal will be echo from the loudspeaker output..

Fast switching (of anything in the system) will not help - the motion of diaphragm of the speaker is your problem here.

If you were to measure the drive voltage from the amplifier and the current in the speaker you could, in theory, separate the signals - certainly you could see the output from the speaker when there was no drive.

Your ears are quite good at this sort of thing, ie they (along with your brain) can pick out a speach signal from a music signal at the same amplitudes. Your signal processing system will need to replicate this and more. You can start with a very precise mdoel of the speaker impedance and use this to estimate the current you expect in the speaker and subtract that from the current you measure. The signal you are left with will contain components form the sound in the room from the speaker, sound which did not originate from the speaker and errors in your model. In a well damped (not much echo) room you might get  a positve signal to noise ratio at low frequencies.

You will need better than 16 bit ADCs for the measurements, a fast DSP and a lot of maths.

Michael Kellett

Hi Michael,

Thank you for your feedback - I understand the point that this is not easy - even if it is doable.

As a first step would the following make sense at all?

- take the signal coming from the pre-amp and split it so that one part goes to the amp and the other part goes to an opamp where it is compared to the (attenuated) signal on the speaker line (assuming the amp is fairly high fidelity). This way I should be able to subtract the amplifier output signal from the total signal present on the loudspeaker line and then use that as a starting point to work with.

Admittedly I would need some way of making sure the amplitude of the signals matches as closely as possible to make sure they cancel each other out as much as possible in the analog part of the circuit before any finer analysis.

ec :-)

Hello Edward,

The problem with that approach is that with a perfect power amplifer you will see nothing, in real life you will see the interaction of the imperfections in the amplifier with the impedance of the speaker. At least give yourself a chance and measure the impedance of the speaker a bit more directly. (Put a resistor in series and measure the voltage drop across that while at the same time measuring the voltage across the speaker.).

You can test these ideas easily enough.

Connect the speaker to the amplifier, turn the volume down, try measuring the signal across the speaker terminals while you make some noise - youll find you get a much better signal if you put a resistor in series with the amplifer. Of course there is an optimum value, too big and you can't drive the speaker, too small and you will not be able to measure the output from the speaker - start with R = nominal speaker impedance.

When you see how small this signal is you should give up - if not, now drive the speaker and put a network of two resistors in series, in parallel with the speaker and series resistor, so you have a bridge driven by the amplifier with two resistors on one side and resistor - speaker on the other. Adjust one of the resistors in the RR side to null the signal between the centres of the two sides. You'll find that you can get a not bad null at some frequencies but that without phase compensation you can't null very well.

Now you see why you need a model of the speaker to ballance the bridge -  and it can't be a simple passive model with R's, Cs and Ls because the speaker is non-linear.

Do you give in yet ?

OK - now you need to transfer this to the maths domain - first measure the complex impedance of the speaker at  several different drive levels, then contruct a mathematical model which can reproduce this to the degree of accuracy required to null out enough of the drive signal for you to see the "speaker as microphone signals".

At this stage you really should give up because you will easily get your PhD on the basis of this work and can start thinking seriously about an academic career.

Michael Kellett

if you ignore the electronics aspect for a minute and just look at the Physics, even if you Could do this (and speakers make lousy Mics at the best of times), any sound (air pressure) arriving at the cone would be dopler shifted out of recognition anyway.

if you tap the speaker at a constant velocity and rate, some taps will take longer to hit the cone if the cone is IN, and a shorter time to hit the cone when it`s pushed Out.

the cone itself has a finite recovery time from either position back to it`s rest place, so even if you did do time slicing Speaker one sec, Mic the next, at a rate above our detection, this Physical time it takes for the cone to move into Listen mode, will modulate the sound received, much in the same way a Vocoder works.

it would sound Well Nasty!