The drift you see is normal because the rubidium oscillator is not locked to any common reference.. Actually, the drift should be more prominent on your GPS disciplined oscillator. To observe the accuracy of the rubidim clock you shoud make it also GPS disciplined as you planned, and then compare both disciplined oscillators (OCXO and rubudium) between update periods from the satelllite and that will tell you which one drifts more, and should confirm the the rubudium one has less drift.
I used these in a positioning system called Starfix. Starfix was a positioning (navigation) system that was designed for providing precise positioning information for offshore work in the Gulf of Mexico. It was similar to a differential GPS system but we used geosynchronous satellites and put up our own spread spectrum signal. GPS was not available at the time. As I remember we built 62 systems. They were great little oscillators, very low phase noise. I still have an Efratom FRT in my lab that has been running since 1987.
Were those made before people realized that general relativity needs to be compensated for? I’ve always heard stories that the first attempts at satellite positioning had accuracy problems relating to forgetting that you’re dealing with relativistic dilation
@@markhaus The relativistic problems were anticipated very early in the development of GPS. With Starfix we did not have that problem because our clocks were all on the ground, not flying clocks as in GPS.
6:50 A-ha! That was on my mind since the beginning of the video - a useful bit of information there, saving us the trouble of finding out ourselves. I guess that makes buying a second hand one a gamble then.
For anyone curious, I did a quick calculation of the frequency difference between the GPSDO and the Rb oscillator. Using the information from www.microfarad.de/blog/comparing-frequencies-with-an-oscilloscope/, I found that at 9:40 the Rb 10 MHz was scrolling at about 1 period every 1m8s, which works out to a difference of about 14 mHz or 1.4 ppb at 10 MHz. I'd say that's pretty damn close!
Nope, 10^-9 is pretty bad. Good OCXO achieve this already , an Rb clock is about 100 times more stable/uncertain. This FRS is capable of regular 10^-11 down to several parts in 10^-12 uncertainty, using an external trim pot, and compared to Cs or a good GPSDO assembly. Comparison time constant is at least 3h (Allan Deviation) The FRS needs a few days until it reaches this kind of stability, as well your GPSDO probably needs 2 days, until its OCXO is trained to low uncertainty. Additionally, GPSDO, depending on the OCXO support of your unit, has a short term instability up to 10^-9, due to the unavoidable varying signal path from the satellites. A Thunderbolt GPSDO with the good OCXO shows better performance. Please! Assemble your Rb into a big case, with good cooling by the case ground plate. Then compare your GPSDO and your FRS via T.I. measurements, hopefully you have a good counter, like the 5370B, 5335A, or the more modern ones as 53132, then you could better use TimeLab by John Miles, KE5FX to monitor those difference on a longterm basis. Visit his homepage for further information. The final stability will be found only after 3h at least. What you show is the usual short term jitter, not the uncertainty of the Rb, or "accuracy", as you call it. I suggest that you leave your Rb as standalone, don't use it inside your GPSDO, as you will sacrifice its very good short term stability. Get a new OCXO for your GPSDO module. Btw.: I also have this FRS, added a 1pps divider (PIC solution by the time-nuts community), a Thunderbolt and both power supplies in one case.
@@DrFrank-xj9bc you're right, I got my exponents crossed. 1 ppb is absolutely not 1E-12. I'll definitely include some longer-duration runtime tests in the follow-up video. The most precision I've been able to measure my GPSDO was down in the 1E-10 range, and it compared well to a commercial HP GPSDO, but beyond that I don't have the gear on hand to dive deeper. The best I can do is log control voltage DAC output of the GPSDO and plot its ADEV over time, which does get down into the 1E-11 to 1E-12 range as you'd expect. I may bring the FRS into work and calibrate it to our Fluke 910R and use it as you say - as a second non-GPS locked reference standard. Thanks for the awesome info! You're obviously a much more informed time nut than myself.
A 50 terminator will definitely help clean up your waveform. My eBay FRS (an even older model) was pretty funky running straight in to the high impedance input of the Oscope. With the termination it was perfect. BTW, back in the '70s I used to work with the HP Cesium standard as shown in the video and have used various Cesium and Rubidium standards over the years.
