1
\$\begingroup\$

Bigger picture:

What do I need to do? I need to send a radio signal and have another device receive that and output a voltage pulse, with the time between starting to send the signal and starting to output the voltage pulse as consistent as possible every time I do that (zero variation is perfect, picosecond or less is good, nanosecond is bad, millisecond probably means I'm wasting my time).

I need to know if it can be done, and if so what the likely timing inconsistency will be, whether due to circuits, or interference, or other - I need the order of magnitude of the inconsistency rather than an exact figure, though exact or a formula would be good.

Note:

  1. The frequency can be any that is listed as 'Amateur' in https://www.ntia.doc.gov/page/2011/united-states-frequency-allocation-chart
  2. While the response time could be slow, faster would be better.
  3. I believe I need that to happen with an analogue AM signal, and no digital clock in this part of the device (as those will add inconsistency in timing based on the digital clock's resolution, and/or require expensive or non-existent ultra fast clocks). I may be wrong about this - I figured I'd find out if it could be done like this, and if not then I would look for other, more complicated, ways that break these two requirements.

Original Question:

Imagine I have an analog AM radio transmitter, frequency 222MHz, which starts transmitting a full strength signal at time 0s, and have a standard analog AM receiver 100 meters away (tuned to 222MHz).

The receiver's output is passed (possibly via an amplifier) to a digital converter circuit that normally outputs 0V but, when the received signal passes (at least) 80% of the transmitted signal strength, outputs 5V. Note: the converter does not use a digital clock or timing signal.

My first question is this: I've been told that the time from the transmission starting (0s) to the time when the converter first outputs 5V (T seconds) will not be a consistent amount of time, because the received wave will not be exactly the same as the transmission - that interference (natural or artificial) will make the received wave have small perturbations that will subtly change the exact timing of when the receiver's input/output hits 80%, making that time different each time. While the effect will be small, it will be non-zero. Is that correct? Can anyone tell me roughly what the effect will be (in terms of time variation)?

My second question is this: is there some kind of circuit or other thing I can do to avoid this? I really need that timing to be entirely consistent (and I mean, down to picoseconds or less if possible).

I don't the circuits laid out yet, but I imagine the transmitter would be along the lines of https://www.electroschematics.com/2522/am-transmitter-circuit/ or http://www.circuitstoday.com/am-transmitter-circuit , and the receiver would be something like https://www.electroschematics.com/601/2-transistor-radio-receiver/ or https://www.electroschematics.com/9043/am-receiver-circuit/ - and I don't have the digital converter circuit but if I recall my electronics lessons of long ago, that's a pretty standard, simple thing.

EDIT: The key here is consistency, over multiple transmissions, within a short period of time (seconds, not hours). I don't mind delay if it is a consistent delay. I should also note that, for reasons that it breaks the basic concept, I can't use digital signals or a clock in the signal analysis (because the clock would add inconsistent delays). Optical or cables are also out.

I have asked the same question in Physics as there is some overlap and I welcome the different viewpoints - https://physics.stackexchange.com/questions/408378/consistent-radio-wave-detection-timing

Please note, this is related to https://physics.stackexchange.com/questions/401512/radio-wave-detection-timing but is not the same question - though I asked a version of the above in the comments.

Thanks in advance.

XY Question:

To a degree you are right. It probably doesn't help that I'm used to the programming stack exchange where a specific, limited question is generally more useful than a vague 'big picture' one, up to a point. However that doesn't mean that this question has no value, nor that I don't need an answer.

1) this is an attempt to do something a new way. So there are existing ways to achieve it but they are not very good and I don't want to replicate them. Getting back to the even bigger question than the above just leaves me with existing methods, and probably kicks it in to whole other areas (physics, for one). If Einstein had asked about gravity in certain circumstances, would you refer him to Newton's Laws? Or would you answer his seemingly XY question? (I'm no Einstein but the point is still valid)

2) there is a limit to how much more information I can give away without laying out my entire idea. Call me selfish but I'd like to avoid that for now.

3) when I asked a more general question I was basically told to be more specific. This was my attempt to do so.

\$\endgroup\$
  • \$\begingroup\$ you want to have Picosecond accuracy with a signal that has a period of 4500 Picoseconds ..... maybe you should use an optical method for your signal transmission \$\endgroup\$ – jsotola May 27 '18 at 2:25
  • 1
    \$\begingroup\$ It sounds like you are describing what is called jitter on the received signal. You can use a PLL to remove jitter. Depending on the PLL and on the characteristics of the transmitted signal, you could potentially get ps level jitter of a local digital clock that has been PLL locked to the transmitted 222MHz carrier. \$\endgroup\$ – crj11 May 27 '18 at 2:54
  • \$\begingroup\$ Hi jsotola, unfortunately, optical is not really an option. I wish is was! \$\endgroup\$ – James Carlyle-Clarke May 27 '18 at 3:25
  • \$\begingroup\$ Hi crj11, I've edited the original question to clarify that I am unable to use a digital clock in these parts of the circuitry. \$\endgroup\$ – James Carlyle-Clarke May 27 '18 at 3:26
  • 3
    \$\begingroup\$ Smells like an XY problem. Please describe what you are trying to achieve functionally without mentioning how you want to acheive it. \$\endgroup\$ – Andy aka May 27 '18 at 9:55
2
\$\begingroup\$

Pico-second level timing is hard when sending signals through air.

