A measurement of 10 MHz down to 10 uHz

[John Heath]

Hi Bob

I touched one of the crystal oscillators to heat it up a pinch to lower it by a cycle or 2. The side bands tracked the frequency change but 100 times greater than the crystal frequency change. I know this is grabbing at straws but if 2 more zeros of resolution can be made from side band information then so be it. I have a few 74hc390 and a 74hc4046 to multiply the hetrodyne beat frequency. Something in then rattling and niose of these gates is causing it. Higher harmonic of a the main square wave beat frequency maybe? I am starting to appreciate your input on the magnitude of this measurement. From only 10 MHz 10 mHz is jumping around like a jelly bean. That is 1 percent of 1 cycle per second of 10 MHz. Surely there is a way to increase stability beyond 10 mHz. I have separated the 74390s and 4046 multiplier from the mixer board in the hopes of cleaning it up a bit.  Also added a graphic EQ to clean up the beat frequency back to a sine wave. Between these two changes possibly  a stable 1 mHz reading can be made. The inertia added to the 4046 PLL for the voltage controlled oscillator is 1 M with 500 U.f. condenser. It takes a good few minutes to creeps up to the target frequency. How could that kind of ineria be jumping around like a jelly bean at 10 mHz. That is a rhetorical question. I am just ventilating Thanks for your interest and I took note of your caution on GPSDOs bought on ebay from another thread. Will need one eventually so it is good to know.

[uncle_bob]

Hi Bob

I touched one of the crystal oscillators to heat it up a pinch to lower it by a cycle or 2. The side bands tracked the frequency change but 100 times greater than the crystal frequency change. I know this is grabbing at straws but if 2 more zeros of resolution can be made from side band information then so be it. I have a few 74hc390 and a 74hc4046 to multiply the hetrodyne beat frequency. Something in then rattling and niose of these gates is causing it. Higher harmonic of a the main square wave beat frequency maybe? I am starting to appreciate your input on the magnitude of this measurement. From only 10 MHz 10 mHz is jumping around like a jelly bean. That is 1 percent of 1 cycle per second of 10 MHz. Surely there is a way to increase stability beyond 10 mHz. I have separated the 74390s and 4046 multiplier from the mixer board in the hopes of cleaning it up a bit.  Also added a graphic EQ to clean up the beat frequency back to a sine wave. Between these two changes possibly  a stable 1 mHz reading can be made. The inertia added to the 4046 PLL for the voltage controlled oscillator is 1 M with 500 U.f. condenser. It takes a good few minutes to creeps up to the target frequency. How could that kind of ineria be jumping around like a jelly bean at 10 mHz. That is a rhetorical question. I am just ventilating Thanks for your interest and I took note of your caution on GPSDOs bought on ebay from another thread. Will need one eventually so it is good to know.

Hi

Grab a couple of YIG oscillators and phase lock their 20 GHz outputs to your 10 MHz OCXO's. You will need some microwave dividers with a good noise floor and an equally good phase detector. Run them into a 20 GHz mixer and you will get a beat note out that moves 2,000X as fast as the 10 MHz note. If you want to go to 10,000X, do it with 100 GHz sources.

Now, when you do, make very sure that the close in noise of your PLL is good enough to still be dominated by the noise of the 10 MHz OCXO's. Your microwave sources will be 60 to 100 db more noisy than the OCXO close in. It's that noise you have to keep down in order to still have a good measurement. A multiplication that increases the noise faster than the resolution is easy to do, but not of much use.

Once you have the beat note that is clean, you still need to do the digital signal processing on it to come up with the data. The OCXO only is stable enough to see what you want to see if the signal is properly processed.

Bob

[zlymex]

I can test 10MHz down to 1uHz by a commercial frequency counter.

[ChunkyPastaSauce]

I can test 10MHz down to 1uHz by a commercial frequency counter.

Time to upgrade

[uncle_bob]

I can test 10MHz down to 1uHz by a commercial frequency counter.

Time to upgrade

Hi

And you can be sure it's good to 26 digits in a second because pictures *never* lie.

Bob

[John Heath]

Stanford Research Systems model SR620. This puppy can park itself on my bench any time it wants. 

Did not know they made a 26 digit model ?? And why would it have not 1 not 2 but 3 mode switches? You can only use one mode at a time so why would there be 3 mode switches almost as if it were 3 frequency counters. This is indeed odd ?

