Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Hi Warren, I arrive a bit late to this discussion, but I hope I can help. I guess the reason for using only one edge is based on the fact that WR is originally designed to measure the phase between a decoded data clock and a system clock. The problem is that this decoded data clock is locked to the incoming data by means of a PFD in the Spartan6/Virtex6 GTP. The PFD normaly only looks at rising edges, so any change in the clock duty cycle will translate in a phase change in the falling edge and not in the rising edge. I am not sure this is really the case, but we certainly had this discussion at the time, but I don't remember if there was any real measurement made. Cheers, Pablo ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
...The problem is that this decoded data clock is locked to the incoming data by means of a PFD in the Spartan6/Virtex6 GTP. The PFD normaly only looks at rising edges, so any change in the clock duty cycle will translate in a phase change in the falling edge and not in the rising edge. I am not sure this is really the case, but we certainly had this discussion at the time, but I don't remember if there was any real measurement made. Well, I don't like correcting myself but this is not right. The PFD is only used in the TX path. In the RX path the clock is decoded using CDR using a VCO which operates at the 1.25Gb/s and then is divided down to 125, so the duty cycle problem is not really a big issue here. In any case it is the typical bug one tries to avoid when doing precise timing. Notice that in this case it would not have been a good idea to sample with the system clock falling edge to increase the performance. pablo ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
In message 9A96CAA5BA7B467D9A106EC858EA0DCE@pc52, Tom Van Baak writes: 3) Every instant on a sine wave is actually a data point, not just the zero crossing(s). So in reality there is near infinite information available. Sorry, but no. If you tell me it is a sine and give me the time of two zero crossings I can tell you everything there has or ever will be to know about any point on that sine-wave. What about phase noise measurements? Yes, what about it ? The CE mark requires that you meet various EMI/EMC rules, one of which is in essence a phase-noise requirement for a 0Hz carrier. But it didn't take long for vendors of clock generator chips to solve that issue: The FM modulate the clock so that if you follow the prescribed measurement procedure there is no spikes above the red line. However, if you measure it without the heavy averaging usually required for phase noise measurements, you find that it is as noisy as always, but at varying frequencies which average out. All statistical treatments of signals are subject to this kind of effect because the entire point of using a statistical treatment is to reduce the amount of information to something we can cope with. ADEV throws a lot of information away (otherwise we wouldn't need phase noise), phase noise throws a lot of information away (otherwise we wouldn't need ADEV) etc. There's no way around thinking critically about what information you throw away and why in each specific application. In the context where this issue came up, the trowing away was about how much of a signal you should feed into a PLL that steers an *XO. There we want to throw as much noise away from the input signal as possible while still keeping the *XO at frequency. Therefore you only want to feed the minimum amount of information about the input frequency necessary for the frequency steering into the PLL, anything above that just adds noise. In this application any information excess to the frequency of the input signal will be noise, and that includes any spurs, harmonic or not. That's why we usually throw all information about amplitude away and focus on one direction of zero crossing, which is (or with a trivial capacitor can be made) well defined for any signal. And to return to the original question: The only reason to look at both zero crossings would be to double the frequency of the input signal to the loop (ie: 2Hz from a 1PPS instead of a 1Hz), at the cost of adding a whole lot of noise in the process. Don't do it. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 p...@freebsd.org | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
There are two ways that both positive and negative slopes could be used, that is, with the input clocks and/or with the reference clock. The PRU on the BBB is not really fast enough to identify the edge direction at a 10mhz rate, so I only collect state changes in real time and then sort it out the direction on the ARM processor afterwards. All transitions are useful for glitch identification, and this does mean I'm already capturing the negative edge of the input signals for free. Admittedly, I'm not doing anything with this data as I currently filter for rising edges fairly early on, but it is 'just a software problem' to utilise data that is already there. _How_ to use the data is the point of discussion, but it will be fairly trivial to implement any ideas and see if they stack up in practice against real data. Regarding the reference clock, utilising the negative edge could be used to double the sample rate (like DDR RAM). I've already been using a 74AC14 Schmitt inverter on the reference clock, primarily as a buffer to distribute the clock to each sampler (flip flop or shift register) and the BBB, but also to try different sample timings (e.g. clocking the samplers on the +ve edge and the BBB capture on the -ve edge for example). Clocking different samplers on alternate edges (but with the same input signal) is therefore relatively straightforward, and feeding more sampling channels to the BBB is not too much of a big deal either. It is not something I have tried though. My initial thought was that doubling the sample rate doesn't buy much, as you could get the same resolution by changing the beat frequency. However, it may help control glitching by obtaining the resolution at a higher beat frequency (greater offset between reference and DUT). Accurately knowing the duty cycle of the reference clock would be essential though so that the time of the -ve edge sample was known. Cheers Simon On 22/10/2014 19:09, WarrenS via time-nuts wrote: The recent discussions about the simple digital mixer got me thinking about the performance vs. complexity trade offs when measuring accurate, high resolution, phase drift differences between two oscillators. It would seem to me, that using both the positive and negative slope edges of the high freq sinewave signal is a better way to go. Is using just one edge, acceptable for a 'state of the art' Phase drift measurements? I am not suggesting the KISS approach is the wrong solution for Simon. I am questioning if the paper posted, is the best way for CERN to make a state of the art femtosecond DDMDT? Here is an extreme example of throwing away useful data for the sake of simplicity: When measuring phase drift of a 10 MHz osc using just a 1PPS signal, 19,999,999 other possible data points are being discarded. Using all possible data points could decrease the noise floor considerably. (by ~5,000 to 1) ws ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Hi On Oct 23, 2014, at 2:01 AM, Poul-Henning Kamp p...@phk.freebsd.dk wrote: In message 9A96CAA5BA7B467D9A106EC858EA0DCE@pc52, Tom Van Baak writes: 3) Every instant on a sine wave is actually a data point, not just the zero crossing(s). So in reality there is near infinite information available. Sorry, but no. If you tell me it is a sine and give me the time of two zero crossings I can tell you everything there has or ever will be to know about any point on that sine-wave. What about phase noise measurements? Yes, what about it ? The CE mark requires that you meet various EMI/EMC rules, one of which is in essence a phase-noise requirement for a 0Hz carrier. But it didn't take long for vendors of clock generator chips to solve that issue: The FM modulate the clock so that if you follow the prescribed measurement procedure there is no spikes above the red line. However, if you measure it without the heavy averaging usually required for phase noise measurements, you find that it is as noisy as always, but at varying frequencies which average out. All statistical treatments of signals are subject to this kind of effect because the entire point of using a statistical treatment is to reduce the amount of information to something we can cope with. ADEV throws a lot of information away (otherwise we wouldn't need phase noise), phase noise throws a lot of information away (otherwise we wouldn't need ADEV) etc. There's no way around thinking critically about what information you throw away and why in each specific application. In the context where this issue came up, the trowing away was about how much of a signal you should feed into a PLL that steers an *XO. There we want to throw as much noise away from the input signal as possible while still keeping the *XO at frequency. Therefore you only want to feed the minimum amount of information about the input frequency necessary for the frequency steering into the PLL, anything above that just adds noise. In this application any information excess to the frequency of the input signal will be noise, and that includes any spurs, harmonic or not. That's why we usually throw all information about amplitude away and focus on one direction of zero crossing, which is (or with a trivial capacitor can be made) well defined for any signal. And to return to the original question: The only reason to look at both zero crossings would be to double the frequency of the input signal to the loop (ie: 2Hz from a 1PPS instead of a 1Hz), at the cost of adding a whole lot of noise in the process. Don't do it. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 p...@freebsd.org | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. Looking at both zero crossings would give you a lot of information about the duty cycle of the input waveforms. If that’s what you are after - there are easier ways to do it. If that’s not what you are after, it’s just going to mess up the readings. Bob ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Poul said; If you tell me it is a sine and give me the time of two zero crossings I can tell you everything there has or ever will be to know ... just to add a bit more nut picking on comment #3. When talking about sub picosecond per second time nut type accuracy, there is no such thing as a pure analog sinewave. so the answer becomes circular in that it takes more than two samples to tell exactly where the two zero crossing points are. Even if there was a near perfect sinewave with under -120 db of bad stuff, and the perfect sine to square wave converter with an exactally known time delay and no zero offset error, and a brick wall low pass cutoff, Given just two data points, it is going to be a real challenge to calculate everything such as Freq, Phase, amplitude, Johnson noise and bandwidth. Add in the typical harmonics, sub harmonics, freq spurs, cycle to cycle changes, cross talk, line noise, ADC