I repaired some rubidium oscillators in the 80s. They were pretty accurate. (10E-11 if memory serves) We compared them to national standard broadcasted during a few hours (they were 10E-12 or -14 at the time. I’m not sure)
Yeah, we get better than 10E-11 on the rubidium we have at work and it is around 20 years old. Every four years (or thereabout) it requires some adjustment from the front panel (C-Field).
Just a thought. There should be a C-Field trim to bring the frequency in. The hyperfine transition is influenced by the magnetic field and so the C-field trim compensates for this. That will likely be the external trim voltage you will be using your GPS disciplining board to control.
The external trim is there to take care of the last few decimals of error. A stable voltage there, whether from a GPS disciplined servo, or just days of manual adjustment, should get you to (or near) the rated accuracy. You may need to do some tweaking of the servo coefficients on the GPSDO board to better suit the stability characteristics of the Rb relative to Quartz oven.
Hello. I have the same model of oscilloscope. It is much better and precise to use 1 and 3 or 2 & 4 channels at the same time. Because of internal architecture of this model of SIGLENT. When you use 1&2 (or 3&4) channels at he same time, ADC works in interlive mode and the real sampling rate is two times lower.
Cool video! I played around a lot with Rubidiums back in the day when the FE-5680As were cheap and plentiful on eBay. As you know and mentioned, the more references you have, the less sure you are about which one is right. :) Rbs aren't perfect, with good enough equipment you can measure the drift and other weirdness that they exhibit, it's just further down in the noise of most people's test equipment. Getting that hooked up to your GPSDO, especially if you put it on a UPS to ride through power outages, should net you really nice long term stability. The OCXO version of your reference might still be useful if it ends up having lower phase noise. I'm looking forward to the follow up!
I've long wanted to build a 10MHz Rubidium frequency standard for my lab, but the formerly plentiful modules from the old GSM stations have turned rather rare by now, and usually go for 250$+, which just isn't worth it for me anymore.
I built a totally digital GPSDO/NTP server around an Rb (Symmetricom X72 or SA.22c - now owned by Microchip, by way of Microsemi) some years back. It did all of the PLL stuff in software, using the Rb's digital frequency adjustment, and it output a whole bunch of stats over the network so you could monitor it in real time. I got the control loop dialed in real nice and it worked great. Ironically, I built it before I became a ham, and disassembled it shortly after becoming a ham, because I'm terrible at hardware design. The thing had RF (10MHz and some other frequencies) flying around all over the place on unshielded wires, and it was in the same room where I was building my HF shack, so I had to kill it to get rid of the interference! For a long time I've wanted to bring it back in a better version without the random mass of wires, but learning PCB design is on my "maybe someday" list.
I started out with an FEI 5650A but I decided that trying to use a DAC to digitally control an analog tuning input wasn't my cup of tea. Too many unknowns and nonlinearities for me.
Your FE5650A didn't have an RS232 port on it? The only way I found to tune the ones I had was over serial, although FEI were famous (notorious?) for making a wide range of products with disjoint feature sets but the same P/N @@hobbified
@@aftbit it was a real "project" device so it changed over the years, but the nicest version was: an Arduino Due clone board (ATSAM3X8E CPU), the X72 connected to external timer inputs, a uBlox M8T GPS, and the uBlox set to timestamp a PPS coming *out* of my board, which offered a bit more resolution than having my board timestamp the GPS's PPS. A 32-bit timer could handle counting up to 10,000 trivially. The timestamp offsets were fed into a dual FLL/PLL (which I think of as the P and I of a PID loop only with a time derivative thrown in), with adaptive time constants so that it would run low tau at startup when still warming up, and increase up to tau=3 hours once stable.
Yes you do. We all do whether we like it or not. Just some of us are inherently sick in the head and it seems like fun. Ps. Your user name is relevant to a problem my aunt had with thermal loss calculations in the early 70s working for dow chemical making closed cell styrofoam roofing insulation. So I like your user name.
