On Tue, 16 Feb 2016 19:55:17 +0200, Tauno Voipio wrote:> On 16.2.16 18:00, Tim Wescott wrote: >> On Tue, 16 Feb 2016 10:03:49 +0200, Tauno Voipio wrote: >> >>> On 15.2.16 22:54, Steve Pope wrote: >>>> <a.turowski@ymail.com> wrote: >>>> >>>>> I think we all arrived to the conclusion that it all should be >>>>> doable. >>>>> The bigger system is now at the very early stage of feasibility >>>>> study. >>>>> That is why I didn't have a complete set of data regarding >>>>> requirements or signal properties. In couple of weeks I will have a >>>>> meeting with people who designed the sensor. This would be the >>>>> opportunity to ask for more data regarding sensor operation, so if >>>>> you have any questions you want me to ask, shoot away. >>>> >>>> One thing you may need is requirements collapse. That is, if your >>>> part of the system requires a timebase good to 0.01 ppm, it may be >>>> that other parts of the system require similarly accurate timebases. >>>> Perhaps there needs to be a system-wide timebase generator, distinct >>>> from your module, since if you "own" the entire timebase problem, >>>> your part of the system might appear disproportionately expensive. >>>> >>>> Another thing you may want to consider is whether you need live >>>> calibration -- that is, when employing the sensor, you alternate >>>> between a calibration phase (which could be infrequent) and >>>> measurement phase. This might make life easier. >>>> >>>> Good luck. >>>> >>>> Steve >>> >>> >>> I just wonder what may be a physical system supplying the frequency to >>> be measured to an accuracy in the ppb class, except a frequency >>> standard. >> >> The 10ppb figure does become suspicious, doesn't it? Some measurement >> system that depends on something like a piezo-electric crystal or a >> mechanical resonator changing it's resonant frequency in response to >> the parameter being measured jumps to mind. >> >> If I'm measuring to "100ppb, 10ppb preferred" then I probably have a >> total error budget of 1ppm or so, and I want the frequency measurement >> part of that error budget to be somewhere between definitely small >> enough and not a worry at all. >> >> I could see such a system being valid if it had temperature >> compensation, >> or if it were measuring temperature. > > > My son, a PhD in measurement technology, likes to say that every > measurement system will be a thermometer, and maybe a barometer, > if sufficiently sensitive. > > I have been suspecting the accuracy spec from the start on.I once sat in a meeting room while a project engineer from a company trying to sell us gyroscopes hold up a soup-can sized gyro and say "This is a gyroscope. It's also a microphone, accelerometer, thermometer, and various other measuring instruments..." We didn't buy the gyroscope, but it was because it was much too big, not because we didn't like the presentation. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Accurate frequency measurement problem
Started by ●February 11, 2016
Reply by ●February 16, 20162016-02-16
Reply by ●February 16, 20162016-02-16
On 16.2.16 20:18, Tim Wescott wrote:> On Tue, 16 Feb 2016 19:55:17 +0200, Tauno Voipio wrote: > >> On 16.2.16 18:00, Tim Wescott wrote: >>> On Tue, 16 Feb 2016 10:03:49 +0200, Tauno Voipio wrote: >>> >>>> On 15.2.16 22:54, Steve Pope wrote: >>>>> <a.turowski@ymail.com> wrote: >>>>> >>>>>> I think we all arrived to the conclusion that it all should be >>>>>> doable. >>>>>> The bigger system is now at the very early stage of feasibility >>>>>> study. >>>>>> That is why I didn't have a complete set of data regarding >>>>>> requirements or signal properties. In couple of weeks I will have a >>>>>> meeting with people who designed the sensor. This would be the >>>>>> opportunity to ask for more data regarding sensor operation, so if >>>>>> you have any questions you want me to ask, shoot away. >>>>> >>>>> One thing you may need is requirements collapse. That is, if your >>>>> part of the system requires a timebase good to 0.01 ppm, it may be >>>>> that other parts of the system require similarly accurate timebases. >>>>> Perhaps there needs to be a system-wide timebase generator, distinct >>>>> from your module, since if you "own" the entire timebase problem, >>>>> your part of the system might appear disproportionately expensive. >>>>> >>>>> Another thing you may want to consider is whether you need live >>>>> calibration -- that is, when employing the sensor, you alternate >>>>> between a calibration phase (which could be infrequent) and >>>>> measurement phase. This might make life easier. >>>>> >>>>> Good luck. >>>>> >>>>> Steve >>>> >>>> >>>> I just wonder what may be a physical system supplying the frequency to >>>> be measured to an accuracy in the ppb class, except a frequency >>>> standard. >>> >>> The 10ppb figure does become suspicious, doesn't it? Some measurement >>> system that depends on something like a piezo-electric crystal or a >>> mechanical resonator changing it's resonant frequency in response to >>> the parameter being measured jumps to mind. >>> >>> If I'm measuring to "100ppb, 10ppb preferred" then I probably have a >>> total error budget of 1ppm or so, and I want the frequency measurement >>> part of that error budget to be somewhere between definitely small >>> enough and not a worry at all. >>> >>> I could see such a system being valid if it had temperature >>> compensation, >>> or if it were measuring temperature. >> >> >> My son, a PhD in measurement technology, likes to say that every >> measurement system will be a thermometer, and maybe a barometer, >> if sufficiently sensitive. >> >> I have been suspecting the accuracy spec from the start on. > > I once sat in a meeting room while a project engineer from a company > trying to sell us gyroscopes hold up a soup-can sized gyro and say "This > is a gyroscope. It's also a microphone, accelerometer, thermometer, and > various other measuring instruments..." > > We didn't buy the gyroscope, but it was because it was much too big, not > because we didn't like the presentation.So, you met a scarcity: a honest salesman. He deserves a prize. -- -TV
