This brings back memories from my childhood. My dad, Dr. Leonard S. Cutler (1928-2006) led the team that developed this clock. He is the man with his back to the camera in the photo at 1:38. Back in the 1960s and early '70s he sometimes took me to Hewlett Packard at 1501 Page Mill Road in Palo Alto, California on the weekends. As a young boy I got to see the research and development of the Hewlett Packard cesium clocks. Later on, development of the cesium clocks was moved to Hewlett Packard's Santa Clara plant. Eventually Agilent sold their atomic clock business to Symmetricom. After my parents passed away, I inherited their house, and in one of my dad's old storage rooms is a four-drawer file cabinet that still has most of his notes and documentation for the 5061 and the later 5071 cesium standards.
@@CuriousMarc I am not actually in Palo Alto, but close. I live in Los Altos Hills fairly close to Stanford. You can send me an email to the address in the "about" section in my channel.
You should do a series interviewing some old engineers from HP, Tektronix, etc. to know the stories behind these designs. Probable a few may still be around. Great work and keep bringing these engineering marvels back to life and for us to enjoy them.
Unfortunately, they probably have non-disclosure agreements that prevent them from talking about anything they did for the company. I think some engineers do interviews for the Computer History Museum (some are posted to YT on the CHM channel), but it is difficult to find them for the same NDA/IP reason. The CHM posts interviews of people labeled as "the verbal history of (insert individual's name)." If you know the person's name, you might find something, but you can't search based on the products, company, position, etc., directly. My favorite interview on there is the one from the Motorola 68k design team (a must watch, but beware it's 3.5hrs long). It's one of the few interviews that actually has the company/product in the title for whatever reason. It's kinda sad really. Someone should silence the lousy lawyers and marketing morons so that the younger generations, like myself, can gain a greater appreciation of the stories of our past. The silencing of company employed inventors and engineers is basically stealing a future generation's foundations in favor of the irrelevant, half baked marketing of the present. NDA's, copyright, and trademark should be like patents, with an expiration date relative to the individual's employment history. A person that spends decades at a company, is a shareholder in the history of the company with as much right to tell their versions of stories as much as any psyco marketing spin doctor. No technology, product, or process from 20+ years ago is financially relevant to the present. If it is, forcing innovation and progress is not a bad thing. Regardless of the criminal state of Right to Repair in the US political clown show, any product currently made can be reverse engineered abroad. Intellectual property is not very valuable in the sphere of the capable. It's inflated valuation in the English speaking world is a devaluation of the inspiration it should instill in future generations. People make companies, companies do not make people. Soapbox...sorry...
@@UpcycleElectronics The verbal histories collected by the Computer History Museum are absolutely priceless. Specifically for the history of semiconductor industry, "History of Semiconductor Engineering" by Bo Lojek also provides some interesting perspective, often very different from conventional corporate history accounts.
@@cogoid Thanks. Funny, I didn't realize the reference is a book. I was trying all kinds of crazy things looking for a CHM upload with Bo Lojek. I gave up and wound up spending the evening watching "Pioneers of Pioneer Computers" Pt 1&2 before searching the web and seeing the book. The 'PoPC' is another good one ...It's funny how many dudes got all bent over von Neumann's name attributions in early digital computing :-)
@@UpcycleElectronics Another really great book is "Crystal Fire: The Invention of the Transistor and the Birth of the Information Age" -- it gives a very detailed behind the scenes picture of what went on in the Bell Labs, and how a series of both mistakes and deliberate efforts have lead to the discovery of transistor. Not quite as technical as Bo Lojec's book, but really well written.
One of the more interesting insights I've ever picked up pertained to LIGO: Being so sensitive, it is an everything-detector and the hard part is filtering out everything except the real signals. Cesium clocks are similar: building the basic clock is "easy," but it is also a detector for the local E and B fields, and their gradients, and the local temperature, etc, etc... The bulk of primary reference clocks largely consists in stabilizing/knowing all these things so that the systematic uncertainty can be reduced to the level of 1e-15.
Wow, I worked at HP Santa Clara Division for 25 years, R&D, Building these Units all the way up-to the smart hp 5071a cesium atomic beam, It was Ethereal! I enjoy every moment. Felipe Espinoza
Wow, that brings back a lot of memories! I was a development engineer in the Precision Frequency Standards (PFS) group at HP Santa Clara (CA) division then. That's where they were developed and produced. It was really interesting. I previously had worked on Cesium clocks at MIT when I was a physics student, then interned at a competitor company in New York and there learned more about both Cesium clocks and crystal oscillators. Worked on the successor product 5062 (? my memory ?) also.
That's actually really cool. I was the Service engineer for the 5334A in 1981 (I think that date is correct) and designed the troubleshooting and self test strategy for the unit! I was responsible for the operations manual and the service manual. At one time I had responsibility as the back up service engineer for the 5061.
@@brianbeasley7270 Yeah, some of those quality and serviceability aspects are still present in the early 2000s HP computer equipment I maintain. Sadly later company practices at HP and HPE have apparently ended this dedication to quality and honesty, making me look elsewhere most of the time.
Nice to see she still works. A year or so back I bought one of its great offspring an Ebay Rb oscillator. Also have a couple of GPS disciplined oven crystal oscillators running at 10Mhz. It cool to put them up on two channels of the scope and watch the phase difference. When you first turn on the Rb oscillator and it's warming up the phase sweeping back and forth looking for the lock. Then it gets closer and closer and finally it just snaps into lock. Just love the idea of owning something containing a "Physics Package". What I found interesting is what I call phase breathing. Over a period of 10's of minutes the phase error may be rock stable then it will breathe and slowly drift then be stable for a while again then drift back. Almost never slips more than one clock cycle at 10Mhz. What I would like to know is how much of this slow drift is from the GPS vs the Rb oscillator. With just two instruments I can't tell who is drifting. Because the drift is relatively short term its most likely the GPS. Got enough coax to run me a 5MHz reference to my Sunnyvale lab? Craig
The comments on these video's are fantastic. Love atomic time. Grew up with friends whose dad's worked at the Dept. of Commerce in Boulder, Colorado. The best was my neighbor, who was an engineer at HP Loveland. This guy unloaded a basement full of electronics on me when I was about 12 years old, because he was moving to a new home. Lifelong radio hobbyist and electronics tech because of it.
The 80 version of the HP Cesium clock was used in military satellite communications terminals to provide both timing and frequency references. I used to daily do comparison time offset measurements between the clock in my terminal and the clock on the distant end terminal. We'd send those values off to the Naval Observatory weekly and periodically receive correction factors to load into our clock. The corrections were intended to eliminate the time offset between clocks. It was a blessing when Techtronics came out with scopes that would allow us to accurately measure the offsets vice having to use multiple freq counters to extrapolate the offset values. Those were great clocks not a lot of trouble with them.
The German equivalent of WWVB (DCF77) apparently uses a bank of 3 of these HP units to keep everything right on the mark, checking against GPS for synchronization between the two alternative transmitter sites.
@@johndododoe1411 These museum clocks may be good for hobbyists but not as a national standard. DCF77 is controlled by the PTB (NIST equivalent) at a precision
My physics lessons at school were one of the more interesting parts of my education - however - this content is on another level - fascinating and I would have thought perfect to trigger new interests from anyone who watches. Should be part of curriculum's everywhere.
I dunno how deep into the technical aspects of physics you like to get, but assuming you preferred the general theories to the hard maths, you should check out physics for future presidents. It's a lecture series from UC Berkeley you can find on RUclips. I loved it because it's not too heavy on the math, so it's easy for anyone with even a rudimentary knowledge of physics to follow. If you prefer something more akin to a standard physics class, I'd recomment MIT's physics 801, 802 and 803 lectures. The ones presented by Walter Lewin are especially good. Though I found a lot of them require at least some knowledge of calculus and algebra to follow. But you can still appreciate them without knowing the math. Enjoy!
I absolutely loved this explanation. I knew spectral lines were related to jumps in the energy level of an electron, but I did not know they came in pairs because of electron spin. I also did not know the nucleus has a spin as well.
@@Fake_Blood I worked on an experiment at LAMPF using the hyperfine and nuclear spin coupling transition in a NMR detector. This brought back a lot of memories.
The nod to _Cody’s Lab_ gets my hearty approval. It’s nice seeing excellent science educators acknowledging each other. Now I need to go catch up on Cody’s recent videos.
@@CuriousMarc yep, tubes only last ~20 years. Mine is from the Datum side of Symmetricom, and I'm still not sure if it's got a dead tube or electronics (my test gear generally stops below 1GHz, haven't yet acquired the bits I need to figure it out). For most of 2018 it lived at my partner team's office in Sunnyvale where I used it as a monitor stand before I got it shipped home.
Coast Guard LORAN-C stations used CAQI-5061A (CAQI=HP, some sort of manufacturer designation) for so many years... Can't tell you how many daily log readings I took on them over the years- Each station had an "Operate" or #1, a "Standby" or #2 feeding an associated timer and other gear that could be "dirty" switched in the event of a failure or "clean" switched by aligning the phase of the output 5mhz, and a "tertiary" that could be patched into either #1 or #2 position in event of a failure. Each freq standard had a log book with daily readings that travelled with it- on the rare occasion of a failure it would be returned and repaired, and the log book would stay with the standard. Each secondary station's standards were compared to the master station's standards and drifts were handled with phase microsteppers, which if I remember correctly adjusted to six decimal places below a nanosecond, and longer term were compensated with "C-Field" adjustments. It was some time after 2003 - 2005?? when the stations timing and control equipment was "modernized" that the 5061s were replaced with 5071s ? I retired from my last station in 2008 and the program ended in 2010. I feel a little bad I don't remember more, but it is amazing the things that stick in one's brain. I believe the story that had been around forever was that free-running, in a perfectly stable environment, the 5061 would drift about one second in 30,000 years. :)
If I’m not mistaken, the advertised stability of the HP5061A is 1E-11, 1 microsecond a day, which translates to 1s every 3,200 years. Baby sitting your 5061 carefully (Zeeman line tweak anyone) or upgrading to the 5071 might have garnered another order of magnitude though that would put it in your range. Current lab clocks are way past 1E-15. Mind-boggling.
