If the OP is lying, the job might be a lot harder, it's true. However, we don't know how far away the transmitters are. If you import a bunch of assumptions that aren't in the problem as given, you can make it as hard as you want.
My parade remains undampened. ;)
The SNR here is good enough to do 100k measurements per second, if the micro has the guts for the job.
Lighting strikes happen at a rate of no more than 0.2 Hz even in a very intense thunderstorm, and average something like 40 Hz, worldwide.
A small amount of sanity checking code ought to be able to toss that out.
The SNR is 50 dB, remember? Those signals will be the kings of the hill.
Nope. 50 dB, remember.
No it won't. You might get some dropouts, but the OP's signals are _big_.
Nah, you just point to the spec, and show where it says "Input CNR at least 50 dB".
As I say, if you change the specs, you can make it arbitrarily hard, but the problem as stated is not difficult at all.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net
I've worked one project where they had a major comeuppance in that respect, that a real intense thunderstorm means no quiet periods at all.
It looked more like in figure 3 here:
formatting link
We've even had storms like that in our (non-tropical) area where the AM radio was in a constant state of loud hash and crackling. It even managed to drown out Rush Limbaugh and that's a strong local transmitter at 1.53MHz. That storm scared the heck out of people. Not so in Puerto Rico where a DC-3 lumbered in right through all the mush.
Ah, so there is a nuclear power station then :-)
But we'd have to know what the OP meant by 50 dB/Hz. To get 50dB without filters would mean over hundreds of kHz, and that is IMHO next to impossible unless the transmitters are in the same town or incredibly powerful.
If the signal really comes in 50dB above anything else the li'l NE602 would completely drown. Those things are ok for garage door openers and stuff but not shortwave, they have no horsepower.
I sort of like NE602s, though you're right, they aren't the strongest mixers in the world. The cross-coupled dual SA614s for I and Q would be a serious possibility if the CNR is really as good as the OP says.
As in most development projects, you stand or fall by the quality of the spec. If the OP goes ahead with this one, Phase 1 will definitely have to be agreeing on a set of specs!
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net
That's the basic problem. The numbers supplied look arbitrary. The numbers needed to make it happen are lacking.
Wait a millisecond... He never specified that the two signals were in any way related. Quoting: So say the two carriers are transmitted from the same location (equal propagation times) but have precisely known different frequencies. Just because they're being transmitted from the same location doesn't mean that they're somehow phase or frequency related.
That was the crux of my rant. If the two signals are not phase or frequency related (asynchronous), then the phase difference (or delay) from any reference point on one signal, to some designated point (probably zero crossing) on the other signal, will change at a rate equal to the differnce between the two frequencies. Because it will be constantly changing, it's useless.
He could do what John Larkin suggested, which is to record the zero crossing of every cycle, and later try to make sense of the numbers. However, that will also produce constantly changing delays.
John Miles contends that all that's necessary is to "convert them to baseband" probably by mixing them down to a lower frequency where an A/D converter can produce more bits in less time and where mixing image problems are minimal. However, the 2nd part of the process "take the phase difference with atan2(Q0,I0) - atan2(Q1,I1)" is over my head. How this process will produce a stable phase difference output is beyond my abilities to decode the suggestion.
That's the instantaneous phase difference at some arbitrary instant in time. Over a longer period, it's a mixer. Assuming a low pass filter on the output, it would produce just the difference between the two RF signals. How does one use the mixed output (something like a sine wave) to produce a single number the represents the phase difference? If the output were DC, as it would if the two input signals were phase or frequency related, one could produce a single number that represents the phase difference. However, with a mixer producing the difference frequency, the phase is constantly changing at the same rate as the mixer output, and is anything but constant.
If you collect a series of instantaneous phase differences, as in a time stamper, each recoreded difference would vary following the difference between frequencies.
I think you just designed the economy time stamper:
I'm sure he would be interested in how it can be done for $20.
--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
Useless to you, maybe, but there are a whole lot of uses of that sort of information. Doppler radar and interferometers, just for a start.
But that's what's actually happening in the system. The frequencies aren't the same, so the phase moves. So what? That doesn't make the measurement ill-defined somehow, it just makes it time-dependent.
