Bruce So why are you saying I need millions of samples? Is it that this method of integration may give the wrong answer one out a million times? And you will not let up until you find that one in a million times that it may error? I don't think you're going to find it, but if you want we can go with that. BTW It does NOT need ANY scale factors, special or otherwise to give the right answers. It uses the same scale factor of ONE for ALL noise sources. If you can't give me an example of a data log that it may fail on, that I can run thru excel to prove otherwise, then We're done here until next time. ws *************** ************* The results have so far only been shown to be useful when white phase noise dominates. When the phase noise is white almost anything can be made to produce a result that differs from ADEV by at scale factor. In practice its sometimes difficult to know over what range of Tau that the phase noise is in fact white. The various tests and comparisons that have been made or are underway are necessary but not sufficient proof of the usefulness of this technique. The phase noise frequency response of the technique is also required so that its limitations can be delineated. 1000 samples of a divergent noise process are insufficient, spreadsheet analysis of the millions of samples that are probably necessary is impossible/impractical. Using something like Matlab is probably necessary to achieve meaningful results. Bruce ******************* WarrenS wrote: > > OK, So, It is not perfect, but its simple and does give answers that > are GOOD enough. > At least you now understand if the integrator works then the tester > works. > > So that we do not go down hill again after all this progress, > If you would like to send me a data file of say 1000 + samples of any > noise type of your choice > I'll send you back an excel spread sheet to show the insignificant > error that this integration method produces. > > ws > > **************** > > You cannot approximate the sinc function frequency response of an ideal > integrator with an arbitrary low pass filter. > Your scheme will tend to misbehave (in that it will produce anomalous > ADEV estimates) when flicker phase noise is significant. > > You actually need to use an analog low pass filter (or its equivalent) > and an integrator to produce useful ADEV measures > > Bruce > > *************** > WarrenS wrote: >> >> (My apologies to all, this is a game Bruce and I play every time I >> bring up my simple tester.) >> >> Bruce wrote: >>> "So you now actually integrate/average the frequency over the sampling >>> interval (Tau) after rejecting the need to do this for months?" >> >> Yes, I integrate/average just the same as I have always done it from >> day one. >> Did you finally understand how the integration works using most any ADC? >> Hint: it's done with oversampling the tau zero time. >> (and a LP filter set to a value above the tau zero but below the >> oversamping rate) >> The VERY SAME thing I have been trying to tell you from day one, >> something that you have chosen to ignore. >> The very original Block diagram that I posted shows it, if you need >> more information. >> >> ws >> >> ******************* >> Warren >> >> So you now actually integrate/average the frequency over the sampling >> interval (Tau) after rejecting the need to do this for months? >> >> Bruce >> >> ***************** >> WarrenS wrote: >>> Bruce >>> >>> Before we go around again and discuses what my simple tester can and >>> can not do and why, >>> It would be helpful if you would take the time to better understand >>> how it works and why it works the way I have done it. >>> You really should try one yourself if you can't see why it works. >>> You are going to be surprised and embarrassed at how good it works. >>> Why you're at it, try the "swing test" with anything you have. Let me >>> know how that goes. >>> >>> I'm not saying that may tester will match someone's Latest ever >>> changing NEW idea of what the "correct AVAR" should be, >>> After all it just Logs correct, integrated, Freq difference data of >>> ANY noise type >>> and does it without adding any dead time or aliasing all by using >>> pretty much using ANY ADC capability of over sampling at the tau Zero >>> rate. >>> If one then uses the data log with something like the classic Stable >>> 32 S/W or Ulrich's Plotter, >>> it gives is the exact same results as other methods costing much much >>> more, over the whole tau range. >>> This is limited only be its reference oscillator (Same way that all >>> others are limited of course, Doesn't get much better than that). >>> If that is not good enough for you, them you need to discuss the >>> results with Symmetricon and others that give the same answer as mine, >>> not me. >>> >>> If for some reason you want to set one up wrong so that it matches the >>> results of some other special instrument, I'd be glad to tell you how >>> to have it add back in the dead time or aliasing artifact problems or >>> whatever else you would like it to do wrong, that it presently does >>> correctly. >>> >>> ws >>> >>> ****************** >>> Bruce wrote >>> >>> As long as one is aware that your method (as implemented by you) >>> doesn't >>> actually measure Allan variance, it may be useful for comparing the >>> relative stability some sources for small Tau (unfortunately the range >>> of Tau for which the method may produce useful results depends on the >>> phase noise characteristics of the sources being compared). >>> To measure AVAR the technique has to have the same response to all >>> phase >>> noise spectral components as does AVAR. >>> Since you do not integrate/average the frequency measures the phase >>> noise response of the method is not identical to that used in >>> calculating AVAR. >>> This technique probably works best when white phase noise dominates the >>> phase noise spectral region of interest (usually for small Tau). >>> >>> For those who can follow the theory, the following paper shows how the >>> above method is affected by aliasing etc: >>> http://hal.archives-ouvertes.fr/docs/00/37/63/05/PDF/alaa_p1_v4a.pdf >>> >>> The paper also shows how the required integration (needed to actually >>> measure AVAR) can be approximated from the discrete sample sequence. >>> Alternatively one could avoid the numerical integration by replacing >>> the >>> ADC with a zero deadtime (ie not a dual slope converter. A multislope >>> algorithm like that used in the 34401A (but not the 3458A) should work >>> as the signal is integrated continuously) integrating ADC. One >>> possibility is to use a VFC as NIST did when they used this technique >>> some decades ago. >>> >>> Of course, the classical DMTD setup undersamples the phase noise >>> spectrum and thus may suffer from aliasing artifacts. >>> Such aliasing artifacts have no significant effect when the phase noise >>> spectrum is flat. >>> >>> Bruce >>> >>> ********************* >>> WarrenS wrote: >>>> For the Really cheap time nuts, >>>> >>>> It sounds like Bert Kehren has done a great Job building a Dual Mixer >>>> tester. >>>> There are other simpler, less standard ways to get good data for Allan >>>> Variance and small frequency differences. >>>> My VERY simple $10.00 analog tight PLL Tester BB (Previously posted) >>>> pretty much accomplishes the same goals as his, >>>> and it can do 1e-13 in a second, and 1e-11 in 10ms (limited of course >>>> by the single reference Oscillator used) >>>> >>>> A simple test