This was very cool! The first thing I noticed when the cover came off was the "Agilent" logo, and I was confused due to the age of the instrument. When the Lissajous figure was stationary, and the scope traces were precisely in phase - wow, that was something. The scientific explanation sounded pretty solid to me, but it has been 37 years since I had physics and 39 since I had chemistry.
This was fantastic. I never knew how these things worked and I believe you did a fantastic job explaining it. Suggest a Kickstarter when this clock runs out of cesium. I'd kick in.
The Navy's AN/BSQ-4 Precision Frequency Standard contained two of these clocks. They were so reliable we almost never had to do anything with them but monitor them.
FYI the power connector shown at 3:15 is a Mil C 5015 connector with what I think is the 14S-1 connector arrangement. Due to their robustness and utility they are still being made and are available from the likes of Digi Key.
Amazing to consider that this box of tricks was built a mere 50 years or so after Rutherford formulated his model of the atom (around 1911). Thanks for the video!
Had one in a military installation. It was set using the audio shown on an oscilloscope to sync it's 1 second pulse with the WWV tick. We had a delay we used for both the delay of the receiver used, and propagation delay from Fort Collins Colorado, so the clock would tick just before the signal on the radio was received. This was in the late 1970's before GPS time distribution became the new standard. Looks like we got one after they were out a while and proven in the field and the price reduced. On one shift, one of the new guys made the mistake of changing it for daylight savings time.. It was supposed to remain on UCT or GMT. When it was first installed, and set, it was installed on the dayshift, and I had the swing shift. I had to call my supervisor as it went into alarm for low battery. It was jumpered for 240v, but was plugged in on 120v, so the battery died on shift. It was New, Expensive, and thus I was not permitted to touch it. On watch just checked for the drift between the clock and WWV. Not touching it, I did check the setting for the line voltage. It did die on shift for low battery because I was not permitted to touch it. I got to work with one when it was cutting edge and just out on the market. Nice piece of history.
Leo Bodnar makes some very good GPS clocks. I treat the other ones as suspect. Now, going back in my personal time machine or at least memory I visit 1966. I was working at a University of Michigan lab. We were measuring partial pressures of various atmospheric constituents at high altitude. We were using an Omegatron, a U of M invention I believe. The measurement was a similar tune for peak. It was at low enough frequencies to use little IRIG electronically variable frequency oscillators. They tuned it to peak in the lab. Um, later on they tuned it off peak a little. More on this later. The use model involved tossing a payload into a Nike Ajax nose and seeing what happens. They quickly learned that their magnets did not take well to the shock of the launch. (It's later.) They attempted to compensate for this by tuning off frequency a little hoping they'd arrive at the peak after launch. They decided that would not work. So they adopted the same sort of strategy as used in the 5061a. They used got fair accuracy, not what they intended. That is where I entered the picture. A little analysis later I figured out their demodulator developed biases that were not nice constants due to the sinewave sweep over the peak. The even harmonics were doing this. Betcha many reading this know what I did next. I changed the 30 Hz (if memory serves) sinewave sweep to a squarewave step to either side of peak. The second harmonic errors vanished, of course. The calculated error of the level of the peak fell to so close to zero I never published the real figure. I published figures for wildly imprecise components. Set it up in the lab and the error in free fall was reduced to parts per million from parts per 100s. I saw applications for this. I may be clever with electronics; but, I could not and still cannot sell everlasting solar powered refrigerators to people living in the Sahara desert. The person I worked for dismissed it with a ho-hum. In found somebody else had finally suggested this in the 1970s while reading a Proceedings of the Frequency Control Symposium. I felt simultaneously vindicated and cheated. I also swore off bothering about patents after that. The patent I do have, under a married name, was an accident. I was blind sided. Later on I was responsible for the frequency synthesizer design for the Phase IIb GPS satellites, the ones with two Rb and one Cs frequency standards. But that's another story. {^_^}
VERY nice! Takes me right back to my career working in electronic calibration laboratories in the 1980's and 1990's. Just discovered this channel and ... Subscribed!
When I was an undergraduate physics student in Chicago in the mid1970's, my (very nice) parents bought me an HP35 calculator. One very hot day, the case melted inside my car. I took it to the nearest HP repair facility (in Niles, Illinois); they swapped shells for me at no charge (but broke one of the battery compartment clips and destroyed the back decal, unfortunately). Afterwards, the repair tech took me into the back to show me their brand-new HP cesium clock. Amazing! A year later, I did a summer internship at Argonne National Laboratory, where I put together a data acquisition system from shipping containers full of fresh HP2016 parts. They had allocated the full summer for me to do so; I had it up and running (BASIC interpreter and TREK73 game) by the end of the week :-)
@@acmefixer1 Its already ionised at that point so the tube in that clock wouldnt work. Unless you mean extracting the pure metal from the salt? You'd need to purify it first. There's better sources for sodium chloride than rock salt.
Rb and Cs clocks have somewhat different operating principles. There is a serendipitous coincidence between energy levels of different isotopes of Rubidium which makes it possible to make a very simple atomic clock. If simplicity is not required, many different elements can and have been be used for making very high performance frequency standards -- but these are research instruments. In commercial use, Cs and Rb dominate, with Hydrogen masers also used for some applications.
Instead of Lissajous, better use Time Interval measurement on your 5334B, between 1pps of Cs clock and GPS clock (or OCXO), and use TimeLab over GPIB control to collect the phase shift. Data collection and averaging over several hours / a whole day only will mitigate any GPS signal jitter, and only then will give you the necessary resolution to calculate the accuracy of the 5061A. GPS receivers usually need several hours, or 1-2 days to really get stable, i.e. to learn the behavior of their internal OCXO (if available) vs the jittered GPS time information. So your momentary snapshots of phase (out of the Lissajous figure) is probably useless. Cheap GPS units probably never manage to achieve a short term stable time signal, (use ADEV analysis) compared to e.g. the Trimble Thunderbolt.
@@edgeeffect Watching (again) Lissajous figures might be fun at first sight, but they become very soon boring, and not practical at all, when you try to adjust atomic clocks. Simply calculate, how fast the figure will rotate once (180° or 50nsec phase shift) for a feasible 1E-9 frequency accuracy adjustment of the OCXO inside the 5334B, that'll be 50sec.. That's still practical, but when you try to adjust an Rb clock, which might be stable / precise to 1E-11, this will take 5000 sec already. If you try to resolve 4° only on the Lissajous figure with your naked eye, equivalent to 1ns, that'll be 2 minutes already.. very boring . When you would try to adjust a hp 5061A Cs standard to 1E-12 (standard tube) or 1E-13 (option 004 tube, as in the video), that minute change of this figure would require up to several hours, and everybody doing this experiment would fall asleep soon, I fear. There's also another problem with such cheap GPS receivers, probably w/o a disciplined OCXO inside, that is the GPS signal jitter on the order of 1E-10 / sec, 3 order of magnitude above the required stability .. you would never come to a conclusion by the Lissajous method.
I have such a hell of time following exactly what is being talked about in these videos. Im a musician, not a scientist. However, I find this stuff super fascinating and despite my issues fully understanding everything, I watch all the way through. Great work :)
We had one of these cesium clocks on our satellite communication's facility back in the eighties. Filled with dozens of cables connecting all the distribution amps together, all of which didn't like to be touched.
Marc, you always pause just briefly before you say "H.P." it's almost like you're preparing to speak a powerful magic word or utter the name of God or something. ;) Thanks for the Quantum Mechanics explaination, you managed to explain it better than a lot of the physicists I have listened to. That oscilloscope was... ... ... too modern! ;)
That particular gpsdo gives a green light before it is locked. It also takes forever to get close to freq. The bigger one with the green pcb front cover only gives the light when it is locked, but the lock is a few millihz off. You cant use any of these units for timing, but they’re ok for freq standards for ham radio which is what they’re designed for. Get a Trimble Thunderbolt.
The man's humor is first class. At 3:19 he remarks on the unique power cord: "It's really annoying if you don't have it" Happens to everyone when they measure orbital time difference with a Cesium Clock.
Great explanation. It is amazing this device was commercially available only a few years after the LASER was invented! Nowadays atomic clocks take advantage of lasers for optical probing of hyperfine transition in Cs vapors. Atomic beams are no longer used. The clock can therefore stay operational for many years without need for "refueling".
Always a pleasure to seen some vintage HP goodness. These guys were the top of the best in groundbreaking lab tech. The fact HP recessed to a second zone home PC and disposable inkjet printers manufacturer puzzles me. Rise and fall of empires... Anyway Agilent was spun off HP in 1999 so that caesium tube is pretty recent.
And then Keysight spun off Agilent. I believe that the Time division was either spun off or purchased to becomes Symmetricon, which was then bought by Microsemi, which was then bought by Microchip. Also, I believe that the crystal oscillator was used to clean up the Cesium clock. Cesium clocks have great accuracy, but they don't have very good phase noise. Crystal oscillator aren't nearly as accurate, but they have good phase noise. By locking them together you get the best of both worlds. At least that is what I remember.
23:07 I can't tell if that pattern is tumbling end over end (rotating on horizontal axis) or if it's spinning on the vertical axis...like one of those mind illusions!
I remember having one of those at Bell Northern Research in the early 90s. IIRC it was used as a reference clock for validating the VCXO frequency used by telephone switches for trunking.
I love the fact that there's an atomic clock and a bench grinder on the table at the same time! Probably not that ususal of a setup 😁 This made me understand how a cesium clock actually works. Amazing explanation!
@@Drew-Dastardly That's very true and usually they are right, I can't afford it! But sometimes my nosy self likes knowing how much things cost so I can dream!
10:10 Totally clear, in a "I could not repeat it, and I could not point at the parts, but it makes sense, and meshes with the tiny amount I understand about Relativity, and the even tinier part I understand about Quantums" way.
Very cool! I've read the manuals for the HP cesium clocks and thought those were fascinating pieces of equipment. It's too bad the cesium tube has a limited life span, but glad to see you found a working one.
FBI Top 10-Things you don't want to mention in a youtube video: "Oh, I like those things! The atomic bomb is on the left, the detonator is on the right and the crypto things over there..." Another great video, thanks! When all this is over, I'd love to come over to the Computer History Museum and take a close look at the IBM 1401 and the other cool stuff there. Maybe next year... Stay well!
I've always wondered how atomic clocks work, and your explanation of how they utilize the Stern-Gerlach apparatus finally put it all together for me, MANY thanks !!!
