No luck on my first attempt at acquiring a cesium frequency standard

[uncle_bob]

Do the Rb standards used by telcos/cell operators fit in this same story somehow?  Cs being of a higher standard than Rb references obviously.

Hi

Back in the "good old days" you had "Stratum" levels for timing. Things have changed a bit, but the ideas still apply. The "master source" needs to be a primary standard. It is intrinsically on frequency / on time. That makes cesium a real good candidate for Stratum one. The next layer down is a standard that mostly locks up to the higher level. If something bad happens, it is required to run for a long time without locking up to the higher level. That's where Rb's come in. They have the stability to do this "holdover" process.

The next level down was Stratum 3. That was an OCXO based approach with less holdover time. The idea was that the further you got from the "center" the less the impact of a fault. The other assumption was that the fault was something you could get to and fix a lot quicker.

After a bit of time went by, you got things that crossed between the various levels. You got 3E and all sorts of other levels that never really existed in the original approach. Fast forward a decade or so and you have stuff that does not fit in any category. Most of these cesium standards went into those sort of systems.

Bob

[jwm_]

I think it is more that the availability of relatively cheap GPS recievers and ubiquitous free NTP servers short circuited most of the strata allowing anyone to get a stratum 2 clock on a relatively inexpensive budget if needed and an okay distributed clock with just IP if that is all you need. strata between the two don't make as much sense.

this app note does a fairly good job explaining the statum levels and their accuracy requirements
http://www.raltron.com/products/pdfspecs/sync_an02-stratumleveldefined.pdf

[uncle_bob]

I think it is more that the availability of relatively cheap GPS recievers and ubiquitous free NTP servers short circuited most of the strata allowing anyone to get a stratum 2 clock on a relatively inexpensive budget if needed and an okay distributed clock with just IP if that is all you need. strata between the two don't make as much sense.

this app note does a fairly good job explaining the statum levels and their accuracy requirements
http://www.raltron.com/products/pdfspecs/sync_an02-stratumleveldefined.pdf

Hi

Once the network guys started in with the Stratum numbers, a lot of others copied the same terms without using the same limits. An NTP server that runs as stratum 1,2,3 is massively less accurate than it's network counterpart with the same name.

Bob

[hydrogen maser]

So I got the right cable (I think) but am not having any luck running the monitor.exe program. I'm on Win 7 Pro 64bit so I have to use an emulator - tried dosdox but not getting anything to work. I just says alarm 52. Alarm 52 is not in the manual and I think that just means the program can't communicate with the unit. Not too surprising given all the weird virtual emulator + usb to serial cable, etc.. Now trying direct terminal programs, but putty won't send the right control characters (02 and 03) that I need to format the messages. Arg... trying some other terminal programs, maybe one will work.

[edpalmer42]

How far back does your DOS experience go?  In general, you can send any character by holding down the ALT key and use the number pad to type in the decimal ASCII code for the character.

e.g. <ALT down>65<ALT release> sends capital 'A'.

I just triied it in putty and the Windows 10 DOS window.  Seems to work.

The other thing to do is create the two or three command strings however you can and then try a copy/paste.  If you want to do any data logging, you should write a little program to send the strings and log the results.

Ed

[hydrogen maser]

I'm not going to bother trying anymore with the serial port, I can't get it to do anything. My guess is that the CPU is not starting up/working because it never goes into initialization mode and won't respond to the serial port, the only thing that comes out of the serial port is a series of 5 nulls (00) - not a good sign. The 10mhz/5mhz/1mhz outputs are all working and read bang on what they should be and the waveforms look like they should on the oscilloscope (granted my test equipment is nowhere near what is needed to measure this device) so at least the oscillator is working and that signal is making it to those outputs. The .1/1/5/10 and 1 PPS outputs don't work - they just show noise across the entire spectrum from near dc up to about 90 mhz. 

edpalmer42 - you think the below should be my next step on this one?
To monitor the ion pump, look at the two high voltage supplies.  Remember that you can do this in either storage or operate mode.  One supply has three low-voltage terminals on the side.  Power, ground, and monitor.  Measure the voltage between ground and monitor.  You should see a negative voltage of a few hundred millivolts or less.  What do you see?

