Hi, Im looking for a long delay line of about 10ms wich can handle a 10mhz sinewave. size is not too important as long as its able to be mobile, a quick calculation reveals I would need about 2000km of cable, wich means its probably far too expensive to do it this way, although I only need 1. One idea I had was to re circulate the signal many times shifting it up in frequency each time, im not sure if this could be done with just one mixer, any ideas ?
other methods of delay such as digisation etc are not suitable.
--
Perhaps convert it to mechanical energy with a piezo transducer,
then run it through something with a low velocity of propagation of
sound and then convert it back to electrical at the other end with
another piezo transducer?
Why can\'t you do it digitally?
Ponder John's suggestion of acoustic transfer. That's how they do the
64usec delay lines on PAL system color TV sets. Ok, you probably don't want to rig 156 of those in series to get your 10msec but it still sounds better than dealing with several flat-bed trucks full with huge cable drums pulling up.
The mixer thing sounds too esoteric. At the end you'd be looking with a magnifier for your signal.
Ten milliseconds is a long time for acoustics at 10 MHz. That's a time-bandwidth product of 10**5, which is a factor of 10 higher than you can usually get, at least in my experience. Digitize the signal and use a big long FIFO buffer--at a 40 MHz sampling rate, 10 ms is only
I had thought of lots of TV delay lines as well lol, as for the cable idea I was thinking more of the type that has lots of very fine wire wound round a ceramic tube with ferrite/copper segments. A big reel of .1mm EC wire would go a long way, I might ony need a few hundred of them ... maybe wind several layers to save having 1000s of such tubes, maybe have to use twisted pair to avoid crosstalk ...
If the delay*bandwidth product is 10 too high maybe I could get away with
1mhz or 1ms, if the bandwidth of the 10mhz signal is only a few hz does that make it easier ?
I cant realy process the signal in anyway like downconvert it or digitise it, as this would not be accurate enough, the idea is to delay it long enough and compare it with the source so any cyclic phase shift shows up, so any such processing would obviate the point of the delay line, also it is on a moving platform.
Im still thinking about using a vector multiplier for a mixer and re circulating it, amplifying it as its recirculated of course so I dont end up looking for a needle in a haystack, might be a nightmare trying to analyse what happens to all the spurious products as they get re circulated and re multiplied together too etc ...
Wouldn't the inductive reactance at 10MHz be multi megaohms?
Your accuracy is limited by the accuracy of the 10ms delay anyway. If it were really made of miles of wire the thermal expansion would probably destroy all the information.
If you want to know phase now vs phase 10ms ago, couldn't you just run the input into two PLLs with different loop bandwidths?
there is a form of tank depth gauge which uses very fine wire wound onto a 2" or so former. Off hand I can't recall the ping time we got from 10 feet of it, but it was much slowed down.
your problem is the 10MHz I'd think - digital would be a much better route, but 50ns samples means hard-wired logic not a processor.
Hi thanks, well although it looks a bit like an inductor the way its wound, it acts like a transmision line so the signal is only attenuated by the losses in the copper/dielectric.
digitise
up,
as long as the expansion isnt cyclic at >1hz it would be cancelled out.
I think you just need 1 PLL with a loop time constant of >10ms, but the ultimatly the VCO limits the performance.
with an accelerating platform XO become rather noisy over a 10ms period. I have a rubidium clock but it is affected by magnetic field and would require significant more sheilding, Im not sure if hydrogen masers would be better but I was looking for an alternative.
Interesting using a delay line as a depth guage, often they are wound on a former with ferrite/copper inserts wich increas ethe propogation delay. like I said, unfortunatly digitisation would not be suitable,
Colin: These days you can easily digitize to 16 bits at >25MSPS if needed. It doesn't cost and arm and a leg anymore. Same for memory, thanks to a few dog fights among RAM manufacturers and MS programs slurping ever more megabytes. What's the problem with digital in this case?
If the delay can be a bit approximate, a 2 ohm resistor will do. It has a conductance of half a mho.
--
OOO - Own Opinions Only. Try www.jmwa.demon.co.uk and www.isce.org.uk
2006 is YMMVI- Your mileage may vary immensely.
John Woodgate, J M Woodgate and Associates, Rayleigh, Essex UK
yes it is a bit of a killer ... I hooked up a tv delay line and it didnt look to pretty at the other end, but nevertheless recoverable. even 50db loss isnt too bad to recover in one go.
As I clearly stated digitising is out, digitising the signal would serve no purpose at all, it would basicaly be like comparing the signal to the digitising reference frequency wich doesnt then even need to be delayed to do this, and its simpler to compare it with a reference frequency directly.
But the objective is to get rid of the need for a reference frequency altogether as this is a very limiting factor.
however looks like this delay line is a non starter, but id thought id ask anyway, you never know what you dont know !
maybe il just pursue the 2 rubidium clock idea some more. how much sheidling would you need to reduce the earths magnetic field by a factor of say 10 or 100 ?
so you're wanting to find a delay line with a propogation delay that's more stable than a crystal oscilator?
I don't know if such even exist.
maybe you just need a better clock?
If delay lines were more stable then crystals people would use them for frequency reference, but the only place I see that done is in analogue musical instruments.
Well it only needs to be more stable averaged repeatedly over 10ms intervals. most oscillators are prone to 1/f noise wich seems to be most pronounced at
10hz they are also suceptable to aceleration, I am assuming delay lines dont suffer from 1/f noise like conducting semiconductor junctions do, although I know they do suffer from microphonics quite badly and also very sensitive thermaly its ok as long as it doesnt vary cyclicly at ~10hz.
I have quite good OCXOs at the moment, ive tried quite a few different types of them. they work fine when stationary. just tilting a few degrees is enough to swamp the result. atomic clocks still have an OXCO as a slave oscillator and so are only marginally better over a 10ms interval plus they are susceptable to magnetic fields.
a reference clock will drift a bit over a 10ms interval, although it can be in a PLL the loop response needs to be over 10ms or else it is just a duplicate of the original and serves no purpose.
I think the reason they are not used is simply that crystals are so cheap and easy and Q is so much higher. there are delay line oscillators at high frequencies, SAW devices wich make use of propogation delay make very low noise oscillators.
Why not a string of inductors and capacitors? That's what a transmission line is, after all - it's just squashed together. If that'd be too lossy, you could build another alongside it, where the nodes match the input impedance of a small buffer, whose output impedance matches its node of the output delay line. That's called a "distributed amplifier". Very broad-band, but I imagine rather noisy.
hmm lets see how many stages, assuming 180' per stage thats 200,000 stages, its very narow band so that might help, assuming pcb capacitance can be used, assuming smallest inductors aranged on 1mm^2 grid thats about 500mmx500mm pcb inductors might be < £0.01 in volume = £2000, not as far out as one might of thought.
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.