5

I've been curious for awhile now about doing some high speed projects, such as measuring time differences of radio wave reception, and was wondering if there exists components that provide clocks much faster than a typical CPU can, such as up to 10 GHz or higher. The fastest of clocks I have used is the CLOCK_MONOTONIC in linux, accurate to 1 nanosecond. This also makes me wonder why there isn't a clock accurate to the actual speed of my processor, churning along at over twice that speed.

Further expanding on that, what approach is used to create such clocks? And how can these clocks be interfaced with other circuitry? Are there really digital or analog circuits that can operate that fast?

FullmetalEngineer
  • 1,254
  • 13
  • 25
krb686
  • 1,215
  • 13
  • 23
  • possible duplicate of [Where/what functions as a clock on a (say intel i3-7) microprocessor?](http://electronics.stackexchange.com/questions/129869/where-what-functions-as-a-clock-on-a-say-intel-i3-7-microprocessor) – Ignacio Vazquez-Abrams Mar 04 '15 at 03:27
  • [Related answer](http://electronics.stackexchange.com/a/77633/7036) about precision oscillators (not a duplicate). – Nick Alexeev Mar 04 '15 at 03:27
  • 1
    @IgnacioVazquez-Abrams Thanks Ignacio, but I think this question has merit in staying open. Your referenced question focuses on the clock present in `Intel i3/i5/i7` processors, and has a single answer, generically stating *a quartz crystal*. I am specifically interested in clocks faster to much faster than what is present in such processors. – krb686 Mar 04 '15 at 03:31
  • That's what the second sentence of the answer covers. – Ignacio Vazquez-Abrams Mar 04 '15 at 03:36
  • @IgnacioVazquez-Abrams So the 2nd sentence of the answer in your linked question is: *Inside the CPU, there is a device called a Phase Locked Loop (PLL) which is used to lock the output of a voltage controlled oscillator (VCO) into a well-defined relationship with the reference clock.* This suggests to me that PLLs and VCOs have something to do with it, however if I already knew what PLLs/VCOs were, I wouldn't have asked this question. Nowhere in that answer does it specfically suggest or relate PLLs/VCOs to clocks much faster than what is present in a standard processor. – krb686 Mar 04 '15 at 03:38
  • @IgnacioVazquez-Abrams Just so you know, no offense intended at all, however I just hope you are not victim to assuming people know more than they do, if you are very knowledged on the subject. Additionally, your comment now provides some insight or a path to follow to learn about this subject, however were it not for that, if I came across your linked answer on my own, I would have no inclination or reason to think that sufficiently answered my original question whatsoever. Hence this question. – krb686 Mar 04 '15 at 03:40
  • A (even somewhat) complete explanation of what are VCOs and PLLs would be too big for a single answer on this site. But now that you know those terms, you can google and find lots of information about them. – The Photon Mar 04 '15 at 04:40
  • I believe in general these devices are called Clock or Frequency Synthesizers and can usually do modulus/division or multiplication of frequencies, and often have a Phase-Lock Loop involved, as seen in the answers given to this question. – KyranF Mar 04 '15 at 05:37
  • A trick I've seen used in some processors is two or more clocks at the same frequency, but phase shifted with respect to each other. Although the frequency is the same, the resulting combination of edges have a higher resolution. – jippie Mar 04 '15 at 06:36
  • @jippie Do you have a link I can read? – krb686 Mar 04 '15 at 06:37
  • @krb686 is your application for the clock or goal for the project to measure a time interval very accurately? If so I have just the thing, I'll write an answer. – tomnexus Mar 04 '15 at 07:00
  • As @jippie mentions, the technique of combining multiple clocks of the same frequency with different phase offsets is called interleaving. It is used in data converter architectures to achieve higher frequencies, but it suffers from spurs caused by timing, gain, dc offset, and bandwidth mismatches between the different clocks. If you want to read more about it here is a good application note about interleaving ADC clocks: http://www.ti.com/lit/wp/snaa111/snaa111.pdf – FullmetalEngineer Mar 04 '15 at 13:59
  • 3
    You seem to be assuming that a high resolution clock will automatically make your overall system and application highly accurate. This is unlikely to be the case. Not directly related, but here is a reasonably good explanation of [Resolution, Precision, Accuracy, and Stability](https://msdn.microsoft.com/en-us/library/windows/desktop/dn553408%28v=vs.85%29.aspx#resolution__precision__accuracy__and_stability). – Bruno Mar 04 '15 at 15:06
  • CSAC http://tycho.usno.navy.mil/ptti/2011papers/Paper27.pdf There is also a "Quantum Clock" http://www.nature.com/nature/journal/vaop/ncurrent/abs/nature12941.html?message-global=remove these would only add bragging rights to research. – Optionparty Mar 04 '15 at 16:08
  • @Bruno I know a high speed clock won't necessarily make a system accurate, but a high speed clock may be necessary to attain accuracy if high speed measurements are required. Thanks for the link! – krb686 Mar 04 '15 at 16:52