Reply by ●February 16, 20162016-02-16
On Tue, 16 Feb 2016 22:38:48 +0200, Tauno Voipio wrote:> On 16.2.16 20:18, Tim Wescott wrote: >> On Tue, 16 Feb 2016 19:55:17 +0200, Tauno Voipio wrote: >> >>> On 16.2.16 18:00, Tim Wescott wrote: >>>> On Tue, 16 Feb 2016 10:03:49 +0200, Tauno Voipio wrote: >>>> >>>>> On 15.2.16 22:54, Steve Pope wrote: >>>>>> <a.turowski@ymail.com> wrote: >>>>>> >>>>>>> I think we all arrived to the conclusion that it all should be >>>>>>> doable. >>>>>>> The bigger system is now at the very early stage of feasibility >>>>>>> study. >>>>>>> That is why I didn't have a complete set of data regarding >>>>>>> requirements or signal properties. In couple of weeks I will have >>>>>>> a meeting with people who designed the sensor. This would be the >>>>>>> opportunity to ask for more data regarding sensor operation, so if >>>>>>> you have any questions you want me to ask, shoot away. >>>>>> >>>>>> One thing you may need is requirements collapse. That is, if your >>>>>> part of the system requires a timebase good to 0.01 ppm, it may be >>>>>> that other parts of the system require similarly accurate >>>>>> timebases. >>>>>> Perhaps there needs to be a system-wide timebase generator, >>>>>> distinct from your module, since if you "own" the entire timebase >>>>>> problem, your part of the system might appear disproportionately >>>>>> expensive. >>>>>> >>>>>> Another thing you may want to consider is whether you need live >>>>>> calibration -- that is, when employing the sensor, you alternate >>>>>> between a calibration phase (which could be infrequent) and >>>>>> measurement phase. This might make life easier. >>>>>> >>>>>> Good luck. >>>>>> >>>>>> Steve >>>>> >>>>> >>>>> I just wonder what may be a physical system supplying the frequency >>>>> to be measured to an accuracy in the ppb class, except a frequency >>>>> standard. >>>> >>>> The 10ppb figure does become suspicious, doesn't it? Some >>>> measurement system that depends on something like a piezo-electric >>>> crystal or a mechanical resonator changing it's resonant frequency in >>>> response to the parameter being measured jumps to mind. >>>> >>>> If I'm measuring to "100ppb, 10ppb preferred" then I probably have a >>>> total error budget of 1ppm or so, and I want the frequency >>>> measurement part of that error budget to be somewhere between >>>> definitely small enough and not a worry at all. >>>> >>>> I could see such a system being valid if it had temperature >>>> compensation, >>>> or if it were measuring temperature. >>> >>> >>> My son, a PhD in measurement technology, likes to say that every >>> measurement system will be a thermometer, and maybe a barometer, >>> if sufficiently sensitive. >>> >>> I have been suspecting the accuracy spec from the start on. >> >> I once sat in a meeting room while a project engineer from a company >> trying to sell us gyroscopes hold up a soup-can sized gyro and say >> "This is a gyroscope. It's also a microphone, accelerometer, >> thermometer, and various other measuring instruments..." >> >> We didn't buy the gyroscope, but it was because it was much too big, >> not because we didn't like the presentation. > > > So, you met a scarcity: a honest salesman. He deserves a prize.He was the engineering manager. I suspect that the sales guy was wishing for a remote with a "voice off" button. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●February 17, 20162016-02-17
On 13.02.2016 20:29, Eric Jacobsen wrote: (snip)> > I tried to run this but there is a term, sgn, near the end that > appears to be undefined? I'm attempting to run it on Octave, so > maybe there's a difference there. > > Cool stuff, though. > > > Eric Jacobsen > Anchor Hill Communications > http://www.anchorhill.com >I could not reproduce the error in Octave. I used the latest Windows release of Octave and loaded three packages (communications, signal and control). And it worked allright, although more slowly than in Matlab. Regards, Evgeny.
Reply by ●February 17, 20162016-02-17
On Wed, 17 Feb 2016 14:04:31 +0300, Evgeny Filatov <filatov.ev@mipt.ru> wrote:>On 13.02.2016 20:29, Eric Jacobsen wrote: > >(snip) > >> >> I tried to run this but there is a term, sgn, near the end that >> appears to be undefined? I'm attempting to run it on Octave, so >> maybe there's a difference there. >> >> Cool stuff, though. >> >> >> Eric Jacobsen >> Anchor Hill Communications >> http://www.anchorhill.com >> > >I could not reproduce the error in Octave. I used the latest Windows >release of Octave and loaded three packages (communications, signal and >control). And it worked allright, although more slowly than in Matlab. > >Regards, >Evgeny.Oh, sorry! I had fixed it and a few other issues in Octave already. It was just an issue with passing the arguments. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
