Datum FTS-4040/A

[tkamiya]

I just bought an used Cesium Standard at a very good price.  Model is FTS-4040/A by Datum Inc., Frequency & Time Systems.

It locks pretty fast (6 minutes) and output is rock stable.  I have it displayed on 2 channel scope, along with GPSDO output.  Neither is moving.

Odd thing is, there is NO manual available online.  I found one for 4040B but there's enough difference.  For example, there is a RS232C connector (labeled so) with no connection in the back.  1 PPS is not there.  So this is probably a unit that was used in very l Limited Applications.

If anyone has any manuals, tech refs, or experience, I'd like to talk to you!  I just can't believe I got this at such a good price.  I just hope the tube lasts for at least for few years.

[tkamiya]

OK, now questions. 

Obviously, now I'm obsessed with accuracy and precision.  So I did the typical dual trace oscilloscope measurement.  Channel A gets 10MHz for GPSDO that has been running for months.  Set a trigger on it.  Channel B has 10MHz output from FTS4040a.  It has been warming up for like 4 hours.  (I know, it's not long enough).  Scope is set for 20ns/div.

In 10 hours, it looks like it moved 1.5 division.  So by extrapolation, to move full cycle (5 division), it'll take approximately 33 hours: which is 118,800 sec, or for ease of calculations 1x10^5  Period is 1x10-5.

10MHz is 10^7.  So movement is 10^-5 / 10^7 = 10^-12.  1 part per billion.

Is this calculations right?  This is within spec but warm up time was insufficient.  So it isn't the final result.

If I want to do this continuously to say get a daily report, how do I automate this?  Since sine wave is a periodic function, there is no way to tell a peak from another peak.  Moving to the right 180 degrees looks exactly like moving to the left 180 degrees. 

I have a chart recorder.  I have counters like HP53131A and HP5335A.  My GPSDO has 1 pps output and 10MHz output, but Datum only  has 10MHz, 5MHz, and 1MHz out.  No pps.  The ideal result will be a continuous recording on chart recorder with 1 day worth of change in 10cm of paper. 

There is no GPIB and there is no serial communication on Datum either.

[tkamiya]

Here are pictures!  Everything looks pristine.  Not much dust either.

[edpalmer42]

Tkamiya and I have exchanged a few private messages.  But nothing was making sense so pictures were necessary.  Now I see the problem.

The 4040A is a wrapper that provides power, user interface, and I/O for a 5045A Cesium Module.  I have a 4065A which is a fancier wrapper for the same 5045A Cesium module.  Unfortunately, Tkamiya, your Cesium Module appears to be an older version which is totally different from my unit.  I've attached a picture that shows the 4040A (date codes - 2000) that I helped another user repair and my 4065A (date codes 1999) which includes the same module.

However, I *have* seen pictures of a unit like yours, but in a Datum 4060 where the Cs module is identified as a 5030 rather than a 5045A  .  There's some descriptive info on this unit here:  http://prc68.com/I/FTS4060.shtml .  Scroll down for the 4060 info.  You may be able to pick up a few nuggets of info there.  But I'm not seeing anything to suggest that the 5030 had RS-232 capability.  Reaching out to the owner wouldn't be a bad thing, but this page dates back to 2005 so who knows.

Ed

Edit:

Datum FTS-5030 Datasheet:  http://www.n4iqt.com/fts5030/5030ds.pdf

[tkamiya]

Thank you.  Thank you.  Thank you!

I really didn't want to start poking into things that I don't understand.  Also, getting a working example with the kind of $$  I paid is probably not going to happen again.

So I will close the cabinet and use what I got.

1 pps; however, is probably a do-able change.  There is a 1 MHz output I'm not going to use.  Perhaps I can make an external module that does this.

Again, thank you.

[texaspyro]

1 pps; however, is probably a do-able change.  There is a 1 MHz output I'm not going to use.  Perhaps I can make an external module that does this.

