1

I'm looking to create several sine waves on a single circuit. All must be under 20Khz frequency and each must be unique. Mostly it will be 5-10 frequencies needed.

As I found - almost all crystal oscillator are in Mhz frequencies and only one kind is 32Khz (which is still too high).

I should be able to get this wave on the other side using FFT.

Ideas? :)

Thanks

roman
  • 409
  • 2
  • 5
  • 9
  • 2
    How about the expected spectral purity of each sine wave? That parameter could lead to different solutions. – Mario Vernari Jul 21 '11 at 11:37
  • 1
    Do you want the frequencies to be fixed or adjustable? – Jim Jul 21 '11 at 15:23
  • 1
    What's important to you? Frequency stability? Frequency accuracy? (5.000000 kHz) Signal to noise ratio? Distortion? Are you doing [multitone testing](http://en.wikipedia.org/wiki/Audio_quality_measurement#Multitone_testing)? – endolith Jul 21 '11 at 16:48

2 Answers2

6

Similar questions were asked here and here.

In this answer I talk about DDS, direct digital synthesis, which has replaced classical analog oscillators like Wien bridge. The DDS technique is crystal-based so has the same stability and accuracy.
Here you'll find a design for a simple DDS. DDSs which use special function ICs typically achieve a wide frequency range with very high frequency resolution.

Community
  • 1
stevenvh
  • 145,145
  • 21
  • 455
  • 667
0

You could easilly divide the 32KHz crystal frequency with a Binary Counter (such as the 4040) to give 16KHz, 8KHz, 4KHz, 2KHz, 1KHz, 500Hz, etc...

Then some clever filtering can create a sine(ish) wave from each of those square waves.

Majenko
  • 55,955
  • 9
  • 105
  • 187
  • 2
    Then you can better start from a **12MHz crystal**. Very common value, and many more dividers. These are the first multiples of 1kHz you can get from 12MHz: 1, 2, 3, 4, 5, 6, 8, 10, 12, 15, 16, 20, 24, 25, 30, 32, 40, 48, 50, 60, 75, 80, 96, 100,... – stevenvh Jul 21 '11 at 10:55
  • Also, a 32kHz crystal is actually 32.768kHz, so you don't get the nicely round numbers from your answer. – stevenvh Jul 21 '11 at 11:57
  • 1
    You get nice round *binary* numbers ;) 16,384Hz, 8,192Hz, 4,096Hz, 2,048Hz, 1,024Hz, 512Hz... – Majenko Jul 21 '11 at 12:03
  • 2
    Yes, but for frequencies binary doesn't make sense like it does in addressing, for instance. Note that most of the time the 32.768kHz is just used to get a 1Hz signal. (yes, I noticed the smiley!) – stevenvh Jul 21 '11 at 12:55
  • I have often wondered why it's 32,767Hz they use. I guess it's easy to divide right down to 1Hz through binary means. But why not start with a lower frequency? – Majenko Jul 21 '11 at 13:18
  • Lower frequencies mean larger crystals, which cost more, but also require more power. The 32kHz crystals use a special construction (tuning-fork) to be able to make them so small. And it's indeed because it require so little hardware to divide down to 1Hz. – stevenvh Jul 21 '11 at 13:23
  • That sound perfect! Do you have any examples on how this can be accomplished? (some circuit examples) - I come from the CS background and not electronics, so complicated stuff is little hard on me for now. Many thanks :) – roman Jul 21 '11 at 13:27
  • 2
    @roman - What Matt didn't tell you is that it's not that easy to filter out a sine from a square wave, especially not when it has to be done for multiple frequencies. I'm going to leave this to him to answer :-). (The DDS would have created a nice sine) – stevenvh Jul 21 '11 at 15:08
  • @stevenvh Gee, thanks. It's more likely Olin will be able to answer that part of it sensibly. – Majenko Jul 21 '11 at 15:40
  • @Matt - Oh no, that won't do, *you* got paid for it! ;-) – stevenvh Jul 21 '11 at 15:44
  • @Stevenvh I did? Then where did the cash go?! – Majenko Jul 21 '11 at 15:50
  • @Matt - [here!](http://electronics.stackexchange.com/users/4245/matt-jenkins?tab=reputation) – stevenvh Jul 21 '11 at 15:53
  • @stevenvh Can I withdraw some and buy a solder rework station? – Majenko Jul 21 '11 at 16:19