that most can do at home, and still challenges the best >>>> high end testers out there is Tom's the swinging Oscillator test. >>>> http://www.leapsecond.com/pages/10811-g/ >>>> (The results from my PLL tester is attached) >>>> >>>> ws >>>> >>>> ****************** >>>> ----- Original Message ----- From: <EWKehren at aol.com> >>>> To: <time-nuts at febo.com> >>>> Sent: Tuesday, May 11, 2010 7:02 AM >>>> Subject: [time-nuts] Dual Mixer >>>> >>>> >>>>> The Dual Mixer project is nearing completion. >>>>> Let me refresh every ones memory as to my goals. >>>>> a) Total cost less than $ 200 >>>>> b) 1 E-13 with a one second offset >>>>> c) use parts attainable by every one >>>>> d) easy to assemble only a few surface mount parts >>>>> e) a five channel counter that yields 1 E 15 resolution and >>>>> interfaces >>>>> directly to a PC via RS232 or USB >>>>> f) A counter that also gives you instant frequency difference at >>>>> the >>>>> sample rate, not only Allan Variance >>>>> g) Modular so one can use only the Dual Mixer >>>>> h) Modular so one can use multiple units to do simultaneous >>>>> comparison of >>>>> more than two oscillators. >>>>> i) Isolation between D/M and counter so that the counter can be >>>>> powered >>>>> by the PC USB port >>>>> >>>>> I am happy to report that all goals have been accomplished, attached >>>>> is a >>>>> picture of the D/M, limitation of the file size does not allow me to >>>>> attach >>>>> an actual board picture, but if you contact me direct I will send >>>>> you one, >>>>> the final board is actually nicer since the first layout had to >>>>> accommodate >>>>> several variances. >>>>> >>>>> The D/M part leans heavy on the original NIST unit with a few >>>>> substitutions >>>>> and recommendations from Bob Camp. Also beside Opto Couplers >>>>> SN65LVDS1's >>>>> have been included for those that want to use other counting >>>>> methods. >>>>> Selection of filter capacitors allow the use at other offset >>>>> frequencies such as >>>>> 10 and 100 Hz. The D/M fits in a standard 74 X 111 X 20 mm Euro case >>>>> and >>>>> the counter can be stacked below or next to it using the Opto >>>>> Isolators as >>>>> the inter connect. The SYPD-1's fit right on the board but >>>>> connections are >>>>> included to use the HP 10514 A. As a matter of fact removing the HP >>>>> mixer >>>>> board from its housing fits it nicely on the board and every thing >>>>> is still >>>>> inside the housing. >>>>> The counter will handle 1 an 10 Hz offset with a 1 E 14 resolution >>>>> at 10 >>>>> Hz. Thanks to Richard Mc Corkle we have great drawings and code, >>>>> available to >>>>> every one. >>>>> Code, drawings, list of material and PC board layouts and its file, >>>>> will >>>>> be available to every one once the project is completed. >>>>> I need help in the following areas >>>>> a) help me create a nice set of drawings that are computer generated >>>>> something I am not able to do >>>>> b) create the computer program that takes the output of the counter >>>>> board >>>>> and allows Allan Variance plots, frequency difference and dual >>>>> temperature >>>>> readings and plots using RS232 and USB. >>>>> c) an independent test by a third party. >>>>> As I said previously, I am not getting in the business of supplying >>>>> parts >>>>> but will work with people that will help achieve the three points >>>>> listed >>>>> above. Presently I have boards on order and will have two >>>>> uncommitted board >>>>> sets and probably also component kits. >>>>> Please contact me directly. >>>>> Again thank you Corby Dawson, Richard Mc Corkle and Bob Camp. >>>>> Bert Kehren Miami >>>>> >>>>>

Spreadsheets can be a snare and a delusion if not carefully applied with full recognition of their limitations. Unfortunately the divergent nature of flicker phase noise etc doesn't become evident until one processes a very large number of samples. 1000 samples is never enough as many tests and simulations both published and unpublished have shown. Your assertion about scale factors is questionable as the equivalent noise bandwidth needs to be identical when comparing measures produced by different systems. If they differ the results will differ by a scale factor. Aliasing will increase the equivalent noise bandwidth somewhat so exact matching may be difficult. Also its not possible to reconstruct the necessary samples that would be produced by an ideal integrator using a spreadsheet from a finite sequence of samples that the spreadsheet can handle as the number of filter coefficients required is too large for the spreadsheet to cope. WarrenS wrote: > Bruce > > So why are you saying I need millions of samples? > Is it that this method of integration may give the wrong answer one > out a million times? > And you will not let up until you find that one in a million times > that it may error? > I don't think you're going to find it, but if you want we can go with > that. > BTW It does NOT need ANY scale factors, special or otherwise to give > the right answers. > It uses the same scale factor of ONE for ALL noise sources. > If you can't give me an example of a data log that it may fail on, > that I can run thru excel to prove otherwise, then > We're done here until next time. > > ws > > *************** > ************* > The results have so far only been shown to be useful when white phase > noise dominates. > When the phase noise is white almost anything can be made to produce a > result that differs from ADEV by at scale factor. > > In practice its sometimes difficult to know over what range of Tau that > the phase noise is in fact white. > > The various tests and comparisons that have been made or are underway > are necessary but not sufficient proof of the usefulness of this > technique. > The phase noise frequency response of the technique is also required so > that its limitations can be delineated. > > 1000 samples of a divergent noise process are insufficient, spreadsheet > analysis of the millions of samples that are probably necessary is > impossible/impractical. > Using something like Matlab is probably necessary to achieve meaningful > results. > > Bruce > > ******************* > WarrenS wrote: >> >> OK, So, It is not perfect, but its simple and does give answers that >> are GOOD enough. >> At least you now understand if the integrator works then the tester >> works. >> >> So that we do not go down hill again after all this progress, >> If you would like to send me a data file of say 1000 + samples of any >> noise type of your choice >> I'll send you back an excel spread sheet to show the insignificant >> error that this integration method produces. >> >> ws >> >> **************** >> >> You cannot approximate the sinc function frequency response of an ideal >> integrator with an arbitrary low pass filter. >> Your scheme will tend to misbehave (in that it will produce anomalous >> ADEV estimates) when flicker phase noise is significant. >> >> You actually need to use an analog low pass filter (or its equivalent) >> and an integrator to produce useful ADEV measures >> >> Bruce >> >> *************** >> WarrenS wrote: >>> >>> (My apologies to all, this is a game Bruce and I play every time I >>> bring up my simple tester.) >>> >>> Bruce wrote: >>>> "So you now actually integrate/average the frequency over the sampling >>>> interval (Tau) after rejecting the need to do this for months?" >>> >>> Yes, I integrate/average just the same as I have always done it from >>> day one. >>> Did you finally