resolution, etc, along with a less than perfect signal to noise ratio, and the number of samples needed to get a good enough answer is increased even further. Nyquist says it takes greater than two samples per cycle to be able to even tell if there is any higher freq content present. These are some of the reasons I believe when starting with a signal in the analog world, it helps to oversample, until you can get the data into the pure digital world where one time-stamped sample per cycle can then give you the signal's freq and phase. ws Poul-Henning Kamp Posted Just to pick a nit here: That depends precisely on what and how you measure. If you measure phase, then no, you probably don't need to measure more often than one phase difference per hour or even day, as long as you can reliably predict (from the frequency including noise) exactly how many periods were in that hour or day. This is basically what timelabs do: They measure against some radio signal (GPS, Two-Way, etc. etc.) every so often, trusting their stability between measurements. If you measure frequency, you MUST measure the frequency continously at all times without any deadtime between the measurements to get the precise result. The advantage is that you make *no* assumptions about the frequency or its stability at all. 3) Every instant on a sine wave is actually a data point, not just the zero crossing(s). So in reality there is near infinite information available. Sorry, but no. If you tell me it is a sine and give me the time of two zero crossings I can tell you everything there has or ever will be to know about any point on that sine-wave. Where looking at the whole curve makes sense is if it is not a sinewave, either because it is a complex signal (Loran's 3rd crossing) or because the sinewave is distorted in a way (ie: non-harmonic) which can be averaged out by looking at the entire curveform (locking onto a received radio signal.) But for pure sine signals or good approximations, measuring the zero crossing tells you all you can ever learn. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Lots of interesting responses, but I did not see any posted that answered the original question: Is the CERN method described in the paper the best way to make a state of the art femtosecond DDMDT? www.ee.ucl.ac.uk/lcs/previous/LCS2011/LCS1136.pdf Restating Assuming it is kept Digital, and not taken to the next level [analog], by sampling more of the 10MHz waveforms than just the zero crossings. If making a digital sampler type of tester to measure Femto second phase differences, Is there (and should there be) a strong preference for using one or two edges on either or both the ref and signal inputs? Whether the second edge helps or hurts in the other cases brought up, depends on where the signals come from and what they are being used for. One extreme is the typical GPS timing pulse output, where the second edge can not be used for timing. On the other hand if a square wave signal is coming from a clean, high freq signal that has been divide by N, then the second edge could have very useful and needed information, such as when applied to an XOR phase detector. additional consideration: Unlike a digital sampler which tend to use a single edge, most if not all high end phase measurers I know of averages the results at both edges of both signals. Also anything that depends on a DMTD mixer for its operation, such as single mixer, dual mixer, TPLL, they are all using at least both edges of the signals, and depending on the degree of overdrive, they may be using a lot more of signal that is near the zero crossing point. ws - Phase, One edge or two? (was Digital mixing with a D Flip Flop) Bob Camp posted Looking at both zero crossings would give you a lot of information about the duty cycle of the input waveforms. If that's what you are after - there are easier ways to do it. If that's not what you are after, it's just going to mess up the readings. --- Poul-Henning Kamp Posted; The only reason to look at both zero crossings would be to double the frequency of the input signal to the loop (ie: 2Hz from a 1PPS instead of a 1Hz), at the cost of adding a whole lot of noise in the process. Don't do it. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Depends on what dominant noises you try to measure. Phase white and phase flicker noise depends on bandwidth, and averaging provides filtering effects that effect those. Filtering will also effect systematic signals, but you should never use ADEV for such noises, it's a bad estimator for them. Cheers, Magnus On 10/23/2014 02:42 AM, Bob Camp wrote: Hi The more you “curve fit” or “average” the more you are filtering the data. Filtering does indeed impact the ADEV both at short tau’s and longer tau’s. You need to be very careful if you filter or you will mess up the data. Bob On Oct 22, 2014, at 7:42 PM, Didier Juges shali...