I believe if you are triggering on your reference frequency and the signal you are observing is drifting left it is high in frequency (shorter in time) with respect to the trigger and the opposite (longer in time) if drifting right. Otherwise a great video. I had a M100 running for about 25 years before it gave out. I need to pull it off the bench and work on it. Sam W3OHM
I have a similar rubidium oscillator and that also lost sync from time to time. When I attached a heatsink and addad a blower to keep the case cool, it didn't do that anymore. Just make sure the cooling is not too much or the heater will have a hard time...
Nice video, but I have couple of notes. 1. You really should have a proper heatsink on the RbO base plate. The base plate temperature should be somewhere around 30-38 deg. C (mine is currently operating at 33 deg. C with passive heatsink). 2. Personally, I am not sure that the idea to put RbO inside the GPS loop is a good idea. I am certainly no time/frequency professional or academic, but I had a rather long e-mail discussion with Brooks Shera (the man, who basically invented the idea of GPS pacification of OCXOs back in 1998) in summer 2000 about this aspect. And his opinion was that RbO is much better to be left alone as a free running frequency normal with excellent short term stability and that the "kicking" of RbO with 1PPS can only make things worse, especially if the time constant of FIR filter is not long enough (typically days or weeks). Same thing is valid when it comes to external steering by some component. Either the multi-turn potenciometer or some DAC. The thermal stability of those components is almost always much worse then the thermal stability of the RbO module without any external stimulus. 3. In my home lab, I am using GPS pacified OCXO (an old HP10811A) with Motorola OnCore UT+ (in position fix mode) built according to the Brooks Shera design - www.qsl.net/n9zia/wireless/QST_GPS.pdf for almost 25 years with only one interruption in 2002 (floods in Czech republic) and even today, the precision is really good. But the short time stability is not so good. So, if I want to measure something with better short term stability, I simply switch on my RbO, wait 3-6 hours or better to next day and do measurements. After this, I switch my RbO off until next measurement. RbO is not like OCXO and it does not need to be switched-on without interruption for decades. It will not become better after ten or twenty years, like good red wine or OCXO. :)
I just checked my data and during last comparison (approx. two months ago) and the difference between my free running RbO (overnight time to stabilise) and my GPSDO was only 20ns in 158 seconds. It means only 0.126 ns or 126ps per second. Because it is only relative measurement, I am not able to say which source is "right". I have neither hydrogen maser nor a cesium fountain at home. But both sources are good enough for my all purposes. BTW, the GPSDO is an old, vintage Brooks Shera design with original PIC program. My only modification was using one isp1032 Lattice CPLD instead of all those counters in original design to save space on PCB. Pretty good result for a 25 y.o. box which is silently working on the bench since 1999.
They're *stable* (once warmed up and locked), and they can be reasonably (extremely) accurate 'out-of-the-box'. But they're not *inherently* exactly 10MHz, just really (really, really...) close. Local effects (e.g. magnetic environment) and shipping might have voided (LOL) any previous calibration. They can be calibrated using the potentiometer input to fine tune some internal magnetic field. And they'd need to be recalibrated in a mandated cycle (e.g. yearly). You calibrate it against the GPSDO, which is (by design) traceable back to government standards. 73
They make a nice basis for a GPSDO. But as far as going out whack... the tuning adjustment is measured in ppb or ppt because even an old, tired, not-calibrated-in-years one is going to be much better than 1ppm. This one looks like it's running at about 9.99999987 MHz as soon as it locks on - about 13ppb slow. More than good enough for SSB on 1296 :)
Ya... I appreciate W3AXL's video; but, it's just propagating a lot of nonsense about these old beat up cell tower pulls. There's a reason they end up at scrappers, and on Ebay... and that reason does not bode well for the end user... if they are doing anything serious with them... other than thinking they have a serious 'frequency standard.' Sometimes... for significant... and possibly complex reasons... really doesn't have any new clothes.