What's the significance of the 7:9 ratio in the definition of the transition states and where does it come from? The ~4.021 and ~5.170 @ 9:28. I imagine that clean ratio comes from quantized particles... but I'd love to know more!
I used to play around with precision clocks. Every time you plot them on a scope together, and see them hardly drift, it's a magical moment. That you can do these things in your own home is incredible. (I was using a rubidium clock, but those have about the same accuracy as a cesium clock from that era.)
So, to sum it up into an elevator pitch, the time signal comes from the oscillator when its output resonates with the lowest cesium spectral lines. Have I got that right?
Thanks for the physics lesson portion of the video. At 3:40 I was surprised to see that the unit on the left spelled the chemical element as Caesium. I've never seen it written with the A before.
I wonder, as a cool project, could a Cesium clock be used (with some processing) to generate a signal that can replace the oscillator in a cheap digital clock to see if it can be made more accurate (not be 1 minute off after a week of use).
I had a couple of friends who worked for HP in the Rohnert Park area, it seems like it was the microwave test equipment division mainly, and I got the tour on one Saturday night, walking through production and testing, into the anechoic chamber, etc. Later they showed up to work one Monday, to the "Company With No Name", because HP had spun them off, but without a name! They were no longer HP, just "...." Not even "The Artist Formerly Know As HP" LOL After some discussion, they decided to let the employees choose a name, and so suggestions were made, and then they started voting, and finally ended up with Agilent. So yes, that was certainly last worked on since then, so that '89 date isn't the date of the tube.
The Rohnert Park (actually Santa Rosa) location of HP was (still is?) the communications test instrument division. It's still an incredible place to visit. They have the same old HP instruments I collect displayed on glass shelves. The engineer's cubes are full of the same instruments, still being used today.
I graduated with a BS in physics and focused on astrophysicist, quantum mechanics and particle physics so I’m not shy in saying your explanation was poetic. Keep that up and you’ll give Neal deGras Tyson some competition.
In US Air Force Precision Measuring Equipment Laboratory we had one. It had to have a yearly touch up on the frequency due to the slight drift, that difference between -11 and -12.
Actually I was wondering myself. The tube is only warranted for 3 years. After you spend $40,000 on it, I find it a bit short. I was guessing it lasts 10-15 years? But I really don’t know, someone more knowledgeable would have to chime in. Also you need to keep running it, so the ion pump is on. If you let it sit for too long you might not be able to restart the tube. On the non high-performance tube, you can run them at lower current to extend their life, at the expense of worse short term stability (i.e. more phase noise). This one is the high performance tube, no such choice.
@@CuriousMarc The high-performance tubes like this one seem to run for 6-8 years at most. It's possible to get 10-15 years out of the standard tubes. The ion pump runs as long as the unit is plugged in, while the cesium oven runs only when the function switch is switched out of standby mode into LOOP OPEN or OPER. So it's best to leave the clock powered up 24/7, but turn on the oven only when you actually want to calibrate something.
By way of coincidence, just today I was reading Paul Forman's excellent IEEE paper on the development of the Atomichron NC-1001 by Natco in the late 1950s - how's that for timing?
I learn so much from you and your friends Marc and with such ease from your superb descriptions. This is the next best thing to being in your lab with you. All the best.
I don't know about the 'usual' PCB-faced GPSDOs that are on ebay like that one, but I have a very slightly fancier one that exhibits instability for a bit when the output 'load' changes significantly, to the point where I could connect/disconnect a device on the timing circuit and watch it hunt on the scope.
Great explanation on the ceasium clock, but I didn't fully understand what you did there at the end with the oscilloscope and several inputs and what led you to not trust the GPS.
He was using the oscilloscope to compare the frequency of the Cesium clock and his GPS disciplined oscillator. They should be very close in frequency, so the pattern on the screen should be very still. The faster it moves, the farther they are off. The Ebay GPSDO he is using is known not to be exactly on frequency with GPS. The way you do the test is you put one frequency on channel 1, the other on channel 2, and put the scope in XY mode. If they match in frequency and phase, you get a circle. This is called a Lissajous pattern. Since he was comparing a 5 MHz Cesium frequency with a 10 MHz GPSDO, the circle had a twist in it making a figure 8 as the frequencies were pretty close to 2:1.
The eBay GPSDOs are usually based on a re-used GPSDO from a well known manufacturer (Trimble, Samsung, Symmetricom). Have you opened your unit to see what's inside? Also it might be worth plugging in to the serial port to see what the TDOP is at when attempting a Lissajous curve with the caesium clock.
The generic ebay GPSDOs aren’t built on reused GPSDOs, they are simply reused TCXOs with a microcontroller and a GPS receiver. They often don’t have good enough software or DACs to keep accurate time.
Awesome. Back in the '60s, even the PLL synthesizer must have been pretty bleeding edge technology. Harmonic generation in PLLs actually has some more down-to-earth applications as well, e.g. there used to be a few high-end FM tuners that would lock their mechanical-varicap-tuned LOs to a 50 or 100 kHz frequency grid this way. This way they would combine the better frontend selectivity of mechanical variable capacitors with the frequency stability of PLL tuned sets while still keeping phase noise at bay (almost all early PLL tuners were afflicted with more or less high levels of phase noise, degrading their maximum S/N).
I also thought that microwave and PLLs and frequency synthesizers were science-fiction in the 1960's. That's until I looked at the telemetry radars and transponders in Apollo. Then my jaw dropped. It's absolutely chock full of these things, some running at 9 GHz. BTW, this HP clock was installed just in time in 1968 to serve as the deep-space network master clock for the Apollo 8 mission.
I really enjoyed this segmanet. The Hafele-Keating experiment was a test of the theory of relativity. In 1971, My law partner's brother, Joseph C. Hafele, a physicist, and Richard E. Keating, an astronomer, took four cesium-beam atomic clocks aboard commercial airliners. They flew twice around the world, first eastward, then westward, and compared the clocks against others that remained at the United States Naval Observatory. When reunited, the three sets of clocks were found to disagree with one another, and their differences were consistent with the predictions of special and general relativity.
There's one of those on display at the Museum of Time in Besançon, France. It's open and you can see the insides. Incredible to think we now fit that technology into a relatively simple MEMS integrated circuit assembly.
@@johndododoe1411 interesting, I haven't heard any of that nor witnessed it on a CSAC. Can you elaborate ? This may have work-related implications for me.
@@TheNefastor I am referring to the original scientific publication of the creating of a cesium clock that was the size of a typical electronic part. The article noted that the device was so temperature sensitive it might be used as a thermometer, which convinced me that it had a less than required inherent accuracy, as temperature should not interfere with the relevant quantum mechanical resonance frequency, and all surrounding circuitry should correctly and exactly reflect the resonance in those atoms. In later years Symmetricom tried using these instead of oxco oscillators in some ready-made time synchronization servers, but it seemed of little relevance compared to the multiple non-government labs still installing HP Cesium clocks like the one in this video.
I was going to mention the tube replacement! Score! Agilent is a pretty recent name...It came about in 2000. The test equipment division was spun off again as Keysight in 2014.
Supposed to be "working" from self-quarantine rn but actually just watching Curiousmarc youtubes 🙏🙏 * 19:15 -cesium clocks- rubidium clocks in orbit (or sometimes hydrogen masers). If I remember right these have smaller Allan deviations for shorter timescales than Cs which is good because the satellites get new correct time standards beamed up every day anyway.
They have both (or used to at least). From what I understand the advantage is that the Cs ones don’t drift and are the master references. The Rb are much cheaper (no high vac involved).
I managed to get my hands on a Rubidium frequency standard used for Omega Navigation back in the day. Similar principles but a bit simpler. It used a rubidium lamp to shine photons through a cavity containing rubidium ions. The cavity had a waveguide subject those ions to GHz frequency but it was dithered at 150 Hz rate by a triagonal wave making it sweep like FM. As that RF sweep went through that hyperfine transistion frequency the ions would then absorb the photons from the lamp and a dip in light intensity was detected after the cavity. This dip was tracked to discipline the 10 MHz crystal oscillator. Apparently it might loose 1 second after 3000 years - Good enough for me
Interesting. It is a model from 1968 HP catalog, but there are stickers inside that say Agilent - the new name of HP's instrument division starting around 2000 or so. They either kept producing them or servicing them for a very long time. The design of the box looks classic, but truthfully the inside of the box looked very clean and free of dust. I"m going to guess this was a later-produced unit.
So do I have this right? Time is the difference between the color of light produced by the spin 1/2 electron and a spin -1/2 electron in the 6s shell of a cesium atom?
More or less. It does not tell you what “time” is, but it gives you the definition of a second. “A second is the duration of 9 192 631 770 periods of the radiation corresponding to the transition between the two hyperfine levels of the ground state of the caesium 133 atom”. You are correct that this transition exists because of the outermost electron, like any electron, has two spin states. It is one of very many caused by that electron and its spin. It can be measured by irradiating the cesium with a microwave frequency close to 9 192 631 770 Hz and tune it until it flips the electron spin. This is what this HP machine does. But you could do an infrared line color difference (which by the way involves 3 states, rather than the direct microwave measurement between just 2 states). For that you’d need a super duper infrared spectrometer and a way to excite the Cesium so it emits light, like in a flame, then you would zoom in on the infrared D4 line, until you see it split in a doublet (caused by the spin). That’s the fine transition. Then you would zoom in waaaay more, and you would eventually see that each line of the doublet again splits in two. Also due to spin. That’s the hyperfine transition. The inverse of the wavelength difference (which is the “color difference”) between these two very close infrared lines should be 9 192 631 770 Hz. Modern clocks use tunable lasers to do essentially that measurement, but in a much more controlled way.
@@CuriousMarc Thank you for the clarification. My posit was a bit of a flippant reduction of a complex and very technical measurement. I'm not a physicist but I'm physics curious so I read and listen to physicists. I think knowing how thing like "spin" and "energy level" are measured has helped my understanding. The Stern-Gerlach measurement is not exactly what physicist mean when they say "spin" but it's the concrete part. Physicist talk a lot about time, they even write whole books about it. A common claim is that our intuitions of time are not always useful. But they then often fail to say exactly what "time" is. But now I know the exact definition of what a unit of "time" is, a 9.192GHz difference. Now I can scratch my head and ask how can they measure GHz without already having measured time?
