According to the datasheet, you should give it a square wave with a duty cycle of 50% at a frequency of 2400 Hz with an amplitude of 5 V. You could make that with a 555 timer chip, 2 resistors, and 2 capacitors.
More details here.
Yes, that was Matt who told you about the filtering to get a sine from a square wave; unfortunately he forgot to tell you how to do this. :-)
Well, it's not easy. A 1kHz square wave will have a 3kHz harmonic with 1/3 amplitude, a 5kHz with 1/5 amplitude, etc. That's a lot of energy in the harmonics which you want to get rid of. You need a sharp low-pass filter to be sure that the 3kHz is attenuated enough. That's possible, but not so easy to make it variable; for the 2kHz sine you want a higher cut-off frequency.
The plot shows that the signal still looks more like a square wave than a sine even with all except the third of the harmonics removed.
All this suggests that this may be an easy approach to generate different crystal-stable frequencies from a single 12MHz oscillator, but not sines. For sines there are analog solutions like the Wien bridge, but these aren't so frequency stable.
Like I said in my other answer DDS really is the way to go; you get the best result for the money. If you don't want to use a microcontroller you can use a lookup ROM programmed with a sine waveform, let a counter run over all addresses, and feed the output to a DAC. That requires a little bit of logic, but no uC. The clock you use for the counter will be a multiple of the sine, so feeding a different clock frequency will give you a different sine frequency.
All this said, this may cost as much as a microcontroller; there's no reason to dismiss a microcontroller because it would be too expensive. uCs are dirt cheap these days and often a more economical solution that analog alternatives. You won't find anything simpler than Jesper's generator (was linked here http://www.myplace.nu/avr/minidds/index.htm, link now broken)
If you want to generate the 5 signals simultaneously and mix them the DDS solution is even more cost-effective: you'll need only one (1), vs 5 oscillators in an analog solution.
edit
If you only need 5 fixed frequencies you might want to switch them on and off selectively, so that you can mix them. Easy with the DSS uC: instead of keeping one phase accumulator (expensive word for "counter"), you just keep 5 of them, and add the sine values before sending them to the external DAC. You could use 5 switches to turn them on and off. You don't need the MAX232 then. Like I said: cheap.
Best Answer
Low frequencies and directionality don't mix very well:
From http://en.wikipedia.org/wiki/Sound_localization#Evaluation_for_low_frequencies
Now here is what I would do (I'll be making a few assumptions). Rather than building a intricate device to determine where the vibrations are coming from do the following when you hear the vibrations:
If the noise is gone you now know it is electrical in origin (just to be sure). Even though you said you switched off everything you did not say exactly how so make sure it is EVERYTHING by doing the above. If you still hear it, go outside, walk around, where do you hear it? By narrowing the parameters down of where you hear the noise will help you to easier locate the source.
If this does not help I would try to use one of these android applications: https://market.android.com/details?id=radonsoft.net.rta&hl=en
https://market.android.com/details?id=com.DanielBach.FrequenSee&feature=also_installed
You should be able to follow the low frequencies depending on the quality and sensitivity of the microphone inside the device. I suggest testing this with your PC first before trying it on the real problem: http://onlinetonegenerator.com/
Here are some devices I have found to cause low frequency vibrations:
Best of luck finding that buzz!