simple frequecny multiplier

wich im

harder to

maybe the

further

The advice Mark gave to use a PLL is pretty sound. What is the phase noise requirement? You could end up with lower phase noise using a PLL than with a multiplier. If you do use a multiplier, consider a step recovery diode if you're trying to do it in one go. Otherwise, consider a two or three stage multiplier.

The best you can do with a 25x multipler is 25x the phase noise offset of the source. The phase noise of the PLL will be dominated by the VCO if you do it right, and that is determined mostly by the loaded Q.

see the following text: Microwave Circuit Design Using Linear and Nonlinear Techniques by Vendelin, Pavio, and Rohde.

Frank Raffaeli

One interesting way to do it would be to use a step-recovery diode to generate harmonics, and use a single high-Q resonator - preferable a coaxial ceramic resonator - to pluck out the 25th. It would only take a few parts. But anything like this takes equipment to evaluate. You really ought to have a several-GHz sampling scope or a spectrum analyzer to do this... this sort of stuff is tough flying blind.

John

Why are picosecond delays interesting here?

If you're trying to characterize the phase change of a 40 MHz signal, multipling may not make things better and could well make things a lot worse. Measuring anything analog to one part in 25,000 will always be tricky.

What's the application?

John

Hi,

I need to multiply the 40mhz output from my nice CMAC TCVCXO up to 1ghz, ie multiply by 25, to feed a gilbert cell mixer.

I know this should be fairly simple and i know the throey of operation but the problem is I wil be doing it blind as i do not have the equipment to do much testing above 100mhz.

already i have a simple test circuit wich takes the output directly and feeds it into a C-C-LC tank and amplifies it with a rf dual gate mosfet bf988 wich drives a coax cable. runing at a low 3v at 3x multiplication i get 0.2 v pk -pk at 120mhz wich is ample and is probably twice as high as this as my scope is only 100mhz bw, and at 5x multiplication i can just detect the 200mhz signal on my scope at about 30mv but ive no real idea how much atenuated it is.

However i am not sure what would be best to provide the next x5 of multiplication, i was thinking of simply putting a C-L-C 1ghz tuned circuit in the source of the mosfet and adding an emiter folower and hoping the mosfet is driven hard enough at 200mhz to give enough 5th harmonics.

As my ability to test it is limited to a RF sniffer probe i would be very grateful if anyone could share anything that is tried and tested. Im wondering if it would be beter to use 1 or 2 cmos inverters in place of the mosfet/emiter folower.

google search throws up so many hits but all i looked at so far arnt that helpfull.

COlin =^.^=

i dont see how it'l be better, it can only add to the phase noise wich im trying hard to avoid, and lot more complication wich is probably harder to debug then a single stage distorting amplifier with tuned output. maybe the fundamental wil be beter supresed but this wil be easy to supres further with another filter.

Colin.

thanks il look into that, arnt SRD hard to get now? i was not trying to do it in one go as such the harmonics from the xtal oscilator contain suficent

subsequent tuned filter would extract the next 5th harmonic of that asuming it it was distorting enough. so it is a 5x 5x rather then 25x but i just wont be able to see the results with my scope. the frequency is fairly flexible i could use more stages with less multiplication. say 3x 3x 2x

i had asumed a good crystal oscilator would outperform a LC type of vco, and i know PLL can wander slightly as they can have dead spots in the edge detector, i hadnt considered the efect of multiplying on the phase noise, i asumed the Q of the tank circuits will have the same efect as the Q of the PLL VCO ? the jitter/phase noise/stability of the oscilator im using is extremly good.

basicaly its one of those projects where it needs to be as good as i can get without going to extremes (atomic clocks are too expensive) and the phase noise is integrated over 100ms or so. but nevertheless needs to be ridiculously small. the reason for multiplying the frequency is to increase the sensitivity of the phase detector to phase changes in the 40mhz signal to detect phase change coresponding to changes in delays of less than a picosecond.

my albeit limited experience of PLL is that it can sometimes take quite a while to get them to work as well as you hope if you are trying for such small jitter.

im not sure exactly of the nature of phase noise and what hapens when you integrate it over such a long period, or how to compare the integreated noise from a multiplied XO compared to a PLL VCO of a higher frequency.

idealy of course i would like a 1ghz crystal oscilator but ...

Colin =^.^=

hmm yes thats interesting, id not heard of one of those resonators, maybe i can just use a length of transmision line/coax instead, shldnt be too long at 1ghz lets see asuming V = 0.7C ... 1/4 Y = 75mm. short circuited at the end i asume.

yes i agree, hence i was hoping for anything tried and tested.

a good chalenge tho ..

thnks for the help

Colin =^.^=

its for an experiment trying detect any relative change in speed of light.

im using the latest VCTCXO from cmac cfpt-9000 series wich are suposed to be very good, certainly from stability anyway although they dont quote phase noise/jitter explicitly, i think its suposed to be one of the lowest, certainly stability is very good. maybe better over short term than an ocxo as it operates at lower temperature and has no thermal feedback loop. i beleive you can also now get these as a oven controled version too. its useful to know the figures you mentioned tho.

i think it might be posible, even if it means integrating over many cycles, although this isnt something i have planed for at this stage.

yes wich is why i was reluctant to go with a PLL i assumed a humble multiplier wouldnt add much jitter at all over 100ms. the voltage output from the phase detector would be very small at 40mhz and hence suceptable to noise at that point. maybe a more moderate multiplaction of x3/x5 that i can easily acheive now would be more optimal, il probably try that for now but leave room for the x25 circuit, or maybe try x3 x3.

there realy isnt a lot of circuitry between the frequency generators and the phase comparator, heck there isnt actualy any ! wel apart from a length of coax/trasnmison line wich i hope isnt going to misbehave. the phase comparator would be a gilbert cell aranged to give a null output on matching phase i think its aranged like this .. (A-B) X J(A+B). the two clocks would be synchronized within a PLL with an exceedingly long response time so as not to cancel out the result.

