Hello Bjoern, that would work well for static acceleration (tilt) but for vibration resistance the crystal must be low-g, or complexly compensated with wide loop bandwidths such as the FEI papers describe. Initial Calibration would also be tricky, and having an algorithm to measure one result (frequency) against five inputs (aging, tempco, X, Y, Z acceleration) and more (crystal jumps, retrace) is also quite sophisticated :) Also, Mems, or other accelerometers have inherent noise, and to compensate a crystal that has say +/-2E-09 per g sensitivity means one would have to add up to +/-2E-09 in offset statically. That's a lot of deviation, and any noise from the mems would find its way into the Allan Variance/phase-noise. For vibration compensation, the compensation could easily go up to +/-1.2E-08 and more (for up to +/-6G vibration to be canceled). Very interesting topic, and I would love to hear what folks think about this, or have come up with in terms of solutions. At the high-end of the spectrum of the technology is the gun-barrel launched artillery shell with crystal oscillator built-in, that has to withstand and operate with 10,000 to 20,000 g acceleration! One caveat for the artillery shell: commercial GPS would likely not work due to the 1000 Knots verlocity limit. bye, Said In a message dated 7/10/2009 16:53:23 Pacific Daylight Time, bg@lysator.liu.se writes: Hi Said & Tom, The below url links some "low-g"-osc papers. http://www.freqelec.com/tech_lit.html Said, did you contemplate adding a cheap 3d-accelerometer and try to teach your holdover algorithms use the accelerometer measurements in the same way as your temperature measurements? -- Björn > Hello Tom, > > this plot looks very similar to our standard double oven units. We > have our low-g option, which reduces the deviation to about 2- 3E-10 > per g, they work great but do cost more than standard units.. > Coincidentally they also reduce sensitivity to vibration and "tapping" > by 5x to 10x... I wish we could offer them at the same price, but they > are very difficult to manufacture. That's why no one uses them by > default in their product. > > Bye, Said > > > > From iPhone > > On Jul 10, 2009, at 15:51, "Tom Van Baak" <tvb@LeapSecond.com> wrote: > >>> One is do crystal oscillators change frequency when they >>> are turned. The answer to that is yes. This gravitational >>> acceleration effect is rather huge, parts in ten to the 9th >>> or so, and anyone can see this. This is why you never >>> touch, bump, or move, or rotate a laboratory frequency >>> standard (this includes GPSDO and cesium standards). >> >> And to give you a *picture* instead of just numbers... Here is >> a plot showing frequency changes in an OCXO (this from a >> free-running Thunderbolt GPSDO) over the span of one hour. >> Every 5 minutes or so I rotated the rectangular box on some >> axis by 90 degrees. >> >> <http://www.leapsecond.com/pages/ocxo-2g/TBolt-2g-6axis.gif> >> >> You can see that the sudden frequency jumps due to change >> in g-force on the crystal are about -0.5e-9 to +1.5 e-9, which >> is 100x the normal frequency noise for this oscillator (about >> 2e-11 pk-pk or about 2e-12 adev). >> >> Hopefully this result won't come as a big surprise to anyone; the >> so-called "2g turn-over" spec is common for quality oscillators. >> Again, this is why when you enter the world of precision timing >> at 1e-10 and below you tend not to ever touch your standards. >> >> Now if one of you happened to have a fully-programmable 3-axis >> turntable and a couple of hours you could slowly create a most >> beautiful high-resolution 3D color plot showing the precise shift >> in frequency as a function of axis. >> >> /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. > _______________________________________________ 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.

