Satellite Clock adjustment to earth time, how is it done?

[Alan McDougall]

Satellites above our head experience a little more time because they are high, thus under lesser gravity and a little less time because they are moving at relatively high speeds.

 

 

How do they do the necessary adjustment to compensate for differences between clocks on satellites and earth?

 

Edited August 12, 2012 by Alan McDougall

[dimreepr]

Satellites above our head experience a little more time because they are high, thus under lesser gravity and a little less time because they are moving at relatively high speeds.

 

 

How do they do the necessary adjustment to compensate for differences between clocks on satellites and earth?

 

 

 

IIRC they sent the first GPS satellite up with the option not to use Einstein's equations to compensate for the time difference.

 

http://www.brighthub...cles/32969.aspx

 

Edit/ Just in case he was wrong.

Edited August 12, 2012 by dimreepr

[granpa]

escape velocity from earth surface is 11.2 km/s

 

lorentz factor for 11.2 km/s is 1.000000000698

 

therefore a clock on earths surface ticks 1,000,000,000,698 times while a clock in deep space only ticks 1,000,000,000,000 times

 

a satellite in very low earth orbit has a velocity of 11.2/sqrt[2] km/s = 7.92 km/s

 

lorentz factor for 7.92 km/s is 1.000000000349

Edited August 13, 2012 by granpa

[swansont]

escape velocity from earth surface is 11.2 km/s

 

lorentz factor for 11.2 km/s is 1.000000000698

 

therefore a clock on earths surface ticks 1,000,000,000,698 times while a clock in deep space only ticks 1,000,000,000,000 times

 

a satellite in very low earth orbit has a velocity of 11.2/sqrt[2] km/s = 7.92 km/s

 

lorentz factor for 7.92 km/s is 1.000000000349

 

Which means a fractional frequency shift of 3.49 x 10^-10. Now multiply by 86400 and you get 30 microseconds. That's how far off such a satellite would be. But GPS satellites are not in LEO. They're halfway to geosynchronous (at an altitude of ~20,000km) and the kinematic dilation is about 7 microseconds. The gravitational term is about 45 microseconds/day fast.

 

——

 

The adjustment is made in the oscillators in the clocks.

[John Cuthber]

 

The adjustment is made in the oscillators in the clocks.

Moot point.

The rate of the (primary) oscillators is is indeed adjusted (from our point of view) by the fact that they are in low gravity and moving fast. And there's nothing you can do about that. It's an atomic clock- you don't get to choose the frequency.

 

The satellites seem to keep time because the counters which on earth would say "when I count 9,192,631,770 ocillations, it's a second" are set up to count to a different number.

 

I imagine that there was once a slightly odd discussion between Hewlett Packard and NASA something along the lines of

NASA "You know those really nice atomic clocks you make?

HP "Yes, what about them?"

NASA " Can you make one that runs slow?"

HP "£$%^* (unprintable)"

NASA-"No, really, we want one that runs at the wrong speed on Earth so it keeps what looks like proper time while it's in space"

[swansont]

Moot point.

The rate of the (primary) oscillators is is indeed adjusted (from our point of view) by the fact that they are in low gravity and moving fast. And there's nothing you can do about that. It's an atomic clock- you don't get to choose the frequency.

 

The satellites seem to keep time because the counters which on earth would say "when I count 9,192,631,770 ocillations, it's a second" are set up to count to a different number.

And the way they do that is by adjusting the local oscillator. The atomic system is not adjusted, but you can't count 10 billion cycles a second. The microwaves used to interrogate the atoms are mixed and multiplied up from lower-frequency oscillators. On the ground, the frequency should be 10.23 MHz, but for the satellites 10.22999999543 MHz is used.

 

(Also, while 9,192,631,770 oscillations defines the SI second, there aren't any clocks that realize this, because there are frequency shifts, from e.g. blackbody, Zeeman and relativistic effects. What you do is measure or estimate these and adjust your result accordingly)

[John Cuthber]

For a start, I can get a 10 GHz counter from ebay.

For a finish, I'm so glad I included the word primary in my earlier post.

The fact remains that you don't change the frequency of the atomic clock, you alter the fraction of that frequency that you call 1Hz (or 10 MHz).

 

BTW, where do you get a 10.2999.... Mhz quartz crystal from?

Never mind the particular frequency- how do you get one with a frequency that's known to 12 digits or so?

 

I'm pretty sure that, somewhere tucked away inside the computers that run these clocks is an adjustable constant which amounts to "How many do I count up to to get a second?"

Edited August 14, 2012 by John Cuthber

[swansont]

Getting a 10GHz counter isn't the issue, it's that these clocks are passive devices — you are not getting a 10 GHz signal from them that you count — you're getting absorbing photons — and you couldn't do it with the precision necessary anyway. The frequency information is contained in the relative populations of the atoms, which depends on how close your microwave signal is to the atomic resonance.