understand how the integration works using most any >>> ADC? >>> Hint: it's done with oversampling the tau zero time. >>> (and a LP filter set to a value above the tau zero but below the >>> oversamping rate) >>> The VERY SAME thing I have been trying to tell you from day one, >>> something that you have chosen to ignore. >>> The very original Block diagram that I posted shows it, if you need >>> more information. >>> >>> ws >>> >>> ******************* >>> Warren >>> >>> So you now actually integrate/average the frequency over the sampling >>> interval (Tau) after rejecting the need to do this for months? >>> >>> Bruce >>> >>> ***************** >>> WarrenS wrote: >>>> Bruce >>>> >>>> Before we go around again and discuses what my simple tester can and >>>> can not do and why, >>>> It would be helpful if you would take the time to better understand >>>> how it works and why it works the way I have done it. >>>> You really should try one yourself if you can't see why it works. >>>> You are going to be surprised and embarrassed at how good it works. >>>> Why you're at it, try the "swing test" with anything you have. Let me >>>> know how that goes. >>>> >>>> I'm not saying that may tester will match someone's Latest ever >>>> changing NEW idea of what the "correct AVAR" should be, >>>> After all it just Logs correct, integrated, Freq difference data of >>>> ANY noise type >>>> and does it without adding any dead time or aliasing all by using >>>> pretty much using ANY ADC capability of over sampling at the tau Zero >>>> rate. >>>> If one then uses the data log with something like the classic Stable >>>> 32 S/W or Ulrich's Plotter, >>>> it gives is the exact same results as other methods costing much much >>>> more, over the whole tau range. >>>> This is limited only be its reference oscillator (Same way that all >>>> others are limited of course, Doesn't get much better than that). >>>> If that is not good enough for you, them you need to discuss the >>>> results with Symmetricon and others that give the same answer as mine, >>>> not me. >>>> >>>> If for some reason you want to set one up wrong so that it matches the >>>> results of some other special instrument, I'd be glad to tell you how >>>> to have it add back in the dead time or aliasing artifact problems or >>>> whatever else you would like it to do wrong, that it presently does >>>> correctly. >>>> >>>> ws >>>> >>>> ****************** >>>> Bruce wrote >>>> >>>> As long as one is aware that your method (as implemented by you) >>>> doesn't >>>> actually measure Allan variance, it may be useful for comparing the >>>> relative stability some sources for small Tau (unfortunately the range >>>> of Tau for which the method may produce useful results depends on the >>>> phase noise characteristics of the sources being compared). >>>> To measure AVAR the technique has to have the same response to all >>>> phase >>>> noise spectral components as does AVAR. >>>> Since you do not integrate/average the frequency measures the phase >>>> noise response of the method is not identical to that used in >>>> calculating AVAR. >>>> This technique probably works best when white phase noise dominates >>>> the >>>> phase noise spectral region of interest (usually for small Tau). >>>> >>>> For those who can follow the theory, the following paper shows how the >>>> above method is affected by aliasing etc: >>>> http://hal.archives-ouvertes.fr/docs/00/37/63/05/PDF/alaa_p1_v4a.pdf >>>> >>>> The paper also shows how the required integration (needed to actually >>>> measure AVAR) can be approximated from the discrete sample sequence. >>>> Alternatively one could avoid the numerical integration by replacing >>>> the >>>> ADC with a zero deadtime (ie not a dual slope converter. A multislope >>>> algorithm like that used in the 34401A (but not the 3458A) should work >>>> as the signal is integrated continuously) integrating ADC. One >>>> possibility is to use a VFC as NIST did when they used this technique >>>> some decades ago. >>>> >>>> Of course, the classical DMTD setup undersamples the phase noise >>>> spectrum and thus may suffer from aliasing artifacts. >>>> Such aliasing artifacts have no significant effect when the phase >>>> noise >>>> spectrum is flat. >>>> >>>> Bruce >>>> >>>> ********************* >>>> WarrenS wrote: >>>>> For the Really cheap time nuts, >>>>> >>>>> It sounds like Bert Kehren has done a great Job building a Dual Mixer >>>>> tester. >>>>> There are other simpler, less standard ways to get good data for >>>>> Allan >>>>> Variance and small frequency differences. >>>>> My VERY simple $10.00 analog tight PLL Tester BB (Previously posted) >>>>> pretty much accomplishes the same goals as his, >>>>> and it can do 1e-13 in a second, and 1e-11 in 10ms (limited of >>>>> course >>>>> by the single reference Oscillator used) >>>>> >>>>> A simple test that most can do at home, and still challenges the best >>>>> high end testers out there is Tom's the swinging Oscillator test. >>>>> http://www.leapsecond.com/pages/10811-g/ >>>>> (The results from my PLL tester is attached) >>>>> >>>>> ws >>>>> >>>>> ****************** >>>>> ----- Original Message ----- From: <EWKehren at aol.com> >>>>> To: <time-nuts at febo.com> >>>>> Sent: Tuesday, May 11, 2010 7:02 AM >>>>> Subject: [time-nuts] Dual Mixer >>>>> >>>>> >>>>>> The Dual Mixer project is nearing completion. >>>>>> Let me refresh every ones memory as to my goals. >>>>>> a) Total cost less than $ 200 >>>>>> b) 1 E-13 with a one second offset >>>>>> c) use parts attainable by every one >>>>>> d) easy to assemble only a few surface mount parts >>>>>> e) a five channel counter that yields 1 E 15 resolution and >>>>>> interfaces >>>>>> directly to a PC via RS232 or USB >>>>>> f) A counter that also gives you instant frequency difference at >>>>>> the >>>>>> sample rate, not only Allan Variance >>>>>> g) Modular so one can use only the Dual Mixer >>>>>> h) Modular so one can use multiple units to do simultaneous >>>>>> comparison of >>>>>> more than two oscillators. >>>>>> i) Isolation between D/M and counter so that the counter can be >>>>>> powered >>>>>> by the PC USB port >>>>>> >>>>>> I am happy to report that all goals have been accomplished, attached >>>>>> is a >>>>>> picture of the D/M, limitation of the file size does not allow me to >>>>>> attach >>>>>> an actual board picture, but if you contact me direct I will send >>>>>> you one, >>>>>> the final board is actually nicer since the first layout had to >>>>>> accommodate >>>>>> several variances. >>>>>> >>>>>> The D/M part leans heavy on the original NIST unit with a few >>>>>> substitutions >>>>>> and recommendations from Bob Camp. Also beside Opto Couplers >>>>>> SN65LVDS1's >>>>>> have been included for those that want to use other counting >>>>>> methods. >>>>>> Selection of filter capacitors allow the use at other offset >>>>>> frequencies such as >>>>>> 10 and 100 Hz. The D/M fits in a standard 74 X 111 X 20 mm Euro >>>>>> case >>>>>> and >>>>>> the counter can be stacked below or next to it using the Opto >>>>>> Isolators as >>>>>> the inter connect. The SYPD-1's fit right on the board but >>>>>> connections are >>>>>> included to use the HP 10514 A. As a matter of fact removing the HP >>>>>> mixer >>>>>> board from its housing fits it nicely on the board and every thing >>>>>> is still >>>>>> inside the housing. >>>>>> The counter will handle 1 an 10 Hz offset with a 1 E 14 resolution >>>>>> at 10 >>>>>> Hz. Thanks to Richard Mc