@gmail.com wrote: Even more effective would be to sample the entire 10MHz waveform instead of just the zero crossing. By doing a best fit of the entire waveform, you should be able to estimate the zero crossing with much greater precision because now the noise is averaged over the entire waveform instead of a single point at the zero crossing. I wish my signal processing were better than they are and that I had some time to evaluate that. Didier KO4BB On October 22, 2014 1:09:11 PM CDT, WarrenS via time-nuts time-nuts@febo.com wrote: The recent discussions about the simple digital mixer got me thinking about the performance vs. complexity trade offs when measuring accurate, high resolution, phase drift differences between two oscillators. It would seem to me, that using both the positive and negative slope edges of the high freq sinewave signal is a better way to go. Is using just one edge, acceptable for a 'state of the art' Phase drift measurements? I am not suggesting the KISS approach is the wrong solution for Simon. I am questioning if the paper posted, is the best way for CERN to make a state of the art femtosecond DDMDT? Here is an extreme example of throwing away useful data for the sake of simplicity: When measuring phase drift of a 10 MHz osc using just a 1PPS signal, 19,999,999 other possible data points are being discarded. Using all possible data points could decrease the noise floor considerably. (by ~5,000 to 1) ws --- Tom Posted Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Hi Simon, Some additional info. I first heard about the D-FF method of frequency comparison in the late 90's (from Rick Hambly, I think) on the old gps mailing list. It sounded really interesting. Since then, the subject has turned up every few years on this list. But each time, the topic seems to go away quietly with little or no data, plots or explanation. In addition, none of the commercial products I've taken apart appear to use this approach. Hmm. So that begs the question -- what's really going on, and why. I'm enjoying this thread because you've shown both technical competence and optimistic persistence. Perhaps once and for all, with your efforts, we can settle this matter. You will either find a working combination with excellent performance, or you will uncover enough uncontrolled variables that you never want to try it again. Either way, we all learn a lot. Keep the photos, data, and plots coming. Thanks, /tvb -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Bruce posted http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/36903/1/01-2617.pdf among other things illustrates a modified approach to the offset generator by replacing the intermediate phase locked VCXO with a bandpass filter. -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Simon posted www.ee.ucl.ac.uk/lcs/previous/LCS2011/LCS1136.pdf ... The idea is based on the following article which describes creating a digital DMTD with an FPGA for clocks @ 125mhz: ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. -- Sent from my Motorola Droid Razr HD 4G LTE wireless tracker while I do other things. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Hi Back in the early days of ADEV, the standard HP gear had a 60 KHz bandwidth. The question that came up *every* FCS and PTTI was “why does it change with bandwidth / should we spec the bandwidth?”. This went on for at least 15 years before anybody really came up with a “use a narrow bandwidth” answer. Bob On Oct 23, 2014, at 5:40 PM, Magnus Danielson mag...@rubidium.dyndns.org wrote: Depends on what dominant noises you try to measure. Phase white and phase flicker noise depends on bandwidth, and averaging provides filtering effects that effect those. Filtering will also effect systematic signals, but you should never use ADEV for such noises, it's a bad estimator for them. Cheers, Magnus On 10/23/2014 02:42 AM, Bob Camp wrote: Hi The more you “curve fit” or “average” the more you are filtering the data. Filtering does indeed impact the ADEV both at short tau’s and longer tau’s. You need to be very careful if you filter or you will mess up the data. Bob On Oct 22, 2014, at 7:42 PM, Didier Juges shali...@gmail.com wrote: Even more effective would be to sample the entire 10MHz waveform instead of just the zero crossing. By doing a best fit of the entire waveform, you should be able to estimate the zero crossing with much greater precision because now the noise is averaged over the entire waveform instead of a single point at the zero crossing. I wish my signal processing were better than they are and that I had some time to evaluate that. Didier KO4BB On October 22, 2014 1:09:11 PM CDT, WarrenS via time-nuts time-nuts@febo.com wrote: The recent discussions about the simple digital mixer got me thinking about the performance vs. complexity trade offs when measuring accurate, high resolution, phase drift differences between two oscillators. It would seem to me, that using both the positive and negative slope edges of the high freq sinewave signal is a better way to go. Is using just one edge, acceptable for a 'state of the art' Phase drift measurements? I am not suggesting the KISS approach is the wrong solution for Simon. I am questioning if the paper posted, is the best way for CERN to make a state of the art femtosecond DDMDT? Here is an extreme example of throwing away useful data for the sake of simplicity: When measuring phase drift of a 10 MHz osc using just a 1PPS signal, 19,999,999 other possible data points are being discarded. Using all possible data points could decrease the noise floor considerably. (by ~5,000 to 1) ws --- Tom Posted Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Hi Simon, Some additional info. I first heard about the D-FF method of frequency comparison in the late 90's (from Rick Hambly, I think) on the old gps mailing list. It sounded really interesting. Since then, the subject has turned up every few years on this list. But each time, the topic seems to go away quietly with little or no data, plots or explanation. In addition, none of the commercial products I've taken apart appear to use this approach. Hmm. So that begs the question -- what's really going on, and why. I'm enjoying this thread because you've shown both technical competence and optimistic persistence. Perhaps once and for