Which measurement techniques should I use to evaluate different crystal oscillators? I have access to some pretty decent spectrum analyzers at work, but I have little experience using them for clock source measurements.
I'd like to compare things like the SiLabs' Si530 "digital" XO to more traditional hi-spec analog XOs.
Thanks,
Borge
I'd like to compare things like the SiLabs' Si530 "digital" XO to more traditional hi-spec analog XOs.
Thanks,
Borge
I would say, measure phase noise.
I've compared the specifications of several clocks here: H I F I D U I N O: Clock in Buffalo II DAC
I've compared the specifications of several clocks here: H I F I D U I N O: Clock in Buffalo II DAC
Hi glt,
which spectrum analyzer parameters have you used for phase noise measurements? Having consistent window/bandwidth etc is important. And I'd like my measurements to be compatible with the charts you refer to.
Thanks for the link to the Buffalo II. The NEL clock looks like it has fabulous performance. Any experience with it? Also, it would be very interesting to see a Tent clock on the same chart.
In my case I'm clocking 4xPCM1704 and one input on a Spartan3 FPGA. I'm currently using only series resistors (with a resistive divider option for the FPGA when I use 5V XOs).
I was also recomended to consider the NB3N551 distributor:
http://www.onsemi.com/pub_link/Collateral/NB3N551-D.PDF
What do you think?
Thanks,
Borge
which spectrum analyzer parameters have you used for phase noise measurements? Having consistent window/bandwidth etc is important. And I'd like my measurements to be compatible with the charts you refer to.
Thanks for the link to the Buffalo II. The NEL clock looks like it has fabulous performance. Any experience with it? Also, it would be very interesting to see a Tent clock on the same chart.
In my case I'm clocking 4xPCM1704 and one input on a Spartan3 FPGA. I'm currently using only series resistors (with a resistive divider option for the FPGA when I use 5V XOs).
I was also recomended to consider the NB3N551 distributor:
http://www.onsemi.com/pub_link/Collateral/NB3N551-D.PDF
What do you think?
Thanks,
Borge
Borge,
I wish I could have access to those equipment. All I've done is a theoretical analysis relying on data sheets and compare each clock on an equal footing (so the comparison can be meaningful).
But please share your findings here.
Regarding the part, the noise figures look decent, but you need to compare with the same frequency and range. I would guess it could double the jitter of a clock like the Crystek 950 in Buffalo II
I wish I could have access to those equipment. All I've done is a theoretical analysis relying on data sheets and compare each clock on an equal footing (so the comparison can be meaningful).
But please share your findings here.
Regarding the part, the noise figures look decent, but you need to compare with the same frequency and range. I would guess it could double the jitter of a clock like the Crystek 950 in Buffalo II
If I were to look at one specific measurement it would be phase noise, particularly in the 1 to 100Hz region. This is really the critical area, and the area that manufacturer's specifications (purposefully) don't usually cover.
Lucky you having access to the equipment needed to do these measurements! Please do post your results.
Are we talking square or sinewave oscillators?
You can measure jitter directly using a suitable oscilloscope (high end digital). It's jitter you're really interested in since it affects the sampling instant.
Here's a link to a Maxim article:- Clock (CLK) Jitter and Phase Noise Conversion - Maxim It shows how phase noise relates to jitter.
w
You can measure jitter directly using a suitable oscilloscope (high end digital). It's jitter you're really interested in since it affects the sampling instant.
Here's a link to a Maxim article:- Clock (CLK) Jitter and Phase Noise Conversion - Maxim It shows how phase noise relates to jitter.
w
What I have access to is mainly high-frequency spectrum analyzers. They should have no problem covering the frequency of the oscillator. (I usually run mine at 44.1x256.) However, they tend to not go below 9kHz. And they are usually 50Ohm terminated, an impedance the ~11MHz oscillator is probably not designed for.
But I picked up a gem that I haven't really had the time to get very familiar with, an HP 3580A 5Hz-50kHz spectrum analyzer. This one would capture low-frequency stuff (assuming its noise floor goes deep enough) but it would not be able to show the main frequency. And it might have folding issues.
Unfortunately, I don't have much equipment in terms of 'scopes. The meanest thing here is an Agilent Infiniium 1Gsps with a Windows user interface.
Here's an idea: if we could find a unified method for measuring and characterizing oscillators, I'm more than willing to do measurements and feed them back to the forum. That unified method should include a power supply (which accepts potentially dirty lab power 12V in) and the settings for the relevant instruments.
