JohnSwenson Posted June 2, 2020 Share Posted June 2, 2020 5 hours ago, MartinT said: The output connectors are 50 ohm. I'm thinking of sourcing PCB mount 75 ohm sockets and swapping them. Using short 75 ohm cables into 75 ohm ER and Mutec in sinewave mode doesn't seem to be harming the signal too much. Just swapping the connectors to 75 ohm is not sufficient, the output circuit has to be adjusted to also be 75 ohm. It may just be just changing a resistor, it depends on what the output circuit is. John S. MartinT 1 Link to comment
JohnSwenson Posted June 3, 2020 Share Posted June 3, 2020 That was the right choice. What the GPSDO brings is much better stability over very long time frames, something that is not needed for audio and adds a LOT of complexity and potentially more noise. I have a good GPSDO (well actually two!) but that is for use as a time base for my very sensitive time measuring equipment, not as a clock for digital audio. Yep getting a good signal to the GPSDO is a major pain. I just bought a new one because it has a MUCH better GPS receiver than the old one, which meant I didn't need an outdoor antenna, one in the window works fine. John S. Link to comment
Popular Post JohnSwenson Posted June 9, 2020 Popular Post Share Posted June 9, 2020 58 minutes ago, zerung said: @JohnSwenson@Superdad How much difference would you find - with the impedance matching? Thank you I can tell you what happens technically with the waveforms on the wire, but how that shows up as sound difference is a completely different issue. Every time you have an impedance mismatch there will be a reflection, the bigger the mismatch the bigger the reflection. The worst thing is a mismatch at both ends, this causes multiple reflections, each edge bounces back and forth, this is can definitely cause problems with reception of signal. If say the source is matched (50 ohm electronics, 50 ohm connector, 50 ohm cable) but the receive side is mismatched, each edge will reflect at the mismatch back towards the source. But since the source is properly matched that reflection gets absorbed by the proper impedance match. So if a clock box has proprer electronics and connectors and cable, a mismatch at the receive end probably will not make too much of a difference. The edge bouncing off the receive end and winding up at the source can cause some distortions at the source itself, but how much and what they look like vary wildly from design to design. John S. Encore, zerung and ZeusOdin 2 1 Link to comment
JohnSwenson Posted June 10, 2020 Share Posted June 10, 2020 8 hours ago, MartinT said: Thanks - I will try it once I've got the impedance sorted. I hope I'm right in thinking an impedance mismatch affects a squarewave worse than a sinewave. Yes, a mismatch has a larger affect on a square wave than it does on a sine wave, BUT the sinewave is already significantly worse than the squarewave. i still think a single mismatch on the squarewave will probably be better than the mismatch on the sinewave. But that is just a guess, I haven't actually done that test and actually measured the output of the clock circuit. John S. Link to comment
Popular Post JohnSwenson Posted June 11, 2020 Popular Post Share Posted June 11, 2020 9 hours ago, Confused said: Out of interest, what is the technical reason for this? The specific reason: the clock input goes into a very low additive phase noise clock synthesizer and all its inputs are designed for square waves. More general reason: all the digital circuits that use clocks actually use squarewaves so feeding a sinewave into the system either doesn't work at all or gives much higher jitter. Technical explanation of the above: All clock receivers have some form of threshold circuit, it changes state when the clock voltage goes through that threshold voltage. NO signals are ever perfect, there is ALWAYS some form of amplitude noise on the signal, AND the receiver itself always has some form of fluctuations on the threshold voltage. So think of the clock signal rising towards the threshold, as it gets nearer to the threshold the amplitude of the noise comes into play, the threshold might get passed when the noise is at its peak or at its lowest point, or some place in between. This noise on the signal causes an uncertainty as to when it will actually get to the threshold, otherwise known as jitter. The faster the voltage rise of the signal the lower the time uncertainty for a given noise amplitude. Thus a square wave with very fast rise and fall times will have much lower jitter than a sinewave which has a much slower changing voltage. This is the reason a circuit designed for square wave input MAY still work with a sinewave, but the jitter in the circuit clocked by the signal will be higher. There ARE some ways to convert sinewaves into square