At RF frequencies, the index of refraction of air is about 1.003, corresponding to an additional delay (w.r.t. vacuum) of about a nanosecond. But that can easily change due to air temperature, pressure and humidity. Getting the propagation time stable to 50 psec over 100m air would be tough.

An on/off AM signal is unlikely to reach close to that, though. It’s very hard to get a stable leading edge by just powering up an oscillator.

If you really need precision, consider a stabilized oscillator emitting continuous wave signal that you can phase-lock at the far end. Then you can use a lower-precision signal, like your AM pulse, to convey “use the next zero-crossing of the precision signal”.

But, if you can do it, the simplest approach might be a digital timing signal via fiber.

CERN knows a lot about how to this at multiple levels of accuracy. See this page for an overview. There’s also a branch of geodesy that’s been doing work at this level, for example in this Metrologia paper.

If you edit your question with more info on your requirements (or add a comment) there are probably papers we can find that would help.

Once you have a consistent time reference between the two points, you still have to pass the arbitrary event time (assuming you can’t stimulate that from the common clock). See the Metrologia paper for discussion.

One approach is to record the \$\Delta t\$ since a reference, i.e. a zero-crossing, and pass that to the other end. You can do that digitally or in analog. Precision voltage-to-time circuits are available (though there’s a practical limit to the accuracy in t-V-t due to noise, etc).

\$\endgroup\$
  • \$\begingroup\$ Hi Bob, you've given me a lot too look at, thanks for that. The key here is consistency. I don't mind delay if it is a consistent delay. I should also note that, for reasons that it breaks the basic concept, I can't use digital signals or a clock in the signal analysis (because the clock would add inconsistent delays). I'll look over what you've said and see if I can add more. \$\endgroup\$ – James Carlyle-Clarke May 27 '18 at 3:25
2
\$\begingroup\$

What you are describing is the transmission of a signal through a channel with noise. It's a fundamental problem, and a lot of research has been done on it.

The timing variation at the receiver is a function of your signal to noise ratio (SNR) in your processing bandwidth. You need to consider :-

a) The strength of the received signal
b) The strength of the received noise
c) The receiver processing bandwidth

So you could ...

  • increase transmitter power, limited by funds and local regulations
  • lose less signal, low loss antenna cables (especially receive cable)
  • use directional antennae, limited by size or need for omnidirectional operation
  • choose an operating frequency where there's less environmental noise
  • choose a channel where there's less signal loss or less susceptibility to environmental noise (optical versus radio for instance)
  • use a low noise receiver, more $$$ allows you to approach the inviolable thermal limit
  • don't throw signal away once you've got it, use a coherent receiver rather than envelope detection

And finally, the most subtle consideration is the receiver bandwidth. Here you can trade off latency, how long the receiver takes to respond to changes in the signal, against noise bandwidth. Thermal noise and much environmental noise is broad spectrum, a wider bandwidth receives more noise power, so degrades the SNR.

In the worst case scenario (1), your signals come at random, and your receiver must output within (say) 10uS of getting an input. You need a processing bandwidth of the order of 100kHz, and will not be able to reduce it.

In the next better case (2), your signals still come at random, but you are allowed (say) 1 second before outputting. This allows you to integrate the signal coherently for a while before making the decision, which reduces your effective noise bandwidth, in this case to the order of 1Hz. You can make this easier for the receiver by designing your transmitted signal for easy detection, perhaps by using chirp or other spread spectrum modulation.

In the best possible case (3), your signals come regularly. You can then put a low jitter clock at your receiver, and adjust its timing slightly as each signal is received. This is how remote timing in the cellular network operates. Of course here if the transmitter makes an alteration to its timing, it will take many signals before the receive clock has adjusted completely to the new timing, it could take minutes. But the tradeoff is if it takes 100s to adjust to the new timing, your noise bandwidth is now in the order of 0.01Hz.

\$\endgroup\$
0
\$\begingroup\$

For 1 nanosecond timing resolution, you'll need about 1GigaHertz bandwidth. Your antenna will not support that bandwidth.

\$\endgroup\$
  • \$\begingroup\$ I don't know whether this would affect your answer, but please note that I don't need to time the transmission and response time to one nanosecond - I need the transmission and response time to be consistent for each time it is used, that consistency to be within 1 nanosecond or less (preferably less). \$\endgroup\$ – James Carlyle-Clarke May 28 '18 at 9:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.