Google PICDIV   Also look into the TAPR TADD2 Mini

[tkamiya]

On front panel, it says "Frequency and Time Systems Inc, a Datum company".  This kind of naming is used in transitional period when one company acquires another.  Maybe mine is a "transition" machine when changes were made and never got documented.

Yes, I was going to TAPR site.  I'm just not sure about PIC doing the division.  How much uncertainty does software driven hardware add?  Is this a credible thing to do for something is done in nano-second order?

[texaspyro]

PICDIV jitter is in the low picrosecond range...

[edpalmer42]

Yes, I was going to TAPR site.  I'm just not sure about PIC doing the division.  How much uncertainty does software driven hardware add?  Is this a credible thing to do for something is done in nano-second order?

Quote from the Picdiv page: 

Output accuracy is equal to input accuracy. Jitter is extremely low since the PIC is a synchronous device. Recent measurement suggest it is less than 2 ps.

Ed

[tkamiya]

That's nice to know.  Thank you both. 

I've also resumed GPSDO Rb project.  Now I have Cs, it would be helpful.

[edpalmer42]

By the way, if your 4040A runs out of Cs, be sure to salvage that OCXO.  I'm not sure what model it is, but that family of oscillators is up at the top end of the OCXO world.  Treat it gently.  Inside it is a rather fragile glass Dewar flask (Vacuum Bottle).  One good thump and it will shatter!

Ed

[tkamiya]

I am having real hard time believing there is no RS232C or some communication port.  It must be needed during manufacturing and aligning prior to shipping, isn't it?

[edpalmer42]

I am having real hard time believing there is no RS232C or some communication port.  It must be needed during manufacturing and aligning prior to shipping, isn't it?

Not at all!  Old Cesium and Rubidium standards were totally dumb - no microprocessor at all, although they might have some TTL logic chips.  Monitoring was accomplished by turning a rotary switch and checking the reading on an analog meter!  The HP 5065A is one of the top Rb Standards ever and that's how it works.  It includes some digital circuitry, but no processor.  I've got pictures of the FTS 4050 and 4060 that still have a switch and meter.  Efratom FRK and M100 Rb standards were also dumb.  They provide two DC voltages and one switched output for the 'lock' indicator.

It's odd that at least some 4040A units have the 'smart' 5045A module.  You'd think that the lower model number would mean a 'dumb' Cs module.  The data sheet for the 4040A talks about a microprocessor, but doesn't mention external communications, yet your unit has a connector labelled as RS232.  The manual for the 4040B talks about the RS-232 connections.

Does your Cs module have a label on it to identify it as a 5030?  The data sheet for the 5030 talks about a microprocessor and software monitoring, but doesn't mention external communications.  It also talks about an FTS 6006 "Control and Monitor Module".  I also see that the manual for that module is on ebay.  Do you have such a thing in your unit?  I don't even know if it's internal or external!

We just don't have enough info to be sure.  All I can suggest is to find the processor board and see if you can find anything interesting.  ICs labelled as MC1488 or MC1489 would be a dead giveaway since those are the standard old-school RS-232 driver/receiver chips.  If you can find the main connector between the Cs module and the rest of the unit, probe with a scope and see if you find any RS-232 leads.

Ed

[tkamiya]

I looked at everything I can see without taking this thing apart.  No voltage converter chips for RS232.  No ports that resemble DB9 or DB25 on CPU board.  I think the cpu was 1802?  I vaguely remember this one.  So THIS ONE is processor controlled.  AND there is an empty port on back panel.  Hum....

By the way, there are coffee makers, rice cookers, and vacuum cleaners that has CPU in it.  It's funny how market driven some of the simplest things are.  I'm waiting for CPU controlled can opener.

I'm still tickled pink that I now have a Cesium standard in my lab.  A lab technician (who is military trained at highest level) told me to not turn it on and off too often.  Best thing to do is to keep it on.  His reasoning was heat cycling will cause more harm than keeping it hot and steady.  For me, I'll have to keep it off because of its power consumption.  Now I have this, I am so supervised how stable my GPSDO is.  It's certainly enough for my needs.  But who's talking NEEDS?  Hydrogen Maser, here I come.  (not!)