@@willthecat3861 I work in the satellite industry and would be interested in hearing more about the reason(s) as I'm not as close to the cellular BS systems. Certainly one would be that they're aged and the bulb's limited lifetime and non-serviceability make them potential issues. The other is that as cell systems are being replaced, it is far less expensive to go with a new GPSDO than a new Rubidium standard. A decent GPSDO with a disciplined OCXO provides a fine frequency/time standard that will freewheel through any short term GPS reception issues.
Anything is serviceable just not all companies respect you enough to let you buy parts. If I understand your explanation right it may be possible to replace the IR lamp with something else that can generate the correct light.
Jumping to conclusions, are we? Have anyone actually worked assembling something this complicated? Pretty sure not - one thing you have to make it serviceable, another thing tuning and assembly can be quite involved, requiring special tools and jigs. Place I've worked did rotary encoders - how hard can it be to replace the source in those? Turns out pretty involved - you have to position the LED correctly, trim its light with adjustment screws, take the thing to the comparator to confirm. This one would be no different.
Efratom were in fact really good like that - you could order anything that had a part number. The reason there was no lamp replacement procedure in the manual was that the lamp was projected to last for the operating life of the unit because it had a PTFE inner coating that eliminated the diffusion issues that limited lamp life in the earlier units. That diffusion issue problem (Rb migrating into the bulb) was the reason that the earlier models (like the FRK series) had lamps that were easily replaceable. And yes, you can replace the light source with anything that creates light of the correct wavelength - in fact, if you use a monochromatic light source like a VCSEL then you can even get rid of the isotopic filter, since that's only needed to eliminate the unwanted emission peak of the discharge lamp. You still typically need temperature stabilization of the optical source because although VCSELs exhibit low sensitivity to wavelength vs temperature it's still non-zero.
@@VEC7ORlt I never said it was easy, optical systems are pockey as all heck but the key word here is POSSIBLE. Also don't assume anything else is jumping to conclusions.
@@deltacx1059 Yes... possible. Pragmatic, practical, doable... probably not. But, someone will make a video about doing it with 'scotch tape and bubblegum.'
The rubidium cell is sealed so ambient pressure outside it shouldn't affect it. The pressure of any residual gas including the Rb vapor within will vary with temperature according to well known gas laws. This is one reason why a stable temperature environment is desirable. In general frequency standards can serve as thermometers, magnetometers, manometers, seismometers, etc. unless precautions are taken.
what a fascinating video. i have now learned some things. thank you sir. i will hit that subscribe button. hmm my thoughts are that the drift is too much. the rate of rotation of that circle is pretty fast. my thoughts are that a magic cesium clock and gps satellite would not be off by that much. a little bit yeah but not that much
Getting some cheap chinese 50 ohm in-line terminators for your scope might be a cheap investment. Will they be flat, capacitance-less response up to 1Ghz? lol no. Will they do the job for 99.9% of cases up to 100Mhz? Yeah. That slow of a lissajous plot for a 20 year old, completely free-running, clock is amazing. Atoms, man.
![กี่หมื่นฟ้า | Your Sky Series EP.1 [1/4]](http://i.ytimg.com/vi/vLGjxFL6U_I/mqdefault.jpg)
The drift you see is normal because the rubidium oscillator is not locked to any common reference.. Actually, the drift should be more prominent on your GPS disciplined oscillator. To observe the accuracy of the rubidim clock you shoud make it also GPS disciplined as you planned, and then compare both disciplined oscillators (OCXO and rubudium) between update periods from the satelllite and that will tell you which one drifts more, and should confirm the the rubudium one has less drift.
I used these in a positioning system called Starfix. Starfix was a positioning (navigation) system that was designed for providing precise positioning information for offshore work in the Gulf of Mexico. It was similar to a differential GPS system but we used geosynchronous satellites and put up our own spread spectrum signal. GPS was not available at the time. As I remember we built 62 systems. They were great little oscillators, very low phase noise. I still have an Efratom FRT in my lab that has been running since 1987.
Were those made before people realized that general relativity needs to be compensated for? I’ve always heard stories that the first attempts at satellite positioning had accuracy problems relating to forgetting that you’re dealing with relativistic dilation
@@markhaus The relativistic problems were anticipated very early in the development of GPS. With Starfix we did not have that problem because our clocks were all on the ground, not flying clocks as in GPS.