It isn't supposed to produce a stable phase difference output, because the phase difference isn't constant. I don't know why that's causing rants and such like.
What? The phase is the phase. It's based on counting, so it's even a Lorentz invariant. Provided you don't lose count, it's perfectly well defined, it's just time dependent.
Assuming a low pass filter
Whoa, there. Do you or don't you agree that my equation gives the phase difference between the two signals as a function of time? So far, this is math and not yet circuits.
Which is what the OP wants to measure, or so I take it. There's nothing in the idea of a phase difference that implies that it has to be constant with time.
Right. Which is why you have to keep doing it often enough that you don't lose any cycles.
What I proposed isn't a time stamper. John's thing associates a number with every single incoming edge, whereas mine is more like a digital-IF radio.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net
Yeah, but over all my consulting years and even the ones as an employee I've learned on thing: Never, ever trust specs 100%. Always ask, always assume it gets worse. For example, I can't even tell you home many times I have discovered a major boo-boo on "professional" RF circuits, a missing inductor to ground as the very first part after the antenna. "So, what's going to happen if lightning strikes a tall lamp post 100ft away?" ... "Ahm, well, you know, oh ... oh dang!"
Guys, your points are all excellent, and I've learned a lot from reading th= em. Didn't mean to be cryptic in my OP. =20
Our propagation is all ground wave; we can rule out skywave externally. We= know carrier frequencies accurately from other near-real time information = sources. We believe propagation paths (ducting) will be stable enough that= phase drift short-term will not be significant.=20
After re-checking my link budget, the nominal C/N is 70 dB in assumed pre-d= etection BW of 50 Hz. Our noise BW may be better than 50 Hz using a PLL de= tector, but on the other side of the coin we will have to tolerate fading. = Also, this C/N includes 50 dB allowance for atmospheric noise above the kT= B noise.
It's interesting that we don't care much about antenna gain, because G/T do= esn't matter if you're swamped by external noise.
Impulse noise is terrible in this band, so I believe we would use slow PLL = detectors and average multiple measurements. =20
A PLL detector will easily lose lock or show excessive noise if you only rely on loop filtering. How bad it'll get will depend on how well your signal stacks up in comparison with other carriers and signals around it, including static.
Fading would somehow mean that there are multiple paths with non-constant attenuation. Sure there is no skywave involved?
If this noise is stronger than 20-30dB below your signal then I'd really consider narrow filters in front of the PLL.
know carrier frequencies accurately from other near-real time information sources. We believe propagation paths (ducting) will be stable enough that phase drift short-term will not be significant.
pre-detection BW of 50 Hz. Our noise BW may be better than 50 Hz using a PLL detector, but on the other side of the coin we will have to tolerate fading. Also, this C/N includes 50 dB allowance for atmospheric noise above the kTB noise.
doesn't matter if you're swamped by external noise.
detectors and average multiple measurements.
That's a reasonable method too, with a few tweaks. ;)
Since the inpulses can be much larger than the signal, I'd want to run a noise blanker ahead of the narrow filter, and use a PLL with a 90 degree phase detector such as a diode bridge or multiplier. That way you don't have to wait for the filter to settle down after the impulse, and turning off the signal doesn't make your PLL immediately lose lock, as a phase-frequency detector would. All still cheap stuff, though as Joerg says, custom crystals can run into a bit of dough.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net
A noise blanker is, as you always say, good medicine. The ham radio community is full of recipes for that. Because they use SSB transmission which is very sensitive to impulse type noise.
If the quantity is high enough crystal prices come down fast. Another trick to push cost down is this: Mix both received signals down to an IF where you can get cheap off-the-shelf crystals or other kinds of filters. In the olden days that were the crystals for the NTSC and PAL carrier frequencies which could be had for a few dimes. Nowadays 8MHz uC crystals are cheap. Also 27MHz which usually are 9MHz crystals run in overtone (which you don't have to).
If you have IEEE library access, try looking at Grove et al.'s paper "Direct Digital Phase Noise Measurement" for a better explanation.
The phase difference isn't stable -- otherwise it wouldn't be of much interest. But if it varies slowly enough -- which it will if you know the carrier frequencies and use that knowledge to convert the two carriers to a common IF -- then it can be measured at extremely high precision within a correspondingly narrow bandwidth. The Grove paper is concerned with broadband measurement using an FFT, but the test set they built also does Allan deviation measurement with band-limited I/Q data.