[uncle_bob]

Stanford Research Systems model SR620. This puppy can park itself on my bench any time it wants. 

Did not know they made a 26 digit model ?? And why would it have not 1 not 2 but 3 mode switches? You can only use one mode at a time so why would there be 3 mode switches almost as if it were 3 frequency counters. This is indeed odd ?

Hi

Save yourself the trouble. I have a number of them, the only way to get a perfect display like that is to loop back the internal standard. That way the noise on the reference cancels out and you get that display.

Yes, there is one other way to get all zeros. It can also be used to expand the display to 26 digits ...

Bob

[KE5FX]

I can test 10MHz down to 1uHz by a commercial frequency counter.

Time to upgrade

 

[danadak]

Basic measurement principles, reciprocal counters, etc..


https://www.dropbox.com/s/dmnyb69ntfgd3o3/Counters.zip?dl=0


Regards, Dana.

[John Heath]

I like the PDF on fundamentals of time and frequency standards. Thanks for posting.

[mrflibble]

Old school TVs found a way to separate noise from frequency for horizontal sync by using high Q resonance with only a 2 to 5 percent composite video sync injection. By doing it this way a 25/75 noise to video ratio could still sync up horizontal scanning rate. I use the same method by a 5 percent coupling of both oscillater inputs to 10 MHz crystals tuned to a slightly lower frequency. 90 % of jitter and noise information is lost. However signal to noise is much improved for a frequency measurement. In your case jitter and noise measurement could be a higher priority. For this a phase detector between high Q resonant crystal and oscillator source would improve jitter and noise measurements but frequency information would be lost. In summary by sacrificing jitter and noise information frequency measurement is improved or by sacrificing frequency information noise and jitter information is improved.

Interesting approach. Do you maybe have a rough schematic of how the oscillator and 10 MHz crystals are connected? I think I get what you mean, with that description, but not sure...

Incidentally, if I understand what you're doing correctly, isn't that more in the realm of clock recovery, and less in the frequency measurement area? What with reducing close in spurs... Also, if I understand the description correctly, aren't you attenuating the lower-than-fundamental-frequency spurs more than you are attenuating the higher-than-fundamental-frequency spurs?

[John Heath]

Old school TVs found a way to separate noise from frequency for horizontal sync by using high Q resonance with only a 2 to 5 percent composite video sync injection. By doing it this way a 25/75 noise to video ratio could still sync up horizontal scanning rate. I use the same method by a 5 percent coupling of both oscillater inputs to 10 MHz crystals tuned to a slightly lower frequency. 90 % of jitter and noise information is lost. However signal to noise is much improved for a frequency measurement. In your case jitter and noise measurement could be a higher priority. For this a phase detector between high Q resonant crystal and oscillator source would improve jitter and noise measurements but frequency information would be lost. In summary by sacrificing jitter and noise information frequency measurement is improved or by sacrificing frequency information noise and jitter information is improved.

Interesting approach. Do you maybe have a rough schematic of how the oscillator and 10 MHz crystals are connected? I think I get what you mean, with that description, but not sure...

Incidentally, if I understand what you're doing correctly, isn't that more in the realm of clock recovery, and less in the frequency measurement area? What with reducing close in spurs... Also, if I understand the description correctly, aren't you attenuating the lower-than-fundamental-frequency spurs more than you are attenuating the higher-than-fundamental-frequency spurs?

Thank you for taking an interest in this.  "more than you are attenuating the higher than fundamental frequency". Well said and I found this out the hard way after building the electronics My alternative solution was to use a graphic equalizer on the beat frequency difference between the two 10 MHz OSCs to try and clean it up a bit. It worked sort of down to 100 m Hz , 10 m Hz on a good day but that is a far cry from a 10 u Hz measurement for the difference between the two. The experiment sits on my bench for the last year with my elisions that some day I will fix it . So far no luck.

[mrflibble]

Thank you for taking an interest in this.  "more than you are attenuating the higher than fundamental frequency". Well said and I found this out the hard way after building the electronics My alternative solution was to use a graphic equalizer on the beat frequency difference between the two 10 MHz OSCs to try and clean it up a bit. It worked sort of down to 100 m Hz , 10 m Hz on a good day but that is a far cry from a 10 u Hz measurement for the difference between the two. The experiment sits on my bench for the last year with my elisions that some day I will fix it . So far no luck.

