Hi all,
Has anybody build an oscillator phase noise measurement system?
There are some threads here dealing with ultra low jitter oscillator designs and it would be great to see some real measurements.
I have read some basics and basically you need a mixer and a low noise amplifier. The amp output could go into a soundcard input or to a spectrum analyzer.
A precision reference oscillator is not needed if you use two identical test oscillators....
Is anybody interested in building such a thing?
Greetings,
Udo
Has anybody build an oscillator phase noise measurement system?
There are some threads here dealing with ultra low jitter oscillator designs and it would be great to see some real measurements.
I have read some basics and basically you need a mixer and a low noise amplifier. The amp output could go into a soundcard input or to a spectrum analyzer.
A precision reference oscillator is not needed if you use two identical test oscillators....
Is anybody interested in building such a thing?
Greetings,
Udo
I thought you guys could hear those 2ps of jitter, otherwise what is all the fuss about! 
You can use the DAC itself as the jitter measuring device, since it translates time errors into amplitude errors. Play a pure tone through the DAC, and the tone will get sidebands that are copies of the master clock's phase noise spectrum. These can be seen on an audio distortion analyser, or a soundcard (if the soundcard's ADC master clock has lower jitter than the one you are trying to measure, and your FFT-fu is strong)
I think the mixer and low noise amplifier approach places equal demands on the spectrum analyser's phase noise performance to just analysing the output of a DAC, but I could be wrong.
You can use the DAC itself as the jitter measuring device, since it translates time errors into amplitude errors. Play a pure tone through the DAC, and the tone will get sidebands that are copies of the master clock's phase noise spectrum. These can be seen on an audio distortion analyser, or a soundcard (if the soundcard's ADC master clock has lower jitter than the one you are trying to measure, and your FFT-fu is strong)
I think the mixer and low noise amplifier approach places equal demands on the spectrum analyser's phase noise performance to just analysing the output of a DAC, but I could be wrong.
Low-frequency oscillators are likely to have far higher phase noise than a ~10 MHz crystal oscillator. A direct measurement is thus far more sensitive.
Basically you need to construct a PLL with very low cutoff frequency. Using two fixed-frequency oscillators doesn't work well, as they need to have a steady long-term phase relation in order to make the phase detector work properly. Conceptually easy, but for good results excellent design practice is needed.
In the < 50 MHz range, a XOR gate is pretty good as phase detector. Unfortunately using this may cause the PLL to never lock. A way around this is to add a phase-frequency detector for the PLL, and use the XOR just for the phase noise measurement.
Samuel
@scopeboy:
Hey, this is a playing field, i don't have pressure to finish things in time here
And if i build a discrete oscillator, I care about the 2ps, even if it may not be audible (2ps are not very ambitious btw).
I like to build things "perfect", or at least understand how they work. I want to know the limitations and be able to measure the performance. It is not my first goal to fit some time tables or have a finished product.
Otherwise i could buy a finished crystal oscillator circuit with specification... or the finished DAC and integrated amplifier... but this is DIY.
Anybody defines for himself how far he want to go.
The trick with the mixer is that the phase noise is translated to DC (zero frequency), so that the spectrum analyzer phase noise is not important.
The spectrum analyer noise is not important too as the low noise amp (after the mixer) does enough amplification to lever the signal above the spectrum analyzers noise.
More details are here:
Low-Cost Phase Noise Measurement | Wenzel Associates, Inc.
Hey, this is a playing field, i don't have pressure to finish things in time here
And if i build a discrete oscillator, I care about the 2ps, even if it may not be audible (2ps are not very ambitious btw).
I like to build things "perfect", or at least understand how they work. I want to know the limitations and be able to measure the performance. It is not my first goal to fit some time tables or have a finished product.
Otherwise i could buy a finished crystal oscillator circuit with specification... or the finished DAC and integrated amplifier... but this is DIY.
Anybody defines for himself how far he want to go.
The trick with the mixer is that the phase noise is translated to DC (zero frequency), so that the spectrum analyzer phase noise is not important.
The spectrum analyer noise is not important too as the low noise amp (after the mixer) does enough amplification to lever the signal above the spectrum analyzers noise.