Amazing video. I see I am not alone in the "one man with two watches never know the exact time" thing. Can't wait for part 2 !!!
i haeve 3 watches on my desk. right now they are prety calibrated 110 secund per day
@@uiopuiop3472 I hope you mean "milliseconds" 😂
6:50 A-ha! That was on my mind since the beginning of the video - a useful bit of information there, saving us the trouble of finding out ourselves. I guess that makes buying a second hand one a gamble then.
For anyone curious, I did a quick calculation of the frequency difference between the GPSDO and the Rb oscillator.
Using the information from www.microfarad.de/blog/comparing-frequencies-with-an-oscilloscope/, I found that at 9:40 the Rb 10 MHz was scrolling at about 1 period every 1m8s, which works out to a difference of about 14 mHz or 1.4 ppb at 10 MHz. I'd say that's pretty damn close!
Yep, I was screaming that at the video. ;-) !
Nice video, just lovely.
Google the Time Nuts crowd, they check Relativity by parking up a mountain.
Nope, 10^-9 is pretty bad. Good OCXO achieve this already , an Rb clock is about 100 times more stable/uncertain.
This FRS is capable of regular 10^-11 down to several parts in 10^-12 uncertainty, using an external trim pot, and compared to Cs or a good GPSDO assembly.
Comparison time constant is at least 3h (Allan Deviation)
The FRS needs a few days until it reaches this kind of stability, as well your GPSDO probably needs 2 days, until its OCXO is trained to low uncertainty.
Additionally, GPSDO, depending on the OCXO support of your unit, has a short term instability up to 10^-9, due to the unavoidable varying signal path from the satellites.
A Thunderbolt GPSDO with the good OCXO shows better performance.
Please! Assemble your Rb into a big case, with good cooling by the case ground plate. Then compare your GPSDO and your FRS via T.I. measurements, hopefully you have a good counter, like the 5370B, 5335A, or the more modern ones as 53132, then you could better use TimeLab by John Miles, KE5FX to monitor those difference on a longterm basis. Visit his homepage for further information.
The final stability will be found only after 3h at least. What you show is the usual short term jitter, not the uncertainty of the Rb, or "accuracy", as you call it.
I suggest that you leave your Rb as standalone, don't use it inside your GPSDO, as you will sacrifice its very good short term stability. Get a new OCXO for your GPSDO module.
Btw.: I also have this FRS, added a 1pps divider (PIC solution by the time-nuts community), a Thunderbolt and both power supplies in one case.
How do you get the Rb 1pps to be in phase with the GPS 1pps? Is that part of the PIC software?
@@DrFrank-xj9bc you're right, I got my exponents crossed. 1 ppb is absolutely not 1E-12. I'll definitely include some longer-duration runtime tests in the follow-up video. The most precision I've been able to measure my GPSDO was down in the 1E-10 range, and it compared well to a commercial HP GPSDO, but beyond that I don't have the gear on hand to dive deeper. The best I can do is log control voltage DAC output of the GPSDO and plot its ADEV over time, which does get down into the 1E-11 to 1E-12 range as you'd expect.
I may bring the FRS into work and calibrate it to our Fluke 910R and use it as you say - as a second non-GPS locked reference standard. Thanks for the awesome info! You're obviously a much more informed time nut than myself.
Oh you brave soul! You’ve gone and rustled the time-nuts. Prepare your brain for the forthcoming onslaught.
A 50 terminator will definitely help clean up your waveform. My eBay FRS (an even older model) was pretty funky running straight in to the high impedance input of the Oscope. With the termination it was perfect.
BTW, back in the '70s I used to work with the HP Cesium standard as shown in the video and have used various Cesium and Rubidium standards over the years.
Great video. Hope the wait for part two isn’t long
Невероятно полезное и уникальное видео. Искренне благодарю Вас за этот труд!
I repaired some rubidium oscillators in the 80s.