waves, but the simple ways actually increase the noise on the signal making the jitter even worse. There are some ways to do it well but they are complex and expensive and take a lot of power. Putting one of those on the clock input of the EtherREGEN would have at least doubled the cost of the device, not worth it in my opinion. BTW these master clocks provide both sinewave and squarewave outputs for different applications. Radio systems that need a very stable frequency reference for running into mixers etc want a sinewave. Digital systems want a square wave. Either system CAN use the other type, but won't work nearly as well. So it is best to get the type of master clock that works well with what you want to use it for. I hope that makes some sense. John S. LowMidHigh, Superdad, richard_crl032 and 2 others 1 1 3 Link to comment
Popular Post JohnSwenson Posted June 26, 2020 Popular Post Share Posted June 26, 2020 5 hours ago, Clockmeister said: Hello Superdad Like your products very much by the way 😎 With regard to the Crystek 575 clock, are you looking to replace this item in any up and coming re-engineering of the Ethernet regen at all? While the specifications are quite good I have noticed a fair amount of harmonic noise does extended well into the 2Ghz territory in numerous products that use this device, which may possibly have an effect on surrounding circuitry? Not a critique' just an observation. Regards DM What signal are you refering to that has 2GHz harmonics? If it is the clock signal itself, then yes, the harmonics go up quite high, that is on purpose, it is after all a square wave, the only way to get a good square wave is lots of high frequency harmonics. The higher the harmonics the faster the rise time and the lower the jitter. If you are referring to high frequencies on other signals, there are circuit implementation details to prevent that from happening. I spent a lot of effort in the EtherREGEN to keep those as low as possible. It is an interesting engineering trade off. The faster the clock edges the lower the jitter, but the higher the ground-plane noise you get from the circuitry operating with those fast edges. I have endeavored to find a good trade off between this that gives the lowest overall jitter in the output stream. John S. PYP and richard_crl032 1 1 Link to comment
Popular Post JohnSwenson Posted June 27, 2020 Popular Post Share Posted June 27, 2020 The 575 does have its own dual regulated supply, and the only thing it drives is the clock synthesizer (which has its own regulators) which only outputs LVDS signals. All clock distribution is carefully impedance matched differential pairs (I had a special board stackup used that allowed continued impedance matching as pairs changed layers). The reclocking flops are fully differential including clock. The chips which have a single ended clock input have a very low additive phase noise LVDS to CMOS converter right next to the clock pin. The result of all this is a system that is much less sensitive to ground noise than is usually seen. That was really the whole reason for the EtherREGEN, produce much less ground plane noise in the first place, and have circuitry much less sensitive to what ground plane noise that is there. Unfortunately I can't do anything about what is inside the chips themselves. I spent 30 years designing power networks inside large high speed chips, I have a VERY good feeling for what that can do to jitter for the circuitry inside a chip. Unfortunately we can't afford to do a full custom chip for every function we need, so all I can do is choose ones where it looks like someone did a fairly decent job on the internal PG networks. John S. richard_crl032 and zerung 2 Link to comment
JohnSwenson Posted August 26, 2020 Share Posted August 26, 2020 2 hours ago, Exocer said: I purchased it because I saw you enjoyed it 😁. Can anyone tell me if this is 75 or 50 ohm? Out1 and 2 are 75 ohm. If the 10MHz looks the same then you are good to go if you use a 75 ohm cable with a 75 ohm plug. John S. Exocer 1 Link to comment
JohnSwenson Posted September 28, 2020 Share Posted September 28, 2020 18 hours ago, sakso136 said: My etherregen is standard one . The guy on starkaudio,hk told me go for 50 ohm bnc I told him I m going for gpsdo clock first then may be for cybershaft. I mentioned that I m using it with er also. So will you confirm I need bnc 75 ohm? Really thankful for answering... If you have a standard ER then you definitely want a 75 ohm one AND a 75 ohm cable. When we were getting close to releasing the ER we did a pole and almost everybody said they wanted 75 ohm, so that became the "standard" ER. Alex makes a few 50 ohm versions every now and them for people that specifically order 50 ohms. John S. Superdad 1 Link to comment