 

The thing is, you do get to choose the output frequency of the device, because that's tied in with the local oscillator and not the transition frequency. Most of the devices we have output a frequency of 5 MHz. It's not exactly 5 MHz. because no two clocks run at exactly the same frequency. You take the outputs and compare them. The clocks on GPS satellites output a frequency that can be transmitted on the GPS signal. It's not 9192 MHz. GPS satellites also have Rb clocks on them, which have a transition frequency of 6.8 GHz. They are designed to have the same output frequency as the Cs clocks.

 

We get our crystals from the same company that make Swatch watches. You can get crystals with a fractional frequency stability of parts in 10^14 at 1 second. They do not exhibit white frequency noise, so they don't integrate down, which means they aren't good for longer-term measurements.

[Alan McDougall]

And the way they do that is by adjusting the local oscillator. The atomic system is not adjusted, but you can't count 10 billion cycles a second. The microwaves used to interrogate the atoms are mixed and multiplied up from lower-frequency oscillators. On the ground, the frequency should be 10.23 MHz, but for the satellites 10.22999999543 MHz is used.

 

(Also, while 9,192,631,770 oscillations defines the SI second, there aren't any clocks that realize this, because there are frequency shifts, from e.g. blackbody, Zeeman and relativistic effects. What you do is measure or estimate these and adjust your result accordingly)

 

It is true that the necessary adjustments to the clocks must be done or the GPS system would simply not work. But if the adjustments are done on the ground they will be wrong by the time they reach space? If the adjustments are made while the satellite is in space how do they compensate for the "speed of light time gap" to reach the satellite with the message of adjustment program, or am I making a mountain out of a molehill?

[John Cuthber]

Getting a 10GHz counter isn't the issue, it's that these clocks are passive devices — you are not getting a 10 GHz signal from them that you count — you're getting photons — and you couldn't do it with the precision necessary anyway. The frequency information is contained in the relative populations of the atoms, which depends on how close your microwave signal is to the atomic resonance.

 

The thing is, you do get to choose the output frequency of the device, because that's tied in with the local oscillator and not the transition frequency. Most of the devices we have output a frequency of 5 MHz. It's not exactly 5 MHz. because no two clocks run at exactly the same frequency. You take the outputs and compare them. The clocks on GPS satellites output a frequency that can be transmitted on the GPS signal. It's not 9192 MHz. GPS satellites also have Rb clocks on them, which have a transition frequency of 6.8 GHz. They are designed to have the same output frequency as the Cs clocks.

 

We get our crystals from the same company that make Swatch watches. You can get crystals with a fractional frequency stability of parts in 10^14 at 1 second. They do not exhibit white frequency noise, so they don't integrate down, which means they aren't good for longer-term measurements.

 

If the GPS satellite's clock isn't giving a signal you can count what is it doing?

I grant you that the way it does it it to multiply up a signal from the quartz clock to give a signal that is optimally attenuated by the Cs (or Rb) gas but, somewhere in that box of tricks is an electronic signal at 9GHz which I could count.

 

If I had £500 to spare I could use this

http://www.ebay.co.uk/itm/HP-5351B-Microwave-Frequency-Counter-26-5GHz-/140823102262?pt=UK_BOI_Electrical_Test_Measurement_Equipment_ET&hash=item20c9b60336#ht_500wt_1203

and get 11 digits

 

I don't think NASA would have any trouble getting the other 2 digits needed to tell 10.22999999543 from 10.23

so I question the assertion that "you couldn't do it with the precision necessary anyway"

 

 

In the end it's a philosophical point.

If I lock a known fixed multiple 10MHz signal to the 9 Ghz then measure the 10MHz signal, have I measured the 9GHz signal?

 

However, if I want the 1MHz to look like 1MHz from earth while the clock is in orbit, I have to change the multiplier one way or another because the 9GHz changes whether I like it or not.

 

Does anyone have a schematic of the clock(s) they use?

[swansont]

If the GPS satellite's clock isn't giving a signal you can count what is it doing?

I didn't say that. I said that there is no signal from the atoms that you can count. All they do is absorb photons.

 

I grant you that the way it does it it to multiply up a signal from the quartz clock to give a signal that is optimally attenuated by the Cs (or Rb) gas but, somewhere in that box of tricks is an electronic signal at 9GHz which I could count.

It comes from a microwave source, which is multiplied up from an RF source. The output signal is at a much lower frequency. Much easier to deal with.

 

If I had £500 to spare I could use this

http://www.ebay.co.uk/itm/HP-5351B-Microwave-Frequency-Counter-26-5GHz-/140823102262?pt=UK_BOI_Electrical_Test_Measurement_Equipment_ET&hash=item20c9b60336#ht_500wt_1203

and get 11 digits

 

I don't think NASA would have any trouble getting the other 2 digits needed to tell 10.22999999543 from 10.23

so I question the assertion that "you couldn't do it with the precision necessary anyway"

Decent atomic clocks have precisions that are at least 100 times better than that, and the best are again 100 times better, but we were talking about measuring the 9192 MHz signal, and there is no EM signal from the atoms to count in a passive standard. The atoms do not radiate anything. You can't count something that doesn't exist. The precision of counters, or lack thereof, is moot.

