crystal filter after crystal oscillator to improve phase noise?

Dear All,
What if I connect a 10 Mhz Crystal filter to the output of my 10 Mhz oscillator?
Will it improve the phase noise at 10 khz (as the filter has a bandwidth of 1.5 khz)? Has this approach a drawback?
Thanx

This is a standard practice. It goes back at least 30 years.

---------- Post added at 17:09 ---------- Previous post was at 17:07 ----------

This is a standard practice. It goes back at least 30 years. You can get much lower bandwidth with another filter topology.

Well, maybe. IF you have a very narrowband crystal bandpass filter, and IF there is not too much mechanical vibration, and IF the bandpass filter is centered right at the center frequency of the oscillator....then you are good to go. Otherwise, you have to carefully evaluate it--it might make it worse.

Rich

yes biff, and you may have to consider Caps with PTC values to correct Xtal or use varicap with thermistor bridge cct. depends on PPM AT cut curve of Xtal.
But knowing how there is a good chance to reduce phase noise, and decouple each stage sharing V+
AT cut Xtals have a Q of 10K but have harmonic overtones but can make good filters if careful. better to use low noise source amp in oscillator. maybe?

What is the source? SC cut OCXO? those are the best. then AT cut VCXO. ... Of course LC XO could not be stable enough .

My point is, if you are on the rejection slope of the xtal filter (where the phase is changing wildly with frequency), you are probably going to cause more phase noise than you filter out.
Rich

THats right Rich.. better to use a linear phase low noise VCO to PLL to the 10Mhz reference. Perhaps check out if this looks familiar https://www.exothink.com/SDR/SDRPN/dds_phase_noise.htm

hey happy guy, what does that have to do with the original poster's question?

Sorry Biff, I guess I was agreeing with you that the steep slope of Xtal filters can add phase noise to XO sources and I was suggesting low Q VCO PLL instead.

HP author wrote here http://www.hparchive.com/seminar_notes/a-117.pdf and I quote his conclusions.

HOW TO MINIMIZE PHASE NOISE
• maximize the Q (with constant device noise source).
• maximize the signal power vs noise power.
• drive the crystal (series) with a controlled, low impedance.
• drive the crystal with a linear amplifier (don't limit in the stage that drives the resonator).
• use a quiet resonator, the oscillator frequency will follow the resonator frequency fluctuations.
• carefully control limiting mechanism so as to not introduce am noise.
• use a quiet device.
• take output power after the resonator if possible to filter noise at large offset frequencies.
• optimize the noise where it is needed, especially consider close in noise vs. large offset noise requirements.
• in a filter application, consider reducing resonator Q if wider bandwidth is acceptable.

I know from my personal experience. the most stable clocks in the world (Rubidium) also have the very poor phase noise, so they phase lock (PLL) a VCXO to this source for low phase noise.

Building ultra low phase noise OCXO 100MHz i used XTAL filter with similar resonator and achieved 2 kHz BW. Oscillator noise floor -165 dBc/Hz after filtering improved down to -182 dBc/Hz.

SunnySkyguy said:

HP author wrote here http://www.hparchive.com/seminar_notes/a-117.pdf and I quote his conclusions.

Click to expand...

Wow! An old article written by Grant Moulton. I met once this guy about 18 years ago, and in 30 min of discussions I learned more about phase noise reductions than in my entire school and career, at that time.

Going back to the original post, as was stated before, filtering the output of an oscillator its an old approach, and can improve the wide band phase noise of an oscillator (> 10kHz from the carrier).
Actually most of the old broadcast transmitters use high-Q filters after their Local Oscillators, and was even a characteristic spec requirement for these filters.

Yah I used to work with guys like Grant. I remember one guy John Boyd whose claim to fame was an RF splitter design with 0.1dB improvement than everyone else for trunk lines in microwave. He was a Potato Farmer in Quebec who liked fast cars and Radar jammers. We worked on CATV ISDN broadband WAN project in the early 80's. He was the who showed me how do proper Noise Figure measurements at 1GHz.

Dr. Drew
How could you get below -174dBc/Hz ? Did you cool the system ?

Do not confuse absolute with relative noise level. The thermal limit is -174dBm/Hz. You can reach -174 dBc/Hz using a carrier power higher than 0 dBm.

Mazz

Right. Absolute noise level at matched load kT-30=-174 dBm/Hz. Put ideal (noise free) sinusoidal signal with power p in through mathced transmission line to the load and you will get at the output a relative noise level -174-p. One half of this noise will be phase noise (i.e. -177-p) and the other - amplitude noise (-177-p too).

flatulent said:

This is a standard practice. It goes back at least 30 years.

---------- Post added at 17:09 ---------- Previous post was at 17:07 ----------

This is a standard practice. It goes back at least 30 years. You can get much lower bandwidth with another filter topology.

Click to expand...

Can you give an example, I would like to go below +-1 khz BW for a 100 Mhz Input.

This a magazine readers contribution. Like all contributions you have to assume they weren't tested by magazine staff. To me it appears to need power supply RF blocking choke and a buffer stage to raise the oscillator output for use.