By the way, the oldest time keeping device I have is a rail road clock from early 1900s.  It was fully serviced and works very accurately.  Also amazing what craftsmen did back then.

[testpoint1]

hope previous user also power down the 4040A, then your kit's life remain will be longer, cs tube life.

[tkamiya]

Yes.  I wondered about that, too.

My unit is AWFULLY CLEAN, inside and out.  Dust accumulation is minimal and only on certain places.  This unit was likely made in 1995 and still locks in 6 minutes.

That leads me to believe it lived a pampered life.  Probably it was in a clean room.  It was probably not turned on all the time.  Otherwise, the tube should be long dead.  It doesn't show any signs of replacement either.

I think I lucked out.

[TheSteve]

Awesome score, doesn't seem like much fun if it isn't always running though.

[edpalmer42]

I looked at everything I can see without taking this thing apart.  No voltage converter chips for RS232.  No ports that resemble DB9 or DB25 on CPU board.  I think the cpu was 1802?  I vaguely remember this one.  So THIS ONE is processor controlled.  AND there is an empty port on back panel.  Hum....

You wouldn't see a voltage converter chip.  I can guarantee that you'll find opamps in this unit, so there will be +15V and -15V supplies.  Add a couple of 12V regulators and you've got +12V and -12V for the RS-232.

It wouldn't surprise me to find that the first generation of processor-controlled standards just replaced the analog loops with digital loops.  Then later they realized that they could add communications.

By the way, there are coffee makers, rice cookers, and vacuum cleaners that has CPU in it.  It's funny how market driven some of the simplest things are.

It makes sense to use a micro.  Replace all the control, sequencing, and timing with a micro rather than complex digital, or worse, mechanical, circuits.

I'm still tickled pink that I now have a Cesium standard in my lab.  A lab technician (who is military trained at highest level) told me to not turn it on and off too often.  Best thing to do is to keep it on.  His reasoning was heat cycling will cause more harm than keeping it hot and steady.

NO, NO, NO, and NO!!!  (Was that emphatic enough?)

Never take advice from someone who works with an unlimited budget.  Yes, under normal circumstances you could argue that it's best to let some random device operate rather than stressing it with turnoff and turnon, particularly if you're going to use it frequently.  But that idea fails if the device will wear out just by being powered up.  Your unit is old and has an unknown amount of life remaining.  It could have thousands of hours of life left.  It could also die next week. Once you exhaust the Cs supply, you have a really effective paperweight!  I only power mine up once or twice a year to keep the tube healthy.  That's a requirement, by the way.  While it's running, I check it against my GPSDO and my 5065A Rb standard.  The 5065A is my house standard.  It runs 24/7 and syncs all my test equipment as well as providing 10 MHz for whatever else I want.  Note that the 5065A actually outputs 5 MHz, but I run it through a frequency doubler to get 10 MHz.

Ed

[tkamiya]

Well, I'm going to have fun with it for sure.

I'm getting set up for a comparison study.  Cs is ready.  GPSDO is ready.  Rb is ready.  I'm working on GPSDRb.  I'm going to run them all at the same time and compare.  So far, I'm impressed with GPDSO's performance.  It is keeping up with Cs.  I know the common wisdom on Cs but I trust these guys also.  They have been really good to me.  I can't afford to keep it on ALL THE TIME, but I have no quorum about having it on weeks at a time if I'm using it.

I'm also seeing periodic nature of fluctuations.  I think it's room temperature.  I'll have to find a spot in my house where changes are minimal.  Lab is actually a worst location.

Will you elaborate on turning on twice being mandatory?  Also, what's this storage mode on front panel?

[edpalmer42]

Well, I'm going to have fun with it for sure.

I'm getting set up for a comparison study.  Cs is ready.  GPSDO is ready.  Rb is ready.  I'm working on GPSDRb.  I'm going to run them all at the same time and compare.  So far, I'm impressed with GPDSO's performance.  It is keeping up with Cs.  I know the common wisdom on Cs but I trust these guys also.  They have been really good to me.  I can't afford to keep it on ALL THE TIME, but I have no quorum about having it on weeks at a time if I'm using it.

