Generating two signals with small difference in frequency using one oscillator

I am building a data transmitter for which I chose FSK modulation. The frequencies should be around 27 MHz. I have single crystal with 27 MHz frequency and I need to get two separate frequencies that will be different.

I thought about amplifying the signal to generate harmonics, from which I will then filter out the frequencies I need. However this doesn't seem like the most elegant idea. I am not sure that this way I will get the frequencies to be close to each other. How is it done conventionally? Do they use two separate oscillators?

• A crystal frequency can be pulled/pushed (very little), so tell us how different your frequencies are, and how fast you intend to switch from one to the other. – glen_geek Aug 18 '17 at 13:39
• The baud rate is about 1000 at the moment, however it can be reduced. The frequencies should be different enough so that the receiver can differentiate between signals using simple filters, maybe several stages. – Anthropomorphous Dodecahedron Aug 18 '17 at 13:41
• A varicap comes to mind – ajeh Aug 18 '17 at 13:53
• Would varicap (basically using LC tank, if I understood you correctly) be viable at these frequencies? – Anthropomorphous Dodecahedron Aug 18 '17 at 13:55
• If you just want a small difference in frequency you can not use harmonics. Harmonics will have a frequency ratio of (more or less small) natural numbers. – Curd Aug 18 '17 at 13:57

I thought about amplifying the signal to generate harmonics, from which I will then filter out the frequencies I need.

Harmonics are always multiples of the base frequency so you'd get 27 MHz, 54 MHz, 81 MHz etc... That's pretty useless for FSK modulation.

For proper FSK modulation at 27 MHz you should be using something like 10 kHz or 100 kHz distance so 27.00 MHz and 27.01 MHz or 27.10 MHz.

That's not something you can easily do with one 27.00 MHz crystal. Crystals can only be tuned in the order of a few hundred Hertz. So you'd need 2 Crystals like 27.00 MHz and 27.10 MHz and select between the two.

There is also another solution and that is to use a mixer. By mixing the 27 MHz with either 0 Hz (DC) or 100 kHz (or -50 kHz and + 50 kHz using quadrature LO signals), after filtering around 27 MHz you'd get 27.00 MHz and 27.10 MHz (or 26.95 MHz and 27.05 MHz for the quadrature solution).

I would call the mixer solution the "classic" way of doing this.

At 27 MHz you might also consider using a DDS. A DDS is a digital waveform generator. There are single-chip DDS solutions. Some DDS chips should be able to "do" FSK I think.

After a comment about a varicap:

Indeed there is another way. You could make an oscillator which can run at 26 MHz to 28 MHz. Using a varicap we can tune the frequency.

Problem is, how to be sure that it will be running at the desired frequency ? Temperature and many other effects will make it run on all frequencies except 27.00 MHz ;-) . So how do we make it stay at 27.00 MHz ?

The solution is a PLL. Using a PLL we can lock the 27 MHz to a reference frequency, for example the 27 MHz from a crystal or some 10 MHz reference from a different crystal. A PLL locks the ratios of these frequencies, it basically wants to multiply the reference frequency by a certain factor, for example 27 (Fref = 1 MHz, Frf = 27 MHz).

But now the FSK modulation. The PLL has a feedback loop which we can influence. By making it too slow to respond to the FSK modulation the influence of that modulation will be filteered out and ignored. On average the oscillator will run at 27 MHz (the PLL and it's slow loop takes care of this) but momentarily it will be at a slightly lower or higher frequency (due to the FSK modulation). So you can just directly FSK modulate on the oscillator (with the varicap).

• Mixing means basically multiplying the two sine waves right? – Anthropomorphous Dodecahedron Aug 18 '17 at 13:52
• Yes a mixer multiplies the signals. Go and study RF mixers and you'll learn that often they do not do a complete linear multiplication. An LO signal can be a square wave so -1 or 1. Then you can multiply (mix) using switches. Yes this will generate more harmonics but these can be filtered out. A linear multiplier at 2.5 GHz (for example) is very difficult to make. Making a stage which multiplies by -1 or +1 at 2.5 GHz is much easier. So that's what's used in almost all consumer RF devices. – Bimpelrekkie Aug 18 '17 at 13:55