@@charlesmanning3454 You can’t define time easily because it’s relative. Only the speed of light is constant. All you can hope to do is find out your local second at your local speed and local gravity. Which is what this HP machine does. You don’t measure the 9 192 631 770 Hz either. The Cesium atom measures it for you. Once the clock locks its probe oscillator on the atomic resonance, you know it is guaranteed to be 9 192 631 770 Hz. You then divide it electronically by 9 192 631 770 and you now you have an exact second. Actually this machine divides it by a smaller number and gives you a perfect 5 MHz or 10 MHz reference output, which you use around the lab. Most HP frequency meters will accept that as their external frequency reference input, become instantly Cesium-referenced, and you do your other measurements with that.
I actually fell asleep at ~7:30 minutes, no joke. I'm currently rewinding to watch again. I actually really appreciate it when people tell the history of science; how things were discovered.
Oh that’s true, because of general relativity we talk about in the beginning of the video. I had not thought about it. But you’d think they would compensate for it. Then, maybe not. It would not be a Hertz like this though, but a tiny barely measurable difference.
The magnitude of the effect is (g*h/c^2), where g=9.81 m/s^2, h= altitude in meters, c = speed of light, 3*10^8 m/s This corresponds to about 10^-16 fractional frequency change per meter of altitude difference. Today this is measurable even on millimeter scale with the most advanced of research clocks, but HP 5061A would require 100 km of altitude to show a clear difference.
and to anyone wondering, yes i am aware of the UHV vacuum requirements for these tubes. And yes i do have the equipment to evacuate them back to UHV from atmospheric pressure.
@@lbochtler in addition to the UHV equipment, you would need something like a Argon-filled glove box to handle the Caesium. And all the equipment to calibrate the tube afterwards (Cs current, oven temperature, ...). In addition, you would need to change and realign the apertures etc. Having worked quite a bit with other UHV devices that had Cs sources, I'd leave that to the manufacturer... :)
@@dsamuelis now, where is the fun in that? All of that sounds like quite manageable to me. (for reference, i have worked on particle accelerators and electron microscopes in the past). I would imagine the apertures are either cleanable the same way that of an electron microscope are, by heating in a high vacuum. As for calibration, yeah that might present a challenge. Though all of this talk of refilling such tubes, is making me want to just build a cesium beam clock from scratch. Oven temperature calibration is relatively trivial in my opinion, as i have and use precision radiometer and contact thermometers. All that would be needed is to calibrate the radiometer to the oven sensitivity. The only challenging thing i can see is the bake out process needed for UHV. As this will evaporate the cesium if im not careful. Also, in case of the HP tube, i would imagine the electron multiplayer is also a part which at the very least would need replating of the dynodes. Hm, if money and time allows it, i might just get myself a hp clock and see if i can refill the tube using my lab equipment. The only thing id need to add is an argon glove box. Though i imagine working under a heavier then air gas with a relatively tight enclosure should suffice as well.
@@lbochtler sounds like a great challenge then :) Most of the "regular" heavier than air gases however would react with Cs (CO2, SF6). You'd be limited to the heavier noble gases, which are extremely expensive and rare. On the other hand, you could buy a reasonable glove box for something that is quite a bit cheaper than a new Cs assembly (40k USD according to the video) - you might be able to get a suitable used box for around 20k. I've been working with alkaline metals in those boxes for quite some time, so please contact me if you'd like to know more about specs and usage of such boxes. Yes, the multiplier would need to be exchanged or serviced as well, as most of the evaporated Cs is deposited there. My experience with Cs sources is mainly from UHV surface/sputter analytics, such as SIMS or ion microscopy. Unfortunately in these devices, the manufacturers try to keep the Cs sources as un-serviceable as possible, so exchanging the source is most often the only way to go. I also have been working at particle accelerators during my academic career (mainly using the parasitic X-rays that are emitted there), but I've never worked "on" an accelerator :)
Wouldn't the high speed of the orbiting GPS satellite mean that its atomic clock is running at a slightly different frequency relative to the atomic clock on Earth? So wouldn't we *expect* there not to be a perfect phase lock between the two clocks? Or have I incorrectly understood?
You bet the GPS system has to make relativistic corrections! Both special relativity for the speed of the satellites and general relativity because they experience less gravity up there. But that’s all corrected for, or the system would not work at all. The cheap GPS DSO was just caught drifting. Might have lost one of the satellites. It corrected itself 30 minutes later. It’s just not good enough, which is why lab quality GPS DSOs also include a rubidium local reference for the short term stability.
Great overview of the clock’s operation. A few years ago, I ran across some open courseware videos, from UC Berkeley as I recall, on an attempt to make an atomic clock on the chip level in MEMS technology. I’m not sure if they managed to make it work. Isaac Asimov’s physics books provide an interesting overview of the basic theory behind quantum mechanics.
The design of the machine is interesting, but perhaps more interesting to me is that it's the first explanation I've seen of the derivation of the standard for defining the second.
Fascinating stuff. How much signal interference do the patch leads and the wiring of the clock give in your measurements, if any, and would shorter one's or different conductor materials and construction make a difference, as in audio equipment, where you can tell between different cables????
When you get into this type of precision, the phase lock of things takes a while... GPS disciplined clocks are a set of things... the GPS receiver which is a whole video or two worth of wonder in itself, which has to know exactly where it is to get the time right (takes a few days if you do it right, with an outdoor amplified antenna), and an oven stabilized clock (good to about 1 part in 10^8 or so by itself over a few days), a numeric controlled oscillator (another video worth), and a computer to feed slight tweeks to lock everything in, compute position, etc. Give the GPS a few days to lock position and time... and they have long term stability as it comes from the national time standards... over the short term, (less than a day) the Cesium will be better, long term the national time standards are better, as they have more, newer clocks in groups all carefully measured against each other. The GPS is the weak link as you've got things whizzing past at relativistic speeds, and random delays from the ionosphere that drift over time.
I can confirm some of the ebay chinesium GPSDO clocks do not ever lock to the correct frequency. They are always a few millihertz off. Not generally a problem for many frequency reference uses, but totally unacceptable for GPS based timekeeping purposes.
Nice to see this old of an atomic clock not seen many of them which is odd considering I've installed a few modern ones over the years. I forget how large many of these older designs can be when they nearly have single chip version of them commercially.
They are high up, so their time goes faster! Plus they move quite fast so there is a special relativity time difference between the satellites too. So they have to do both a special and general relativistic correction of the time for the system to work.
Before GPS was LOng Range Aid to Navigation. The Coast Guard manned xmitter sights that were all controlled by Cesium clocks. LORAN stations were all over the world and they each had 3 of these in the rack. Triple redundancy and all.
This brings back memories from my childhood. My dad, Dr. Leonard S. Cutler (1928-2006) led the team that developed this clock. He is the man with his back to the camera in the photo at 1:38. Back in the 1960s and early '70s he sometimes took me to Hewlett Packard at 1501 Page Mill Road in Palo Alto, California on the weekends. As a young boy I got to see the research and development of the Hewlett Packard cesium clocks. Later on, development of the cesium clocks was moved to Hewlett Packard's Santa Clara plant. Eventually Agilent sold their atomic clock business to Symmetricom. After my parents passed away, I inherited their house, and in one of my dad's old storage rooms is a four-drawer file cabinet that still has most of his notes and documentation for the 5061 and the later 5071 cesium standards.
Well, kudos to your Dad! Can you contact me through the link under the video? Are you still in Palo Alto?
@@CuriousMarc I am not actually in Palo Alto, but close. I live in Los Altos Hills fairly close to Stanford. You can send me an email to the address in the "about" section in my channel.
Cool,
Measure the time an electron's circuit around the nucleus of a cesium Atom. That equals 1 second.
😊
You should do a series interviewing some old engineers from HP, Tektronix, etc. to know the stories behind these designs. Probable a few may still be around. Great work and keep bringing these engineering marvels back to life and for us to enjoy them.
Unfortunately, they probably have non-disclosure agreements that prevent them from talking about anything they did for the company.
I think some engineers do interviews for the Computer History Museum (some are posted to YT on the CHM channel), but it is difficult to find them for the same NDA/IP reason. The CHM posts interviews of people labeled as "the verbal history of (insert individual's name)." If you know the person's name, you might find something, but you can't search based on the products, company, position, etc., directly.
My favorite interview on there is the one from the Motorola 68k design team (a must watch, but beware it's 3.5hrs long). It's one of the few interviews that actually has the company/product in the title for whatever reason.
It's kinda sad really. Someone should silence the lousy lawyers and marketing morons so that the younger generations, like myself, can gain a greater appreciation of the stories of our past. The silencing of company employed inventors and engineers is basically stealing a future generation's foundations in favor of the irrelevant, half baked marketing of the present. NDA's, copyright, and trademark should be like patents, with an expiration date relative to the individual's employment history. A person that spends decades at a company, is a shareholder in the history of the company with as much right to tell their versions of stories as much as any psyco marketing spin doctor. No technology, product, or process from 20+ years ago is financially relevant to the present. If it is, forcing innovation and progress is not a bad thing. Regardless of the criminal state of Right to Repair in the US political clown show, any product currently made can be reverse engineered abroad. Intellectual property is not very valuable in the sphere of the capable. It's inflated valuation in the English speaking world is a devaluation of the inspiration it should instill in future generations. People make companies, companies do not make people.
Soapbox...sorry...
@@UpcycleElectronics The verbal histories collected by the Computer History Museum are absolutely priceless. Specifically for the history of semiconductor industry, "History of Semiconductor Engineering" by Bo Lojek also provides some interesting perspective, often very different from conventional corporate history accounts.
@@cogoid
Thanks.
Funny, I didn't realize the reference is a book. I was trying all kinds of crazy things looking for a CHM upload with Bo Lojek. I gave up and wound up spending the evening watching "Pioneers of Pioneer Computers" Pt 1&2 before searching the web and seeing the book. The 'PoPC' is another good one ...It's funny how many dudes got all bent over von Neumann's name attributions in early digital computing :-)
@@UpcycleElectronics Another really great book is "Crystal Fire: The Invention of the Transistor and the Birth of the Information Age" -- it gives a very detailed behind the scenes picture of what went on in the Bell Labs, and how a series of both mistakes and deliberate efforts have lead to the discovery of transistor. Not quite as technical as Bo Lojec's book, but really well written.