JohnSwenson Posted October 1, 2020 Share Posted October 1, 2020 20 hours ago, RickyV said: An other difference I noticed is that my clock uses the LT1764 wich is a 3A version but still with 40uVrms noise. On the other pictures I saw they used LT1963. Now, I am a bit of a modifier so why not replace these regulators with a 2A or 3A Lt3045 ( 3x 1A LDOvr reg). @JohnSwenson would this be a good idea? Yes, clocks ARE susceptible to supply noise. The LT3045s are much lower noise than the ones mentioned, BUT a LT3045 is only good for 0.5A, so you need to put a bunch in parallel to attain the current you mention. Are you thinking about designing your own boards or using someone elses boards? If you use existing boards they either need to be a single board designed for that current, or lower current boards specifically designed to be paralleled. You can't just take two or three 1A boards and parallel them unless they were specifically designed for that. John S. Link to comment
Popular Post JohnSwenson Posted October 1, 2020 Popular Post Share Posted October 1, 2020 18 minutes ago, RickyV said: A few years ago I asked Alexey from ldovr.com if it was possible to parallel his boards and after he did some experimentation he said it was by paralleling all the inputs, all the outputs and all the “set” pins. The 1A boards he sells now have a through hole connected to the set pin, see picture. In the picture a few posts up you see a 1V 5A version I made to power a processor of a certain switch. Heat is a problem as I can’t get a heatsink in there so I stack more boards to lighten the current load of all and lowered the input voltage to 0.8V above output. Now the temperature stays under 40 degrees C approximately. When I used an input voltage of 2.5V they got about 70 degrees C, too hot. So do you think this would be a worthwhile modification? Yes, this is the correct way to go. With paralleled SET pins you CAN parallel the boards. I definitely think it is worthwhile to do. As you noted you just have to make sure you can get rid of the heat generated. John S. RickyV, [email protected] and Exocer 1 2 Link to comment
JohnSwenson Posted October 13, 2020 Share Posted October 13, 2020 On 10/10/2020 at 9:09 AM, R1200CL said: That’s nice to know. (Rember there’s an A after 1963😀) So the clock itself requires 11 V then. Did you find any data sheets for clock saying anything about voltage requirements? I wonder if we’re able to convince the Chinese to exchange that present regulator in use to LT3045. I think he’s reading this tread, but a bit reluctant to participate in any discussion. However there may be other changes that could be done. So maybe if we’re patient John maybe give some suggestions. Remember that the LT3045 is just 0.5A. Even 12V OCXOs take more than that when warming up the oven. So you would have to parallel two to three LT3045s, this CAN be done but is a bit more complicated, and WAY more expensive than the 1963A. John S. Link to comment
Popular Post JohnSwenson Posted October 13, 2020 Popular Post Share Posted October 13, 2020 On 10/10/2020 at 9:32 AM, magnuska said: Well that could be the issue, its a 150cm Audio sensibility signature cable. Not cheap. That length was the preferred length for spdif signal according to AS. There is NO need for a specific cable length IF the impedances are properly matched all along (cable, connectors on cable, connectors on boxes and internal board impedances). The issue did exist for SPDIF because the impedances almost never came close. So for a clock cable where everything is properly impedance matched, there is no need for a specific length, in general shorter is better. BUT you have to take into account the bandwidth of the cable. A longer piece of very high bandwidth cable can be better than a short piece of not so high bandwidth cable. Also shielding can matter. The best are very high bandwidth double shielded cables. These are not cheap but will give the best results. Note NONE of this has anything to do with the clock not working at all, it has to do with subtle differences in the edges of the clock waveform at the clock receiver causing slight differences in jitter in the received clock which you may perceive as better sound. As to the specific cable mentioned, I have no idea, I don't know the specifics of that cable. As long as the cable and connectors on the cable have the correct impedance to match the input on the ER and the output of the clock you can certainly try it. (you can use anything you want, it won't harm anything, it just may not sound as good as it could, heck wire coat hangers will WORK, but probably won't sound that great). John S. magnuska, RickyV, Exocer and 1 other 1 3 Link to comment