5 Answers5

12

I've been curious for awhile now ... and was wondering if there exists components that provide clocks much faster than a typical CPU can, such as up to 10 GHz or higher.

Opto-electronic Oscillators (OEOs) are oscillators that take a photonic signal, like a pump laser, modulate it, and convert it to an electrical signal using a photodiode. The signals generated by these OEOs have extremely high Q factor and thus very low jitter. Here is a diagram of an OEO, taken from this overview of OEOs. The focus here is on ultra-high stability, not a high frequency output. But, there are also OEOs that achieve high frequencies, for instance this dual-loop OEO achieves a tuning range of 32 to 42.7 GHz.

enter image description here

Besides photonic oscillators, frequency synthesizers can provide clocks above 10 GHz. As other answers have mentioned, these can achieve frequencies way above 10 GHz. For instance, Analog Devices makes a frequency synthesizer that generates frequencies up to 13.6 GHz. In addition, synthesizers generate the ferquencies for a signal generators such as this one, which can reach 67 GHz.


Here's a brief overview of synthesizers if you want to read it.

A synthesizer is composed of a PLL (which contains a VCO), and sometimes also a microcontroller as a means of adjusting the PLL digitally.

Quoting from an Analog Devices tutorial on PLLs:

A phase-locked loop is a feedback system combining a VCO and a phase comparator so connected that the oscillator maintains a constant phase angle relative to a reference signal. Phase-locked loops can be used, for example, to generate stable output high frequency signals from a fixed low-frequency signal.

enter image description here

A VCO (Voltage Controlled Oscillator) is a circuit that generates an output frequency controlled by a tuning voltage. One way to implement a VCO is to apply the tuning voltage to varactors, which adjusts the capacitance of the LC tank in the circuit and generates a different frequency.

Basically, a PLL is used to generate an in-phase multiple of a lower reference frequency. They are used to clock data converters, which can go up to multiple GSPS, and CPUs as well.

Besides PLLs, there are a variety of crystal oscillators (TCXOs, OCXOs, Sapphire Oscillators, GPS disciplined Oscillators, etc.). However, unlike synthesizers, they output a fixed frequency. They are usually designed for ultra-low phase noise and long-term stability, not high output frequencies. Due to these characteristics, they are often used as a reference for PLLs.

FullmetalEngineer
  • 1,254
  • 13
  • 25
5

I'm don't know about high frequencies, but there's a neat set of slides on LeapSecond.com that goes on a Powers of Ten-style journey through various levels of accuracy in timekeeping standards. Here's the list with the accuracy for each item in seconds. Perhaps others can edit this answer to fill in other electronic devices.