Corkle we have great drawings and code, >>>>>> available to >>>>>> every one. >>>>>> Code, drawings, list of material and PC board layouts and its file, >>>>>> will >>>>>> be available to every one once the project is completed. >>>>>> I need help in the following areas >>>>>> a) help me create a nice set of drawings that are computer >>>>>> generated >>>>>> something I am not able to do >>>>>> b) create the computer program that takes the output of the counter >>>>>> board >>>>>> and allows Allan Variance plots, frequency difference and dual >>>>>> temperature >>>>>> readings and plots using RS232 and USB. >>>>>> c) an independent test by a third party. >>>>>> As I said previously, I am not getting in the business of supplying >>>>>> parts >>>>>> but will work with people that will help achieve the three points >>>>>> listed >>>>>> above. Presently I have boards on order and will have two >>>>>> uncommitted board >>>>>> sets and probably also component kits. >>>>>> Please contact me directly. >>>>>> Again thank you Corby Dawson, Richard Mc Corkle and Bob Camp. >>>>>> Bert Kehren Miami >>>>>> >>>>>> > > > _______________________________________________ > 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. >

Warren Calculating an integral using a sampled data system when the Nyquist criterion is met is exactly equivalent to filtering albeit using just the right coefficients. Using rectangular approximation to the integral of the underlying continuous function is also equivalent to a filter albeit a very simple one. Unfortunately rectangular integration (which you use) isnt particularly accurate, using trapezoidal integration is far more accurate in most cases. Since this isnt a control system the instability associated with trapezoidal integration and higher order integration algorithms in feedback systems isnt an issue. Whittaker-Shannon-Kotelnikov interpolation allows an exact reconstruction (when the Nyquist sampling criterion is met) of the underlying continuous function from the samples. The result can then be integrated term by term to produce a set of weights/filter coefficients for the data samples. In other words in a sampled data system integration is equivalent to using a filter. Near enough is never good enough if you cant estimate the errors involved in the various approximations. This is particularly true when one is attempting to evaluate the deviation of an approximate method from that achieved using the correct method. Bruce WarrenS wrote: > Bruce > > So why are you saying I need millions of samples? > Is it that this method of integration may give the wrong answer one > out a million times? > And you will not let up until you find that one in a million times > that it may error? > I don't think you're going to find it, but if you want we can go with > that. > BTW It does NOT need ANY scale factors, special or otherwise to give > the right answers. > It uses the same scale factor of ONE for ALL noise sources. > If you can't give me an example of a data log that it may fail on, > that I can run thru excel to prove otherwise, then > We're done here until next time. > > ws > > *************** > ************* > The results have so far only been shown to be useful when white phase > noise dominates. > When the phase noise is white almost anything can be made to produce a > result that differs from ADEV by at scale factor. > > In practice its sometimes difficult to know over what range of Tau that > the phase noise is in fact white. > > The various tests and comparisons that have been made or are underway > are necessary but not sufficient proof of the usefulness of this > technique. > The phase noise frequency response of the technique is also required so > that its limitations can be delineated. > > 1000 samples of a divergent noise process are insufficient, spreadsheet > analysis of the millions of samples that are probably necessary is > impossible/impractical. > Using something like Matlab is probably necessary to achieve meaningful > results. > > Bruce > > ******************* > WarrenS wrote: >> >> OK, So, It is not perfect, but its simple and does give answers that >> are GOOD enough. >> At least you now understand if the integrator works then the tester >> works. >> >> So that we do not go down hill again after all this progress, >> If you would like to send me a data file of say 1000 + samples of any >> noise type of your choice >> I'll send you back an excel spread sheet to show the insignificant >> error that this integration method produces. >> >> ws >> >> **************** >> >> You cannot approximate the sinc function frequency response of an ideal >> integrator with an arbitrary low pass filter. >> Your scheme will tend to misbehave (in that it will produce anomalous >> ADEV estimates) when flicker phase noise is significant. >> >> You actually need to use an analog low pass filter (or its equivalent) >> and an integrator to produce useful ADEV measures >> >> Bruce >> >> *************** >> WarrenS wrote: >>> >>> (My apologies to all, this is a game Bruce and I play every time I >>> bring up my simple tester.) >>> >>> Bruce wrote: >>>> "So you now actually integrate/average the frequency over the sampling >>>> interval (Tau) after rejecting the need to do this for months?" >>> >>> Yes, I integrate/average just the same as I have always done it from >>> day one. >>> Did you finally understand how the integration works using most any >>> ADC? >>> Hint: it's done with oversampling the tau zero time. >>> (and a LP filter set to a value above the tau zero but below the >>> oversamping rate) >>> The VERY SAME thing I have been trying to tell you from day one, >>> something that you have chosen to ignore. >>> The very original Block diagram that I posted shows it, if you need >>> more information. >>> >>> ws >>> >>> ******************* >>> Warren >>> >>> So you now actually integrate/average the frequency over the sampling >>> interval (Tau) after rejecting the need to do this for months? >>> >>> Bruce >>> >>> ***************** >>> WarrenS wrote: >>>> Bruce >>>> >>>> Before we go around again and discuses what my simple tester can and >>>> can not do and why, >>>> It would be helpful if you would take the time to better understand >>>> how it works and why it works the way I have done it. >>>> You really should try one yourself if you can't see why it works. >>>> You are going to be surprised and embarrassed at how good it works. >>>> Why you're at it, try the "swing test" with anything you have. Let me >>>> know how that goes. >>>> >>>> I'm not saying that may tester will match someone's Latest ever >>>> changing NEW idea of what the "correct AVAR" should be, >>>> After all it just Logs correct, integrated, Freq difference data of >>>> ANY noise type >>>> and does it without adding any dead time or aliasing all by using >>>> pretty much using ANY ADC capability of over sampling at the tau Zero >>>> rate. >>>> If one then uses the data log with something like the classic Stable >>>> 32 S/W or Ulrich's Plotter, >>>> it gives is the exact same results as other methods costing much much >>>> more, over the whole tau range. >>>> This is limited only be its reference oscillator (Same way that all >>>> others are limited of course, Doesn't get much better than that). >>>> If that is not good enough for you, them you need to discuss the >>>> results with Symmetricon and others that give the same answer as mine, >>>> not me. >>>> >>>> If for some reason you want to set