[edpalmer42]

I'm not going to bother trying anymore with the serial port, I can't get it to do anything. My guess is that the CPU is not starting up/working because it never goes into initialization mode and won't respond to the serial port, the only thing that comes out of the serial port is a series of 5 nulls (00) - not a good sign. The 10mhz/5mhz/1mhz outputs are all working and read bang on what they should be and the waveforms look like they should on the oscilloscope (granted my test equipment is nowhere near what is needed to measure this device) so at least the oscillator is working and that signal is making it to those outputs. The .1/1/5/10 and 1 PPS outputs don't work - they just show noise across the entire spectrum from near dc up to about 90 mhz. 

edpalmer42 - you think the below should be my next step on this one?
To monitor the ion pump, look at the two high voltage supplies.  Remember that you can do this in either storage or operate mode.  One supply has three low-voltage terminals on the side.  Power, ground, and monitor.  Measure the voltage between ground and monitor.  You should see a negative voltage of a few hundred millivolts or less.  What do you see?

If the processor isn't working, who output 5 nulls?  Remember our manual is for the 4040B.  Check the comm parameters.  Start by scoping the data xmit line from the unit to check the baud rate and look for repeating bursts of data.  You can also 'scope around and see if there's activity on the dataand address lines in and out of the processor.  Also, if you haven't already done so, check power supplies for level and noise and nudge all socketed chips out and back in.  For now, limit things to the main board.

If the processor isn't working, then almost nothing is working.  Yes, the oscillator feeds directly to the divider that generates the 10/5/1 MHz output, but the oscillator is just free-running.  The processor doesn't just monitor and talk to the outside world, it runs the entire process.  When you say the frequencies are bang on, how are you measuring them.  Do you have a GPSDO?

Yes, you should measure the ion pump.  That won't tell you if the tube is healthy, but it's a necessary step before you can do much else that involves the tube.

Ed

[edpalmer42]

I realized that my unit is overdue for a run so I'm going to try to duplicate what you're doing.  Since my unit is basically the same 5045A Cesium module that you have with a fancier user interface, I'll disconnect my user interface and talk directly to the 5045A.  I think that should duplicate what you've got.

First, a correction.  Turns out that the ion pump power supply has 4 terminals rather than 3 like I said.  Mine is labelled as +24, GND, I PUMP, and FAIL SENSE.  Measure the voltage between I PUMP and GND.

I'm just getting ready to start measuring my ion pump.  My unit has been unplugged for 10 months so it should be interesting.  I'll capture the voltage data and post the graph.

Ed

[edpalmer42]

Okay, good news and bad news.  Turns out my unit had an unnoticed fault.  The ion pump won't run in storage mode.  I'll have to investigate that later.  But when I turned it to operate mode, everything was fine.

The attached graph needs some explaining.  The trash at the beginning was before the unit was plugged in.  Then I waited for the ion pump to kick in.  Nothing.  Wiggled the leads and caused that first downward spike at about 22:27.  Damn, it's not working!  Checked my old notes and realized that the same thing happened the last time I did this measurement.  Powered down the unit and rearranged my leads.  That's the 'thin' section.  Powered up again and turned the power to operate.  Waited a few minutes while it did some "pre-flight checks".  Finally the ion pump kicked in and went to ~ -0.09 volts.  This equals about 90 ua of ion pump current.  Pretty good for 10 months on the shelf considering that the supply can source up to 300 ua.  I was making about 4 measurements per second so I'm confident in this measurement.  Over the next 10 minutes it faded down to about -0.005 volts which is where it always sits.  The unit reports the ion pump current as 000.  This test is actually better than my previous one.  After being shut down for only 6 months, the ion pump spiked to 140 ua.

It's getting late.  Tomorrow I'll start on the comm link.

Ed

[edpalmer42]

First phase of the comm challenge.

I connected directly to the 5045A module using a three-wire connection to a USB/serial converter and putty.  Not the results I expected.  I tried 2400 7o2 as stated in the 4040B manual, but the results were gibberish.  But there were results spit out automatically every second.  Changed to 4800 7o2 and saw the following repeated endlessly:

Code: [Select]

XR05-005|RR +0004|Z*-$$$|RZ*-0092Y

XR05+015|RR -0014|Z*-024|RZ -0011Y
^                                ^
|                                |
ASCII 01                         ASCII 17
The capital 'X' and 'Y' are standing in for the control codes.  The first line was before lock, the second line was after lock.

I tried to talk to the unit, but it ignored me.  Since this protocol doesn't match what we're expecting, I'm going to move on.

HM, does the 4040A have another processor board between the 5045A and the comm port?  I'm going to move my connection from the 5045A to the main comm port on the back.  That'll be in the next message.