Hi Said, What GPSDO-products do compensate for tilt? It seem like a major error source -- if the user for some reason want to tilt a unit in holdover. It seems to be a "low hanging fruit" to attenuate this error substantially even with a $2 MEMS accelerometer. Once the ambitions grow -- more complexities can be added. Then again, is there a use-case giving some hope the engineering costs can be regained. -- Björn > Hello Bjoern, > > that would work well for static acceleration (tilt) but for vibration > resistance the crystal must be low-g, or complexly compensated with wide > loop > bandwidths such as the FEI papers describe. > > Initial Calibration would also be tricky, and having an algorithm to > measure one result (frequency) against five inputs (aging, tempco, X, Y, Z > acceleration) and more (crystal jumps, retrace) is also quite > sophisticated :) > > Also, Mems, or other accelerometers have inherent noise, and to > compensate a crystal that has say +/-2E-09 per g sensitivity means one > would have to > add up to +/-2E-09 in offset statically. That's a lot of deviation, and > any noise from the mems would find its way into the Allan > Variance/phase-noise. > > For vibration compensation, the compensation could easily go up to > +/-1.2E-08 and more (for up to +/-6G vibration to be canceled). > > Very interesting topic, and I would love to hear what folks think about > this, or have come up with in terms of solutions. > > At the high-end of the spectrum of the technology is the gun-barrel > launched artillery shell with crystal oscillator built-in, that has to > withstand > and operate with 10,000 to 20,000 g acceleration! > > One caveat for the artillery shell: commercial GPS would likely not work > due to the 1000 Knots verlocity limit. > > bye, > Said > > > In a message dated 7/10/2009 16:53:23 Pacific Daylight Time, > bg@lysator.liu.se writes: > > Hi Said & Tom, > > The below url links some "low-g"-osc papers. > > http://www.freqelec.com/tech_lit.html > > Said, did you contemplate adding a cheap 3d-accelerometer and try to > teach > your holdover algorithms use the accelerometer measurements in the same > way as your temperature measurements? > > -- > > Björn > >> Hello Tom, >> >> this plot looks very similar to our standard double oven units. We >> have our low-g option, which reduces the deviation to about 2- 3E-10 >> per g, they work great but do cost more than standard units.. >> Coincidentally they also reduce sensitivity to vibration and "tapping" >> by 5x to 10x... I wish we could offer them at the same price, but they >> are very difficult to manufacture. That's why no one uses them by >> default in their product. >> >> Bye, Said >> >> >> >> From iPhone >> >> On Jul 10, 2009, at 15:51, "Tom Van Baak" <tvb@LeapSecond.com> wrote: >> >>>> One is do crystal oscillators change frequency when they >>>> are turned. The answer to that is yes. This gravitational >>>> acceleration effect is rather huge, parts in ten to the 9th >>>> or so, and anyone can see this. This is why you never >>>> touch, bump, or move, or rotate a laboratory frequency >>>> standard (this includes GPSDO and cesium standards). >>> >>> And to give you a *picture* instead of just numbers... Here is >>> a plot showing frequency changes in an OCXO (this from a >>> free-running Thunderbolt GPSDO) over the span of one hour. >>> Every 5 minutes or so I rotated the rectangular box on some >>> axis by 90 degrees. >>> >>> <http://www.leapsecond.com/pages/ocxo-2g/TBolt-2g-6axis.gif> >>> >>> You can see that the sudden frequency jumps due to change >>> in g-force on the crystal are about -0.5e-9 to +1.5 e-9, which >>> is 100x the normal frequency noise for this oscillator (about >>> 2e-11 pk-pk or about 2e-12 adev). >>> >>> Hopefully this result won't come as a big surprise to anyone; the >>> so-called "2g turn-over" spec is common for quality oscillators. >>> Again, this is why when you enter the world of precision timing >>> at 1e-10 and below you tend not to ever touch your standards. >>> >>> Now if one of you happened to have a fully-programmable 3-axis >>> turntable and a couple of hours you could slowly create a most >>> beautiful high-resolution 3D color plot showing the precise shift >>> in frequency as a function of axis. >>> >>> /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. >> > > > > _______________________________________________ > 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. >

Hi Most of the GPSDO stuff is made for fixed location use. In that case, paying for acceleration compensation doesn't make much sense. About the only people who try to do this stuff mobile (and have the ability to pay) are the military. Bob KB8TQ On Jul 11, 2009, at 5:55 AM, bg@lysator.liu.se wrote: > Hi Said, > > What GPSDO-products do compensate for tilt? > > It seem like a major error source -- if the user for some reason > want to > tilt a unit in holdover. It seems to be a "low hanging fruit" to > attenuate > this error substantially even with a $2 MEMS accelerometer. > > Once the ambitions grow -- more complexities can be added. > > Then again, is there a use-case giving some hope the engineering > costs can > be regained. > > -- > > Björn > >> Hello Bjoern, >> >> that would work well for static acceleration (tilt) but for vibration >> resistance the crystal must be low-g, or complexly compensated with >> wide >> loop >> bandwidths such as the FEI papers describe. >> >> Initial Calibration would also be tricky, and having an algorithm to >> measure one result (frequency) against five inputs (aging, tempco, >> X, Y, Z >> acceleration) and more (crystal jumps, retrace) is also quite >> sophisticated :) >> >> Also, Mems, or other accelerometers have inherent noise, and to >> compensate a crystal that has say +/-2E-09 per g sensitivity means >> one >> would have to >> add up to +/-2E-09 in offset statically. That's a lot of deviation, >> and >> any noise from the mems would find its way into the Allan >> Variance/phase-noise. >> >> For vibration compensation, the compensation could easily go up to >> +/-1.2E-08 and more (for up to +/-6G vibration to