JohnSwenson Posted October 13, 2020 Share Posted October 13, 2020 5 hours ago, GMG said: Anyone tried or has an opinion on the low priced Finisar modules offered at Aliexpress and marked as manufactured for <XXX> XXX = Macroni, Ericsson, RedBack, and more Are those original modules that are manufactured for network providers? Some are priced well below $40 (some used also at $10) examples: https://www.aliexpress.com/item/32854739380.html?spm=a2g0o.productlist.0.0.c7724cc2Lzx9i7&algo_pvid=d5269e93-5aaf-4deb-a2f1-33232e82592d&algo_expid=d5269e93-5aaf-4deb-a2f1-33232e82592d-8&btsid=0b0a0ae216025962495106651ea8fc&ws_ab_test=searchweb0_0,searchweb201602_,searchweb201603_ https://www.aliexpress.com/item/32854767247.html?spm=a2g0o.productlist.0.0.c7724cc2Lzx9i7&algo_pvid=d5269e93-5aaf-4deb-a2f1-33232e82592d&algo_expid=d5269e93-5aaf-4deb-a2f1-33232e82592d-16&btsid=0b0a0ae216025962495106651ea8fc&ws_ab_test=searchweb0_0,searchweb201602_,searchweb201603_ There are two possible meanings of that "Manufactured for", one is as you mentioned that Finisar made them and manufacturer listed sold them as their own. But the more likely possibility is that they contain that particular manufacturer's code. The SFP modules contain a code for a specific manufacturer, the equipment (switches etc) from many manufacturers will only work if you insert a module with THEIR code embedded in the module. So Manufactured for means they contain that company's code. None of this matters for the ER since it doesn't look at that code at all. John S. Link to comment
JohnSwenson Posted October 13, 2020 Share Posted October 13, 2020 6 minutes ago, RickyV said: What do you think about the 75ohm adapters I used and the two extra contact passes, it is short. 9 posts up. As long as they are the correct impedance I think it is a great idea. Those adapters almost always use teflon dielectric and are machined to tight tolerances. There is a high probability what you have will be better that an equivalent piece of short cable of any type. John S. RickyV 1 Link to comment
JohnSwenson Posted October 13, 2020 Share Posted October 13, 2020 14 minutes ago, RickyV said: @JohnSwenson I was wondering do you know of modules or small pcb’s for sale that have a 10Mhz input and have a relatively easy selectable output frequency, 24Mhz, 25Mhz etc audio frequencies? I don't know of anything specifically for audio frequency use. You could do it with the Si5340-D-EVB which is the evaluation board for the synthesizer used in the ER. BUT you need to have it connected to a computer via USB to program the chip every time it powers up. It is great for testing and exploring but not good for a board you put in a product. And it is a pretty big board as well. John S. RickyV 1 Link to comment
Popular Post JohnSwenson Posted October 29, 2020 Popular Post Share Posted October 29, 2020 5 hours ago, GMG said: Since leakage is due of the power supply I don't see how using 2 identical power supplies will prevent the leakage to each of the powered components Vs. using 1 power supply of the same nature but with double the power. Maybe I'm missing some understanding here, but my logic is that if you use a quality PS with low leakage then you will not suffer from leakage, regardless of the number of power supplies you use. Hi GMG, there are two completely independent things that can happen, ground loops and leakage loops. They are caused by completely different mechanisms, ground loops are caused induced current in safety grounds, and leakage loops are a property of power supplies. To make things really complicated these two can interact in strange ways. For example a leakage loop starts at a power supply, it can go through another power supply OR a safety ground. There are no simple "one size fits all" rules with this. As to the question above, having one low leakage supply does not prevent OTHER higher leakage supplies from causing leakage loops. Say you have a low leakage supply with a safety ground connection, the leakage from another supply can travel over the signal cables to get to that safety ground. Again no simple rule. Another example is the EtherREGEN, it is designed specifically to stop leakage current from cheap network requipment from getting into you equipment. If you use the same supply which has the negative of the two outputs connected, and one drives the EtherRegen and the other a streamer or DAC downstream of the ER, you short the isolation in the ER and leakage from the network equipment goes through the single power supply right into the audio equipment. That is why two completely separate, isolated supplies would be better. The whole subject is important, but very difficult to figure out by just thinking about it. Both ground loops and leakage loops can go through BOTH power supply and signal (both digital and analog) cables, so any attempt at figuring them out takes drawings that include all power and signal connections in your system and then try and work things out. With a complex system it becomes very difficult to do. That doesn't mean give up, start with isolating the strongest sources of loops, which usually come from cheap home network equipment, then once you get rid of that start working on lesser sources. John S. Superdad and R1200CL 1 1 Link to comment