@@UpcycleElectronics Do NDAs really last that long? That's pretty shitty.
23:02: Ultra-precision instruments in the foreground, bench grinder in the background.
OK, so...?
I don't know what kind of adjustments that bench grinder has but...
coarse.
Sometimes brute Force is the solution
You do notice Marc lined the table with paper and taped all the edges prior to setting down this incredible device 😉love this channel
You're saying a bench grinder is NOT an ultra-precision device?
Really interesting and well explained. I never knew how complicated the process is!
Thanks Ben!
@@CuriousMarc next step ben will do a DIY version of this clock
All hacking requires convolution. We are hacking time
One of the more interesting insights I've ever picked up pertained to LIGO: Being so sensitive, it is an everything-detector and the hard part is filtering out everything except the real signals. Cesium clocks are similar: building the basic clock is "easy," but it is also a detector for the local E and B fields, and their gradients, and the local temperature, etc, etc... The bulk of primary reference clocks largely consists in stabilizing/knowing all these things so that the systematic uncertainty can be reduced to the level of 1e-15.
@@CuriousMarc your video's background music isn't loud and annoying enough
Wow, I worked at HP Santa Clara Division for 25 years, R&D, Building these Units all the way up-to the smart hp 5071a cesium atomic beam, It was Ethereal! I enjoy every moment. Felipe Espinoza
Wow, that brings back a lot of memories! I was a development engineer in the Precision Frequency Standards (PFS) group at HP Santa Clara (CA) division then. That's where they were developed and produced. It was really interesting. I previously had worked on Cesium clocks at MIT when I was a physics student, then interned at a competitor company in New York and there learned more about both Cesium clocks and crystal oscillators. Worked on the successor product 5062 (? my memory ?) also.
That's actually really cool. I was the Service engineer for the 5334A in 1981 (I think that date is correct) and designed the troubleshooting and self test strategy for the unit! I was responsible for the operations manual and the service manual. At one time I had responsibility as the back up service engineer for the 5061.
@@brianbeasley7270 Yeah, some of those quality and serviceability aspects are still present in the early 2000s HP computer equipment I maintain. Sadly later company practices at HP and HPE have apparently ended this dedication to quality and honesty, making me look elsewhere most of the time.
I could listen for hours, to what you did back then.
Nice to see she still works. A year or so back I bought one of its great offspring an Ebay Rb oscillator. Also have a couple of GPS disciplined oven crystal oscillators running at 10Mhz. It cool to put them up on two channels of the scope and watch the phase difference. When you first turn on the Rb oscillator and it's warming up the phase sweeping back and forth looking for the lock. Then it gets closer and closer and finally it just snaps into lock.
Just love the idea of owning something containing a "Physics Package".
What I found interesting is what I call phase breathing. Over a period of 10's of minutes the phase error may be rock stable then it will breathe and slowly drift then be stable for a while again then drift back. Almost never slips more than one clock cycle at 10Mhz. What I would like to know is how much of this slow drift is from the GPS vs the Rb oscillator. With just two instruments I can't tell who is drifting. Because the drift is relatively short term its most likely the GPS. Got enough coax to run me a 5MHz reference to my Sunnyvale lab?
Craig
That's gold 😂
@Craig S
You don't need a coax. Just shine a laser beam across town at night. Switch it on and off at 10 MHz. 👍
@@acmefixer1 If its a big enough beam, we can run it in the daytime too =)
Back when HP made the best instruments in the world.
I always thought when they spun off the test equipment it should have kept the HP name, not Agilent.
@@dfmayes Same here. I tend to think of two eras: real HP and fake HP
*Looking at my printer with disdain…
The comments on these video's are fantastic. Love atomic time. Grew up with friends whose dad's worked at the Dept. of Commerce in Boulder, Colorado.
The best was my neighbor, who was an engineer at HP Loveland. This guy unloaded a basement full of electronics on me when I was about 12 years old, because he was moving to a new home. Lifelong radio hobbyist and electronics tech because of it.
The 80 version of the HP Cesium clock was used in military satellite communications terminals to provide both timing and frequency references. I used to daily do comparison time offset measurements between the clock in my terminal and the clock on the distant end terminal. We'd send those values off to the Naval Observatory weekly and periodically receive correction factors to load into our clock. The corrections were intended to eliminate the time offset between clocks. It was a blessing when Techtronics came out with scopes that would allow us to accurately measure the offsets vice having to use multiple freq counters to extrapolate the offset values.
Those were great clocks not a lot of trouble with them.
The German equivalent of WWVB (DCF77) apparently uses a bank of 3 of these HP units to keep everything right on the mark, checking against GPS for synchronization between the two alternative transmitter sites.
@@johndododoe1411 These museum clocks may be good for hobbyists but not as a national standard. DCF77 is controlled by the PTB (NIST equivalent) at a precision
My physics lessons at school were one of the more interesting parts of my education - however - this content is on another level - fascinating and I would have thought perfect to trigger new interests from anyone who watches.
Should be part of curriculum's everywhere.
I agree, a great "show and tell", not just the theory but the practical application of the process..👍👍👍
I dunno how deep into the technical aspects of physics you like to get, but assuming you preferred the general theories to the hard maths, you should check out physics for future presidents. It's a lecture series from UC Berkeley you can find on RUclips. I loved it because it's not too heavy on the math, so it's easy for anyone with even a rudimentary knowledge of physics to follow.
If you prefer something more akin to a standard physics class, I'd recomment MIT's physics 801, 802 and 803 lectures. The ones presented by Walter Lewin are especially good. Though I found a lot of them require at least some knowledge of calculus and algebra to follow. But you can still appreciate them without knowing the math.
Enjoy!
I absolutely loved this explanation. I knew spectral lines were related to jumps in the energy level of an electron, but I did not know they came in pairs because of electron spin. I also did not know the nucleus has a spin as well.
@@Fake_Blood I worked on an experiment at LAMPF using the hyperfine and nuclear spin coupling transition in a NMR detector. This brought back a lot of memories.
The nod to _Cody’s Lab_ gets my hearty approval. It’s nice seeing excellent science educators acknowledging each other. Now I need to go catch up on Cody’s recent videos.
He did a good video on ion pumps also iirc.
Fascinating video. That's the first explanation of a caesium clock I've ever seen that I fully understood from beginning to end.
Was surprised to see the name Agilent inside that box. I guess they serviced this sometime after 1999?
We think that’s the case, and the reason why it still works.
@@CuriousMarc yep, tubes only last ~20 years. Mine is from the Datum side of Symmetricom, and I'm still not sure if it's got a dead tube or electronics (my test gear generally stops below 1GHz, haven't yet acquired the bits I need to figure it out). For most of 2018 it lived at my partner team's office in Sunnyvale where I used it as a monitor stand before I got it shipped home.
@@laptop006 if you don't want it and just want to give it away I'll gladly take it off your hands ::)
Oh wait, that's a high perf tube, they're normally more like 10 year lifetimes IIRC, very good luck with that.
Agilent is what happened to the _real_ part of HP that meant something. It is where the true heart of HP went.
Lived with these as a Loran technician for the U. S. Coast Guard. I was the project manager in the early 2000s for the upgrade to the Agilent clock.
Coast Guard LORAN-C stations used CAQI-5061A (CAQI=HP, some sort of manufacturer designation) for so many years... Can't tell you how many daily log readings I took on them over the years- Each station had an "Operate" or #1, a "Standby" or #2 feeding an associated timer and other gear that could be "dirty" switched in the event of a failure or "clean" switched by aligning the phase of the output 5mhz, and a "tertiary" that could be patched into either #1 or #2 position in event of a failure. Each freq standard had a log book with daily readings that travelled with it- on the rare occasion of a failure it would be returned and repaired, and the log book would stay with the standard. Each secondary station's standards were compared to the master station's standards and drifts were handled with phase microsteppers, which if I remember correctly adjusted to six decimal places below a nanosecond, and longer term were compensated with "C-Field" adjustments. It was some time after 2003 - 2005?? when the stations timing and control equipment was "modernized" that the 5061s were replaced with 5071s ? I retired from my last station in 2008 and the program ended in 2010. I feel a little bad I don't remember more, but it is amazing the things that stick in one's brain. I believe the story that had been around forever was that free-running, in a perfectly stable environment, the 5061 would drift about one second in 30,000 years. :)
I should have said a clean switch was done by phase matching the 5mhz output of the two timers, not the two cesiums.
You are correct, the adjustments were made in femtoseconds :)
CAQI was BuSHIPS (Bureau of Ships) code for HP.
If I’m not mistaken, the advertised stability of the HP5061A is 1E-11, 1 microsecond a day, which translates to 1s every 3,200 years. Baby sitting your 5061 carefully (Zeeman line tweak anyone) or upgrading to the 5071 might have garnered another order of magnitude though that would put it in your range. Current lab clocks are way past 1E-15. Mind-boggling.
This was very cool! The first thing I noticed when the cover came off was the "Agilent" logo, and I was confused due to the age of the instrument. When the Lissajous figure was stationary, and the scope traces were precisely in phase - wow, that was something. The scientific explanation sounded pretty solid to me, but it has been 37 years since I had physics and 39 since I had chemistry.
This was fantastic. I never knew how these things worked and I believe you did a fantastic job explaining it. Suggest a Kickstarter when this clock runs out of cesium. I'd kick in.
The Navy's AN/BSQ-4 Precision Frequency Standard contained two of these clocks. They were so reliable we almost never had to do anything with them but monitor them.
They lose one second every million years....
FYI the power connector shown at 3:15 is a Mil C 5015 connector with what I think is the 14S-1 connector arrangement. Due to their robustness and utility they are still being made and are available from the likes of Digi Key.
Amazing to consider that this box of tricks was built a mere 50 years or so after Rutherford formulated his model of the atom (around 1911). Thanks for the video!
Had one in a military installation. It was set using the audio shown on an oscilloscope to sync it's 1 second pulse with the WWV tick. We had a delay we used for both the delay of the receiver used, and propagation delay from Fort Collins Colorado, so the clock would tick just before the signal on the radio was received. This was in the late 1970's before GPS time distribution became the new standard. Looks like we got one after they were out a while and proven in the field and the price reduced.
On one shift, one of the new guys made the mistake of changing it for daylight savings time.. It was supposed to remain on UCT or GMT.