be canceled). >> >> Very interesting topic, and I would love to hear what folks think >> about >> this, or have come up with in terms of solutions. >> >> At the high-end of the spectrum of the technology is the gun-barrel >> launched artillery shell with crystal oscillator built-in, that has >> to >> withstand >> and operate with 10,000 to 20,000 g acceleration! >> >> One caveat for the artillery shell: commercial GPS would likely >> not work >> due to the 1000 Knots verlocity limit. >> >> bye, >> Said >> >> >> In a message dated 7/10/2009 16:53:23 Pacific Daylight Time, >> bg@lysator.liu.se writes: >> >> Hi Said & Tom, >> >> The below url links some "low-g"-osc papers. >> >> http://www.freqelec.com/tech_lit.html >> >> Said, did you contemplate adding a cheap 3d-accelerometer and try to >> teach >> your holdover algorithms use the accelerometer measurements in the >> same >> way as your temperature measurements? >> >> -- >> >> Björn >> >>> Hello Tom, >>> >>> this plot looks very similar to our standard double oven units. We >>> have our low-g option, which reduces the deviation to about 2- >>> 3E-10 >>> per g, they work great but do cost more than standard units.. >>> Coincidentally they also reduce sensitivity to vibration and >>> "tapping" >>> by 5x to 10x... I wish we could offer them at the same price, but >>> they >>> are very difficult to manufacture. That's why no one uses them by >>> default in their product. >>> >>> Bye, Said >>> >>> >>> >>> From iPhone >>> >>> On Jul 10, 2009, at 15:51, "Tom Van Baak" <tvb@LeapSecond.com> >>> wrote: >>> >>>>> One is do crystal oscillators change frequency when they >>>>> are turned. The answer to that is yes. This gravitational >>>>> acceleration effect is rather huge, parts in ten to the 9th >>>>> or so, and anyone can see this. This is why you never >>>>> touch, bump, or move, or rotate a laboratory frequency >>>>> standard (this includes GPSDO and cesium standards). >>>> >>>> And to give you a *picture* instead of just numbers... Here is >>>> a plot showing frequency changes in an OCXO (this from a >>>> free-running Thunderbolt GPSDO) over the span of one hour. >>>> Every 5 minutes or so I rotated the rectangular box on some >>>> axis by 90 degrees. >>>> >>>> <http://www.leapsecond.com/pages/ocxo-2g/TBolt-2g-6axis.gif> >>>> >>>> You can see that the sudden frequency jumps due to change >>>> in g-force on the crystal are about -0.5e-9 to +1.5 e-9, which >>>> is 100x the normal frequency noise for this oscillator (about >>>> 2e-11 pk-pk or about 2e-12 adev). >>>> >>>> Hopefully this result won't come as a big surprise to anyone; the >>>> so-called "2g turn-over" spec is common for quality oscillators. >>>> Again, this is why when you enter the world of precision timing >>>> at 1e-10 and below you tend not to ever touch your standards. >>>> >>>> Now if one of you happened to have a fully-programmable 3-axis >>>> turntable and a couple of hours you could slowly create a most >>>> beautiful high-resolution 3D color plot showing the precise shift >>>> in frequency as a function of axis. >>>> >>>> /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. >>> >> >> >> >> _______________________________________________ >> 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. >

bg@lysator.liu.se said: > What GPSDO-products do compensate for tilt? > It seem like a major error source -- if the user for some reason want > to tilt a unit in holdover. I would assume most GPSDOs are rack mounted and are unlikely to get tilted. If tilt is this interesting, it should show up in data sheets. I just took a look at the Thunderbolt. It's got a max temperature change spec on the holdover but no mention of tilt. I didn't find any tilt type specs in the Z3801A manual. -- These are my opinions, not necessarily my employer's. I hate spam.

Said and Björn, SAIDJACK@aol.com wrote: > Hello Bjoern, > > that would work well for static acceleration (tilt) but for vibration > resistance the crystal must be low-g, or complexly compensated with wide loop > bandwidths such as the FEI papers describe. It could work for low-frequency compensation, as it would take a bit more analysis to figure out the impulse response needing of compensation/equalisation. > Initial Calibration would also be tricky, and having an algorithm to > measure one result (frequency) against five inputs (aging, tempco, X, Y, Z > acceleration) and more (crystal jumps, retrace) is also quite sophisticated :) But hey, we are time-nuts, arn't we? :) > Also, Mems, or other accelerometers have inherent noise, and to > compensate a crystal that has say +/-2E-09 per g sensitivity means one would have to > add up to +/-2E-09 in offset statically. That's a lot of deviation, and > any noise from the mems would find its way into the Allan > Variance/phase-noise. Which is scaled by the g-sensitivity of the crystal. If you have a g-sensitivity of 2E-6 of the crystal and then compensate that with a noise of 2E-9 you would end up with a 4E-15 noise contribution. Right? Now, that was assuming white-noise... the 1/f noise sources needs to be estimated for real performance, as they behave differently in ADEV. > For vibration compensation, the compensation could easily go up to > +/-1.2E-08 and more (for up to +/-6G vibration to be canceled). > > Very interesting topic, and I would love to hear what folks think about > this, or have come up with in terms of solutions. > > At the high-end of the spectrum of the technology is the gun-barrel > launched artillery shell with crystal oscillator built-in, that has to withstand > and operate with 10,000 to 20,000 g acceleration! Hence the military and related industry have worked alot on figuring out how to handle it. Using two crystal blanks in opposite direction to subtract the vector-field of the g-sensitivity is one way. The patents for that is available. > One caveat for the artillery shell: commercial GPS would likely not work > due to the 1000 Knots verlocity limit. Not an issue for military receivers, but they need to be built for it not to break at the moment of fire rather than the moment of impact. Cheers, Magnus