BFSK implementation with two oscillators

[JKJoy]

Hello,

Can I generate BinaryFSK with two different oscillators (one for mark frequency and other for space) switching with analog switch or multiplexer? How about the switching noise?

Thanks

 

[jiripolivka]

Hello,

Can I generate BinaryFSK with two different oscillators (one for mark frequency and other for space) switching with analog switch or multiplexer? How about the switching noise?

Thanks

Yes, you can but it is not a reasonable way to do.
The usual way is to use one oscillator and key the control voltage of the varicap which switches the frequency.

I saw you method many years ago in one test signal generator. It wasted power and the spectrum was not clean due to what you name "switching noise".

 

[betwixt]

Expanding on that: If you use two oscillators, their output waveforms will not necessarily be aligned (not phase locked) so switching between one and the other may result in cycles being cut short (making a new higher frequency) or extended (making a lower frequency). To some extent you can remove these additional frequencies by filtering but the better method as suggested by Jiripolivka is to frequency modulate a single oscillator. If the oscillators are producing logic level signals, you can also get some improvement by synchronizing the frequency change with the oscillator output waveform so it always completes a full cycle before the changeover occurs.

Brian.

 

[JKJoy]

Thanks Jiripolivka and Brian.
I understand it as the ContinuousPhaseFSK.
I already tried TI's CD74HC4046. But the frequency is not found stable for TSSOP package.
I had a prototype with DIP package and that worked quite fine. Also in datasheet, 0.11%/degreeC drift is mentioned.
Anyone know any FSK generation IC which is stable for industrial grade temperature?

 

[betwixt]

What are the two frequencies and the switching rate?

Brian.

 

[JKJoy]

What are the two frequencies and the switching rate?

Brian.

30KHz-33KHz and switching rate is 2400bps

 

[betwixt]

You could try starting with 330KHz (the lowest common multiple) and switching a divider between 10 and 11. That should give you a clean change-over and precise frequency generation as long as the 330KHZ is accurate.

Brian.

 

[D.A.(Tony)Stewart]

Where are your radio specs for Level, SNR, BER , Channel characteristics, Environmental and power specs?

What reasons for your choice of carrier or centre frequency, deviation ratio. Also complexity factor, size, interface & cost limitations were not given.

These requirements ought to stated first, as they affect implementation greatly.

 

[JKJoy]

What reasons for your choice of carrier or centre frequency, deviation ratio. Also complexity factor, size, interface & cost limitations were not given.

These requirements ought to stated first, as they affect implementation greatly.

My aim is to send both voice and fskdata in a single twisted pair cable. So I decided fsk above audible range, thereby center freq around 30KHz. It may travel more than a mile to reach another node.

- - - Updated - - -

You could try starting with 330KHz (the lowest common multiple) and switching a divider between 10 and 11. That should give you a clean change-over and precise frequency generation as long as the 330KHZ is accurate.

Brian.

Thanks Brian for the better idea. I think that will eliminate any need of filters also. But to implement this requires more components and thereby space.
Now what I have is 2400bps TTL out which is given to 4046 after level shifting for FSK generation.
If you can share any idea of any single chip solution for the above mentioned divide by 10 and 11 switching
would be helpful.

 

[betwixt]

I'm "between workshops" at the moment while my old house is being rebuilt and all my data sheets and parts are in temporary storage so I have to go from memory/intuition.

I can't think of a single chip divider like that although it might be possible to use a digital divider and feedback gating to make one out of two ICs.

Other possible solutions which would have a low part count would be:
1. to use something like a 12F1822 PIC in PWM mode. Using it's internal oscillator you can make hundreds of frequencies by programming the two PWM registers. I *think* (no data sheet here) the nearest you can make is 29.85KHz and 33.057KHz.
2. use a PIC with an NCO module which will let you set precise frequencies.
See **broken link removed**

Note that these PICs have internal clock oscillators so in theory you can simply feed your data to one of the pins and have the frequency pop out of another. They are 8-pin and very inexpensive. Some even have built in UARTs although that might just change your problem to one of getting the data to it instead.

Brian.