As in that case, the job that George describes could likely be handled entirely on the digital side, using an ADC at the antenna followed by a couple of DDS cores on an FPGA. The basic technique yields good results down to the tens of femtoseconds when used to measure a strong signal from an OCXO or something like that. Off-air measurement won't be as good, of course, but it sounds like there would be plenty of margin.
expect to sub-contract the design work but looking for an opinion first = about practicality.=20
range, and report their relative phase with at least 50 nsec accuracy.
advance and are assumed to be perfectly stable. Nominal C/No >50 dB/Hz = each carrier.=20
want to know is how much one leads or lags the other in time, like = measuring the relative zero-crossing times of the two sinusoids. For = measurement accuracy it is acceptable to time-average each measurement = over as long as 100 msec.
assuming this concept is practical. Any recommendations for someone to = do the design are appreciated. We are located in Southern Calif.
Well, whether anybody else says so or not i have say that phase is a null concept for different carrier frequencies. Perhaps there can be a definition iff they are rationally related frequencies and you can = provide some definition of phase for different frequencies.
I expect to sub-contract the design work but looking for an opinion first= about practicality.
2.5 MHz range, and report their relative phase with at least 50 nsec = accuracy.
advance and are assumed to be perfectly stable. Nominal C/No >50 dB/Hz = each carrier.
really want to know is how much one leads or lags the other in time, like= measuring the relative zero-crossing times of the two sinusoids. For = measurement accuracy it is acceptable to time-average each measurement = over as long as 100 msec.
assuming this concept is practical. Any recommendations for someone to = do the design are appreciated. We are located in Southern Calif.
dB,
MHz
just
a curve fit of the difference tone amplitude against time? How do we get= relative phase of the two RF carriers from the curve fit?
So you would measuring the minimum zero crossing phase between the two signals???
I expect to sub-contract the design work but looking for an opinion first= about practicality.
2.5 MHz range, and report their relative phase with at least 50 nsec = accuracy.
advance and are assumed to be perfectly stable. Nominal C/No >50 dB/Hz = each carrier.
really want to know is how much one leads or lags the other in time, like= measuring the relative zero-crossing times of the two sinusoids. For = measurement accuracy it is acceptable to time-average each measurement = over as long as 100 msec.
assuming this concept is practical. Any recommendations for someone to = do the design are appreciated. We are located in Southern Calif.
dB,
MHz
just
do a curve fit of the difference tone amplitude against time? How do we = get relative phase of the two RF carriers from the curve fit?
if
and
You
is
and
In a sense measuring the phases noise of the beat signal?
It may help if you stop thinking of them as two different frequencies. If you have an FM receiver, you can still recover the audio with a discriminator after downconversion to a fixed IF, right? Changes in the carrier phase will survive the downconversion process because those changes are ultimately just time differences.
In the OP's case, he can downconvert each of the two signals to a nominal IF of 0 Hz, using both sin and cos DDS outputs to generate complex baseband representations. Now the two signals are at the same "frequency," namely DC. Since they are complex-valued signals, their phase in radians can be obtained with the arctan function.
The key is to understand that phase _differences_ are what George is after. Presumably the two signals' phases will change over time, so you can take the differences between successive phase readings in each channel to obtain a pair of phase slopes (which can also be thought of as each channel's frequency deviation from 0 Hz). At any given moment, the difference between the two phase slopes represents the phase difference between the two channels.
expect to sub-contract the design work but looking for an opinion first about practicality.
range, and report their relative phase with at least 50 nsec accuracy.
advance and are assumed to be perfectly stable. Nominal C/No >50 dB/Hz each carrier.
want to know is how much one leads or lags the other in time, like measuring the relative zero-crossing times of the two sinusoids. For measurement accuracy it is acceptable to time-average each measurement over as long as 100 msec.
assuming this concept is practical. Any recommendations for someone to do the design are appreciated. We are located in Southern Calif.
curve fit of the difference tone amplitude against time? How do we get relative phase of the two RF carriers from the curve fit?
I don't know what that means.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.