When it was first installed, and set, it was installed on the dayshift, and I had the swing shift. I had to call my supervisor as it went into alarm for low battery. It was jumpered for 240v, but was plugged in on 120v, so the battery died on shift. It was New, Expensive, and thus I was not permitted to touch it. On watch just checked for the drift between the clock and WWV. Not touching it, I did check the setting for the line voltage. It did die on shift for low battery because I was not permitted to touch it.
I got to work with one when it was cutting edge and just out on the market. Nice piece of history.
Leo Bodnar makes some very good GPS clocks. I treat the other ones as suspect.
Now, going back in my personal time machine or at least memory I visit 1966. I was working at a University of Michigan lab. We were measuring partial pressures of various atmospheric constituents at high altitude. We were using an Omegatron, a U of M invention I believe. The measurement was a similar tune for peak. It was at low enough frequencies to use little IRIG electronically variable frequency oscillators. They tuned it to peak in the lab. Um, later on they tuned it off peak a little. More on this later.
The use model involved tossing a payload into a Nike Ajax nose and seeing what happens. They quickly learned that their magnets did not take well to the shock of the launch. (It's later.) They attempted to compensate for this by tuning off frequency a little hoping they'd arrive at the peak after launch. They decided that would not work. So they adopted the same sort of strategy as used in the 5061a. They used got fair accuracy, not what they intended. That is where I entered the picture. A little analysis later I figured out their demodulator developed biases that were not nice constants due to the sinewave sweep over the peak. The even harmonics were doing this. Betcha many reading this know what I did next. I changed the 30 Hz (if memory serves) sinewave sweep to a squarewave step to either side of peak. The second harmonic errors vanished, of course. The calculated error of the level of the peak fell to so close to zero I never published the real figure. I published figures for wildly imprecise components. Set it up in the lab and the error in free fall was reduced to parts per million from parts per 100s.
I saw applications for this. I may be clever with electronics; but, I could not and still cannot sell everlasting solar powered refrigerators to people living in the Sahara desert. The person I worked for dismissed it with a ho-hum. In found somebody else had finally suggested this in the 1970s while reading a Proceedings of the Frequency Control Symposium. I felt simultaneously vindicated and cheated. I also swore off bothering about patents after that. The patent I do have, under a married name, was an accident. I was blind sided. Later on I was responsible for the frequency synthesizer design for the Phase IIb GPS satellites, the ones with two Rb and one Cs frequency standards. But that's another story.
{^_^}
VERY nice! Takes me right back to my career working in electronic calibration laboratories in the 1980's and 1990's. Just discovered this channel and ... Subscribed!
I can't upvote this enough. Thank you for the clear and concise explanation!
When I was an undergraduate physics student in Chicago in the mid1970's, my (very nice) parents bought me an HP35 calculator. One very hot day, the case melted inside my car. I took it to the nearest HP repair facility (in Niles, Illinois); they swapped shells for me at no charge (but broke one of the battery compartment clips and destroyed the back decal, unfortunately). Afterwards, the repair tech took me into the back to show me their brand-new HP cesium clock. Amazing!
A year later, I did a summer internship at Argonne National Laboratory, where I put together a data acquisition system from shipping containers full of fresh HP2016 parts. They had allocated the full summer for me to do so; I had it up and running (BASIC interpreter and TREK73 game) by the end of the week :-)
Since both Rb and Cs are usable for making time references, could we go further up the group and make less accurate clocks with K, Na, etc?
Don’t know about accuracy, but hydrogen frequency standards are well used on board of navigation and scientific satellites.
@@bakagaijin7452 Hydrogen masers are the gold standard for short-term stability in the RF domain but they slowly drift over time (
Maybe using rock salt for the sodium??
@@acmefixer1 Its already ionised at that point so the tube in that clock wouldnt work. Unless you mean extracting the pure metal from the salt? You'd need to purify it first. There's better sources for sodium chloride than rock salt.
Rb and Cs clocks have somewhat different operating principles. There is a serendipitous coincidence between energy levels of different isotopes of Rubidium which makes it possible to make a very simple atomic clock. If simplicity is not required, many different elements can and have been be used for making very high performance frequency standards -- but these are research instruments. In commercial use, Cs and Rb dominate, with Hydrogen masers also used for some applications.
Instead of Lissajous, better use Time Interval measurement on your 5334B, between 1pps of Cs clock and GPS clock (or OCXO), and use TimeLab over GPIB control to collect the phase shift.
Data collection and averaging over several hours / a whole day only will mitigate any GPS signal jitter, and only then will give you the necessary resolution to calculate the accuracy of the 5061A.
GPS receivers usually need several hours, or 1-2 days to really get stable, i.e. to learn the behavior of their internal OCXO (if available) vs the jittered GPS time information. So your momentary snapshots of phase (out of the Lissajous figure) is probably useless. Cheap GPS units probably never manage to achieve a short term stable time signal, (use ADEV analysis) compared to e.g. the Trimble Thunderbolt.
Yes, but watching Lissajous figures is much more fun.
I was really pleased to see lissajous being used for a practical purpose rather than just "ooooh pretty" which is all I've ever done with them.
@@edgeeffect Watching (again) Lissajous figures might be fun at first sight, but they become very soon boring, and not practical at all, when you try to adjust atomic clocks.
Simply calculate, how fast the figure will rotate once (180° or 50nsec phase shift) for a feasible 1E-9 frequency accuracy adjustment of the OCXO inside the 5334B, that'll be 50sec..
That's still practical, but when you try to adjust an Rb clock, which might be stable / precise to 1E-11, this will take 5000 sec already.
If you try to resolve 4° only on the Lissajous figure with your naked eye, equivalent to 1ns, that'll be 2 minutes already.. very boring .
When you would try to adjust a hp 5061A Cs standard to 1E-12 (standard tube) or 1E-13 (option 004 tube, as in the video), that minute change of this figure would require up to several hours, and everybody doing this experiment would fall asleep soon, I fear.
There's also another problem with such cheap GPS receivers, probably w/o a disciplined OCXO inside, that is the GPS signal jitter on the order of 1E-10 / sec, 3 order of magnitude above the required stability .. you would never come to a conclusion by the Lissajous method.
How do I set the Cessium 133 clock on my VCR so the Lissajous pattern stops flashing '12:00'? (so many things wrong with that statement)
I’m literally glowing from all this lovely new knowledge in my head 🤓
I have such a hell of time following exactly what is being talked about in these videos. Im a musician, not a scientist. However, I find this stuff super fascinating and despite my issues fully understanding everything, I watch all the way through.
Great work :)
"in soviet Russia you do not measure atom, atom measures you"
I never understood the principles of the caesium clock until today. What a wonderful, clear, explanation.
Thanks, mate, from Chris, in Tasmania.
We had one of these cesium clocks on our satellite communication's facility back in the eighties. Filled with dozens of cables connecting all the distribution amps together, all of which didn't like to be touched.
Marc, you always pause just briefly before you say "H.P." it's almost like you're preparing to speak a powerful magic word or utter the name of God or something. ;)
Thanks for the Quantum Mechanics explaination, you managed to explain it better than a lot of the physicists I have listened to.
That oscilloscope was... ... ... too modern! ;)
H.P. Use to be The GOD of test equipment. The Name is more then special. LOL Still is in my heart.
I agree, in the 1970s HP was mentioned in reverent tones. Similar to the milspec HF radio used by amateurs, the Collins... (sinks to knees).
First result for cesium clock, 100% the exact level of detail i was looking for!
That particular gpsdo gives a green light before it is locked. It also takes forever to get close to freq. The bigger one with the green pcb front cover only gives the light when it is locked, but the lock is a few millihz off. You cant use any of these units for timing, but they’re ok for freq standards for ham radio which is what they’re designed for.
Get a Trimble Thunderbolt.
The man's humor is first class. At 3:19 he remarks on the unique power cord: "It's really annoying if you don't have it" Happens to everyone when they measure orbital time difference with a Cesium Clock.
I am astonished by the sheer amount of cool toys you guys get to play with! Good work!
Great explanation. It is amazing this device was commercially available only a few years after the LASER was invented! Nowadays atomic clocks take advantage of lasers for optical probing of hyperfine transition in Cs vapors. Atomic beams are no longer used. The clock can therefore stay operational for many years without need for "refueling".
Cesium clocks run for decades between tube swaps.
Always a pleasure to seen some vintage HP goodness.
These guys were the top of the best in groundbreaking lab tech. The fact HP recessed to a second zone home PC and disposable inkjet printers manufacturer puzzles me. Rise and fall of empires...
Anyway Agilent was spun off HP in 1999 so that caesium tube is pretty recent.
I used to work for HP. It’s beyond saddening to see what it has become. Carly Fiorina flooded the company with cheap 3rd worlders. Damn her to hell.
@@trainingtheworld5093 :(
And then Keysight spun off Agilent. I believe that the Time division was either spun off or purchased to becomes
Symmetricon, which was then bought by Microsemi, which was then bought by Microchip.
Also, I believe that the crystal oscillator was used to clean up the Cesium clock. Cesium clocks have great accuracy, but they don't have very good phase noise. Crystal oscillator aren't nearly as accurate, but they have good phase noise. By locking them together you get the best of both worlds. At least that is what I remember.
23:07 I can't tell if that pattern is tumbling end over end (rotating on horizontal axis) or if it's spinning on the vertical axis...like one of those mind illusions!
I remember having one of those at Bell Northern Research in the early 90s. IIRC it was used as a reference clock for validating the VCXO frequency used by telephone switches for trunking.
Love the hyperfine transition discussion. So fun. Thank you.
I love the fact that there's an atomic clock and a bench grinder on the table at the same time! Probably not that ususal of a setup 😁
This made me understand how a cesium clock actually works. Amazing explanation!
What a fantastic video Marc! Excellent discussion and explanation of this type of clock
wow supplier catalogues that list prices, that's novel. I'm used to "SEE DEALER FOR A QUOTE" on everything
"SEE DEALER FOR A QUOTE" -- I hate seeing that so much, always so disappointing to run into that.
Whenever I see that I move on, it's an instant "Look, if you need to ask you just can't afford it!"
@@Drew-Dastardly That's very true and usually they are right, I can't afford it! But sometimes my nosy self likes knowing how much things cost so I can dream!
I remember as a kid picking up an HP catalog at a Hamfest. A hardbound book as a catalog! All those treasures that were buried within it..
