I have made an a fm communication system to send audio through power lines. The center frequency is 100khz or 200 khz and maximum frequency deviation is ±10khz(with no IF stage).
The modulator and demodulator work fine when connected directly with wires. Now I want to use 220v power lines as transmission medium.
If the receiver board is connected to the input and powerlines using a small ferrite toroid as an isolated transformer.
The powerline-side is 2 turns series with a 100nf capacitor and the receiver-side is 12 turns paralleled with a capacitor(the resonant frequency of 12turn side paralleled with capacitor was 200khz) I tested the filter in lab with a spectrum analyzer, unfortunately the Q factor was too low. The spectrum analyzer was 50ohm input/output put a 1k resistor in series with the input port of the analyzer to enhance Q with little effect.
The transmitter device was already off, I connected the output of the above BPF to an audio amplifier directly (without fm demodulator) and I was surprised with an irritating noise in the frequency of 4-6khz range.
After some time I discovered that the source of the noise is a nearby CFL lamp powered by the same powerline! I designed a 3 stage RC HPF with cutoff frequency of 50khz but it couldn't eliminate the 4khz noise completely and the noise power was still high.
There are multiple choices with this design:
1-I can use an active filter. I designed it using nu-hertz filter solution program.but it needed some 4.7pF or 6.8pF capacitors to which I think opamp's parasitic capacitances are comparable and I worried it may cause error in center frequency and filter response.
2- I can use 455 khz ceramic filters(and change system frequency to 455khz). but unfortunately they are hard to buy for me.another problem is that most of these filters have low BW (less than 20khz).
3- I can use a crystal filter as used in SSB communication circuits , but they are hard to design for low frequency.
Can you please help me which one to proceed?
Best Answer
With TI’s free active filter design for any filter type Bessel , Linear Phase , maximally flat Chebychev etc up to 10th order where 8th order is just a quad OpAmp and 8RC’s with choice for nearest tolerance options but you MUST DEFINE;
If you define the noise level and BW then signal level and BW and probably want Bessel for flat group delay then you will know SNR from which you can determine BER from related curves.