I don't know how much this will help you towards your end-goal. But concentrating on the sub-problem "Reduce close in spurs, and reduce them symmetrically", there's a fairly straightforward modification you can try.

Get two more crystals that are also detuned, but these two are tuned to a slightly higher frequency. No doubt there will be mismatches in both frequency location and "spur absorbance" (for lack of a better term). But as long as the frequency differences of the fundamental & low detuned frequency is about the same as  for the fundamental & high detuned frequency. You can tune out the mismatches to some degree by adjusting the coupling factors and the dissipation factors. Basically adding a few extra Rs & Cs at the relevant locations.

If you characterize your 4 detuned crystals in advance, you can even try to make some educated guesses. Then measure and see how well reality matches your predictions, and adjust model accordingly. I have no idea if the amount of extra work is justified because I cannot predict the results, nor can I predict how useful they are for getting you to your end goal. But like I said, if you just want to see if things improve enough if only that darn spur attenuation was symmetrical, this is a low cost way of finding out. Not taking time spent into account for cost, because the idea would be you pay some time, and you gain some insight. Well, at least that's what I always tell myself in those situations.

[John Heath]

I agree. You have to fail to learn. I can control the beat frequency between the two 10 MHz oscillators to around 10 Hz with temperature or 100 Hz if need be by selecting different crystals. The main problem is jitter noise. Both 10 MHz oscillators have 7805 regulated power supplies coming off a 9 volt rechargeable battery to eliminate ground issues. One of the oscillators comes in through a 20 foot toslink fiber cable. This was necessary to burn off 10 K volts dc from the Faraday cage it lives in. I believe most of the noise is coming from the conversion of TTL to fiber then back to TTL.   I tried a brute force fiber direct mixer to recover the beat frequency. With toslink fiber this is easy by using a y connector. This has an advantage of 10 MHz beat plus a 500 THz beat of the two fiber red LEDs. The bandwidth is too broad for it to be useful but maybe with match laser type sources it could pan out. Who needs 10 MHz clock reference if there is the possibility of using a 500 THz source , red light . Much better if it can be made to work. It does work for 10 MHz beat but can not find the 500 THz beat frequency. It could be up in the 200 MHz range putting it out of range my scope to see it in FFT mode? 

[mrflibble]

I suspect that going the laser mixer route will open the external cavity can of worms (due to the narrow linewidth requirements). The ever popular project within a project.

[John Heath]

I got it done by pulling one of the fiber connectors out a little from it's normal snap in position. . By fiddling a bit there is a sweet spot where both laser inputs in the Y adapter are making a 50 / 50 contribution. As luck would have it there was enough friction for the fiber plug to stay put ounce the sweet spot was found. A 500 THz mixer , not bad for of the self parts.

[mrflibble]

Oh wait, I think I understand now what your intention is. So you have two separate laser diodes, shine those into separate fibers, have those fibers meet in the middle aka Y adapter. Then one of following two:
1) have something happen at "the sweet spot"
2) hope that something happens at "the sweet spot"

then have photons propagate through some more fiber towards the detector (presumably a photodiode?). Have photons interact with detector. Do clever things with electronics. Watch scope display.

Is that a reasonable discription of the signal chain? Because if yes, then I have some bad news. Photons only do cool stuff when the surrounding matter is really convincing. Otherwise they are lazy bastards and will just ignore their fellow photons. Even when those fellow photons are really close at Y-sections. But you should have known that, because hey, boson. Want something more? Hire fermions (electrons).

(no photon-photon interaction at the Y-section)

[John Heath]

I understand what you are saying about the boson quantification issue. Photons in itself is quantification in it purest form and there is not much that can be done about it. However in this case it is thousands of photons mixing together so it will average out to our friendly Newtonian world. By using a Y adapter for the fiber acts as an adder to mix the two 10 MHz oscillators. With a little luck it will also mix the two red lasers for the 500 THz frequency difference.

My goal is to measure the voltage coefficient of a crystal oscillator. Not the voltage driving the crystal but the voltage frame of reference voltage coefficient. For example when flying in an air plane at 32,000 feet your voltage frame of reference is around +100 K volts. Does a crystal oscillator run at the same speed at 100 K volt frame of reference . This is the question answer. This is the reason for my 10 K volt Faraday cage to change the voltage frame of reference of the crystal every 10 seconds by 10 K volts and test for a small difference. My predicted frequency change is 50 u Hz for a 10 MHz crystal for a 10 K volt change in voltage frame of reference. As you can imagine a 50 u Hz change of 10 MHz is not easy to measure. 