The degree that humanity has come to understand the physical world never ceases to amaze!
10:10 Totally clear, in a "I could not repeat it, and I could not point at the parts, but it makes sense, and meshes with the tiny amount I understand about Relativity, and the even tinier part I understand about Quantums" way.
Very cool! I've read the manuals for the HP cesium clocks and thought those were fascinating pieces of equipment. It's too bad the cesium tube has a limited life span, but glad to see you found a working one.
FBI Top 10-Things you don't want to mention in a youtube video: "Oh, I like those things! The atomic bomb is on the left, the detonator is on the right and the crypto things over there..."
Another great video, thanks! When all this is over, I'd love to come over to the Computer History Museum and take a close look at the IBM 1401 and the other cool stuff there. Maybe next year... Stay well!
I've always wondered how atomic clocks work, and your explanation of how they utilize the Stern-Gerlach apparatus finally put it all together for me, MANY thanks !!!
What's the significance of the 7:9 ratio in the definition of the transition states and where does it come from? The ~4.021 and ~5.170 @ 9:28. I imagine that clean ratio comes from quantized particles... but I'd love to know more!
I used to play around with precision clocks. Every time you plot them on a scope together, and see them hardly drift, it's a magical moment. That you can do these things in your own home is incredible. (I was using a rubidium clock, but those have about the same accuracy as a cesium clock from that era.)
The most satisfying 8 zeros I've ever seen on RUclips.
Thank you, no really, thank you for showing us this your channel is now officially my favourite channel
So, to sum it up into an elevator pitch, the time signal comes from the oscillator when its output resonates with the lowest cesium spectral lines. Have I got that right?
Thanks for the physics lesson portion of the video. At 3:40 I was surprised to see that the unit on the left spelled the chemical element as Caesium. I've never seen it written with the A before.
Caesium is the original latin name of the element. It is still spelled that way in German.
@@fabianschmid2850 And when you spell it like this it feels more special and you can sell it for more money ;-)
23:02 you can change the rotation of the image on the scope clockwise/anti clockwise just by thinking about it, try it!
I wonder, as a cool project, could a Cesium clock be used (with some processing) to generate a signal that can replace the oscillator in a cheap digital clock to see if it can be made more accurate (not be 1 minute off after a week of use).
I built a digital clock driven from my rubidium oscillator. It keeps proper time within a few dozen milliseconds per year.
The GPSDO normally needs a couple of hours, normally a day, to get its precision as it uses averaging.
Indeed it settled down after one hour.
Thats only true of good GPSDOs, many of these generic ebay units do not lock exactly on frequency.
I had a couple of friends who worked for HP in the Rohnert Park area, it seems like it was the microwave test equipment division mainly, and I got the tour on one Saturday night, walking through production and testing, into the anechoic chamber, etc.
Later they showed up to work one Monday, to the "Company With No Name", because HP had spun them off, but without a name! They were no longer HP, just "...."
Not even "The Artist Formerly Know As HP" LOL
After some discussion, they decided to let the employees choose a name, and so suggestions were made, and then they started voting, and finally ended up with Agilent.
So yes, that was certainly last worked on since then, so that '89 date isn't the date of the tube.
The Rohnert Park (actually Santa Rosa) location of HP was (still is?) the communications test instrument division. It's still an incredible place to visit. They have the same old HP instruments I collect displayed on glass shelves. The engineer's cubes are full of the same instruments, still being used today.
I graduated with a BS in physics and focused on astrophysicist, quantum mechanics and particle physics so I’m not shy in saying your explanation was poetic. Keep that up and you’ll give Neal deGras Tyson some competition.
In US Air Force Precision Measuring Equipment Laboratory we had one. It had to have a yearly touch up on the frequency due to the slight drift, that difference between -11 and -12.
Ah, that C-field to adjust the Zeeman line. We have not done that yet.
This chanel is getting better and better! :)
Stupid Question: How long does the 6g Cesium last? What's the "consumption rate" of the clock?
Actually I was wondering myself. The tube is only warranted for 3 years. After you spend $40,000 on it, I find it a bit short. I was guessing it lasts 10-15 years? But I really don’t know, someone more knowledgeable would have to chime in. Also you need to keep running it, so the ion pump is on. If you let it sit for too long you might not be able to restart the tube. On the non high-performance tube, you can run them at lower current to extend their life, at the expense of worse short term stability (i.e. more phase noise). This one is the high performance tube, no such choice.
@@CuriousMarc The high-performance tubes like this one seem to run for 6-8 years at most. It's possible to get 10-15 years out of the standard tubes.
The ion pump runs as long as the unit is plugged in, while the cesium oven runs only when the function switch is switched out of standby mode into LOOP OPEN or OPER. So it's best to leave the clock powered up 24/7, but turn on the oven only when you actually want to calibrate something.
By way of coincidence, just today I was reading Paul Forman's excellent IEEE paper on the development of the Atomichron NC-1001 by Natco in the late 1950s - how's that for timing?
Pretty sure the end word of your post was unintentional. Brilliant!
Thanks Karl - have located that and printed out for leisurely reading.
23:05 is the 8 on the oscilloscope turning right, or left side ?
It's an existential question that nobody ever answered!
Locked to atomic time. Huzzah! Coolest sound effect in a RUclips video ever. Kudos!
Thanks for posting Marc
I learn so much from you and your friends Marc and with such ease from your superb descriptions. This is the next best thing to being in your lab with you. All the best.
when you say that your mind will be blown and when I watch it it's like it makes total sense to me
I don't know about the 'usual' PCB-faced GPSDOs that are on ebay like that one, but I have a very slightly fancier one that exhibits instability for a bit when the output 'load' changes significantly, to the point where I could connect/disconnect a device on the timing circuit and watch it hunt on the scope.
Great explanation on the ceasium clock, but I didn't fully understand what you did there at the end with the oscilloscope and several inputs and what led you to not trust the GPS.
He was using the oscilloscope to compare the frequency of the Cesium clock and his GPS disciplined oscillator. They should be very close in frequency, so the pattern on the screen should be very still. The faster it moves, the farther they are off. The Ebay GPSDO he is using is known not to be exactly on frequency with GPS. The way you do the test is you put one frequency on channel 1, the other on channel 2, and put the scope in XY mode. If they match in frequency and phase, you get a circle. This is called a Lissajous pattern. Since he was comparing a 5 MHz Cesium frequency with a 10 MHz GPSDO, the circle had a twist in it making a figure 8 as the frequencies were pretty close to 2:1.
@@stargazer7644 Thx!
The eBay GPSDOs are usually based on a re-used GPSDO from a well known manufacturer (Trimble, Samsung, Symmetricom). Have you opened your unit to see what's inside? Also it might be worth plugging in to the serial port to see what the TDOP is at when attempting a Lissajous curve with the caesium clock.
The generic ebay GPSDOs aren’t built on reused GPSDOs, they are simply reused TCXOs with a microcontroller and a GPS receiver. They often don’t have good enough software or DACs to keep accurate time.
Awesome. Back in the '60s, even the PLL synthesizer must have been pretty bleeding edge technology.
Harmonic generation in PLLs actually has some more down-to-earth applications as well, e.g. there used to be a few high-end FM tuners that would lock their mechanical-varicap-tuned LOs to a 50 or 100 kHz frequency grid this way. This way they would combine the better frontend selectivity of mechanical variable capacitors with the frequency stability of PLL tuned sets while still keeping phase noise at bay (almost all early PLL tuners were afflicted with more or less high levels of phase noise, degrading their maximum S/N).
I also thought that microwave and PLLs and frequency synthesizers were science-fiction in the 1960's. That's until I looked at the telemetry radars and transponders in Apollo. Then my jaw dropped. It's absolutely chock full of these things, some running at 9 GHz. BTW, this HP clock was installed just in time in 1968 to serve as the deep-space network master clock for the Apollo 8 mission.
That looks like a day spent well!
I really enjoyed this segmanet. The Hafele-Keating experiment was a test of the theory of relativity. In 1971, My law partner's brother, Joseph C. Hafele, a physicist, and Richard E. Keating, an astronomer, took four cesium-beam atomic clocks aboard commercial airliners. They flew twice around the world, first eastward, then westward, and compared the clocks against others that remained at the United States Naval Observatory. When reunited, the three sets of clocks were found to disagree with one another, and their differences were consistent with the predictions of special and general relativity.
There's one of those on display at the Museum of Time in Besançon, France. It's open and you can see the insides. Incredible to think we now fit that technology into a relatively simple MEMS integrated circuit assembly.
Last I heard, the MEMS equivalent had precision problems.
@@johndododoe1411 interesting, I haven't heard any of that nor witnessed it on a CSAC. Can you elaborate ? This may have work-related implications for me.
@@TheNefastor I am referring to the original scientific publication of the creating of a cesium clock that was the size of a typical electronic part. The article noted that the device was so temperature sensitive it might be used as a thermometer, which convinced me that it had a less than required inherent accuracy, as temperature should not interfere with the relevant quantum mechanical resonance frequency, and all surrounding circuitry should correctly and exactly reflect the resonance in those atoms. In later years Symmetricom tried using these instead of oxco oscillators in some ready-made time synchronization servers, but it seemed of little relevance compared to the multiple non-government labs still installing HP Cesium clocks like the one in this video.
I was going to mention the tube replacement! Score!
Agilent is a pretty recent name...It came about in 2000. The test equipment division was spun off again as Keysight in 2014.
Supposed to be "working" from self-quarantine rn but actually just watching Curiousmarc youtubes 🙏🙏
* 19:15 -cesium clocks- rubidium clocks in orbit (or sometimes hydrogen masers). If I remember right these have smaller Allan deviations for shorter timescales than Cs which is good because the satellites get new correct time standards beamed up every day anyway.
They have both (or used to at least). From what I understand the advantage is that the Cs ones don’t drift and are the master references. The Rb are much cheaper (no high vac involved).
I managed to get my hands on a Rubidium frequency standard used for Omega Navigation back in the day. Similar principles but a bit simpler.
It used a rubidium lamp to shine photons through a cavity containing rubidium ions. The cavity had a waveguide subject those ions to GHz frequency but it was dithered at 150 Hz rate by a triagonal wave making it sweep like FM. As that RF sweep went through that hyperfine transistion frequency the ions would then absorb the photons from the lamp and a dip in light intensity was detected after the cavity. This dip was tracked to discipline the 10 MHz crystal oscillator.
