Hi, Im looking at using a parametric method for a phase detector, The source (~4mhz) is already very strong but the sideband, @ 1-10hz, is extremely low, I am considering using a quartz crystal as the most sensitive slope detector but to get a significant signal to noise ratio I would need a crystal with a Q of over a million. I dont expect acheiving such a high Q is practical, unless cooling it drasticaly would help ?
The main noise problem is 1/f noise especialy as a high Q circuit would need a high impedance amplifier and a mosfet would have too high 1/f noise at
I understand that you intend to apply a 4MHz signal with low sidebands to a high Q crystal offset slightly so that pahse modulation will appear as amplitude modulation and then perhaps employ a low frequency speactrum analyser to look at the phase noise. Is this the idea? If so: A good quality 4MHZ crystal wcan have a Q approaching 1000000. All depends on the diameter/contour surface polish and the quality of quartz used. The crystal could be operated at its series resonance and terminated in
well it is actualy a modulation signal im trying to detect rather than just noise but it is very similar problem, of course the phase noise of the reference oscillator is a big issue but I intend to try and null most of that out I hope.
I'l go looking for some nice shiny quartz rocks... I chose 4mhz as it seemed they had the lowest series resistance for the frequency coresponding to highest Q although manafacturers dont list Q on a per frequency basis, its generally 20k for off the shelf types, im not sure if its advantagous to use an overtone or not. Im not sure if i can get any better ones, but the ones i bought seem to demonstrate a higher Q than anticipated from the simulation.
aha, I have looked at using it in parrellel mode trying to use the considerably high voltage generated from resonance to improve snr before detection, although I was woried about destroying the crystal I can easily see 50v pk-pk. I will experiment with the series mode in ltspice, the problem with most
1nv/hz^-2 amplifiers is 1/f noise is stil quite high as much as 1uv total below 10hz, not that many manafacturers seem to specify noise below 10hz, but certainly it has meant i have had to rethink things, I was hoping a parametric amp would avoid the 1/f problem unless i am mistaken here ? I have managed to get significant gain using the reactance of a varactor in a tuned circuit, but there still the problem of the preceeding detector diode.
so could a 400mhz cavity have a higher Q ?
Ive seen test aplication where the same type of oscillator is used as the reference if it is a VCXO of course. Ive tried using a PLL with a OCXO but this proved disapointing, of course one never fuly knows if its the topology used or just unlucky construction with such low snr, from what I can gather I think 1/f noise can be degraded with handling/soldering etc.
Intrinsic acoustic loss is per cycle so ltimate q reduces from about
10000000 at 1 MHz to 100000 at 100MHz. Modern crystals use grown quartz which can be much lower Q however the best grown approaches the ultimate. A 5 MHz 5th overtone will have a q of 1 to 2oooooo but cannot be driven hard.
I looked at using series resoance with a 50ohm resistor but the much lower voltage means a much smaler signal even tho the slope is higher, I did consider a curent transformer aproach but think it wpould be pushing things a bit to actualy get more SNR.
I looked at auto zero amplifiers and found they auto cancel the 1/f noise along with the dc offest, the AD8552 looks superb, very low voltage noise
1-10hz and very low input curent/noise too, so i can use a highly tuned high impedance/high voltage section to give maximum signal, so i got one of these now, (I already had one in my parts box !) and boarding it up now.
I couldnt find much on error multipliers, maybe il look some more ...
I looked and did not find, the technique was to multiply a 1MHz test signal to 10MHz mix with 9 MHz clean reference filter the resultant 1MHz and multiply again to 10MHz mix with the reference again can be done 4
0r 5 times.with increasing sidebands the exact opposite to dividing. Various manufactrurers in the 60s
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.