They were pretty accurate. (10E-11 if memory serves)
We compared them to national standard broadcasted during a few hours (they were 10E-12 or -14 at the time. I’m not sure)
Yeah, we get better than 10E-11 on the rubidium we have at work and it is around 20 years old. Every four years (or thereabout) it requires some adjustment from the front panel (C-Field).
Just a thought. There should be a C-Field trim to bring the frequency in. The hyperfine transition is influenced by the magnetic field and so the C-field trim compensates for this. That will likely be the external trim voltage you will be using your GPS disciplining board to control.
The external trim is there to take care of the last few decimals of error. A stable voltage there, whether from a GPS disciplined servo, or just days of manual adjustment, should get you to (or near) the rated accuracy. You may need to do some tweaking of the servo coefficients on the GPSDO board to better suit the stability characteristics of the Rb relative to Quartz oven.
Hello. I have the same model of oscilloscope. It is much better and precise to use 1 and 3 or 2 & 4 channels at the same time. Because of internal architecture of this model of SIGLENT. When you use 1&2 (or 3&4) channels at he same time, ADC works in interlive mode and the real sampling rate is two times lower.
lmao, loved the pokemon lilycove "elevator music" thrown in!
Cool video! I played around a lot with Rubidiums back in the day when the FE-5680As were cheap and plentiful on eBay. As you know and mentioned, the more references you have, the less sure you are about which one is right. :) Rbs aren't perfect, with good enough equipment you can measure the drift and other weirdness that they exhibit, it's just further down in the noise of most people's test equipment. Getting that hooked up to your GPSDO, especially if you put it on a UPS to ride through power outages, should net you really nice long term stability. The OCXO version of your reference might still be useful if it ends up having lower phase noise. I'm looking forward to the follow up!
I've long wanted to build a 10MHz Rubidium frequency standard for my lab, but the formerly plentiful modules from the old GSM stations have turned rather rare by now, and usually go for 250$+, which just isn't worth it for me anymore.
I built a totally digital GPSDO/NTP server around an Rb (Symmetricom X72 or SA.22c - now owned by Microchip, by way of Microsemi) some years back. It did all of the PLL stuff in software, using the Rb's digital frequency adjustment, and it output a whole bunch of stats over the network so you could monitor it in real time. I got the control loop dialed in real nice and it worked great. Ironically, I built it before I became a ham, and disassembled it shortly after becoming a ham, because I'm terrible at hardware design. The thing had RF (10MHz and some other frequencies) flying around all over the place on unshielded wires, and it was in the same room where I was building my HF shack, so I had to kill it to get rid of the interference!
For a long time I've wanted to bring it back in a better version without the random mass of wires, but learning PCB design is on my "maybe someday" list.
I started out with an FEI 5650A but I decided that trying to use a DAC to digitally control an analog tuning input wasn't my cup of tea. Too many unknowns and nonlinearities for me.
Your FE5650A didn't have an RS232 port on it? The only way I found to tune the ones I had was over serial, although FEI were famous (notorious?) for making a wide range of products with disjoint feature sets but the same P/N @@hobbified
Very interesting! I have an SA.22c sitting on a shelf waiting to get built into a GPSDO. Can you give some more details on your setup?
@@TrimeshSZ you're completely right and it turns out that my memory is way faulty. I never had a FE-5650A. I had an LPRO-101. :)
@@aftbit it was a real "project" device so it changed over the years, but the nicest version was: an Arduino Due clone board (ATSAM3X8E CPU), the X72 connected to external timer inputs, a uBlox M8T GPS, and the uBlox set to timestamp a PPS coming *out* of my board, which offered a bit more resolution than having my board timestamp the GPS's PPS. A 32-bit timer could handle counting up to 10,000 trivially. The timestamp offsets were fed into a dual FLL/PLL (which I think of as the P and I of a PID loop only with a time derivative thrown in), with adaptive time constants so that it would run low tau at startup when still warming up, and increase up to tau=3 hours once stable.
Looks like you need a third reference ;)
Nice start screen on the scope btw.
I don't really "do RF" but I really like your presentation style.
Yes you do. We all do whether we like it or not. Just some of us are inherently sick in the head and it seems like fun.