Ed

[hydrogen maser]

edpalmer42 thanks so much for all the time and effort you have put into helping me, I really appreciate it. I pulled a "real" serial port card out of a different computer. Going to try connecting through that and see if I see anything - I will also try different baud rates. As to the oscillator "working",  all I know is that my frequency counters and oscilloscope say it is exactly what it should be, but they are not really up to the task of measuring something like this. So the oscillator is not grossly broken. I'm guessing it is just free running and not locked to the cesium. If I don't get anything from the serial port on the computer I will see if anything is on the tx line with the scope, I have the rs232 trigger and decoder options in my scope so that is helpful. I will also measure the ion pump.

[edpalmer42]

One suggestion.  When you run this unit, keep it horizontal.  Some time ago, I was running it while it was standing on its side.  I had a couple of instances where the unit lost lock.  I was logging the data at the time and some of the parameters were very odd.  I put it flat on the bench and things smoothed out.  No problems since.  I don't know if it's supposed to be position sensitive - the manual doesn't mention it - but I decided that from then on, I'd only run it horizontal.

By the way, does your unit have anything on the back about options?  Since your Cesium tube is different from mine, I was wondering if you got the high performance tube.  Cross your fingers and hope you didn't.  The high performance tube lives fast and dies young!

Ed

[hydrogen maser]

Nothing on the back about any options and no extra telco options boards installed. The serial port on the 5045A is completely dead. I finally did what I should have done to begin with and unplugged the serial cable from the 5045 that went to the connector on the case and probed the pins directly on the board. Nothing. It is really a shame that there is no service manual for this thing because there are a dozen test points just on the top board. Measuring some voltages at various places now. 

[edpalmer42]

So it's just wires from the 5045A to the back panel?  Odd.  Mine goes to the processor board for the front panel.  By the way, I checked the firmware on my main board.  You and I have identical versions so we should see the same data.

I see that U3 and U5 in the corner of the main board near the RS-232 connector are a 1488 and 1489.  Can't get much more 'old school' than that!

As for the service manual, I guess that companies that buy this kind of unit don't mind paying for repairs.  Rather than have their own techs muck about with it, they'd rather pay the bucks to ensure that the unit is in perfect condition.  I suspect that repairs to something like this weren't cheap - no matter how simple the fix.

Ed

[uncle_bob]

So it's just wires from the 5045A to the back panel?  Odd.  Mine goes to the processor board for the front panel.  By the way, I checked the firmware on my main board.  You and I have identical versions so we should see the same data.

I see that U3 and U5 in the corner of the main board near the RS-232 connector are a 1488 and 1489.  Can't get much more 'old school' than that!

As for the service manual, I guess that companies that buy this kind of unit don't mind paying for repairs.  Rather than have their own techs muck about with it, they'd rather pay the bucks to ensure that the unit is in perfect condition.  I suspect that repairs to something like this weren't cheap - no matter how simple the fix.

Ed

Hi

Equally likely:

The service manual has stuff in it that a competitor might be interested in. That's what killed off schematics of test gear in general. Now if you want the manual, you must have a "need to know". Next step is to get approved ( = yes you buy a lot of them).  Then your boss's boss signs the NDA and the manual gets sent over.

Bob

[edpalmer42]

Oh, I hate serial ports soooooooooooooooo much. 

Fighting with the leads, straight through, swapped, CTS/RTS, friggin' non-standard standard, crap!  Turns out the serial/USB converter I was using was FUBAR.  Anyway, from the port on the back of my box, 9600 7o2, using putty to send the D*1 and D*2 commands, here's the output.

Code: [Select]

=~=~=~=~=~=~=~=~=~=~=~= PuTTY log 2016.03.16 20:39:32 =~=~=~=~=~=~=~=~=~=~=~=
X
ID00000 XXX XXhXXmnXXs 2 R+Z ALM:01(16,00,00,00,00)C+000 F+000000 +25.35V Ct05.0
R-047RR +0043Z+012RZ -0060AR+5029PR0037AZ-4417PZ0062A0-0002GN*5.00LA+0007Pu-4499
+5.08V T+31.8 +14.9V -15.3V OLc FO14.4 VS17.5 VF0.75 IC15.3 HT09.0 IP000 +020 mV
00:00:00 00:00:00 O O O O O AC  33.3 00:00:00 00:00:00 M2m0 Y SYS:0000
XX
ID00000 XXX XXhXXmnXX Vs 17.5V Vf 0.75V GDAC 002 -15:+00 +15:+00 +45:+00 +75:+00
Pi+0000 Oz+0089 Fo:10.0 MHZ CDU:UK TXT:  09-14-96 ver 1.9
XX
The first block is the output of the D*1 command.  Second block is D*2 command.  The Capital 'X's at the beginning and ends of the blocks are control characters.  Notice the ID00000.  That shows that the 'serial number' is 00000.  On my unit, you enter the serial number, but it's volatile.  Since it defaults to 00000, I just leave it there.