one up wrong so that it matches the >>>> results of some other special instrument, I'd be glad to tell you how >>>> to have it add back in the dead time or aliasing artifact problems or >>>> whatever else you would like it to do wrong, that it presently does >>>> correctly. >>>> >>>> ws >>>> >>>> ****************** >>>> Bruce wrote >>>> >>>> As long as one is aware that your method (as implemented by you) >>>> doesn't >>>> actually measure Allan variance, it may be useful for comparing the >>>> relative stability some sources for small Tau (unfortunately the range >>>> of Tau for which the method may produce useful results depends on the >>>> phase noise characteristics of the sources being compared). >>>> To measure AVAR the technique has to have the same response to all >>>> phase >>>> noise spectral components as does AVAR. >>>> Since you do not integrate/average the frequency measures the phase >>>> noise response of the method is not identical to that used in >>>> calculating AVAR. >>>> This technique probably works best when white phase noise dominates >>>> the >>>> phase noise spectral region of interest (usually for small Tau). >>>> >>>> For those who can follow the theory, the following paper shows how the >>>> above method is affected by aliasing etc: >>>> http://hal.archives-ouvertes.fr/docs/00/37/63/05/PDF/alaa_p1_v4a.pdf >>>> >>>> The paper also shows how the required integration (needed to actually >>>> measure AVAR) can be approximated from the discrete sample sequence. >>>> Alternatively one could avoid the numerical integration by replacing >>>> the >>>> ADC with a zero deadtime (ie not a dual slope converter. A multislope >>>> algorithm like that used in the 34401A (but not the 3458A) should work >>>> as the signal is integrated continuously) integrating ADC. One >>>> possibility is to use a VFC as NIST did when they used this technique >>>> some decades ago. >>>> >>>> Of course, the classical DMTD setup undersamples the phase noise >>>> spectrum and thus may suffer from aliasing artifacts. >>>> Such aliasing artifacts have no significant effect when the phase >>>> noise >>>> spectrum is flat. >>>> >>>> Bruce >>>> >>>> ********************* >>>> WarrenS wrote: >>>>> For the Really cheap time nuts, >>>>> >>>>> It sounds like Bert Kehren has done a great Job building a Dual Mixer >>>>> tester. >>>>> There are other simpler, less standard ways to get good data for >>>>> Allan >>>>> Variance and small frequency differences. >>>>> My VERY simple $10.00 analog tight PLL Tester BB (Previously posted) >>>>> pretty much accomplishes the same goals as his, >>>>> and it can do 1e-13 in a second, and 1e-11 in 10ms (limited of >>>>> course >>>>> by the single reference Oscillator used) >>>>> >>>>> A simple test that most can do at home, and still challenges the best >>>>> high end testers out there is Tom's the swinging Oscillator test. >>>>> http://www.leapsecond.com/pages/10811-g/ >>>>> (The results from my PLL tester is attached) >>>>> >>>>> ws >>>>> >>>>> ****************** >>>>> ----- Original Message ----- From: <EWKehren at aol.com> >>>>> To: <time-nuts at febo.com> >>>>> Sent: Tuesday, May 11, 2010 7:02 AM >>>>> Subject: [time-nuts] Dual Mixer >>>>> >>>>> >>>>>> The Dual Mixer project is nearing completion. >>>>>> Let me refresh every ones memory as to my goals. >>>>>> a) Total cost less than $ 200 >>>>>> b) 1 E-13 with a one second offset >>>>>> c) use parts attainable by every one >>>>>> d) easy to assemble only a few surface mount parts >>>>>> e) a five channel counter that yields 1 E 15 resolution and >>>>>> interfaces >>>>>> directly to a PC via RS232 or USB >>>>>> f) A counter that also gives you instant frequency difference at >>>>>> the >>>>>> sample rate, not only Allan Variance >>>>>> g) Modular so one can use only the Dual Mixer >>>>>> h) Modular so one can use multiple units to do simultaneous >>>>>> comparison of >>>>>> more than two oscillators. >>>>>> i) Isolation between D/M and counter so that the counter can be >>>>>> powered >>>>>> by the PC USB port >>>>>> >>>>>> I am happy to report that all goals have been accomplished, attached >>>>>> is a >>>>>> picture of the D/M, limitation of the file size does not allow me to >>>>>> attach >>>>>> an actual board picture, but if you contact me direct I will send >>>>>> you one, >>>>>> the final board is actually nicer since the first layout had to >>>>>> accommodate >>>>>> several variances. >>>>>> >>>>>> The D/M part leans heavy on the original NIST unit with a few >>>>>> substitutions >>>>>> and recommendations from Bob Camp. Also beside Opto Couplers >>>>>> SN65LVDS1's >>>>>> have been included for those that want to use other counting >>>>>> methods. >>>>>> Selection of filter capacitors allow the use at other offset >>>>>> frequencies such as >>>>>> 10 and 100 Hz. The D/M fits in a standard 74 X 111 X 20 mm Euro >>>>>> case >>>>>> and >>>>>> the counter can be stacked below or next to it using the Opto >>>>>> Isolators as >>>>>> the inter connect. The SYPD-1's fit right on the board but >>>>>> connections are >>>>>> included to use the HP 10514 A. As a matter of fact removing the HP >>>>>> mixer >>>>>> board from its housing fits it nicely on the board and every thing >>>>>> is still >>>>>> inside the housing. >>>>>> The counter will handle 1 an 10 Hz offset with a 1 E 14 resolution >>>>>> at 10 >>>>>> Hz. Thanks to Richard Mc Corkle we have great drawings and code, >>>>>> available to >>>>>> every one. >>>>>> Code, drawings, list of material and PC board layouts and its file, >>>>>> will >>>>>> be available to every one once the project is completed. >>>>>> I need help in the following areas >>>>>> a) help me create a nice set of drawings that are computer >>>>>> generated >>>>>> something I am not able to do >>>>>> b) create the computer program that takes the output of the counter >>>>>> board >>>>>> and allows Allan Variance plots, frequency difference and dual >>>>>> temperature >>>>>> readings and plots using RS232 and USB. >>>>>> c) an independent test by a third party. >>>>>> As I said previously, I am not getting in the business of supplying >>>>>> parts >>>>>> but will work with people that will help achieve the three points >>>>>> listed >>>>>> above. Presently I have boards on order and will have two >>>>>> uncommitted board >>>>>> sets and probably also component kits. >>>>>> Please contact me directly. >>>>>> Again thank you Corby Dawson, Richard Mc Corkle and Bob Camp. >>>>>> Bert Kehren Miami >>>>>> >>>>>> > > > _______________________________________________ > 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. >