Colin =^.^=

TCXOs have a temperature transient problem: the temp comp sensor never has the same thermal time constant as the crystal itself, so whereas the compensation averages very good, a millikelvin delta-t over 100 ms can cause a lot of phase shift. A good OCXO will have a huge thermal isolation system and also operates the crystal at its "turning temperature" where the inherent TC is zero. All that makes a huge difference in close-in phase noise. If you use a TCXO, put it in a heavy aluminum can to slow down temperature transients; that alone can cut thermally-induced phase noise 10:1.

A good SC-cut OCXO is a few hundred dollars new and is the best you can do without going atomic. You can get a used rubidium for about the same, but I'm not sure the short-term stability is necessarily better than the SC.

Considered an optical interferometer?

John

Any practical transmission line will have a bad temperature-frequency coefficient. Coaxial cable and pcb traces are usually rotten. So tiny temperature shifts will cause big (by your standards) phase shifts.

What's the physics here? Can you just use baseband (40 MHz) processing along with a lock-in sort of technique to correct for noise and drift?

John

A tuned filter is a rather hard to achieve with almost no tools. This frequency range you intend to work in is the crossover. It is above where discrete L & C behave as they should, and below the range where microstripline is feasible. I was calculating a 1 GHz stripline filter recently and the achievable Q is not that brathtaking.

Rene

I'd recommend borrowing a ready synthesizer with a low phase noise 1GHz output. You'll never get below a

Rene

so you are trying to measue the phase of the 40 MHz and you are multiplying it up to increase the gain or accuracy of the measurment?

so you don't really care about 1 GHz, but just want to measure the phase of the 40 MHz?

Why 40 MHz then, why not start out with a 1 GHz or whatever oscilator?

Mark

If the problem is to measure a path length by timing the speed of light, any mechanical distortions of the path make the same noise, whatever the detector mechanism.

Is this a ring gyro? Lots of work has been done on them! And the detectors are usually optical interferance.

John

I need to replace a piston in my Ford engine. But all I have is a beer can and a used ice cream stick. Can someone please provide a ready-made solution?

For any endeavor, there's some minimum set of tools required. (props to McGuyver not withstanding) We can argue all day about what that minimum is for this case, but I don't think your sniffer is gonna hack it. You need to buy something already configured and tested...or borrow some tools. mike

I gues ive been looking at to much of the sales blurb, .. yes a good thermal slug and wind proofing might be a good idea, also the device is basicaly on its own exept for decoupling and is lightly loaded. its also all on a single SMD chip.

i had considered OCXO or puting a peltier device to keep it at constant temp, maybe even at its lower temp turning point. even the idea of having a temp sensitive oscilator as a sensor and using a PLL to keep the temp constant. however low power batery operation would make things a lot easier. however thinking about it i might wel have to resort to a ocxo. the whole thing should fit into an aluminium or steel tube.

lots of things i havnt realy considered seem to be poping up i gues il just have to try it now and see just how much everything adds up. theres not actualy a great deal to it.

yes i am thinking about that, however i cant simply use the same light source and bounce it back to meet the original as this would cuase an imeasurable result, the Sagnac effect is measured with a single source, but i would need 2 lasers and be able to tune one of them and keep it in synch with the other one. im not sure how feasable this is. maybe two closly matched laser diodes cld be kept in synch by controling the temperature, i gues they must have some temperature sensitivity.

Colin =^.^=

im realy relying on the whole device remaining at a constant temperature at least throughout each cycle, it only needs to be operated at standard room temperature.

the upshifting is purly to increase the sensitvity of the phase detector to any time diference so its a compromise of phase noise to detector noise.

the cycle time is less than 1 second, averaging over as many cycles as necessary is posible and probably is necessary, but would be realy nice not to have to.

of course any mechanical flexing would severly afect the result, and is probably my bigest concern as it might flex more at one point in the cycle. (the whole thing rotates at a slowish speed).

Colin =^.^=

Ok, so we're gonna argue all day about it. How do you propose to determine the frequency coming out of your multiplier with that beer can and popsicle stick? Tweek on it and you get more...more what???...who knows...maybe you got lucky...maybe... Of course, you can build a resonant cavity from the beer can. Good luck. I gots a piston to change. My beer can is almost empty and the popsicle stick is sharp.

On a more serious note, there are 100MHz x 5 and 500MHz x 2 multipliers in the Tektronix TG501 Time Mark Generator. Sounded like a good idea at the time...30 years ago. But it was a nightmare. You really do need to control the drive waveform and have idler paths for the undesired harmonics. Amplitudes can get very high and you loose all your energy in losses at the frequencies of undesired harmonics. And we had the BEST equipment available to work on it. mike

If i was using light I could use fiber optics wich i would assume would largly overcome this accoustical problem.

Colin.