Single vs cascading op amp

message

message

gain

is to

G_1*...*G_2

handle it and

stage adds and

Cascaded stages

should use as

it.

advisable to have

the signal

level)

Stanford Research makes an equivalent, with an SC-cut crystal. We have one, which we put into a rack-mount chassis with a power supply and some buffers and dividers. The osc alone cost about $600.

I wonder how good the rubidiums and GPSs are for jitter in the seconds sort of range. I'm guessing not much better than the basci undisciplined XOs inside, since both the rubidium and the GPS are very slow corrections.

The SRS has jitter of a few picoseconds per second of delay, about

100x better than I need, and about 10x our budget. And it's huge.

We can get an SC-cut crystal with a oven-compatible turning-point temperature for around $20. But designing low-noise overtone crystal oscillators is a nuisance.

John

message news: snipped-for-privacy@4ax.com...

also

that

more

knee.

and

text -

And how many posters in this NG (s.e.d) do you think have regular access to all these articles? For that matter any NG.

Anyone can point to any amount of papers, theses, articles or any other publications not generally available and claim very much. This is USENET. Do try to properly report the reasoning of these articles as well as reference them.

lso

to

that

more

en

e.

and

to

xt -

rom

he

y

Oops. Please ignore the data in the above post. I realized driving in today that at low currents I was just measuring the input impedance of the voltmeter. (Idiot! why didn't the 1Meg ohm tip me off)

George H.

genews: snipped-for-privacy@4ax.com...

so

t .

t -

om

e

Excellent Howard I'll look up the references.

George H.

agenews: snipped-for-privacy@4ax.com...

lso

at

e

e.

d

xt -

rom

he

I appreciate the references. I=92m an alumina member of my college and as such can go into the library and look up any references that they have access to. I think it is the main reason that I give them money every year.

George H.

message

message

news: snipped-for-privacy@r33g2000yqn.googlegroups.com...

gain

to

G_1*...*G_2

it and

adds and

stages

use as

to have

the signal

Rubidium and hydrogen are both better at short times than caesium. If the volume would support the engineering required, how about using N cell phone VCXOs and getting rid of the phase noise via closure? (N oscillators give you (N**2-N)/2 phase measurements, which for N>2 is overdetermined.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net

Think of this as a sonar-type thing. We ping the process with an impulse and digitize the received signal. The frequency range is up to

20 MHz and the total echo time can be up to 10 seconds. The signals can be very weak, so we time-average a lot of shots, over up to an hour maybe. So in order that the shots dum nicely, we need the sample clock to change less than a few ns per 10-second event, over an hour. That's a few hundred PPB effective phase noise+drift.

We took two of our benchtop digital delay generators and triggered them together. One was set to make a 10 second delay, the other 10 seconds+200ns. Then we used an HP 5370 time-interval counter to measure the 200 ns difference and the standard deviation of same.

Using the standard, pretty-good TCXO versions of the DDGs, we got around 38 ns RMS jitter on the 200 ns delta. Not good, considering that a 20 MHz signal has a period of 50 ns. Signal averaging would be wiped out.

We sent the units down to manufacturing and they upgraded them to OCXOs; I've always wanted an OCXO in mine anyhow. That dropped the jitter to about 2 ns RMS. Divide by root 2, and it looks pretty good.

People do a similar thing to measure oscillator phase noise: get two identical units, tune them to slightly different frequencies, run them into a mixer, and sic a spectrum analyser onto the resulting difference signal.

So it looks like we need to buy an OCXO. OCXOs run the crystal at the flat spot on its parabolic frequency/temperature curve. TCXOs probably aren't as good because the thermal time constant of the crystal isn't perfectly matched to the tau of the temperature sensor. Putting a cover over a TCXO helps a lot by lowpass filtering millikelvin temperature fluctuations.

John

Yep. OCXOs are definitely better for long times, but they aren't particularly cheap. You used to be able to get ovens for just the crystal, which were a big help though not as good as ovenizing the whole works.

GPS-disciplined oscillators are a bit like the AC grid--not great at short times, really good at long times--which isn't a good match to the requirement. Closure is one way to lock a group of oscillators together such that the random parts of their variations are reduced--obviously since there are only relative phase measurements, there's an overall phase term exp(j omega t + phi) that you can't get rid of. Still, if you have the FPGA space, it would be N times better, and would at least meet the cost target.

Cheers,

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net

I think N would be unreasonable for cheap VCXOs... ballpark 100.

Lap-Tech will make us an oven-specified SC-cut, glass case, vacuum sealed overtone crystal for $20 or so. SC crystals used to cost over $100, so I guess the processing has evolved a lot. So maybe I'll do my own ovenized oscillator. We have one in production already, a surface-mount board on a heated aluminum block, with the crystal buried in the block. I might need a new circuit and board, but at least we have the mechanics and thermals and test fixtures worked out.

Dang, I don't like designing overtone crystal oscillators.

The Lap-Tech crystals are beautiful.

ftp://jjlarkin.lmi.net/Lap-Tech.JPG

John

"Think of this as a sonar-type thing. We ping the process with an

Sounds like a free induction decay from an NMR. 10 seconds would mean darn good magnetic field homogeneity or a really small sample.

George H.