More details are here:
Low-Cost Phase Noise Measurement | Wenzel Associates, Inc.
Where is the low frequency oscillator? Any tone you can play through a DAC, you can generate digitally. The only oscillator involved in that process is the DAC's master clock, the very thing we are trying to test.
I guess the issue might be the low frequency reference oscillator to compare the tone against. If you used a notch-type THD analyser, again there is no oscillator involved. But maybe the notch filter in that has some sort of noise analogous to phase noise in an oscillator?
The problem with this is finding a VCO for your PLL with audiophile grade jitter specs.
The lowest jitter VCO I can think of would be something like a Tent Labs VCXO. You could lock this to incoming MCLK from a SPDIF receiver, and get a very good readout of the link jitter from an XOR gate. But could you measure anything useful by locking it to another crystal oscillator?What is the Tent VCXO phase noise specification?
A good reference is the Morion MV89. It is affordable and has very good specifications which is verified independently.
(-145dBc@100Hz http://www.morion.com.ru/catalog_pdf/MV89-OCXO.pdf)
But the trick with the PLL method is that you can use two test oscillators for the measurement (one must be pullable).
You do not need an expensive reference for the measurement. If the two oscillators are identical, your result is 3 dB too high.
A good reference is the Morion MV89. It is affordable and has very good specifications which is verified independently.
(-145dBc@100Hz http://www.morion.com.ru/catalog_pdf/MV89-OCXO.pdf)
But the trick with the PLL method is that you can use two test oscillators for the measurement (one must be pullable).
You do not need an expensive reference for the measurement. If the two oscillators are identical, your result is 3 dB too high.
Tent specify "less than 3ps RMS (3 sigma)". VCXO
Other VCXOs are available, but Tent are the only company I know who stock them in the usual digital audio master clock frequencies.
A pullable oscillator will always have worse jitter and phase noise than an ordinary XO. The wider the pulling range, the worse the phase noise. If I am reading the Morion spec right, you are paying for the extraordinary phase noise with a pulling range of only 0.25ppm. Not to mention a sine wave output that will need squared before it could drive a DAC. At the levels we are discussing, the noise floor of the squaring circuit would probably dominate the jitter performance. (the squaring operation transforms amplitude noise into phase noise)
Other VCXOs are available, but Tent are the only company I know who stock them in the usual digital audio master clock frequencies.
A pullable oscillator will always have worse jitter and phase noise than an ordinary XO. The wider the pulling range, the worse the phase noise. If I am reading the Morion spec right, you are paying for the extraordinary phase noise with a pulling range of only 0.25ppm. Not to mention a sine wave output that will need squared before it could drive a DAC. At the levels we are discussing, the noise floor of the squaring circuit would probably dominate the jitter performance. (the squaring operation transforms amplitude noise into phase noise)
Last edited:
Look at the amazing work Martein Baker PA3AKE did on measuring phase noise for his high performance receiver design.
The quadrature method works, and can see close in, where notch filtering is limited to about a 2KHz offset from the carrier.
It is overkill for audio, but low noise sources are a fascinating area.
Regards, Dan.
The quadrature method works, and can see close in, where notch filtering is limited to about a 2KHz offset from the carrier.
It is overkill for audio, but low noise sources are a fascinating area.
Regards, Dan.
Well, I implemented already the SW part supporting a (cross spectrum) dual channel conjugate complex averaging (as Demian requested) on the power spectrum density (with 1/Hz normalization) to get the last possible noise floor.
I know other(s) who deal with power noise & amplifier noise issues while this part is also critical.
You are invited to deal with the evaluation version
Hp
Both DAC and ADC will add considerable phase noise. Also by looking at FFT amplitude spectra you can't distinguish amplitude from phase noise.
The problem with this is finding a VCO for your PLL with audiophile grade jitter specs.
The standard approach is to use cross-correlation, i.e. build two fully independent PLLs and cross-correlate their outputs in the frequency domain. Theoretically you can average out the noise contribution of the PLL as much as you like, but the measurement time increases drastically (by a factor of 2 for each 3 dB).