Ps. Your user name is relevant to a problem my aunt had with thermal loss calculations in the early 70s working for dow chemical making closed cell styrofoam roofing insulation. So I like your user name.
Waiting.....waiting....waiting
For part 2.
I believe if you are triggering on your reference frequency and the signal you are observing is drifting left it is high in frequency (shorter in time) with respect to the trigger and the opposite (longer in time) if drifting right. Otherwise a great video. I had a M100 running for about 25 years before it gave out. I need to pull it off the bench and work on it. Sam W3OHM
I have a similar rubidium oscillator and that also lost sync from time to time. When I attached a heatsink and addad a blower to keep the case cool, it didn't do that anymore. Just make sure the cooling is not too much or the heater will have a hard time...
You can put a bandpass filter on the 10 MHz output. That would give you a clean sine wave.
That's a great deal for the efratom part, well explained.
Very nice video, thank you 😊
Nice video, but I have couple of notes.
1. You really should have a proper heatsink on the RbO base plate. The base plate temperature should be somewhere around 30-38 deg. C (mine is currently operating at 33 deg. C with passive heatsink).
2. Personally, I am not sure that the idea to put RbO inside the GPS loop is a good idea. I am certainly no time/frequency professional or academic, but I had a rather long e-mail discussion with Brooks Shera (the man, who basically invented the idea of GPS pacification of OCXOs back in 1998) in summer 2000 about this aspect.
And his opinion was that RbO is much better to be left alone as a free running frequency normal with excellent short term stability and that the "kicking" of RbO with 1PPS can only make things worse, especially if the time constant of FIR filter is not long enough (typically days or weeks).
Same thing is valid when it comes to external steering by some component. Either the multi-turn potenciometer or some DAC. The thermal stability of those components is almost always much worse then the thermal stability of the RbO module without any external stimulus.
3. In my home lab, I am using GPS pacified OCXO (an old HP10811A) with Motorola OnCore UT+ (in position fix mode) built according to the Brooks Shera design - www.qsl.net/n9zia/wireless/QST_GPS.pdf for almost 25 years with only one interruption in 2002 (floods in Czech republic) and even today, the precision is really good. But the short time stability is not so good.
So, if I want to measure something with better short term stability, I simply switch on my RbO, wait 3-6 hours or better to next day and do measurements. After this, I switch my RbO off until next measurement. RbO is not like OCXO and it does not need to be switched-on without interruption for decades. It will not become better after ten or twenty years, like good red wine or OCXO. :)
I just checked my data and during last comparison (approx. two months ago) and the difference between my free running RbO (overnight time to stabilise) and my GPSDO was only 20ns in 158 seconds.
It means only 0.126 ns or 126ps per second. Because it is only relative measurement, I am not able to say which source is "right". I have neither hydrogen maser nor a cesium fountain at home. But both sources are good enough for my all purposes.
BTW, the GPSDO is an old, vintage Brooks Shera design with original PIC program. My only modification was using one isp1032 Lattice CPLD instead of all those counters in original design to save space on PCB. Pretty good result for a 25 y.o. box which is silently working on the bench since 1999.
They're *stable* (once warmed up and locked), and they can be reasonably (extremely) accurate 'out-of-the-box'.
But they're not *inherently* exactly 10MHz, just really (really, really...) close.
Local effects (e.g. magnetic environment) and shipping might have voided (LOL) any previous calibration.
They can be calibrated using the potentiometer input to fine tune some internal magnetic field. And they'd need to be recalibrated in a mandated cycle (e.g. yearly).
You calibrate it against the GPSDO, which is (by design) traceable back to government standards.
73
They make a nice basis for a GPSDO. But as far as going out whack... the tuning adjustment is measured in ppb or ppt because even an old, tired, not-calibrated-in-years one is going to be much better than 1ppm. This one looks like it's running at about 9.99999987 MHz as soon as it locks on - about 13ppb slow. More than good enough for SSB on 1296 :)
Ya... I appreciate W3AXL's video; but, it's just propagating a lot of nonsense about these old beat up cell tower pulls. There's a reason they end up at scrappers, and on Ebay... and that reason does not bode well for the end user... if they are doing anything serious with them... other than thinking they have a serious 'frequency standard.' Sometimes... for significant... and possibly complex reasons... really doesn't have any new clothes.