Bruce Good, It does seem like we are finally making some good progress. You now seem to acknowledge that my tester could work if I integrate. You now seem to acknowledge that I am integrating by using a filter. I acknowledge that my integration method is not perfect, BUT it is simple and good enough. It would seem the only issue left is to show you just how good of answers my integration method gives. At least now we are JUST talking about what the S/W needs to do. Hopefully you now see that the hardware is adequate. What would you consider an acceptable error band, 3 dB, 1 dB, 0.1 dB? Pick a number >> zero. For a typical high speed data log taken at say 1 K samples per second, one would generally run a quick test with maybe a minute's worth of data. That would provide enough data to give a good tau plot up to about 10 seconds. Now if you can supply me with a 60K data log with any type of reasonably typical noise that you want to include in it I'll show you how close my approximate Integration comes to your perfect integration. I can set this up to do as many times as you want, until I have demonstrated by example that it is close enough, for every data log case that you will provide. Near enough IS good enough for me and most Nuts. As John pointed out, this is measuring noise. One is not going to get the exact same answer twice in a row anyway. My answer will not be perfect, but it will be simple and fast and easy and below the noise uncertainty band. Your turn to put a data log where your math is. Do try and remember I'm working with Frequency and not phase. BTW. just a heads-up warning to be fair. I have set up this situation so that I can not loose. If you want to setup your own situation go for it. I'll see if I can do it. Only requirement is that it should be broken down into no more than 60K sample sizes max for each test at the start. After I pass that, if you want to go for millions of samples or whatever, fine as long as I can read the text data log file. ws **************** ----- Original Message ----- From: "Bruce Griffiths" <bruce.griffiths@xtra.co.nz> To: "Discussion of precise time and frequency measurement" <time-nuts@febo.com> Sent: Tuesday, May 11, 2010 9:12 PM Subject: Re: [time-nuts] Dual Mixer > Warren > > Calculating an integral using a sampled data system when the Nyquist > criterion is met is exactly equivalent to filtering albeit using just the > right coefficients. > Using rectangular approximation to the integral of the underlying > continuous function is also equivalent to a filter albeit a very simple > one. > Unfortunately rectangular integration (which you use) isnt particularly > accurate, using trapezoidal integration is far more accurate in most > cases. > Since this isnt a control system the instability associated with > trapezoidal integration and higher order integration algorithms in > feedback systems isnt an issue. > > Whittaker-Shannon-Kotelnikov interpolation allows an exact reconstruction > (when the Nyquist sampling criterion is met) of the underlying continuous > function from the samples. > The result can then be integrated term by term to produce a set of > weights/filter coefficients for the data samples. > > In other words in a sampled data system integration is equivalent to using > a filter. > > Near enough is never good enough if you can't estimate the errors involved > in the various approximations. > This is particularly true when one is attempting to evaluate the deviation > of an approximate method from that achieved using the correct method. > > Bruce > ***************************************** > > WarrenS wrote: >> Bruce >> >> So why are you saying I need millions of samples? >> Is it that this method of integration may give the wrong answer one out a >> million times? >> And you will not let up until you find that one in a million times that >> it may error? >> I don't think you're going to find it, but if you want we can go with >> that. >> BTW It does NOT need ANY scale factors, special or otherwise to give the >> right answers. >> It uses the same scale factor of ONE for ALL noise sources. >> If you can't give me an example of a data log that it may fail on, >> that I can run thru excel to prove otherwise, then >> We're done here until next time. >> >> ws >> >> *************** >> ************* >> The results have so far only been shown to be useful when white phase >> noise dominates. >> When the phase noise is white almost anything can be made to produce a >> result that differs from ADEV by at scale factor. >> >> In practice its sometimes difficult to know over what range of Tau that >> the phase noise is in fact white. >> >> The various tests and comparisons that have been made or are underway >> are necessary but not sufficient proof of the usefulness of this >> technique. >> The phase noise frequency response of the technique is also required so >> that its limitations can be delineated. >> >> 1000 samples of a divergent noise process are insufficient, spreadsheet >> analysis of the millions of samples that are probably necessary is >> impossible/impractical. >> Using something like Matlab is probably necessary to achieve meaningful >> results. >> >> Bruce >> >> ******************* >> WarrenS wrote: >>> >>> OK, So, It is not perfect, but its simple and does give answers that >>> are GOOD enough. >>> At least you now understand if the integrator works then the tester >>> works. >>> >>> So that we do not go down hill again after all this progress, >>> If you would like to send me a data file of say 1000 + samples of any >>> noise type of your choice >>> I'll send you back an excel spread sheet to show the insignificant >>> error that this integration method produces. >>> >>> ws >>> >>> **************** >>> >>> You cannot approximate the sinc function frequency response of an ideal >>> integrator with an arbitrary low pass filter. >>> Your scheme will tend to misbehave (in that it will produce anomalous >>> ADEV estimates) when flicker phase noise is significant. >>> >>> You actually need to use an analog low pass filter (or its equivalent) >>> and an integrator to produce useful ADEV measures >>> >>> Bruce >>> >>> *************** >>> WarrenS wrote: >>>> >>>> (My apologies to all, this is a game Bruce and I play every time I >>>> bring up my simple tester.) >>>> >>>> Bruce wrote: >>>>> "So you now actually integrate/average the frequency over the sampling >>>>> interval (Tau) after rejecting the need to do this for months?" >>>> >>>> Yes, I integrate/average just the same as I have always done it from >>>> day one. >>>> Did you finally understand how the integration works using most any >>>> ADC? >>>> Hint: it's done with oversampling the tau zero time. >>>> (and a LP filter set to a value above the tau zero but below the >>>> oversamping rate) >>>> The VERY SAME thing I have been trying to tell you from day one, >>>> something that you have chosen to ignore. >>>> The very original Block diagram that I posted shows it, if you need >>>> more information. >>>> >>>> ws >>>> >>>> ******************* >>>> Warren >>>> >>>> So you now actually integrate/average the frequency over the sampling >>>> interval (Tau) after rejecting the need to do this for months? >>>> >>>> Bruce >>>> >>>> ***************** >>>> WarrenS wrote: >>>>> Bruce >>>>> >>>>> Before we go around again and discuses what my simple tester can and >>>>> can not do and why, >>>>> It would be helpful if you would take the time to better understand >>>>> how it works and why it works the way I have done it. >>>>> You really should try one yourself if you can't see why it works. >>>>> You are going to be surprised and embarrassed at how good it works. >>>>> Why you're at it, try the "swing test" with anything you have. Let me >>>>> know how that goes. >>>>> >>>>> I'm not saying that may tester will match someone's Latest ever >>>>> changing NEW idea of what the "correct AVAR" should be, >>>>> After all it just Logs correct, integrated, Freq difference data of >>>>> ANY noise type >>>>> and does it without adding any dead time or aliasing all by using >>>>> pretty much using ANY ADC capability of over sampling at the tau Zero >>>>> rate. >>>>> If one then uses the data log with something like the classic Stable >>>>> 32 S/W or Ulrich's Plotter, >>>>> it gives is the exact same results as other methods costing much much >>>>> more, over