Popular Post JohnSwenson Posted December 28, 2020 Popular Post Share Posted December 28, 2020 OK, here goes a bit of a primer on clocks, clock transmission and digital audio. First the use: digital circuits require square wave clock signals. If fed a sine wave the signal would have a very slow ramp at the threshold that noise on the wire would produce large amounts of jitter, so much so that in many circumstances the circuit would not work AT ALL. BUT square waves are hard to send long distances over wires. If you look at the specs for coax cables you will find that the attenuation through the cable increases as the frequency goes up, otherwise known as a low pass filter. Putting a square wave through this attempts to convert it into a sine wave. The longer the cable the more like a sine wave it looks at the other end. This is why all traditional clock distribution networks use sine waves. It takes very expensive cable to keep a square wave square over long distances. Over short distances (say a few feet) this effect is very small. The reason most "master clocks" you find out there use sine waves is they were designed to drive large distribution networks where a sine wave was the only way to do it. If you send a sine wave into a digital audio device, there has to be something that converts the sine wave into a square wave for use by the digital circuits. ALL such circuits add some jitter to the clock signal in doing this conversion, some circuits a lot, some a little and some a VERY little amount. There are three common ways to do this: Comparitors are cheap and easy, but because they are essentially very high gain amplifiers, they also amplify the noise on the sine wave, increasing jitter by a large amount. Cheap digital inverters are also frequently used, these do quite a good job actually: because they are not very fast they tend to limit the high frequency noise thus add quite a bit less jitter. The best way is a multi-stage low gain amplifier with a filter between stages designed to just amplify the specific frequency. This gets rid of almost all of the noise which results in a VERY small amount of jitter added to the signal. It should be obvious that this is neither cheap nor easy, thus very rarely done. When dealing with a clock synthesizer (such as in the ER and SOtM products), things are a little different, you are not directly converting the sine wave into a square wave. The input is used as a reference signal of a PLL. You can use either a sine or square, but will get different performance depending on how the PLL was designed. In the case of the Silicon Labs synthesizers they were specifically designed for a square wave reference, they WILL work with a sine wave, but the jitter on the output will be lower with a square wave. We designed the EtherREGEN making the assumption that the user will have the "master clock" a few feet away rather than hundreds of feet away. In this configuration it is better to assume a square wave going directly into a clock synthesizer. If we had decided to assume a sine wave the extra circuitry would have degraded the signal if you fed it a square wave. Thus for a master clock a short distance away the best result is obtained with a square wave going directly into the synthesizer. It is also way cheaper, and saved months off the development time. John S. Qstik, austinpop, ZeusOdin and 16 others 2 11 6 Link to comment
Popular Post JohnSwenson Posted December 29, 2020 Popular Post Share Posted December 29, 2020 21 hours ago, manueljenkin said: Regarding the analog design, is it like the inverse of notch filters at the fundamental and harmonics of the square wave to be obtained? What are the recommended cabling/signal routing choices for this approach? Do we have a litz or uspcb equivalent for this connection for both better hf performance and lower spurious noise. The hf roll off issue is because of skin effect which is because of eddy currents which is because of electromagnetism (governed by Maxwell's laws) which is because nature just seems to work that way right? (Apart from self inductance, and line-to-line or line to ground inductance or capacitance too). The type of filter used in the multistage squarers are classed as bandpass, they let a narrow range of frequencies pass through without attenuation but attenuate everything else. The HF roll-off of the cables is actually primarily due to the dielectric used and the size of the cable. This means that the lowest roll-off cables are large diameter and stiff (they use teflon). Thin flexible cables have high attenuation at high frequencies. My analysis of the cable situation shows that probably the best square wave clock cables would be custom flat cables. There are some extremely good high frequency dielectrics available for flat cables. Unfortunately they are very expensive and the NRE charges for implementing such a cable are very high. John S. Superdad, manueljenkin, James Stephens and 1 other 2 2 Link to comment