 

[JKJoy]

Other possible solutions which would have a low part count would be:
1. to use something like a 12F1822 PIC in PWM mode. Using it's internal oscillator you can make hundreds of frequencies by programming the two PWM registers. I *think* (no data sheet here) the nearest you can make is 29.85KHz and 33.057KHz.
2. use a PIC with an NCO module which will let you set precise frequencies.
Brian.

Thank you Brian for your valuable suggestions.
Let me check the feasibility of implementing these.

 

[D.A.(Tony)Stewart]

My aim is to send both voice and fskdata in a single twisted pair cable. So I decided fsk above audible range, thereby center freq around 30KHz. It may travel more than a mile to reach another node.

- - - Updated - - -

Thanks Brian for the better idea. I think that will eliminate any need of filters also. But to implement this requires more components and thereby space.
Now what I have is 2400bps TTL out which is given to 4046 after level shifting for FSK generation.
If you can share any idea of any single chip solution for the above mentioned divide by 10 and 11 switching
would be helpful.

FSK may work, but not without highly compressed mp3 audio it will be poor. Telephony uses 64kbps with an 8 bit u or A Law log ADC/DAC that gives > 60dB SNR over microwave or satellite or even paired wires using duo binary biPhase encoding using +24V,0,-24V

Raised Cosine filtering improves the Reception from ISI even more.

When using FSK with low SNR, you also want as wide a deviation as possible on a higher carrier. This yields an improvement on CNR to SNR by root of deviation ratio of modulation to bandwidth in receiver.

 

[betwixt]

I think the idea is to send the audio as analog and the data as FSK just above audible frequencies. On a dedicated cable it should be possible although over a mile I would expect the higher frequencies to be quite heavily attenuated. With a 10% frequency shift the effects of phase distortion along the line can probably be ignored, especially as such low data rate. I would guess nothing more than LPF/HPF and maybe some amplification should allow recovery of both voice and data quite easily.

Brian.

 

[JKJoy]

I think the idea is to send the audio as analog and the data as FSK just above audible frequencies. On a dedicated cable it should be possible although over a mile I would expect the higher frequencies to be quite heavily attenuated. With a 10% frequency shift the effects of phase distortion along the line can probably be ignored, especially as such low data rate. I would guess nothing more than LPF/HPF and maybe some amplification should allow recovery of both voice and data quite easily.
Brian.

Yes, The idea is to send the audio as analog and the data together in a single cable. And data as FSK just above audible frequencies. I used 74HC4046A, eventhough it has some frequency drift over time.
With driving section it transmit the FSK with around 350mV rms. Output has an isolation transformer.
I just tested how far it can send data using 305mtr CAT5 cable bundle. I joined the 4 pairs to become 1220mtrs. It communicates successfully.
Since the cable is coiled as bundle, is there any chance that this 30Khz signal can induce among twists, or one pair to another?

 

[D.A.(Tony)Stewart]

How much jitter do you have? Can you reduce signal level until errors ?
Repeat with twice the deviation then 4x.

Yes coiled cable is not the same impedance and coupling will occur.

 

[JKJoy]

How much jitter do you have? Can you reduce signal level until errors ?
Repeat with twice the deviation then 4x.

Thanks.
But I dont have much analog understanding to measure out jitter and all. At receiving end I can hardly recognize FSK plunged in some noise signals.

 

[betwixt]

Although jitter can be an issue, I think it is unlikely to cause problems at such low frequencies and data rates.

The best way to do this with CAT5 is to send the signals differentially, in other words send it 'normal' on one wire of a twisted pair and 'inverted' on the other. The idea is that if you call your combined audio/data stream 'AD' and interference picked up by the cable 'N' , one wire carries AD and the other -AD. At the receiving end you pick up AD+N and -AD+N so inverting the signal on the wire you previously inverted then adding them gives (AD+N) + (AD-N) = 2AD. You get twice the signal and the interference which will be the same on both wires is cancelled out. The balanced magnetic fields on the wires also minimizes signal radiation from the wires to other wires/equipment.

I've seen analog signals up to 10MHz carried over several hundred metres this way and it works very well at a fraction of the cost of co-axial cables or wireless links.

Brian.

 

[JKJoy]

Yea, I understand the advantages of differential lines.

Thanks for the valuable suggestions.