all, with your efforts, we can settle this matter. You will either find a working combination with excellent performance, or you will uncover enough uncontrolled variables that you never want to try it again. Either way, we all learn a lot. Keep the photos, data, and plots coming. Thanks, /tvb -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Bruce posted http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/36903/1/01-2617.pdf among other things illustrates a modified approach to the offset generator by replacing the intermediate phase locked VCXO with a bandpass filter. -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Simon posted www.ee.ucl.ac.uk/lcs/previous/LCS2011/LCS1136.pdf ... The idea is based on the following article which describes creating a digital DMTD with an FPGA for clocks @ 125mhz: ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. -- Sent from my Motorola Droid Razr HD 4G LTE wireless tracker while I do other things. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. ___
[time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
The recent discussions about the simple digital mixer got me thinking about the performance vs. complexity trade offs when measuring accurate, high resolution, phase drift differences between two oscillators. It would seem to me, that using both the positive and negative slope edges of the high freq sinewave signal is a better way to go. Is using just one edge, acceptable for a 'state of the art' Phase drift measurements? I am not suggesting the KISS approach is the wrong solution for Simon. I am questioning if the paper posted, is the best way for CERN to make a state of the art femtosecond DDMDT? Here is an extreme example of throwing away useful data for the sake of simplicity: When measuring phase drift of a 10 MHz osc using just a 1PPS signal, 19,999,999 other possible data points are being discarded. Using all possible data points could decrease the noise floor considerably. (by ~5,000 to 1) ws --- Tom Posted Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Hi Simon, Some additional info. I first heard about the D-FF method of frequency comparison in the late 90's (from Rick Hambly, I think) on the old gps mailing list. It sounded really interesting. Since then, the subject has turned up every few years on this list. But each time, the topic seems to go away quietly with little or no data, plots or explanation. In addition, none of the commercial products I've taken apart appear to use this approach. Hmm. So that begs the question -- what's really going on, and why. I'm enjoying this thread because you've shown both technical competence and optimistic persistence. Perhaps once and for all, with your efforts, we can settle this matter. You will either find a working combination with excellent performance, or you will uncover enough uncontrolled variables that you never want to try it again. Either way, we all learn a lot. Keep the photos, data, and plots coming. Thanks, /tvb -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Bruce posted http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/36903/1/01-2617.pdf among other things illustrates a modified approach to the offset generator by replacing the intermediate phase locked VCXO with a bandpass filter. -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Simon posted www.ee.ucl.ac.uk/lcs/previous/LCS2011/LCS1136.pdf ... The idea is based on the following article which describes creating a digital DMTD with an FPGA for clocks @ 125mhz: ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Which is why the new style instruments sampling the waveforms with a common clock and then downsampling digitally until churning out phase data for further processing can achieve such a good measurement floor. See Sam Steins papers. For some applications the DDMTD approach is pretty amazing precision for it's simplicity. For some you can get more. Cheers, Magnus On 10/22/2014 08:09 PM, WarrenS via time-nuts wrote: The recent discussions about the simple digital mixer got me thinking about the performance vs. complexity trade offs when measuring accurate, high resolution, phase drift differences between two oscillators. It would seem to me, that using both the positive and negative slope edges of the high freq sinewave signal is a better way to go. Is using just one edge, acceptable for a 'state of the art' Phase drift measurements? I am not suggesting the KISS approach is the wrong solution for Simon. I am questioning if the paper posted, is the best way for CERN to make a state of the art femtosecond DDMDT? Here is an extreme example of throwing away useful data for the sake of simplicity: When measuring phase drift of a 10 MHz osc using just a 1PPS signal, 19,999,999 other possible data points are being discarded. Using all possible data points could decrease the noise floor considerably. (by ~5,000 to 1) ws --- Tom Posted Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Hi Simon, Some additional info. I first heard about the D-FF method of frequency comparison in the late 90's (from Rick Hambly, I think) on the old gps mailing list. It sounded really interesting. Since then, the subject has turned up every few years on this list. But each time, the topic seems to go away quietly with little or no data, plots or explanation. In addition, none of the commercial products I've taken apart appear to use this approach. Hmm. So that begs the question -- what's really going on, and why. I'm enjoying this thread because you've shown both technical competence and optimistic persistence. Perhaps once and for all, with your