  • \$10^{-1}\$ (10%): Human heart beat
  • \$10^{-2}\$ (1%): Tuning fork oscillator
  • \$10^{-3}\$ (0.1%): Precision tuning fork
  • \$10^{-4}\$ (100 ppm): Mechanical oscillator
  • \$10^{-5}\$ (10 ppm): Mains AC electricity (over 1 second, long-term average is better)
  • \$10^{-6}\$ (1 ppm): Quartz watch crystal oscillator
  • \$10^{-7}\$ (100 ppb): 1940s Navy Chronometer (looks like a fancy wall clock)
  • \$10^{-8}\$ (10 ppb): High-tech pendulum clock
  • \$10^{-9}\$ (1 ppb): Earth's rotation
  • \$10^{-10}\$: Oven-controlled crystal oscillator (OCXO)
  • \$10^{-11}\$: Good OCXO
  • \$10^{-12}\$: Excellent OCXO
  • \$10^{-13}\$: Rubidium oscillator
  • \$10^{-14}\$: Cesium oscillator (short-term) or BVA quartz (extreme short-term)
  • \$10^{-15}\$: Hydrogen maser (short-term) or cesium oscillator (long-term)

Of course, accuracy is not the same as precision. The hydrogen maser sounds really exciting until you realize that it only oscillates at 1.4 GHz. An accurate frequency standard is only part of the picture. Also, some of these oscillators only achieve their best performance after a long warm-up period. Some suffer from long-term drift.

Adam Haun
  • 21,331
  • 4
  • 50
  • 91
5

measuring time differences of radio wave reception

This is related to interferometry, isn't it? That tends to be done not so much by measuring the arrival time of signals against a fast stopwatch of some sort, but by measuring phase differences. If you have a 1GHz signal and can measure its phase to within 1%, that's actually more useful than a 10GHz sample clock.

pjc50
  • 46,540
  • 4
  • 64
  • 126
4

there exists components that provide clocks much faster than a typical CPU can, such as up to 10 GHz or higher.

As mentioned in other answers, it's currently possible to get semiconductor VCOs with output frequency into the 20-30 GHz range. For stability, these oscillators typically need to be used in a phase locked loop (PLL) referenced to a high stability crystal oscillator at a lower frequency (50-200 MHz is common).

high speed projects, such as measuring time differences of radio wave reception

Unfortunately, if you want to measure the time difference between two events, a high-frequency oscillator is not necessarily the biggest challenge. Just designing your measurement system so that the clock signal arrives at two measurement circuits with the same delay (or with a known difference in delays) is harder than finding a 20 GHz oscillator. Designing your sample circuits to react to the input stimulus with consistent delay is another challenge.

The Photon
  • 126,425
  • 3
  • 159
  • 304
3

You can buy commercial oscillators that run up to 6+ Ghz without too much trouble.

Higher frequency oscillators can be made, but they generally have to be designed for a specific use, because just the device packaging starts to become problematic at very high frequencies.

In general, these sorts of oscillators aren't that precise, at least by themselves. They're usually employed in something called a phase locked loop, which uses a lower frequency, high-precision oscillator to "discipline" the higher frequency oscillator, by comparing the phase of the two clocks, and using that comparison to modulate the control voltage to the VCO.

It's also possible to multiply a lower frequency up to a higher frequency, which can allow the more complex portions of a oscillator system (the VCO) to run at a more approachable (and testable) frequency, while still having a high-frequency output.


As the frequency gets even high, the oscillator systems become even more exotic:

Connor Wolf
  • 31,938
  • 6
  • 77
  • 137
  • You can also buy higher-frequency VCOs. Macom has catalog models up to 14.4 GHz. Analog (formerly Hittite Microwave) has a catalog model for 23.8-24.8 GHz. Just for example. – The Photon Mar 04 '15 at 04:44
  • @ThePhoton - Yeah, I'm adding stuff at the moment. Hittite has multipliers that go up to 46 GHz. – Connor Wolf Mar 04 '15 at 04:49
  • I was just about to add, "and for frequencies up to 20 GHz with really wide tuning range, there's the YIG oscillator" – The Photon Mar 04 '15 at 04:50