JohnSwenson Posted December 29, 2020 Share Posted December 29, 2020 2 hours ago, manueljenkin said: Thank you very much John. Is there any possibility of mitigating some of the cable limitations by using a carefully designed PCB (controlled parasitics) instead of a wire? Curious about this because the uspcb sounds much better than any cable at similar price to me. The causation I guess is different though, the latter being more due to better ground plane right? What I mentioned above ARE PCBs, to be precise flexible multilayer PCBs using very high quality thin flexible dielectrics between copper layers. With multi layers they can be shielded very well and have very high precision impedance control. They just cost a lot of money. Back in my days as a custom chip designer we used such cables for very high speed connections between boards, they cost in the range of $900 to $1400 per cable and were about 7 inches long. Not exactly cheap cables. John S. manueljenkin 1 Link to comment
JohnSwenson Posted December 29, 2020 Share Posted December 29, 2020 34 minutes ago, R1200CL said: I think @JohnSwenson is supposed to have one, if Alex has sent his over for evaluation. I guess not a prioritized task 😀 I have not received anything yet. John S. Link to comment
Popular Post JohnSwenson Posted January 1, 2021 Popular Post Share Posted January 1, 2021 1 hour ago, manueljenkin said: Happy new year to all! Came across this recently: https://en.antelopeaudio.com/products/10mx/ (rubidium transition based clock). Not sure if it can trickle down to consumer levels but interesting nevertheless. I've said this many times before but I'll say it again. What matters for digital audio is the phase noise, the absolute accuracy, and the aging (drift over time) and temperature coefficient do not matter. These are all so good that they will not even come into the picture for digital audio. The lowest phase noise is produced by the best crystal oscillators, period. All these other types of clocks (mostly "atomic clocks" such as rubidium, cesium, hydrogen maser, cesium fountain) all have lousy phase noise, but great long term stability (fractions of a part per billion), which is completely meaningless for digital audio. The time it will take to get a one second error is greater than the age of the universe, but that is meaningless for digital audio! Even when you want to very accurately measure time a rubidium oscillator is not a good choice any more. For example I have a GSPDO, GPS Disciplined Oscillator, which cost a whopping $200 and has about 1000 times better frequency stability than even very good rubidiums. A GPSDO takes a good OCXO (which has quite low phase noise, very good short term stability) and SLOWLY tweaks it with the timing results coming out of a GPS timing receiver. GPS works by knowing the time very accurately and comparing that with signals from satellites (which come from very good atomic clocks IN the satellites) which have extremely accurately known orbits, from which your location can be accurately computed. But it also works the other way, if you know your location extremely well you can compute the time extremely well. GPS timing receivers assume they are stationary and over a period of many hours compute their location extremely well, thus they know the time extremely well. Thus such a device provides you time accuracy which is actually better than what is in the satellites, because it uses the signals from many satellites, which gives you an average which is better than any of them alone. This is weird but an $80 GPS timing receiver gives you more accurate time than a $100,000 atomic clock! So never EVER, think about getting a rubidium clock, there is no reason for them no matter what you are trying to do. But that doesn't mean you need a GPSDO for digital audio. That tweaking of the OCXO that happens in the GPSDO increases the phase noise of the OCXO, so for digital audio you get better results with just the OCXO. John S. ZeusOdin, manueljenkin, lwr and 2 others 1 1 3 Link to comment