efforts, we can settle this matter. You will either find a working combination with excellent performance, or you will uncover enough uncontrolled variables that you never want to try it again. Either way, we all learn a lot. Keep the photos, data, and plots coming. Thanks, /tvb -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Bruce posted http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/36903/1/01-2617.pdf among other things illustrates a modified approach to the offset generator by replacing the intermediate phase locked VCXO with a bandpass filter. -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Simon posted www.ee.ucl.ac.uk/lcs/previous/LCS2011/LCS1136.pdf ... The idea is based on the following article which describes creating a digital DMTD with an FPGA for clocks @ 125mhz: ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Here is an extreme example of throwing away useful data for the sake of simplicity: When measuring phase drift of a 10 MHz osc using just a 1PPS signal, 19,999,999 other possible data points are being discarded. Using all possible data points could decrease the noise floor considerably. (by ~5,000 to 1) Nice posting. A couple of comments that might help: 1) Depending on the resolution or quantization of your measurement system, more samples don't necessarily give you more information. Higher sample rates may help when the samples are statistically independent. When there is redundancy, you can fool yourself with more data. More data does not automatically imply more precision. Imagine a very fast 1 GHz based counter which measures not 1PPS but all 10 million edges of a 10 MHz signal. It's quite likely, over a second, that all 10 million readings will be the same. So there is no 10,000,000 to 1 or even sqrt(10,000,000) = 3100 to 1 advantage here. More averaging != more precision, except in very rare cases. The sqrt(N) trick you're thinking of works only if you have clean white noise (Gaussian distribution) and a static process. In general oscillators are more complex than this. 2) For long-term analysis, even 1 PPS is overkill. Having more data may not improve your oscillator drift plot at all. This is because the frequency is a moving target. Ever more precise measurements of a moving target are wasted; they don't add any clarity to the overall trend. Consider measuring a 10811 for a year. Do you need to follow its phase or frequency every 100 ns? Or second? Or minute? Maybe as little as one data point per day is more than enough to make a perfectly accurate long-term frequency drift plot. 3) Every instant on a sine wave is actually a data point, not just the zero crossing(s). So in reality there is near infinite information available. So is sample rate the limitation? Or is it sample resolution? Or something else too. How would having trillions of data points differ from 10 million data points from 100 data points from just 1, per second? What more would you know about the oscillator with more data? Are you after frequency domain measurements, or just time domain plots. /tvb ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
In message 0D2DB2B131E5461BB087713B3E49BEA9@pc52, Tom Van Baak writes: Consider measuring a 10811 for a year. Do you need to follow its phase or frequency every 100 ns? Or second? Or minute? Maybe as little as one data point per day is more than enough to make a perfectly accurate long-term frequency drift plot. Just to pick a nit here: That depends precisely on what and how you measure. If you measure phase, then no, you probably don't need to measure more often than one phase difference per hour or even day, as long as you can reliably predict (from the frequency including noise) exactly how many periods were in that hour or day. This is basically what timelabs do: They measure against some radio signal (GPS, Two-Way, etc. etc.) every so often, trusting their stability between measurements. If you measure frequency, you MUST measure the frequency continously at all times without any deadtime between the measurements to get the precise result. The advantage is that you make *no* assumptions about the frequency or its stability at all. 3) Every instant on a sine wave is actually a data point, not just the zero crossing(s). So in reality there is near infinite information available. Sorry, but no. If you tell me it is a sine and give me the time of two zero crossings I can tell you everything there has or ever will be to know about any point on that sine-wave. Where looking at the whole curve makes sense is if it is not a sinewave, either because it is a complex signal (Loran's 3rd crossing) or because the sinewave is distorted in a way (ie: non-harmonic) which can be averaged out by looking at the entire curveform (locking onto a received radio signal.) But for pure sine signals or good approximations, measuring the zero crossing tells you all you can ever learn. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 p...@freebsd.org | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
t...@leapsecond.com said: 2) For long-term analysis, even 1 PPS is overkill. Having more data may not improve your oscillator drift plot at all. This is because the frequency is a moving target. Ever more precise measurements of a moving target are wasted; they don't add any clarity to the overall trend. Consider measuring a 10811 for a year. Do you need to follow its phase or frequency every 100 ns? Or second? Or minute? Maybe as little as one data point per day is more than enough to make a perfectly accurate long-term frequency drift plot. You need more than 1 sample per day for ADEV plots left of 100,000 K seconds. Suppose you have lots and lots of data at, say, 1 second samples. You can turn that into an ADEV plot. Does anybody scan the data in clumps, say a day, to see if the short-tau picture changes over time? -- These are my opinions. I hate spam. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