the whole tau range. >>>>> This is limited only be its reference oscillator (Same way that all >>>>> others are limited of course, Doesn't get much better than that). >>>>> If that is not good enough for you, them you need to discuss the >>>>> results with Symmetricon and others that give the same answer as mine, >>>>> not me. >>>>> >>>>> If for some reason you want to set one up wrong so that it matches the >>>>> results of some other special instrument, I'd be glad to tell you how >>>>> to have it add back in the dead time or aliasing artifact problems or >>>>> whatever else you would like it to do wrong, that it presently does >>>>> correctly. >>>>> >>>>> ws >>>>> >>>>> ****************** >>>>> Bruce wrote >>>>> >>>>> As long as one is aware that your method (as implemented by you) >>>>> doesn't >>>>> actually measure Allan variance, it may be useful for comparing the >>>>> relative stability some sources for small Tau (unfortunately the range >>>>> of Tau for which the method may produce useful results depends on the >>>>> phase noise characteristics of the sources being compared). >>>>> To measure AVAR the technique has to have the same response to all >>>>> phase >>>>> noise spectral components as does AVAR. >>>>> Since you do not integrate/average the frequency measures the phase >>>>> noise response of the method is not identical to that used in >>>>> calculating AVAR. >>>>> This technique probably works best when white phase noise dominates >>>>> the >>>>> phase noise spectral region of interest (usually for small Tau). >>>>> >>>>> For those who can follow the theory, the following paper shows how the >>>>> above method is affected by aliasing etc: >>>>> http://hal.archives-ouvertes.fr/docs/00/37/63/05/PDF/alaa_p1_v4a.pdf >>>>> >>>>> The paper also shows how the required integration (needed to actually >>>>> measure AVAR) can be approximated from the discrete sample sequence. >>>>> Alternatively one could avoid the numerical integration by replacing >>>>> the >>>>> ADC with a zero deadtime (ie not a dual slope converter. A multislope >>>>> algorithm like that used in the 34401A (but not the 3458A) should work >>>>> as the signal is integrated continuously) integrating ADC. One >>>>> possibility is to use a VFC as NIST did when they used this technique >>>>> some decades ago. >>>>> >>>>> Of course, the classical DMTD setup undersamples the phase noise >>>>> spectrum and thus may suffer from aliasing artifacts. >>>>> Such aliasing artifacts have no significant effect when the phase >>>>> noise >>>>> spectrum is flat. >>>>> >>>>> Bruce >>>>> >>>>> ********************* >>>>> WarrenS wrote: >>>>>> For the Really cheap time nuts, >>>>>> >>>>>> It sounds like Bert Kehren has done a great Job building a Dual Mixer >>>>>> tester. >>>>>> There are other simpler, less standard ways to get good data for >>>>>> Allan >>>>>> Variance and small frequency differences. >>>>>> My VERY simple $10.00 analog tight PLL Tester BB (Previously posted) >>>>>> pretty much accomplishes the same goals as his, >>>>>> and it can do 1e-13 in a second, and 1e-11 in 10ms (limited of >>>>>> course >>>>>> by the single reference Oscillator used) >>>>>> >>>>>> A simple test that most can do at home, and still challenges the best >>>>>> high end testers out there is Tom's the swinging Oscillator test. >>>>>> http://www.leapsecond.com/pages/10811-g/ >>>>>> (The results from my PLL tester is attached) >>>>>> >>>>>> ws >>>>>> >>>>>> ****************** >>>>>> ----- Original Message ----- From: <EWKehren at aol.com> >>>>>> To: <time-nuts at febo.com> >>>>>> Sent: Tuesday, May 11, 2010 7:02 AM >>>>>> Subject: [time-nuts] Dual Mixer >>>>>> >>>>>> >>>>>>> The Dual Mixer project is nearing completion. >>>>>>> Let me refresh every ones memory as to my goals. >>>>>>> a) Total cost less than $ 200 >>>>>>> b) 1 E-13 with a one second offset >>>>>>> c) use parts attainable by every one >>>>>>> d) easy to assemble only a few surface mount parts >>>>>>> e) a five channel counter that yields 1 E 15 resolution and >>>>>>> interfaces >>>>>>> directly to a PC via RS232 or USB >>>>>>> f) A counter that also gives you instant frequency difference at >>>>>>> the >>>>>>> sample rate, not only Allan Variance >>>>>>> g) Modular so one can use only the Dual Mixer >>>>>>> h) Modular so one can use multiple units to do simultaneous >>>>>>> comparison of >>>>>>> more than two oscillators. >>>>>>> i) Isolation between D/M and counter so that the counter can be >>>>>>> powered >>>>>>> by the PC USB port >>>>>>> >>>>>>> I am happy to report that all goals have been accomplished, attached >>>>>>> is a >>>>>>> picture of the D/M, limitation of the file size does not allow me to >>>>>>> attach >>>>>>> an actual board picture, but if you contact me direct I will send >>>>>>> you one, >>>>>>> the final board is actually nicer since the first layout had to >>>>>>> accommodate >>>>>>> several variances. >>>>>>> >>>>>>> The D/M part leans heavy on the original NIST unit with a few >>>>>>> substitutions >>>>>>> and recommendations from Bob Camp. Also beside Opto Couplers >>>>>>> SN65LVDS1's >>>>>>> have been included for those that want to use other counting >>>>>>> methods. >>>>>>> Selection of filter capacitors allow the use at other offset >>>>>>> frequencies such as >>>>>>> 10 and 100 Hz. The D/M fits in a standard 74 X 111 X 20 mm Euro >>>>>>> case >>>>>>> and >>>>>>> the counter can be stacked below or next to it using the Opto >>>>>>> Isolators as >>>>>>> the inter connect. The SYPD-1's fit right on the board but >>>>>>> connections are >>>>>>> included to use the HP 10514 A. As a matter of fact removing the HP >>>>>>> mixer >>>>>>> board from its housing fits it nicely on the board and every thing >>>>>>> is still >>>>>>> inside the housing. >>>>>>> The counter will handle 1 an 10 Hz offset with a 1 E 14 resolution >>>>>>> at 10 >>>>>>> Hz. Thanks to Richard Mc Corkle we have great drawings and code, >>>>>>> available to >>>>>>> every one. >>>>>>> Code, drawings, list of material and PC board layouts and its file, >>>>>>> will >>>>>>> be available to every one once the project is completed. >>>>>>> I need help in the following areas >>>>>>> a) help me create a nice set of drawings that are computer >>>>>>> generated >>>>>>> something I am not able to do >>>>>>> b) create the computer program that takes the output of the counter >>>>>>> board >>>>>>> and allows Allan Variance plots, frequency difference and dual >>>>>>> temperature >>>>>>> readings and plots using RS232 and USB. >>>>>>> c) an independent test by a third party. >>>>>>> As I said previously, I am not getting in the business of supplying >>>>>>> parts >>>>>>> but will work with people that will help achieve the three points >>>>>>> listed >>>>>>> above. Presently I have boards on order and will have two >>>>>>> uncommitted board >>>>>>> sets and probably also component kits. >>>>>>> Please contact me directly. >>>>>>> Again thank you Corby Dawson, Richard Mc Corkle and Bob Camp. >>>>>>> Bert Kehren Miami >>>>>>> >>>>>>> >> >> >> _______________________________________________ >> 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. >> > > > > >