[mrflibble]

I admire your optimism, but that's not what I was getting at.

With a little luck it will also mix the two red lasers for the 500 THz frequency difference.

Snowball's chance in hell. Just no way. But tell you what. If you bought that fiber + splitter at regular consumer prices at the magic emporium, and it does indeed do proper mixing in the fiber, I will buy said fiber by the truckload for double the price. What the hell, make it triple! People have to do actual work with temperature controlled lasers, gratings + control loops just to get lasers that enable that. And after that they have to expend actual effort in getting fiber optics with juuust the right properties. How likely is it that "stick it in at an oblique angle" is a substitute for that?

You can even employ logic... What makes the Y-section different from a straight edge of fiber? What makes a photon more likely or less likely to mix in the y-section than a straight fiber stretch? Given these two hints, take the situation of a single laser A, laser B is off. With just laser A in a straight piece of fiber, or the magic y-section for that matter, why would those photons not suddenly selfmix in equivalent circumstances?
Because the photons of just laser A passing through the magic Y-section with "properties" have sufficiently different energies that if they were doing what you hope they were doing, then you'd also be able to see that (the selfmixing). And then you would heat/cool the laser, and check if the change in measured signal agrees with expectations for the selfmixing scenario.

The 10 MHz beat you are seeing is not photon-photon "mixing" interaction. What you are seeing is the equivalent of this setup:
- take photodiode
- divide it in left and right part (just for sake of argument, in reality people don't give a shit because same results no matter where you shine)
- laser A of some nominal wavelength, use 10 MHz signal to modulate laser A supply current, voltage, whatever really.
- laser B of some nominal wavelength, use another 10 MHz signal to modulate laser B supply current, voltage, whatever really.
- shine laser A on the left part of the photodiode
- shine laser B on the right part of the photodiode
Tadaaaaaa, "mixing". For suffiently correct use of the word mixing. In that setup you will indeed be able to see the beat frequency that is the difference between the two 10 MHz signals. But alas, no 500 THz mixer on the cheap.

Couple of random links:
https://en.wikipedia.org/wiki/Laser_linewidth
https://en.wikipedia.org/wiki/Tunable_laser#Narrowband_tuning
https://en.wikipedia.org/wiki/Four-wave_mixing#Applications_of_FWM
google this one: "laser diode external cavity wavelength", loads of articles related to one of the required ingredients.
or google this: "four-photon mixing fiber". Yes, your hoped for case would be two-photon mixing, but the info density is higher for the 4-photon google.

Anyways, I hope that makes it clear what you are seeing. If not ...  that was my best shot. 

[T3sl4co1l]

Incidentally: for a long time a lot of people believed that independent light sources couldn't produce interference fringes (or coherent RF mixing products -- same thing, temporal rather than spacial).  Of course, with no justification for that -- on the contrary, theory clearly predicts such.  This is a consequence of belief-based-on-experience versus belief-based-on-proof.  Both belief systems have their downsides regarding new information: the experimentalist assumes nothing new will ever happen, except when it does; the theorist assumes that that theory is complete, that nothing new will be added to it.

As it turns out, lasers just needed to be cleaned up a little bit.  After all, we're talking parts per billion stability here, either to obtain a reasonably clean beat frequency, or to obtain an interference pattern that can be photographed with a high speed camera, if perhaps not observed by eye.

Tim

[John Heath]

I would call your best shot a double side band in the side pocket. Well done. You put your finger right on the problem. A 500 THz beat frequency is unlikely. It was a quick test so it did not hurt to try. Results were negative in agreement with your thoughts. Why would I even try for the 500 THz beat ? An act of desperation as you never know when god will cut you some slack. Not in this case.

It is worse than you think. In order of best to least best fiber cable there is single mode , multi mode and then there is toslink fiber. I am using tosling fiber. It is a standard developed by Toshiba to replace audio cables with fiber. Workable band width is 10 MHz driven by a red LED not solid state laser. On the positive side it is as cheap as dirt. I can buy toslink fiber cables terminated with their plugs at the local dollar store. Toslink TTL to fiber and fiber to TTL adapters are only 30 bucks each. The original reason was to burn off 10 K volts from the Faraday cage. The idea of a 500 THz red beat frequency with Y adapters came to mind later.