You need more than 1 sample per day for ADEV plots left of 100,000 K seconds. Correct. What I sometimes do is collect data for just a few minutes at 1000 samples per second. That's enough to make an ADEV plot for tau 0.001 to 1 or 10 seconds. Then I'll collect data for a couple of days at 1 sample per second. That gives you an ADEV plot for tau 1 to 100 k seconds. For really long-term tracking you don't need samples every millisecond or every second. This approach means you don't tie up expensive high-resolution instruments with long-term measurements. The key point here is you don't need to use the same instrument or the same sampling rate or the same resolution for short- and long-term measurements. Suppose you have lots and lots of data at, say, 1 second samples. You can turn that into an ADEV plot. Does anybody scan the data in clumps, say a day, to see if the short-tau picture changes over time? Yes, Stable32 implements DAVAR (dynamic AVAR) which gives an impressive 3D ADEV plot. This is useful for timing sources where the stability, for whatever reason, changes over time. John's TimeLab software implements another form of this: you can specify a trace history value. For example, instead of computing ADEV of a million points, you compute 10 ADEV's on successive groups of 100,000 points each. This is a great way to show the confidence in your statistics. It also works well on oscillators that are warming up, over hours and days. /tvb ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
3) Every instant on a sine wave is actually a data point, not just the zero crossing(s). So in reality there is near infinite information available. Sorry, but no. If you tell me it is a sine and give me the time of two zero crossings I can tell you everything there has or ever will be to know about any point on that sine-wave. What about phase noise measurements? It's the distortion from a mathematically ideal sine wave that allows one to make L(f) plots out to MHz and GHz. TimePod also gives you AM noise plots; you don't get that from just looking at zero crossings. /tvb ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Even more effective would be to sample the entire 10MHz waveform instead of just the zero crossing. By doing a best fit of the entire waveform, you should be able to estimate the zero crossing with much greater precision because now the noise is averaged over the entire waveform instead of a single point at the zero crossing. I wish my signal processing were better than they are and that I had some time to evaluate that. Didier KO4BB On October 22, 2014 1:09:11 PM CDT, WarrenS via time-nuts time-nuts@febo.com wrote: The recent discussions about the simple digital mixer got me thinking about the performance vs. complexity trade offs when measuring accurate, high resolution, phase drift differences between two oscillators. It would seem to me, that using both the positive and negative slope edges of the high freq sinewave signal is a better way to go. Is using just one edge, acceptable for a 'state of the art' Phase drift measurements? I am not suggesting the KISS approach is the wrong solution for Simon. I am questioning if the paper posted, is the best way for CERN to make a state of the art femtosecond DDMDT? Here is an extreme example of throwing away useful data for the sake of simplicity: When measuring phase drift of a 10 MHz osc using just a 1PPS signal, 19,999,999 other possible data points are being discarded. Using all possible data points could decrease the noise floor considerably. (by ~5,000 to 1) ws --- Tom Posted Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Hi Simon, Some additional info. I first heard about the D-FF method of frequency comparison in the late 90's (from Rick Hambly, I think) on the old gps mailing list. It sounded really interesting. Since then, the subject has turned up every few years on this list. But each time, the topic seems to go away quietly with little or no data, plots or explanation. In addition, none of the commercial products I've taken apart appear to use this approach. Hmm. So that begs the question -- what's really going on, and why. I'm enjoying this thread because you've shown both technical competence and optimistic persistence. Perhaps once and for all, with your efforts, we can settle this matter. You will either find a working combination with excellent performance, or you will uncover enough uncontrolled variables that you never want to try it again. Either way, we all learn a lot. Keep the photos, data, and plots coming. Thanks, /tvb -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Bruce posted http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/36903/1/01-2617.pdf among other things illustrates a modified approach to the offset generator by replacing the intermediate phase locked VCXO with a bandpass filter. -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Simon posted www.ee.ucl.ac.uk/lcs/previous/LCS2011/LCS1136.pdf ... The idea is based on the following article which describes creating a digital DMTD with an FPGA for clocks @ 125mhz: ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. -- Sent from my Motorola Droid Razr HD 4G LTE wireless tracker while I do other things. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Hi There are a number of papers out there that talk about decimation vs averaging for ADEV. They have various data sets processed by both techniques. Bottom line - decimation does a better job than averaging for ADEV. At least that’s true if you want the result to resemble the “real” ADEV of the source(s). Bob On Oct 22, 2014, at 6:45 PM, Tom Van Baak t...@leapsecond.com wrote: 3) Every instant on a sine wave is actually a data point, not just the zero crossing(s). So in reality there is near infinite information available. Sorry, but no. If you tell me it is a sine and give me the time of two zero crossings I can tell you everything there has or ever will be to know about any point on that sine-wave. What about phase noise measurements? It's the distortion from a mathematically ideal sine wave that allows one to make L(f) plots out to MHz and GHz. TimePod also gives you AM noise plots; you don't get that from just looking at zero crossings. /tvb ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Hi The more you “curve fit” or “average” the more you are filtering the data. Filtering does indeed impact the ADEV both at short tau’s and longer tau’s. You need to be very careful if you filter or you will mess up the data. Bob On Oct 22, 2014, at 7:42 PM, Didier Juges shali...@gmail.com wrote: Even more effective would be to sample the entire 10MHz waveform instead of just the zero crossing. By doing a best fit of the entire waveform, you should be able to estimate the zero crossing with much greater precision because now the noise is averaged over the entire waveform instead of a single point at the zero crossing. I wish my signal processing were better than they are and that I had some time to evaluate that. Didier KO4BB On October 22, 2014 1:09:11 PM CDT, WarrenS via time-nuts time-nuts@febo.com wrote: The recent discussions about the simple digital mixer got me thinking about the performance vs. complexity trade offs when measuring accurate, high resolution, phase drift differences between two oscillators. It would seem to me, that using both the positive and negative slope edges of the high freq sinewave signal is a better way to go. Is using just one edge, acceptable for a 'state of the art' Phase drift measurements? I am not suggesting the KISS approach is the wrong solution for Simon. I am questioning if the paper posted, is the best way for CERN to make a state of the art femtosecond DDMDT? Here is an extreme example of throwing away useful data for the sake of simplicity: When measuring phase drift of a 10 MHz osc using just a 1PPS signal, 19,999,999 other possible data points are being discarded. Using all possible data points could decrease the noise floor considerably. (by ~5,000 to 1) ws --- Tom Posted Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Hi Simon, Some additional info. I first heard about the D-FF method of frequency comparison in the late 90's (from Rick Hambly, I think) on the old gps mailing list. It sounded really interesting. Since then, the subject has turned up every few years on this list. But each time, the topic seems to go away quietly with little or no data, plots or explanation. In addition, none of the commercial products I've taken apart appear to use this approach. Hmm. So that begs the question -- what's really going on, and why. I'm enjoying this thread because you've shown both technical competence and optimistic persistence. Perhaps once and for all, with your efforts, we can settle this matter. You will either find a working combination with excellent performance, or you will uncover enough uncontrolled variables that you never want to try it again. Either way, we all learn a lot. Keep the photos, data, and plots coming. Thanks, /tvb -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Bruce posted http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/36903/1/01-2617.pdf among other things illustrates a modified approach to the offset generator by replacing the intermediate phase locked VCXO with a bandpass filter. -- Re: [time-nuts] Digital Mixing with a BeagleBone Black and D Flip Flop Simon posted www.ee.ucl.ac.uk/lcs/previous/LCS2011/LCS1136.pdf ... The idea is based on the following article which describes creating a digital DMTD with an FPGA for clocks @ 125mhz: ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. -- Sent from my Motorola Droid Razr HD 4G LTE wireless tracker while I do other things. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Phase, One edge or two? (was Digital mixing with a D Flip Flop)
Hi On Oct 22, 2014, at 5:57 PM, Hal Murray hmur...@megapathdsl.net wrote: t...@leapsecond.com said: 2) For long-term analysis, even 1 PPS is overkill. Having more data may not improve your oscillator drift plot at all. This is because the frequency is a moving target. Ever more precise measurements of a moving target are wasted; they don't add any clarity to the overall trend. Consider measuring a 10811 for a year. Do you need to follow its phase or frequency every 100 ns? Or second? Or minute? Maybe as little as one data point per day is more than enough to make a perfectly accurate long-term frequency drift plot. You need more than 1 sample per day for ADEV plots left of 100,000 K seconds. Suppose you have lots and lots of data at, say, 1 second samples. You can turn that into an ADEV plot. Does anybody scan the data in clumps, say a day, to see if the short-tau picture changes over time? ADEV most certainly does change with time, even for short tau’s. Bob -- These are my opinions. I hate spam. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
