2
\$\begingroup\$

I am looking to create a high q passive notch filter using a crystal. I am not very familiar with crystals, hoping that someone who's had a similar experience can enlighten. I know crystals are mostly used for generating signals, but I mainly need to filter. My working frequency is 40KHz

I know there are such parameters as series resonance and parallel resonance.

  1. How do I pick the right crystal for the job? The datasheets don't always state whether the crystal is designed for series or for parallel operation.
  2. What is the actual typical difference in frequency between the two?
  3. Is it possible to use the same crystal both to generate my signal and also to notch filter it? I read something about "pulling" the resonant frequency.
  4. How would you build a notch filter with a crystal?

Thank you so much.

\$\endgroup\$
  • \$\begingroup\$ What's the required notch frequency? \$\endgroup\$ – Chu Jun 2 '16 at 9:15
  • \$\begingroup\$ Required frequency is 40KHz (I can work with up to about 1KHz deviation) \$\endgroup\$ – Ilan lewin Jun 2 '16 at 9:17
  • \$\begingroup\$ How high is the Q factor you need? \$\endgroup\$ – Peter Smith Jun 2 '16 at 10:45
  • \$\begingroup\$ I guess somewhere between 1K and 10K, I don't worry much about position of the 3dB point, but I need to remove about 40-50 dB off the carrier \$\endgroup\$ – Ilan lewin Jun 2 '16 at 11:49
  • \$\begingroup\$ For lower Q, consider ceramic resonators/filters. \$\endgroup\$ – user2943160 Jun 2 '16 at 15:45
4
\$\begingroup\$

Crystals have an extremely high quality factor, making them only suitable for very narrow bandpass / bandreject filters.

The quality factor is the ratio Q = Centre frequency / Bandwidth.

For a crystal Q can easily be 100000 so for a 40 kHz crystal that means 0.4 Hz ! Is that what you want ?

1) Crystals aren't designed for parallel or series resonance, they have both ! It depends on how you use them what will be the dominant mode of operation.

2) that depends on the crystal, I was working with a 25 MHz crystal recently and there the frequency difference was less than 1 kHz. Typically series and parallel resonance frequencies are very close together.

3) No, once the signal is generated you cannot "pull" it to a different frequency. The only way to slightly change a crystal's resonance frequency is to add capacitance in parallel with it. But that way you can change the frequency only very slightly. Like +/- 10 kHz for a 25 MHz crystal.

4) If you google "crystal notch filter" and select to see images only you will see lots of examples. Note that they're almost all for frequencies in the MHz range !

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Thank you FakeMoustache. Yes, I think that width could work for me. I have a very strong carrier that I need to get rid of. Let's put "pulling" aside for a moment. I looked it up on Google, but couldn't get a good reference to how I would go about building such a filter with a crystal. What other components would I typically need? \$\endgroup\$ – Ilan lewin Jun 2 '16 at 12:01
  • \$\begingroup\$ If that carrier is EXACTLY at the frequency of the crystal then you will be in luck. If not and/or the carrier is not a single frequency (it might have modulation, it might vary in frequency) then this crystal solution will not work. I would go in the direction of an LC notch filter or an active (opamp) filter. \$\endgroup\$ – Bimpelrekkie Jun 2 '16 at 13:50
2
\$\begingroup\$

No this won't do what you want. The series resonance (low impedance) point is usually in all cases very, very close to the parallel resonance (high impedance) point so what you will get is this: -

http://www.learnabout-electronics.org/Oscillators/images/Crystal.jpg

So at very low frequencies you get nothing through your xtal then, as you approach series resonance you get signals passing through it. Then within a hair's breadth it goes very high impedance and no signals will effectively pass then gradually, as you increase frequency signals start to get through.

Not really a notch filter.

Given the flaws in your idea I don't think it's important to go any further.

| improve this answer | |
\$\endgroup\$
  • 1
    \$\begingroup\$ I know crystals are being used as notch filters, so what am I missing? \$\endgroup\$ – Ilan lewin Jun 2 '16 at 11:20
  • \$\begingroup\$ Can you provide a link to a design - maybe there is something "special" about the crystal? \$\endgroup\$ – Andy aka Jun 2 '16 at 11:30
  • \$\begingroup\$ Yes, here is one thing that I found that is similar to my need ve2azx.net/technical/XTAL_NOTCH_ckt.pdf \$\endgroup\$ – Ilan lewin Jun 2 '16 at 11:33
  • \$\begingroup\$ That's quite a complex circuit and I am confused about "L" - they say "Select L to resonate above and below XTAL frequency" and this makes no sense to me. Are you sure this diagram is a proven design and not just somebody's pipe dream? \$\endgroup\$ – Andy aka Jun 2 '16 at 11:41
  • \$\begingroup\$ Don't know, here is another example la3pna.wordpress.com/2015/02/01/… \$\endgroup\$ – Ilan lewin Jun 2 '16 at 11:44
1
\$\begingroup\$

There are a number of manufacturers of theses explicitly designed for the purpose that you desire. It was not an unusual application in the early days of radio but now a days less so. They were used as bandpass filters in IF stages of super-het receivers. Look also for ceramic filters as well which tend to work at lower frequencies for high q. Vectron is one example of a manufacturer, but if you look at their catalog you can see what your issue is going to be ... these are usually available in the MHz range, not the KHz you desire.

Crystals in the KHz range tend to be tuning fork type (watch crystals - 32,764 Hz) and I know you can custom order those for other frequencies but I can only find 32,764 as an off the shelf device.

Also look for filter ladder designs for the IF stages mentioned above, these are used as a bandpass which of course can be repurposes/transformed into a notch filter through usage or the addition of an op-amp. Perhaps you can gain insight into how to use this technique.

Any circuit you come up with will be very sensitive to loading, so you might end up being quite a challenge to get stability.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Would any crystal work, or does it have to be a specialized crystal? I did find crystals at my desired frequency of 40KHz \$\endgroup\$ – Ilan lewin Jun 2 '16 at 14:02
  • \$\begingroup\$ like I said, at 40KHz you will only get tuning fork type. Or rather, I am only aware of tuning fork type, I may be wrong. This is not a bad thing, High-Q etc. \$\endgroup\$ – placeholder Jun 2 '16 at 14:04
0
\$\begingroup\$

Given your frequency requirements and the need for filter Q/bandwidth control independent of crystal Q, I'm going to suggest a slightly....quirky setup: use a crystal ladder (Cohn) bandpass filter and a subtractor (differential amplifier) to create a bandreject filter, as seen in the schematic below. You will need to adjust the values of the filter capacitors (and add/subtract filter poles) to get the correct skirt response (which may not be symmetrical because of the series-resonance stuff Andy mentioned).

schematic

simulate this circuit – Schematic created using CircuitLab

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Interesting. I will try that \$\endgroup\$ – Ilan lewin Jun 3 '16 at 1:15
  • \$\begingroup\$ this is very interesting, however I wonder. The crystals I find for this frequencies (40KHz) are tuning fork type, and all have a very high series resistance (around 50KOhm). Could this concept (crystal ladder) work at all for such frequencies? \$\endgroup\$ – Ilan lewin Jun 14 '16 at 10:13
  • \$\begingroup\$ @Ilanlewin -- never tried, but don't see why that'd break things outright (you may have to experiment more with cap values to get things to match though) \$\endgroup\$ – ThreePhaseEel Jun 14 '16 at 11:36
  • \$\begingroup\$ great, thanks. Is there a tool you can recommend for designing such a thing? I found that circuitlab doesnt simulate crystals \$\endgroup\$ – Ilan lewin Jun 14 '16 at 11:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.