Borge
But I picked up a gem that I haven't really had the time to get very familiar with, an HP 3580A 5Hz-50kHz spectrum analyzer. This one would capture low-frequency stuff (assuming its noise floor goes deep enough) but it would not be able to show the main frequency. And it might have folding issues.
Unfortunately, I don't have much equipment in terms of 'scopes. The meanest thing here is an Agilent Infiniium 1Gsps with a Windows user interface.
Here's an idea: if we could find a unified method for measuring and characterizing oscillators, I'm more than willing to do measurements and feed them back to the forum. That unified method should include a power supply (which accepts potentially dirty lab power 12V in) and the settings for the relevant instruments.
Borge
Borge,
My opinion is that phase noise is more important directly at the input of the DAC chip than the XO parameters. For example, in the case of TDA1541A, the BCK signal controls the timing of conversion, and the clean master clock signal might get degraded until it gets to the DAC input. I would divide the master clock to 44.1 kHz with a fast, low-jitter synchronous divider that performs re-clocking at the same time. Then you can measure the fundamental frequency of the 44.1 kHz squarewave and its noise "skirt" +/- 100 Hz.
Laszlo
My opinion is that phase noise is more important directly at the input of the DAC chip than the XO parameters. For example, in the case of TDA1541A, the BCK signal controls the timing of conversion, and the clean master clock signal might get degraded until it gets to the DAC input. I would divide the master clock to 44.1 kHz with a fast, low-jitter synchronous divider that performs re-clocking at the same time. Then you can measure the fundamental frequency of the 44.1 kHz squarewave and its noise "skirt" +/- 100 Hz.
Laszlo
XOs fro DAC
There are some good comments here about clock sources, but I'd like to add my own random thoughts
For the SI530, this is a PLL using a crystal at some convenient frequency and PLL to translate to some other. Great if you need an oddball frequency or one higher than available with a crystal, but for a DAC its no problem to have the specific crystal needed.
Things like Rubidium are specifically for long term frequency accuracy and use the atomic reference frequency to stabilize the frequency of a crystal oscillator, can't imagine it could be much use here.
It is hard to compare different oscillators when their loads, frequencies, etc are different. For example a 10MHz oscillator has to have phase noise 18dB better than the 80MHz oscillator to have the same jitter.
One could go even further with the comment about noise level at the dac input. The question is what is the actual jitter internal to the DAC, by the time it has gone through lots of CMOS circuitry it will be degraded. This is probably worse in a DAC with a high level of processing like sigma delta than more straight forward designs like 1704.
The question I have is, what is good enough? At some point it is not the major issue and other factors dominate. Perhaps this has been studied. Any references?
Crystal oscillators are tremendously microphonic. In the real world this may be a huge factor.
Last, some of those basic oscillators are not so bad, I've attached a measurement I made on 24.576MHz XO from Kyocera (via digi-key). By the way a spectrum analyzer is not so good for offset frequency close to the carrier. This measurement is on true phase noise tester from Agilent.
Dave
There are some good comments here about clock sources, but I'd like to add my own random thoughts
For the SI530, this is a PLL using a crystal at some convenient frequency and PLL to translate to some other. Great if you need an oddball frequency or one higher than available with a crystal, but for a DAC its no problem to have the specific crystal needed.
Things like Rubidium are specifically for long term frequency accuracy and use the atomic reference frequency to stabilize the frequency of a crystal oscillator, can't imagine it could be much use here.
It is hard to compare different oscillators when their loads, frequencies, etc are different. For example a 10MHz oscillator has to have phase noise 18dB better than the 80MHz oscillator to have the same jitter.
One could go even further with the comment about noise level at the dac input. The question is what is the actual jitter internal to the DAC, by the time it has gone through lots of CMOS circuitry it will be degraded. This is probably worse in a DAC with a high level of processing like sigma delta than more straight forward designs like 1704.
The question I have is, what is good enough? At some point it is not the major issue and other factors dominate. Perhaps this has been studied. Any references?
Crystal oscillators are tremendously microphonic. In the real world this may be a huge factor.
Last, some of those basic oscillators are not so bad, I've attached a measurement I made on 24.576MHz XO from Kyocera (via digi-key). By the way a spectrum analyzer is not so good for offset frequency close to the carrier. This measurement is on true phase noise tester from Agilent.
Dave
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.