The 50 u Hz measurement is still out of range. I am only down to 10 m hz range on a good day for a 10 MHz crystal. There is a third option. Increase the voltage frame of reference change from 10 K volts to 200 K volts with a Van de Graph generator. They are on ebay for about 300 bucks. This would change the predicted frequency change from 50 u Hz to 1 m Hz. Still just outside the range of measurement but getting warmer.

 

[John Heath]

Incidentally: for a long time a lot of people believed that independent light sources couldn't produce interference fringes (or coherent RF mixing products -- same thing, temporal rather than spacial).  Of course, with no justification for that -- on the contrary, theory clearly predicts such.  This is a consequence of belief-based-on-experience versus belief-based-on-proof.  Both belief systems have their downsides regarding new information: the experimentalist assumes nothing new will ever happen, except when it does; the theorist assumes that that theory is complete, that nothing new will be added to it.

As it turns out, lasers just needed to be cleaned up a little bit.  After all, we're talking parts per billion stability here, either to obtain a reasonably clean beat frequency, or to obtain an interference pattern that can be photographed with a high speed camera, if perhaps not observed by eye.

Tim

A laser interference pattern is what was used to detect a gravity waves. As you say if you clean it up a bit it could be possible for a beat frequency.  Air craft navigation laser ring to replace gyros is another example of beat frequency . Interference pattern from a double slit experiment is a third example. As long as there are enough photons for them to average out to a wave.

[T3sl4co1l]

Well, those are all correlated sources (a single laser).  You can do that with sunlight, if your experiment has a short enough path length to maintain (spacial) coherence.  You can't do it with, say, two separate incandescent bulbs.  But you can do it with two independent radio antennas on the same frequency, or two independent lasers tuned to the same frequency.

Tim

[John Heath]

Well, those are all correlated sources (a single laser).  You can do that with sunlight, if your experiment has a short enough path length to maintain (spacial) coherence.  You can't do it with, say, two separate incandescent bulbs.  But you can do it with two independent radio antennas on the same frequency, or two independent lasers tuned to the same frequency.

Tim

The coherence of a red laser pointer is surprisingly high. All photons lined up in phase and polarity 0 degrees. On my red pointer a 90 degree polarizing sheet will blank out the laser to the point that it can not be seen. However if you start to take a closer look at the photon and width it is not as perfect as it seems. The LED photon source that is driving the laser resonate cavity sets the range of photons available.  It is the resonate light cavity in front of the LED that is narrowing the band width for the photons. How narrow is that? I googled into the wee hours of the night and found it was narrow but not really narrow like you would expect. The problem is in the cavity walls where the photons resonate. Tiny imperfections in it's flatness causes little side bands on the laser light. With all these little imperfections in the resonate cavity a LED type laser ends up looking like a bell curve with 3 DB points at 200 MHz for 500 THz. That is a high Q for for 500 THz but under a magnifying glass it is a dirty carrier. A Helium laser is better but that is mainly because it has a glass resonate cavity not plastic one like the red LED pointer.

I am going to google up some laser band widths so you can see what I am getting at and some nice lectures on the subject.



 

[mrflibble]

A Helium laser is better but that is mainly because it has a glass resonate cavity not plastic one like the red LED pointer.

Mmmh? The resonant cavity of that red laser diode would be made of semiconductor. The edges (mirrors) are formed by cleaving the crystal.

Quick reference, third paragraph here: https://en.wikipedia.org/wiki/Laser_diode#Failure_mechanisms

At the edge of a diode laser, where light is emitted, a mirror is traditionally formed by cleaving the semiconductor wafer to form a specularly reflecting plane. This approach is facilitated by the weakness of the [110] crystallographic plane in III-V semiconductor crystals (such as GaAs, InP, GaSb, etc.) compared to other planes. A scratch made at the edge of the wafer and a slight bending force causes a nearly atomically perfect mirror-like cleavage plane to form and propagate in a straight line across the wafer.

When designing experiments: more precision, less assumptions based on suboptimally-evolved-monkey-brain intuitions. I should know, I have one of those suboptimally evolved things as well. Intuition for avoiding sable tooth tigers? Check! Intuition for avoiding locations strewn with fellow primate bones? Check! Intuition for dealing with 100+ dimensional subspaces? No habla 4+-d senor!


But seriously, don't just assume stuff.