AFAIK, GPS is actually better than Cs.  The people in charge of GPS monitor each satellite individually and nudge its clock as necessary to keep it dead on.

Are you set up to capture data to a computer?  Long term measurements without that capability are of limited value because you'll miss so much.  Did I see somewhere that you've got a reasonably good counter?

I'm also seeing periodic nature of fluctuations.  I think it's room temperature.  I'll have to find a spot in my house where changes are minimal.  Lab is actually a worst location.

Could be room temperature and/or GPS shift due to the ionosphere changing through the day.  I know that temperature changes can affect a standalone Rb.  I'm not sure how much effect temperature would have on a GPSDO.  I'd think it was minimal.  Ionospheric disturbances are worst at sunrise and sunset.

Will you elaborate on turning on twice being mandatory?  Also, what's this storage mode on front panel?

Every Cs manual talks about this.  The Cs tube requires a really, really good vacuum.  Every tube leaks a bit and/or has some outgassing from the internal parts.  Over time, these will eventually kill the tube.  Every tube includes an ion pump to grab any gas that it finds and take it out of harms way.  It's not a pump in the conventional sense.  It's a high voltage source that ionizes any gas it touches and then accelerates it so it slams into a target and sticks there.  You need to run the unit occasionally so that the pump can purge the tube.  If you let it get ahead of you, the ion pump won't be able to deal with it.  When you first powered up your unit, the first thing that happened was the ion pump kicked in and pumped down the tube before it even turned on the Cs oven.  Depending on how bad the tube was, that could have taken days.  The ion pump current is one of the main ways to estimate the health of the tube.  A good tube will drop to 0 ua.  On my 4065A, a current of > 240 ua causes a fatal error that shuts the unit down.  I can measure mine either in software or with a meter.  You'll have to hunt for yours.

You'll have to do some checking, but in my 4065A, 'storage mode' powers everything down except for the ion pump.  That means that as long as the unit has power, it will be continuously purged of anything leaking in.  I'm a little surprised that your unit has that feature.  Seems like that unit is just full of surprises!

Ed

[tkamiya]

No, I don't have a computer capture setup.  But I do own a GPIB adapters and I have lots of computers.  Yes.  I have some good HP counters.  But none are good enough without something being an external reference clock.  Which comes right back to what's being compared to which, and what's the reference?

Right now, I have everything fed to 3 traces of O'scope.  Data is really hard to interpret but if I were to assume Cs being absolute, everything else is moving.  Conversely, if I were to pick anything else as absolute, Cs is is moving, too. 

I am experiencing a bad case of recursive home brew (DIY) feature creep.  I got to finish Rb based GPDSO.  Once that's done, I'll need to somehow find a way to compare everything at the same time and see who's the winner.  I will also have to determine what's better what on WHAT.  Say for short term drift, long term stability, accuracy, etc.  I'm sure that will send me to a quest for finding a way to control the environment.  Yes, Rb moves quite a bit on case on/off. 

Just last week I calibrated Rb and OCXO.  Had cabinet open so I can access the trim pot.  Get it agree with standard.  Close the cabinet, and drifting starts.  OCXO is very susceptible to temperature change.  I thought oven WAS supposed to eliminate that to good degree.  Not so!

This is going to be a frustrating fun.  I'm already having too much fun!

[FransW]

Hi tkamiya,

It seems that you are stuck in your uncertainty deliberations.
Therefore it is nice to contemplate the shortest possible time step, or
the Planck time = T = (G ℏ / C5)½ = 5.39⋅10-44 seconds.
Which indicates that 12 or 14 digits is not enough.
This unit or time step is far from what we can measure today.
Our lack of completely understanding physics puts us (and you, with many others)
in this situation.
The conclusion is that you just have to decide for yourself which 10 MHz standard to use.
Whether you are right or wrong is not a matter for today (and/or tomorrow).
Making measurements is working with uncertainty. Be happy!