Then I went back to the comm link on the 5045A itself and eavesdropped on the conversation between the 5045A and the front panel.  4800, 7o2 with putty.  Here's the results at startup.  I've reformatted it, removed the control characters, and provided translation.

Code: [Select]

D*1 00030           // First main status data
W07 00030 400000000 // Set programmable output frequency to 10 MHz
T*2 00030           // Enable Remote Control (I think this means programming through this interface)
W01 00030           // Frequency Adjust ??           \ These should have numbers with them.
W02 00030           // C-field Adjust ??              |  Maybe no numbers = reset to zero
W10 00030           // Filter Order & Time Constant  /
W06 00030 000000000 // Zeeman Servo On
D*1 00030           // First main status data
D*2 00030           // Second main status data
D*1 and D*2 commands repeat forever
It's odd.  Why unit serial number 00030?  That's not the serial number of my 5045A.  But it works.

I then sent the 'D*1 00030' command from putty using the ALT key trick for the control characters and the 5045A gave the expected response.

So basically, it does work as documented.  Imagine that.

Ed

[hydrogen maser]

I got the wiring harnesses separated out and yes: the DB25 serial output from the 5045A is wired directly to the DB25 connector on the back and does not go into any of the 4040A electronics. There are only 8 wires total from the 4040 to the 5045A - 2 RF's and 6 other wires. Of the 6 other wires at least 2 of them have to be ground and power so unless there is some serial data going over the 4 remaining wires (perhaps major and minor alarm and initialization signals?) this thing is pretty simple in concept - just a power supply and RF distribution board combined with the 5045A in a box. Still mapping out what wires go where but it is a lot easier now that I have the 5045A out of the case.   

[hydrogen maser]

OK, now we are getting somewhere. The input voltage to to 5045A is 29V from the 4040 power supply. The high voltage power supply on the 5045A has two input terminal sets: 24V and 10V. The 24V input measures: 8.1V and the 10V input measures 3.5V. Yeah, that might be a small problem. Starting to look it is has come full circle - a power supply problem, but not in the 4040 in the 5045. Time to tear more things apart

[edpalmer42]

I don't know what you mean by 'the high voltage power supply on the 5045A'.  Do you mean the ion pump supply, the EM supply, or something else?

Notice that both power supplies are at 1/3 of their proper value.  I'm suspicious of coincidences like that.

I don't think you'll learn anything new, but there's a datasheet on the 5045A here:

http://www.gigatest.net/datum/5045txt2.pdf

Here's an interesting page that shows some data from a 5045A.

http://www.delphelectronics.co.uk/gps/datum.htm

Ed

[hydrogen maser]

I mean the ion pump and EM, and the below pic would explain why those voltages were so low.


 

[hydrogen maser]

If it is just the capacitors, then this will be an easy fix.

[Vgkid]

I see at least 6 bad caps. The one barely in frame is leaking from the bottom 
 If the yellow axials are not glued in ... then there will be some issues.

[hydrogen maser]

The yellow axials are glued in, and I don't think what looks like corrosion coming from that top cap is corrosion, I think it is glue because it is very hard and looks like glue. All the blue electrolytic caps have to go and I think they are glued on the base. Anybody have any tips for how to get glued caps out without damaging the board? Hot air?

[Vgkid]

Xacto knife after desoldering, wear safety glasses!
That is if the glue is still holding..

[edpalmer42]

The fact that it's still putting out voltage is a good sign that nothing else has fried.

The three yellow axial components that I can see in the picture are inductors.  I can see one is labelled 20uh.  Are there others I'm not seeing?

One minor nuisance in sourcing new capacitors is that with improvements in capacitor technology, you'll find that 220uf 50v capacitors today are smaller than the old ones and won't fit in the holes.  You might have to step up in voltage to get ones that fit.  Actually, that's not a bad thing.  Usually, the glue holding the capacitors isn't very good.  But in a piece of equipment at this quality level - who knows - epoxy?

I've never looked inside that power supply section.  Maybe I should! 

Ed