Popular Post JohnSwenson Posted January 2, 2021 Popular Post Share Posted January 2, 2021 16 hours ago, manueljenkin said: Thank you. Sorry for the unintended diversion. One question though, is why do atomic clocks have worse phase noise? Aha, THAT as they say is a good question! The understand that you have to understand how they both work. The crystal oscillator is easy, quartz actually changes shape when an electric field is applied. It has a sharp "resonance" frequency where the expansion and contraction works extremely well. Same thing as a swing, if you "pump" a little energy into it at the right frequency you get large excursions. If you pump it at some other frequency not much happens. Same thing with the crystal. The oscillator circuit is just a high gain amplifier fed back into itself, which naturally oscillates. Then throw the crystal into the loop and it oscillates at that one frequency. This gives very low phase noise, but the exact frequency can move around with temperature, electrical parameters of the circuit etc. You don't get something that is always exactly at the precise frequency no matter what. Atomic clocks are very different. They make use of "spectral lines" of certain elements. You are probably aware of optical spectral lines, say the green and blue lines of mercury (common in older street lights). These are caused by atoms which are pumped by electrical current running through a plasma made of the element in question, these cause the electrons in the atoms to go to a higher energy stage (An incorrect, but useful visualization are electrons in orbits around a nucleus, the reality is much more complex, but this is a useful quick and dirty visualization). They eventually loose that energy by emitting light of a specific frequency. When you pass that light through a prism you can see the "spectral lines" of those specific light emissions. But you can't use those directly, the lines are actually not one frequency but many frequencies called "fine structure". The fine lines are much more precise but harder to determine the frequencies. It turns out the fine lines can even be sub-divided into hyperfine structure if you look REALLY carefully. The frequency of a fine line is determined by the electron distribution in an atom, so it still depends on interactions of one atom to the next so they vary somewhat. BUT the hyperfine lines are determined by the interactions of the nucleus with the electrons, and the nucleus is FAR more stable and less affected by what is around it, thus ideal for time measurement. The energy difference between these lines is very small so the electromagnetic waves are in the microwave region. An atomic clock works by pumping a gas of the particular atoms so that they give off light, including the line of interest, then microwaves of the particular frequency are added to the chamber. When the microwaves are tuned to the exact frequency a little bit of it is absorbed by the hyperfine line producing a very small dip in the intensity of the microwaves. The atomic clock works by slowly moving the microwave frequency across the region of interest and slowly "narrowing in" on the frequency of the dip. That microwave frequency is then measured and THAT gives the timing of the "atomic clock". I hope it is obvious that this process is very "fiddly". You are trying to make extremely sensitive measurements of very small things. Such measurements are always quite noisy, thus the phase noise is high. BUT when you average it out over months (or years!) The frequency is very precise. It always fascinates me that engineers have been able to put this process into a small box that can be put in satellites! Ok, way more than you ever wanted to know. But that is why atomic clocks by themselves have terrible phase noise. John S. MartinT, richard_crl032, Superdad and 5 others 1 3 4 Link to comment
Popular Post JohnSwenson Posted January 2, 2021 Popular Post Share Posted January 2, 2021 33 minutes ago, manueljenkin said: Thank you John. Bringing it back on topic, few more questions. Would this resonance frequency also depend on the pressure and temperature of air around it and the angular alignment of the crystal (parallel to gravitational force vs perpendicular to gravitational force, ie shear stress vs normal stress). We had piezoelectric effect in undergraduate course but I don't remember reading about the causation. Having a read at Wikipedia on the topic, it looks rigorous but very interesting. Ignoring high precision circuits where things might be more complex, I assume a basic oscillator can make use of feedback network and amplifier that satisfies barkhausen criterion. Thank you very much for the detailed explanation. I would require some more time to understand this completely, but a few doubts (clarifications) here too. 1. We are using the frequency of light in a particular way to get our clock? (Light is also an electromagnetic wave so it should interact with other electromagnetic waves). 