Low Noise Direct Coupled Preamp

On a sunny day (Sun, 28 Feb 2021 09:02:46 +0200) it happened snipped-for-privacy@downunder.com wrote in :

In 2012 I bought, for 72$, a 10MHz Rubidium reference on ebay. Still working to this day, takes a minute or 2 to warm up... I have many GPS modules, Glonas, but GPS reception is in no way guaranteed here especially indoors.

When I search ebay now for FE-5680A Rubidium Atomic Frequency Standard 10MHz out

I see prices have now risen to $419.00 ! So was a good investment :-)

I'm not doing time domain. Everything is FFT.

Every GPS except old Trimbles uses a clock that is asynchrononous with the carrier. This causes a periodic advance and retard of the 1 PPS by the period of the clock.

For example, a 40 MHz GPS clock produces +/- 25 ns jitter of the 1 PPS.

Some GPS produce an output that shows how much error the previous 1 PPS signal had. You need to generate a synthetic 1 PPS with this error removed.

The ic timing circuits that can do this have serious errors of their own, plus stupendous phase noise.

General purpose GPS receivers have a large diurnal time error, so the time constant of the filter needs to be greater than 24 * 3600 = 86,400 seconds.

This is impractical with analog components, so the filtering must be done digitally. The quantization noise is added to the error signal and increases phase noise.

The diurnal error also has variations in ionospheric delays, so the error cannot be simply canceled out in software.

Surveying and military GPS receivers use the two GPS carrier signals to cancel these errors, but they don't need a stable 1 PPS signal and may not even make it available.

A Rubidium clock could, in principle, be used as a secondary standard and only peroidically compared against the GPS 1 PPS for calibration. However, the Rubidium clock has large phase noise so it would have to be followed by a low noise OCXO locked to the clock. I suppose the combination could be called a GPSDROCXO.

The Rubidium has a limited life due to depletion of the Rb cell. Stanford Research offers the PRS10 which has significantly lower phase noise and 20 year lamp life at only USD$1695:

Apparently you are referring to the internal GPS receiver clock, since the received carrier frequency of each individual satellite varies due to Doppler.

So 10 uHz. The OCXO medium time stability must be good for a day or two, since you can determine the average frequency just after one or two days.

Is DGPS correction signals available at your location ? This contains the pseudo code errors for each satellite and hence the amount of ionospheric and tropospheric errors, which could then be used to calculate the timing error.

If no DFPS corrections available but the GPSDO is stationary (and hence known location after a few days), you can calculate how much the location differs from GPS reported current location and known long time averaged location and hence adjust timing.

This will help compensate for ionospheric errors at much shorter periods than a day.

LT had a proposal to use only the follower output section of the 3045 and use a heavily filtered LT6655 a a reference. It is a band gap, but they managed somehow to make it look like a Zener. On that flickr gallery, there is a plot of an ElCheapo LT6655 solution.

The FFT is the saloon door between time and frequency. It works in both directions. JL does not like that. :-)

That's complaining at a quite high level. It still dives to adev = 4E-13 at one Hz. You would not place electrolytics into an SC cut oven.

I'm not sure about my Lucent KS24361, built by HP/Agilent/Symmetricom for Lucent. At least, it has 2 MTI-260 ovens (redundant). The MTI-260 can invade Rubidium land running free. But the Lucent probably has not the top notch selection.

Yes, on the early Oncores, it was just a port pin of the CPU and you could get an estimate of the error after the fact if you were lucky. A sound solution is different.

That's what a multi frequency receiver is for. When the military designed GPS, they did not anticipate that one could simply lock to the carrier phase, even without knowing the secret polynomial.

I have a board for a MTI-260, Morion or HP10811A that can be locked to an incoming 5/10 MHz that does produce a local 1pps. There is hardware such as a PP-cap and FET op amps for locking to an incoming

A Ru clock also relies on a crystal oscillator as a short term flywheel. It's only real advantage is the longer hold over when GPS goes away. I still have to see GPS going away.

Gerhard

I upgraded the -50dB ripple filter I posted earlier.

The new one gives -110dB at 1mHz. Here is the frequency list:

I think this beats the LT3042. There is no zener or bandgap reference, but preferably green LEDs for low noise. I couldn't find any green leds in LTspice and had to use whatever was available. The op amp is RRIO so you can use any LED you prefer.

You can download the LTspice files at