Apparently it might loose 1 second after 3000 years - Good enough for me
Cesium clock is the ultimate surface plate. Our world is built on precision references. Thank you Marc for explaining the 5061A operation to us.
Interesting. It is a model from 1968 HP catalog, but there are stickers inside that say Agilent - the new name of HP's instrument division starting around 2000 or so. They either kept producing them or servicing them for a very long time. The design of the box looks classic, but truthfully the inside of the box looked very clean and free of dust. I"m going to guess this was a later-produced unit.
The tube has obviously been replaced at least once. The unit is the 1968 “A” model, manufactured in 1981.
So do I have this right? Time is the difference between the color of light produced by the spin 1/2 electron and a spin -1/2 electron in the 6s shell of a cesium atom?
More or less. It does not tell you what “time” is, but it gives you the definition of a second. “A second is the duration of 9 192 631 770 periods of the radiation corresponding to the transition between the two hyperfine levels of the ground state of the caesium 133 atom”. You are correct that this transition exists because of the outermost electron, like any electron, has two spin states. It is one of very many caused by that electron and its spin. It can be measured by irradiating the cesium with a microwave frequency close to 9 192 631 770 Hz and tune it until it flips the electron spin. This is what this HP machine does. But you could do an infrared line color difference (which by the way involves 3 states, rather than the direct microwave measurement between just 2 states). For that you’d need a super duper infrared spectrometer and a way to excite the Cesium so it emits light, like in a flame, then you would zoom in on the infrared D4 line, until you see it split in a doublet (caused by the spin). That’s the fine transition. Then you would zoom in waaaay more, and you would eventually see that each line of the doublet again splits in two. Also due to spin. That’s the hyperfine transition. The inverse of the wavelength difference (which is the “color difference”) between these two very close infrared lines should be 9 192 631 770 Hz. Modern clocks use tunable lasers to do essentially that measurement, but in a much more controlled way.
@@CuriousMarc Thank you for the clarification. My posit was a bit of a flippant reduction of a complex and very technical measurement. I'm not a physicist but I'm physics curious so I read and listen to physicists. I think knowing how thing like "spin" and "energy level" are measured has helped my understanding. The Stern-Gerlach measurement is not exactly what physicist mean when they say "spin" but it's the concrete part.
Physicist talk a lot about time, they even write whole books about it. A common claim is that our intuitions of time are not always useful. But they then often fail to say exactly what "time" is. But now I know the exact definition of what a unit of "time" is, a 9.192GHz difference. Now I can scratch my head and ask how can they measure GHz without already having measured time?
@@charlesmanning3454 You can’t define time easily because it’s relative. Only the speed of light is constant. All you can hope to do is find out your local second at your local speed and local gravity. Which is what this HP machine does.
You don’t measure the 9 192 631 770 Hz either. The Cesium atom measures it for you. Once the clock locks its probe oscillator on the atomic resonance, you know it is guaranteed to be 9 192 631 770 Hz. You then divide it electronically by 9 192 631 770 and you now you have an exact second. Actually this machine divides it by a smaller number and gives you a perfect 5 MHz or 10 MHz reference output, which you use around the lab. Most HP frequency meters will accept that as their external frequency reference input, become instantly Cesium-referenced, and you do your other measurements with that.
I actually fell asleep at ~7:30 minutes, no joke. I'm currently rewinding to watch again. I actually really appreciate it when people tell the history of science; how things were discovered.
Lol around 16:00 minutes you read out two separate Hz measurement numbers wrong in a row. Lol. Must have been a long night hahaha.
Isnt it normal that the 2 are a bit off because of the height diffrence
Oh that’s true, because of general relativity we talk about in the beginning of the video. I had not thought about it. But you’d think they would compensate for it. Then, maybe not. It would not be a Hertz like this though, but a tiny barely measurable difference.
The magnitude of the effect is (g*h/c^2), where g=9.81 m/s^2, h= altitude in meters, c = speed of light, 3*10^8 m/s
This corresponds to about 10^-16 fractional frequency change per meter of altitude difference.
Today this is measurable even on millimeter scale with the most advanced of research clocks, but HP 5061A would require 100 km of altitude to show a clear difference.
If would be interesting if you could setup a NTP server using this clock and a vintage mainframe or microcomputer. Cheers
Commercial NTP servers often have rubidium atomic clocks available as an option for their holdover oscillators.
i wonder if the tubes are refillable
and to anyone wondering, yes i am aware of the UHV vacuum requirements for these tubes. And yes i do have the equipment to evacuate them back to UHV from atmospheric pressure.
@@lbochtler in addition to the UHV equipment, you would need something like a Argon-filled glove box to handle the Caesium. And all the equipment to calibrate the tube afterwards (Cs current, oven temperature, ...). In addition, you would need to change and realign the apertures etc. Having worked quite a bit with other UHV devices that had Cs sources, I'd leave that to the manufacturer... :)
@@dsamuelis now, where is the fun in that? All of that sounds like quite manageable to me. (for reference, i have worked on particle accelerators and electron microscopes in the past). I would imagine the apertures are either cleanable the same way that of an electron microscope are, by heating in a high vacuum.
As for calibration, yeah that might present a challenge. Though all of this talk of refilling such tubes, is making me want to just build a cesium beam clock from scratch.
Oven temperature calibration is relatively trivial in my opinion, as i have and use precision radiometer and contact thermometers. All that would be needed is to calibrate the radiometer to the oven sensitivity.
The only challenging thing i can see is the bake out process needed for UHV. As this will evaporate the cesium if im not careful.
Also, in case of the HP tube, i would imagine the electron multiplayer is also a part which at the very least would need replating of the dynodes.
Hm, if money and time allows it, i might just get myself a hp clock and see if i can refill the tube using my lab equipment. The only thing id need to add is an argon glove box. Though i imagine working under a heavier then air gas with a relatively tight enclosure should suffice as well.
@@lbochtler sounds like a great challenge then :) Most of the "regular" heavier than air gases however would react with Cs (CO2, SF6). You'd be limited to the heavier noble gases, which are extremely expensive and rare. On the other hand, you could buy a reasonable glove box for something that is quite a bit cheaper than a new Cs assembly (40k USD according to the video) - you might be able to get a suitable used box for around 20k. I've been working with alkaline metals in those boxes for quite some time, so please contact me if you'd like to know more about specs and usage of such boxes.
Yes, the multiplier would need to be exchanged or serviced as well, as most of the evaporated Cs is deposited there.
My experience with Cs sources is mainly from UHV surface/sputter analytics, such as SIMS or ion microscopy. Unfortunately in these devices, the manufacturers try to keep the Cs sources as un-serviceable as possible, so exchanging the source is most often the only way to go. I also have been working at particle accelerators during my academic career (mainly using the parasitic X-rays that are emitted there), but I've never worked "on" an accelerator :)
No, because you not only have to put new cesium in, but you somehow have to clean out the old cesium crud that is splattered all over the inside
I saw your interview on Gizmodo, very cool! I also saw the 1401 you restored at the CHM, really awesome.
Wouldn't the high speed of the orbiting GPS satellite mean that its atomic clock is running at a slightly different frequency relative to the atomic clock on Earth? So wouldn't we *expect* there not to be a perfect phase lock between the two clocks? Or have I incorrectly understood?
You bet the GPS system has to make relativistic corrections! Both special relativity for the speed of the satellites and general relativity because they experience less gravity up there. But that’s all corrected for, or the system would not work at all. The cheap GPS DSO was just caught drifting. Might have lost one of the satellites. It corrected itself 30 minutes later. It’s just not good enough, which is why lab quality GPS DSOs also include a rubidium local reference for the short term stability.
Is the dial on the right where you wind it up at?
Great overview of the clock’s operation. A few years ago, I ran across some open courseware videos, from UC Berkeley as I recall, on an attempt to make an atomic clock on the chip level in MEMS technology. I’m not sure if they managed to make it work. Isaac Asimov’s physics books provide an interesting overview of the basic theory behind quantum mechanics.
The design of the machine is interesting, but perhaps more interesting to me is that it's the first explanation I've seen of the derivation of the standard for defining the second.
can you check against the wwv time signal? wiki lists a telephone service number if no rc clock is at hand.
Fascinating stuff. How much signal interference do the patch leads and the wiring of the clock give in your measurements, if any, and would shorter one's or different conductor materials and construction make a difference, as in audio equipment, where you can tell between different cables????
Did he say the GPS clock was a few cycles off?. I have two GPS clocks but I haven’t compared them at all. Something I should do.
When you get into this type of precision, the phase lock of things takes a while... GPS disciplined clocks are a set of things... the GPS receiver which is a whole video or two worth of wonder in itself, which has to know exactly where it is to get the time right (takes a few days if you do it right, with an outdoor amplified antenna), and an oven stabilized clock (good to about 1 part in 10^8 or so by itself over a few days), a numeric controlled oscillator (another video worth), and a computer to feed slight tweeks to lock everything in, compute position, etc.
Give the GPS a few days to lock position and time... and they have long term stability as it comes from the national time standards... over the short term, (less than a day) the Cesium will be better, long term the national time standards are better, as they have more, newer clocks in groups all carefully measured against each other. The GPS is the weak link as you've got things whizzing past at relativistic speeds, and random delays from the ionosphere that drift over time.
He said it settled down later, but unfortunately, some Chinese GPS clocks are indeed inaccurate.
I can confirm some of the ebay chinesium GPSDO clocks do not ever lock to the correct frequency. They are always a few millihertz off. Not generally a problem for many frequency reference uses, but totally unacceptable for GPS based timekeeping purposes.
Nice to see this old of an atomic clock not seen many of them which is odd considering I've installed a few modern ones over the years. I forget how large many of these older designs can be when they nearly have single chip version of them commercially.
Hardly.
How is GPS time affected by the satellites being higher in the gravity well?
They are high up, so their time goes faster! Plus they move quite fast so there is a special relativity time difference between the satellites too. So they have to do both a special and general relativistic correction of the time for the system to work.
Before GPS was LOng Range Aid to Navigation. The Coast Guard manned xmitter sights that were all controlled by Cesium clocks. LORAN stations were all over the world and they each had 3 of these in the rack. Triple redundancy and all.