Hi

On the lightweight Rb’s they feed the correction into a DDS. The DDS steps are big enough to show up above the ADEV. You get a “hump” in the ADEV as a result. The solution is to disconnect the temperature sensor that feeds the correction circuit.

Well, there’s this list :)

The things that need to be done are a “that depends” based on the approach you take. A full listing of all you might do down all of the roads you could take would be a wall full of books.

GPSDO is a good place to start.

First step is to be able to extract timing data from individual sat’s. Not all GPS modules do this correctly.

Thus the “daily” statement. There has been discussion on the list that interchange with other list members at a more rapid rate might make sense. Given the floor of most setups, daily updates are not a bad thing.

1. Measure your local time offset by GPS sat number every hour and save the data.

2. When NIST posts there data, pull it down.

3. Compare your data to theirs

4. Do a fit

5. Feed that into your control loop equation.

The missing element is the per sat data….

Bob

On Sun, 7 Dec 2014 19:34:02 -0500, you wrote:

That depends a lot on the Rb. With a temperature controlled LPRO it's
easy - just logging air pressure against frequency get you most of the
way. With the LPRO's I've tested that gets the variation with pressure
to down under +/- 2E-13, some well under, over 60mbars pressure range.
I have one that had almost no residual left after correction, though
the others had a little. Getting past that is harder.

There is also some time lag - looked like something around 3/4Hr, but
with GPS as my reference and air pressure moving so slowly, it was
hard to tell.

Of course the lower the numbers, the more error sources start to
become significant, but since the LPRO's I've test are all around
8E-14/mbar, it's not exactly hard to measure. It is time consuming
though, since it normally took 3-6 weeks to do each test, depending on
the weather.

Interestingly, the both the FE5680A's  I tested had similar responses
to pressure variations - very variable compared to the LPRO's, so
impossible to correct simply and well. Seeing the comments about the
temp correction in FE5680A's causing problems, I did wonder if that
might be part of the problem, but have not got around to testing that
yet.

Angus.

Hi

….. and I’ve been playing with FE5680’s. They (for what ever reason) do not correct easily.

OK, Thanks

Yes, I was afraid of that!

I searched google with the following:

https://www.febo.com/pipermail/time-nuts/ gpsdo

This gave 68 hits. I then used

site:www.febo.com/pipermail/time-nuts/ gpsdo

This gave about 2,310 results. Going to page 2 reduced the count to 136.

I had previously downloaded the archives from 2004 to 2013. A Boyer-Moore
search gave the following:

Files Searched : 63404
Matches Found  : 6198
Bytes Read    : 329,255,183
Elapsed Time  : 31,350.793 ms

So it seems google is not very useful.

It would appear the best approach is to simply start reading the posts at
the beginning and save the interesting ones.

Can you tell me some of the ones that do?

OK, that makes sense.

There's another term I need to research!

OK, so I figure out how to do this. How do I tell if this is making the
gpsdo more accurate? In other words, how do I get the ADEV without having
an H-Maser?

Thanks for your help!

Mike

Hi

Consider that a team of people likely work for a few years to bring out a fully tested and working design like this when they first start out. That is with a fully equipped lab and full time to focus on it.

There are lots of interesting threads on Time Nuts ...

I have yet to see one for under $2K that does it correctly. I don’t have the cash to buy ones at those sort of prices. Some have reported that the old Motorola UT’s will do it. The samples I have tried have not done very well. I may have not had them running right - who knows….

It’s not a simple control process, but it’s not all that terrible either. It just takes a bit of work to optimize. Figure a few months to a few years for the optimization depending on what you have for issues along the way.

You get a Cs (or other atomic standard) or you compare several different GPSDO’s against each other. The preference would be for groups of three so you can rule out ones that are not doing what they should.

You also need the measurement gear to resolve frequency down in the 1x10^-13 range. A normal counter will not do that.

—————————

Backing off a bit - why do this?

You can buy commercial GPSDO’s on eBay.

You can build published GPSDO designs that are known to work.

What is the objective here?

Bob

Bob,

On 12/11/2014 02:25 AM, Bob Camp wrote:

GPS receivers can have considerable common mode effects, which won't
show up in the three-cornered hat setup. You can work with a rubidium
clock and a crystal clock if you don't have a cesium standing by.