@@willthecat3861 I work in the satellite industry and would be interested in hearing more about the reason(s) as I'm not as close to the cellular BS systems. Certainly one would be that they're aged and the bulb's limited lifetime and non-serviceability make them potential issues. The other is that as cell systems are being replaced, it is far less expensive to go with a new GPSDO than a new Rubidium standard. A decent GPSDO with a disciplined OCXO provides a fine frequency/time standard that will freewheel through any short term GPS reception issues.
Never heard of these before, but anything with an ‘ium’ at the end of its name (apart from pentium) tends to pique my interest, so here I am.
Anything is serviceable just not all companies respect you enough to let you buy parts.
If I understand your explanation right it may be possible to replace the IR lamp with something else that can generate the correct light.
Jumping to conclusions, are we?
Have anyone actually worked assembling something this complicated?
Pretty sure not - one thing you have to make it serviceable, another thing tuning and assembly can be quite involved, requiring special tools and jigs.
Place I've worked did rotary encoders - how hard can it be to replace the source in those? Turns out pretty involved - you have to position the LED correctly, trim its light with adjustment screws, take the thing to the comparator to confirm.
This one would be no different.
Efratom were in fact really good like that - you could order anything that had a part number. The reason there was no lamp replacement procedure in the manual was that the lamp was projected to last for the operating life of the unit because it had a PTFE inner coating that eliminated the diffusion issues that limited lamp life in the earlier units. That diffusion issue problem (Rb migrating into the bulb) was the reason that the earlier models (like the FRK series) had lamps that were easily replaceable.
And yes, you can replace the light source with anything that creates light of the correct wavelength - in fact, if you use a monochromatic light source like a VCSEL then you can even get rid of the isotopic filter, since that's only needed to eliminate the unwanted emission peak of the discharge lamp. You still typically need temperature stabilization of the optical source because although VCSELs exhibit low sensitivity to wavelength vs temperature it's still non-zero.
@@VEC7ORlt I never said it was easy, optical systems are pockey as all heck but the key word here is POSSIBLE.
Also don't assume anything else is jumping to conclusions.
@@deltacx1059 Yes... possible. Pragmatic, practical, doable... probably not. But, someone will make a video about doing it with 'scotch tape and bubblegum.'
Cool vid, hope to see you at a TNG soon!!!
Due to pressure broadening of the absorption spike, I wonder how much shift is measurable for various elevations from sea level reference?
The rubidium cell is sealed so ambient pressure outside it shouldn't affect it. The pressure of any residual gas including the Rb vapor within will vary with temperature according to well known gas laws. This is one reason why a stable temperature environment is desirable. In general frequency standards can serve as thermometers, magnetometers, manometers, seismometers, etc. unless precautions are taken.
what a fascinating video. i have now learned some things. thank you sir. i will hit that subscribe button. hmm my thoughts are that the drift is too much. the rate of rotation of that circle is pretty fast. my thoughts are that a magic cesium clock and gps satellite would not be off by that much. a little bit yeah but not that much
Whoa, that thing surely gets hot.
me when my dealer named 'god' gives 1 mega ohm impendance: yess
me when 'god' gives me apple 2gs with dead capcitor again: nooooooo
Getting some cheap chinese 50 ohm in-line terminators for your scope might be a cheap investment. Will they be flat, capacitance-less response up to 1Ghz? lol no. Will they do the job for 99.9% of cases up to 100Mhz? Yeah.
That slow of a lissajous plot for a 20 year old, completely free-running, clock is amazing. Atoms, man.
CSAC when? 😉
I think it's 32 years old :P
wait til the modular music nerds get ahold of this
Cool video
$50.00 each for 3? Every time I look on eBay for these they are selling for $175.00 for 1. If you are lucky.
$50 bucks... these things going for $150 at least on eBay...