Samuel
Udo,
Apart from my interest in your AP1 USB APIB interface pod, maybe I can help you with this project - I have various Jitter / Phase noise measurement systems in my lab (HP / Wavecreast), while they are pretty poor within the Audio Bandwidth (Most systems are designed from 12KHz upwards due to some Telco Spec), atleast we can calibrate our DIY effort (which will be optimised for LF measurements) against known "industry standard" units.
Once calibrated at higher levels phase noise levels, its safe to presume that the design is linear and lower level Phase noise results should still be reliable until we hit the noise floor.
My HP systems are limited by the Reference signal generator (HP8662A), although with a bit of playing we could use a Second pullable crystal (VCXO) reference oscillator in place of the HP8662A).
I don't have much faith in the Wavecreast systems as they have too much LF noise....
As per my earlier post on the APIB thread, I'm located 50km or so from Brno - Czech Rep. if your located near Vienna then its only a couple of hours drive...
I recall from CZ to the Osterreich boarder, there is a
beautiful drive that goes to the train station there in Austria.
For kilometers and kilometers each side of the two lane street
has cherry trees spaced out 10 meters I guess. These trees are
hundreds of years old (I'm guessing) and were planted such
that folks walking or in a carriage would have shade and
plenty of cherries to eat on their trip. A very thoughtful
thing.
I can't remember which street it was because I wasn't driving,
but the thought of planting the tress and seeing them has
stuck with me all these years.
Enjoy your trip to Wien if it shall so be and enjoy
a Wiener Wurst auf Mich.
Thanks John for the offer! I would be glad to visit Brno. It is about 2 hours away. At the moment i want to finish the AP interface. The phase noise measurement setup is in the "idea collection" state.
But i have good schematics of a state of the art low noise oscillator with automatic gain control. The circuit is not too difficult to build.
Greetings, Udo
But i have good schematics of a state of the art low noise oscillator with automatic gain control. The circuit is not too difficult to build.
Greetings, Udo
I'd be happy to design the hardware with you for the Phase noise measurement system, I have some experience in this area - measuring phase noise has been a constant struggle for me as there's little hardware available that's "optimised" for measuring the Audio Bandwidth from DC to say 40Khz offsets, most are designed for 12KHz upwards missing the important "In band" frequency span for our audio requirements....
The Wavecrest's have far too high noise floor towards DC - I also have a 11729C with HP8662A carrier noise measurement system, but the HP8662 Ref Sig Gen is the limiting component.
Its possible to use a high quality VCXO in place of the HP8662, but I'm concerned if the HP11729C PLL Loop will remain stable as the HP8662A and VCXO will have very different "Frequency Tuning characteristic", but it would be something we can play with on our way to develop the DIY version.
Great news
I have been following this guy's posts on YouTube about LeCroy scopes. But here in this post he describes his phase noise measurement with diy made discriminator. I am sure you might find bits and pieces of interest for your project:
https://www.youtube.com/watch?v=GP6B_ImnoII
Thank you for the link, I felt alittle dizzy with the cameras rapid movement!
The problem with his system is his noise floor, about -140dBc at 1MHz and only -100dBc at 100Hz - I would expect a high performance audio clock to measure better then this, Measurement noisefloor is the problem I have always faced
Looking forward to seeing how low we can go with our DIY solution.
I have a HP3589 - but its a horrid machine to use, the earlier HP3585 is much nicer (but without the digital based RBW fillters) - so its limited to a 3Hz RBW.
The problem with his system is his noise floor, about -140dBc at 1MHz and only -100dBc at 100Hz - I would expect a high performance audio clock to measure better then this, Measurement noisefloor is the problem I have always faced
Looking forward to seeing how low we can go with our DIY solution.
I have a HP3589 - but its a horrid machine to use, the earlier HP3585 is much nicer (but without the digital based RBW fillters) - so its limited to a 3Hz RBW.
Last edited:
After watching AR2's link above, I found this Symmetricom 3120A :-
https://www.youtube.com/watch?v=9PG9KNcmRPk
Looks VERY interesting, wonder how much it costs?
https://www.youtube.com/watch?v=9PG9KNcmRPk
Looks VERY interesting, wonder how much it costs?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.