Indeed.

Cheers,
Magnus

Hi

Which is why you want several different GPSDO designs with different “stuff" inside them….

The GPS is likely to have issues that are 12 / 24 / 48 hour cycle related. Rb’s and crystals have temperature dependance that likely follows the same sort of pattern. That will indeed limit your ability to check things at 1x10^-14 at 400,000 seconds with a 40,000,000 second run. In most basement design cases, keeping that > 1 year run going is unlikely. Most of the debugging and testing work will get done inside the range of the better auction site GPSDO’s.  Working that part out on the cheap first is the best way to do it.

It’s also a matter of the other gear on your test bench. There is a balance between design, reference standards, and measurement gear. You inevitably move one forward at a time. When you do, you need to revisit the other two and bring them forward as well. Trying to measure to 1x10^-14 with a “Bob’s of Bulgaria (TM pending)” frequency meter is probably not a real good idea …..

Bob

Yes. It is very easy to get distracted.

OK, that pretty much eliminates the NIST data.

I see the issues people have here getting their cs to work. Maybe the best
approach is different gpsdos in a N-corner hat.

I have already bought stuff in groups of 3 to prepare - gps receivers,
rubidium and morion oscillators, a trimble, etc.

Yes, I have a new invention for a DMTD that may work.

Yes, I also see the difficulties some have here trying to get them to run
or even talk to them.

Yes, I have seen most of them. I think I can do better.

I'm retired, kids are married, wife moved back to France, and I have no
girlfriend or interest in getting one. I have a small pension so I can
afford to buy parts and pcbs.

I have been involved in precision instrumentation since 1960 and have a few
patents and some new inventions I think could be applied to time and
frequency applications.

There are many new dacs, op amps, and bipolar microwave devices that offer
much lower noise than current designs use. I think the performance can be
improved in some areas with new components and design techniques, and I
have  equipment and time to explore.

As usual, I will have to build much of my own test equipment to measure the
improvements, but I have been doing that for 5 decades so I'm used to the
idea. Some of the items may be good enough to start a new company and bring
to market. So I can call it R&D and get tax benefits.

Most of all, it's an interesting challenge. Lots to learn, many new ideas
and concepts, and a ready source of information in the archives and online.

What could be better?

Thanks,

Mike

On 2014-12-09 04:10, Mike Monett wrote:

IGS http://igscb.jpl.nasa.gov/components/data.html has pointers to
get 15 minute samples from high rate IGS stations around the world:
http://cddis.gsfc.nasa.gov/Data_and_Derived_Products/GNSS/high-rate_data.html

USNO info tells you which SVNs/PRNs have Cs (few) and Rb (most) clocks:
might be interesting to compare accuracy of Cs vs Rb SVs in view.

Take care. Thanks, Brian Inglis

Hi

or means that more research is needed on single sat boards. There are a lot of them out there.

A Cs is normally a very limited time item. If you are doing the sort of long term tests that they really are useful for, you will wear one out pretty fast. They are not a good reference for short tau.

The ideal would be > 3 TBolts, > 3 same model Rb’s, > 3 same model OCXO’s and more than three groups overall. Each has their own ADEV curve  Comparing devices with vastly different ADEV is not the best way to go.

Well, make sure it works. Testing is at least as critical as any of this.

It’s roughly 100X easier to get an eBay GPSDO running in 8 hours than it is to design a GPSDO in under a year.

That’s fine.

There are only two points that system noise really comes into the GPSDO design:

1. The TDC must have enough resolution

2. The DAC on the EFC must have noise below the OCXO

The TDC is limited by the basic resolution of the GPS system. It’s easy to build one that has far more resolution than needed / useful.

The DAC issue is normally solved with a < $4 part. Unless you have an OCXO with a very wide EFC range, it’s noise is unlikely to be an issue. DAC resolution can be addressed to any desired level with two DAC’s (fine and coarse). (Once you get going, you only use the fine DAC).

The real “fun and games” revolves around the software used to implement the filtering / control loop between the GPS and the OCXO (or Rb, or TCXO, or MEMS, or VCXO, or Cs or …)

I like the idea of tax benefits from doing this …

Focus on the software.

Bob