WarrenS wrote: > Bruce > > Good, It does seem like we are finally making some good progress. > You now seem to acknowledge that my tester could work if I integrate. > You now seem to acknowledge that I am integrating by using a filter. In a sampled data system integration is equivalent to a filter but not just any arbitrary low pass filter. The errors in your method are explicitly spelled out in the paper I gave the link to: http://hal.archives-ouvertes.fr/docs/00/37/63/05/PDF/alaa_p1_v4a.pdf In this paper xi is a phase sample and yi is a frequency sample. > I acknowledge that my integration method is not perfect, BUT it is > simple and good enough. Not yet proven nor quantified. > It would seem the only issue left is to show you just how good of > answers my integration method gives. > At least now we are JUST talking about what the S/W needs to do. > Hopefully you now see that the hardware is adequate. > What would you consider an acceptable error band, 3 dB, 1 dB, 0.1 dB? > Pick a number >> zero. > The answer depends on how long one is willing to spend making the measurements. Certainly 0.1dB or better would require heroic efforts to demonstrate. Since the error will also depend on the phase noise spectra of the oscillators being compared a single figure answer isnt feasible. However for the case where white phase noise dominates the error should be not more than 1dB but potentially much less. The errors due to digital signal processing should be at least an order of magnitude lower. > For a typical high speed data log taken at say 1 K samples per second, > one would generally run a quick test with maybe a minute's worth of data. > That would provide enough data to give a good tau plot up to about 10 > seconds. That's a rather sweeping statement given that no estimates of the contribution to measurement noise due to the finite number of samples has been made. The maximum usable tau for a given record length depends on the maximum acceptable error due to the finite number of samples. > Now if you can supply me with a 60K data log with any type of > reasonably typical noise that you want to include in it > I'll show you how close my approximate Integration comes to your > perfect integration. > You can't because your method of perfect integration isnt and its errors cannot be made sufficiently small with so few samples. > I can set this up to do as many times as you want, until I have > demonstrated by example that it is close enough, > for every data log case that you will provide. Near enough IS good > enough for me and most Nuts. Quantify near enough else all is just noise. > > As John pointed out, this is measuring noise. One is not going to get > the exact same answer twice in a row anyway. > My answer will not be perfect, but it will be simple and fast and easy > and below the noise uncertainty band. > Your turn to put a data log where your math is. Do try and remember > I'm working with Frequency and not phase. > Thats idle speculation as you havent quantified anything at all. The repeatability of the measurements needs to be quantified. > BTW. just a heads-up warning to be fair. I have set up this situation > so that I can not loose. Its actually almost trivial to produce a set of samples for which any given method will fail. Doing so is an unproductive exercise. > If you want to setup your own situation go for it. I'll see if I can > do it. > Only requirement is that it should be broken down into no more than > 60K sample sizes max for each test at the start. > After I pass that, if you want to go for millions of samples or > whatever, fine as long as I can read the text data log file. > > ws > Bruce

Sounds like someone is grandstanding to me! Steve On 12 May 2010 22:26, Bruce Griffiths <bruce.griffiths@xtra.co.nz> wrote: > WarrenS wrote: >> >> Bruce >> >> Good, It does seem like we are finally making some good progress. >> You now seem to acknowledge that my tester could work if I integrate. >> You now seem to acknowledge that I am integrating by using a filter. > > In a sampled data system integration is equivalent to a filter but not just > any arbitrary low pass filter. > The errors in your method are explicitly spelled out in the paper I gave the > link to: > http://hal.archives-ouvertes.fr/docs/00/37/63/05/PDF/alaa_p1_v4a.pdf > In this paper xi is a phase sample and yi is a frequency sample. >> >> I acknowledge that my integration method is not perfect, BUT it is simple >> and good enough. > > Not yet proven nor quantified. >> >> It would seem the only issue left is to show you just how good of answers >> my integration method gives. >> At least now we are JUST talking about what the S/W needs to do. >> Hopefully you now see that the hardware is adequate. >> What would you consider an acceptable error band, 3 dB, 1 dB, 0.1 dB? >>  Pick a number >> zero. >> > The answer depends on how long one is willing to spend making the > measurements. > Certainly 0.1dB or better would require heroic efforts to demonstrate. > Since the error will also depend on the phase noise spectra of the > oscillators being compared a single figure answer isnt feasible. > However for the case where white phase noise dominates the error should be > not more than 1dB but potentially much less. > The errors due to digital signal processing should be at least an order of > magnitude lower. >> >> For a typical high speed data log taken at say 1 K samples per second, one >> would generally run a quick test with maybe a minute's worth of data. >> That would provide enough data to give a good tau plot up to about 10 >> seconds. > > That's a rather sweeping statement given that no estimates of the > contribution to measurement noise due to the finite number of samples has > been made. > The maximum usable tau for a given record length depends on the maximum > acceptable error due to the finite number of samples. >> >> Now if you can supply me with a 60K data log with any type of reasonably >> typical noise that you want to include in it >> I'll show you how close my approximate Integration comes to your perfect >> integration. >> > You can't because your method of perfect integration isnt and its errors > cannot be made sufficiently small with so few samples. > >> I can set this up to do as many times as you want, until I have >> demonstrated by example that it is close enough, >> for every data log case that you will provide. Near enough IS good enough >> for me and most Nuts. > > Quantify near enough else all is just noise. >> >> As John pointed out, this is measuring noise. One is not going to get the >> exact same answer twice in a row anyway. >> My answer will not be perfect, but it will be simple and fast and easy and >> below the noise uncertainty band. >> Your turn to put a data log where your math is.  Do try and remember I'm >> working with Frequency and not phase. >> > Thats idle speculation as you havent quantified anything at all. > The repeatability of the measurements needs to be quantified. >> >> BTW. just a heads-up warning to be fair. I have set up this situation so >> that I can not loose. > > Its actually almost trivial to produce a set of samples for which any given > method will fail. > Doing so is an unproductive exercise. >> >> If you want to setup your own situation go for it. I'll see if I can do >> it. >> Only requirement is that it should be broken down into no more than 60K >> sample sizes max for each test at the start. >> After I pass that,  if you want to go for millions of samples or whatever, >> fine as long as I can read the text data log file. >> >> ws >> > Bruce > > > _______________________________________________ > 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. > -- Steve Rooke - ZL3TUV & G8KVD A man with one clock knows what time it is; A man with two clocks is never quite sure.