2. Are these hyperfine levels related to the wave equations etc? In that case I believe both the temperature and crystal structure (formed after the long processes natural or artificial which could involve heating, pressure etc so the new stable structure may not completely go hand in hand with the general assumed Paulis exclusion principle), and impurities would cause changes to this hyperfine structure frequencies? (Like different structures of carbon having different thermal, optical and mechanical properties) But in any case, I guess for a well defined crystal lattice it'll be far more stable over time unlike oscillators which drift due to changes in mechanical, crystal and chemical structure. Edit: I think I got it now. The patterns do change from one produced variant to another and also at one point to another within the same device due to the effects mentioned above but the hyperfine "delta" between two levels remain almost constant (or is less varying as compared to the other effect since nucleus is very far away and the influences are similar). Is that correct? 3. We use some circuit to generare a microwave that has an energy that is equal to the energy difference between the two hyperfine states so the electron absorbs, moves to the higher orbit and then after a short time jumps back to the old orbit for stability emitting back this light (would this emitted light be in the same direction as incident light? Not necessarily right). And we have some form of feedback loop that changes the parameters of the circuit generating the em wave to be incident on the atoms. It should ideally try to maximize intensity of the absorbed wave (or minimize considering that the atom absorbs the light, but then where to we get the output? Do we take two outputs from the wave generation circuit, one going to the feedback loop that has this atomic interactions, and the other used as output?). 4. If the hyperfine levels are so close won't we have the possibility of our input frequency also exciting the other levels and making the circuit design harder? Wow, someone who gets this stuff! I'll make an attempt to answer this stuff. First off I forgot to mention magnetic fields. The whole process takes place in a strong magnetic field which tends to align spins making these lines easier to detect. #1, this is a gas at low pressure, it doesn't have to be ultra high purity, but that does help. Visual light is used to pump the gas to the particular line in question. So the gas is glowing at that color (line) that contains the hyperfine line in question. #2, YES! This is all about the wave equations for the atoms. But not just the electron cloud but of the nucleus as well. Your second part is correct. The fine lines ARE changed by interactions from one atom to the nearby atoms in the gas. But the hyperfine lines are nucleus to its own electrons which are much closer than the electrons in other atoms due to the low pressure of the gas. That is actually one of design parameters, lower gas pressure give less interaction with other atoms, but fewer atoms to look at! The energy difference between hyperfine lines stays constant, even though temperature, pressure etc are changing the fine structure. #3, Good question, I'm not quite sure. It always seems that when you are using a microwave field to tune for such lines, when you are in tune the field gets lower (atoms absorb the microwaves). I'm not sure what happens with the re-emitted em waves. There are two ways to do the feedback, one has a separate coil of wire going to an amplifier to detect the amplitude of the field for the feedback loop, the other just measures the electrical signal going into the chamber. The electrical signal going in is always what is used to measure the frequency. #4, you bet! That is always one of the big issues with the designs. Really good deigns these days use hydrogen masers for that purpose! (they cost $250,000 and are the size of a refrigerator). The REALLY top of the line ones these days are called "fountains" they let the atoms fall in free space so the interactions between atoms is extremely small. I think we better stop this now, Alex is probably going to get annoyed with this digression from EtherREGEN stuff. John S. manueljenkin, richard_crl032 and lwr 1 2 Link to comment
JohnSwenson Posted January 6, 2021 Share Posted January 6, 2021 4 hours ago, GMG said: The Cybershafts and Afterdark clocks are sine wave (as far as I could gather from the translation) I thought the Etherregen prefers square wave I just got an AfterDark clock that Alex sent me. It doesn't say what model it is, but the output is definitely a sine wave with 1.3V peak to peak. The output is 50 ohms so I loaded it with 50 ohms and used a 50 ohm cable. I have not had a chance to run it with an ER or do phase noise measurements. John S. soares 1 Link to comment
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now