[John Cuthber]

OK. There's no disagreement here about the fact that the atoms absorb photons (and emit them in order to avoid problems with the conservation of energy).

 

This document is a bit old, but I'm sure we can agree that NIST know about clocks.

 

http://tf.nist.gov/general/pdf/1916.pdf

 

On page 305 there's a nice simple schematic (another reason for choosing an old paper) of a clock and there's a box labelled " frequency multiplier and modulator control chain" which is shown as having an output to the electrodes in the gas cell.

 

When the clock is running properly that µwave frequency is the resonance frequency of the atoms in the gas cell.

 

I can count that.

If I wait for it to tick 9,192,631,770 times, that's a second of local time.

Whether I count that directly or use a phase locked loop to run a quartz clock at some known fraction of that rate and count the ticks of the quartz clock is not the point.

 

The atoms ring at exactly the frequency they choose and if I want the clock to look like it ticks at 1Hz from down here I have to change the division ratio, not the frequency of the microwave oscillator.

[swansont]

OK. There's no disagreement here about the fact that the atoms absorb photons (and emit them in order to avoid problems with the conservation of energy).

They emit them much later. If they emitted the photons during the clock measurement cycle, you'd get an error because that atom would be in the wrong state when it was counted. See below.

 

This document is a bit old, but I'm sure we can agree that NIST know about clocks.

Yes, as do people working on atomic clocks at the Naval Observatory.

 

http://tf.nist.gov/general/pdf/1916.pdf

 

On page 305 there's a nice simple schematic (another reason for choosing an old paper) of a clock and there's a box labelled " frequency multiplier and modulator control chain" which is shown as having an output to the electrodes in the gas cell.

 

When the clock is running properly that µwave frequency is the resonance frequency of the atoms in the gas cell.

 

I can count that.

If I wait for it to tick 9,192,631,770 times, that's a second of local time.

A few things about that. It's an atomic beam, not a gas cell. The beam goes through a magnet (3) in the beginning to separate out the proper (m=0) state. The rectangular dotted line area has a constant magnetic field (C-field) so the atoms have a quantization axis, meaning that 9,192,631,770 oscillations won't be a second, but you can adjust for that. The magnet (7) at the end separates the beam, so that atoms that have completed the transition hit a hot-wire detector (10) and be counted. If the microwaves are on resonance, that signal will be maximized, and you can feed back on the microwave frequency chain — that's the error signal control circuit going back to the quartz oscillator. The information about the frequency is encoded in the number of atoms detected. That's what is counted (or, technically, a current from the ionized atoms, which is proportional to the number of atoms). This is why you don't want the atoms to emit photons while this measurement is going on.

 

Nowhere in there is anything counting 9,192,631,770 oscillations. (There's no point, since these days you want to determine time and compare clocks down to the ~picosecond level if you can, i.e. some small fraction of an oscillation. You do that by comparing the phase of the output analog signals.)

 

The frequency divider gets you to a signal that is the clock output. It's often an RF signal, but not a number. No counter. You compare that signal to other clocks. In many clock systems, the time is determined by integrating the frequency of that output. It's not the output of a single clock — it's something you do after you've gotten the signal from your clock, with an external measurement system.

 

Whether I count that directly or use a phase locked loop to run a quartz clock at some known fraction of that rate and count the ticks of the quartz clock is not the point.

Then why bring it up? I mentioned that difficulty of counting directly as an aside. I was rebutting the issue of whether you actually have a counter hooked up to anything in an atomic clock, much less one counting to 9,192,631,770, and you generally don't.

 

The atoms ring at exactly the frequency they choose and if I want the clock to look like it ticks at 1Hz from down here I have to change the division ratio, not the frequency of the microwave oscillator.

Fine. You are free to try and do that. (Good luck, BTW. Changing the multiplier by such a tiny amount? You'd give your RF engineer apoplexy.) But the question was how do the GPS clocks do it, and the answer to that question is that the frequency of the (nominally) 10.23 MHz oscillator is adjusted. What they probably do is not straight division — they have whole number dividers and then mix in another frequency — the 10.23 MHz (which may or may not be multiplied or divided) to get to the number they want. But its'a schematic, and that detail wouldn't be shown.

[John Cuthber]

When all is said and done, the master atomic clock that runs the satellite will continue to oscillate at 9.192...GHz (give or take a defined and calculable offset due to the field) in its frame of reference. That's a little bit "wrong" from our point of view.

 

As I said in the 7th post here

"For a finish, I'm so glad I included the word primary in my earlier post."

Edited August 16, 2012 by John Cuthber