3
\$\begingroup\$

I need an ultra low jitter/phase noise 100Mhz crystal oscillator with phase noise of less than 70dbc @ 1 hz and less than 100dBc @ 10Hz and the lowest jitter I can manage and I need a differential output which makes things difficult because the best XOs I can find are VCXOs with single ended output and discreet XOs will never be that good.

Maybe this is a stupid question but can I use two single ended XOs and offset the phase of one of them by 180 degrees to make a pseudo differential output?

\$\endgroup\$
2
  • \$\begingroup\$ Look into transformer coupling, or clock buffers with differential outputs. \$\endgroup\$ Jan 14 at 22:53
  • \$\begingroup\$ The ones I've been able to find don't have less than70dbc @1Hz and less than 100dBc @ 10Hz which is what I'm looking for. \$\endgroup\$
    – coinmaster
    Jan 14 at 23:23
5
\$\begingroup\$

Maybe this is a stupid question but can I use two single ended XOs and offset the phase of one of them by 180 degrees to make a pseudo differential output?

You could. However, at 100 MHz, small RF transformers will easily handle that frequency and give you a well balanced output with none of the extra jitter you would get from active devices.

If you are after ultra low phase noise, then having found the best you can find, use two of them and phase lock them to the same frequency, add their voltage outputs, and reap an additional 3dB improvement in phase noise. Double the number of oscillators again for another 3dB. What's your budget and board size?

\$\endgroup\$
3
  • \$\begingroup\$ Budget and size is whatever it takes. I did not know that about parallelling oscillators. How do I phase lock them? How do RF transformers affect jitter and phase noise? \$\endgroup\$
    – coinmaster
    Jan 14 at 23:19
  • 1
    \$\begingroup\$ As I said in the answer, transformers don't add to the jitter and phase noise, they're passive, unlike active buffers that always increase it. I assumed that this was a 'clean-up' oscillator phase locked to a lower frequency reference, so would be phase locked anyway. If it's free-running, then phase locking two together is fairly straightforward, at 100 MHz I'd use an RF mixer, or a logic XOR gate as the PSD. Alternatively you may find injection locking them is sufficient. \$\endgroup\$
    – Neil_UK
    Jan 14 at 23:42
  • \$\begingroup\$ Unfortunately it seems the transformers add to the phase unbalance though which is problematic for me since I'm trying to get them to cancel at the output of the circuit. \$\endgroup\$
    – coinmaster
    Jan 15 at 0:21
4
\$\begingroup\$

You can't synchronize independent crystal oscillators, they will drift rapidly. You can instead use a differential clock buffer / driver with a single-ended input. Diodes Inc has a bunch of different types in their Pericom line, for example. They go down to 0.01ps RMS of added jitter.

\$\endgroup\$
1
\$\begingroup\$

A 180 degrees signal is just the inverted signal so you could take the signal and invert it. Inverting does add a small time delay though. Also keeping the amplitudes of the signal and its inverted version the same could be a challenge.

I would therefore consider a balanced crystal oscillator, for example something like this:

enter image description here Source

Making that work properly at 100 MHz on a PCB might be challenging though. You will need some RC-circuit and PCB design experience for this.

\$\endgroup\$
1
  • \$\begingroup\$ This would be ideal but It would take some wizardry to make it anywhere near as good as a commercial chip. \$\endgroup\$
    – coinmaster
    Jan 14 at 23:27

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.