Best regards, Frans

[edpalmer42]

No, I don't have a computer capture setup.  But I do own a GPIB adapters and I have lots of computers.  Yes.  I have some good HP counters.  But none are good enough without something being an external reference clock.  Which comes right back to what's being compared to which, and what's the reference?

Right now, I have everything fed to 3 traces of O'scope.  Data is really hard to interpret but if I were to assume Cs being absolute, everything else is moving.  Conversely, if I were to pick anything else as absolute, Cs is is moving, too. 

I am experiencing a bad case of recursive home brew (DIY) feature creep.  I got to finish Rb based GPDSO.  Once that's done, I'll need to somehow find a way to compare everything at the same time and see who's the winner.  I will also have to determine what's better what on WHAT.  Say for short term drift, long term stability, accuracy, etc.  I'm sure that will send me to a quest for finding a way to control the environment.  Yes, Rb moves quite a bit on case on/off. 

Just last week I calibrated Rb and OCXO.  Had cabinet open so I can access the trim pot.  Get it agree with standard.  Close the cabinet, and drifting starts.  OCXO is very susceptible to temperature change.  I thought oven WAS supposed to eliminate that to good degree.  Not so!

This is going to be a frustrating fun.  I'm already having too much fun!

First, turn off the scope.  It's the wrong tool for the job.  You need a time interval counter or analyzer.  You can do it by measuring frequencies, but that makes no sense to me.  I always measure time intervals.  Collect the data on a computer.  Now you can start to see real numerical data that show how much drift and/or frequency difference there is between the two devices under test.  If your data shows drift, think about the devices.  If you're comparing Cs and Rb, the drift will be due to the Rb.  With Rb and Quartz, the drift is almost always due to the Quartz.  A GPSDO always averages out to perfect, but there will be jitter and perhaps strange anomalys in the short term so there will be situations where it's not clean enough to use as a standard.

Second, accept the fact that there will always be differences between your devices.  Cesium doesn't drift, but the frequency can be nudged a bit.  Modern standards do a marvelous job of compensating for all the conflicting forces that affect the frequency of a Cs standard, but they still aren't perfect, so there's a method to nudge the frequency as necessary.  For most users, even the professionals, the frequency error is too small to worry about.  Rubidium does drift and so, needs a method to adjust the frequency.  Quartz is worst of all so it has a relatively large adjustment range as compared to the other two.  The larger the drift, the more frequently you have to calibrate the device to remain within your desired accuracy limits.  But drift is relentless so, as soon as you finish calibrating your device, it's off frequency, but still within specifications.

Ed

[tkamiya]

Is it that simple? 

All Cesium, Rubidium, and whatever "ium" standards have OCXO inside.  Any of them can lose a lock and go wild.  I have seen GPSDO-OCXO do this.  3 of 4 clocks are new to my lab.  All of them were purchased used and none of them were professionally calibrated.  One of them is home brew out of used parts.  So far, all of them have performed admirably but I'm still doing test runs.  If I can somehow convince myself Cs is as good as it should be, I can then move to more quantitative testing.  I do have interval counters.  But I have to get GPIB working first.  I have FOUR identical frequency counter / interval meter so once it's all set up, it would be an ideal test environment.

To my surprise though, least expensive selection, GPSDO has performed well.  Certainly well enough. 

It would be funny if I end up with GPSDO as my standard after all this. 

[tkamiya]

Bad news....  (that's why I wanted to do this test)

My homebrew Rubidium GPSDO did not lock overnight.  I fed pps from another source and it locked right away.

It worked when they were on my desk, so either temperature or noise must be an issue.  Half success, I guess.  I will have to try with a larger case.  I was a bit uncomfortable after seeing how cramped it was.  Heat dissipation also seem to be an issue.  Heatsink mass (not size) is about half of what commercially made rubidium standard is, so it may not have sufficient radiant cooling effect.

Datum FTS-4040/A is still working beautifully.