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

[edpalmer42]

I can't think of anything else I'd like you to check, but thanks for asking.

As for the restart error, mine does that too.  Startup is always with a major alarm.  You reset it after lock.  I guess it's just to warn you that the power was off - otherwise you might not realize it.  That could cause major confusion in your analysis until you figured it out.  Right up there with having the power cord fall out.   

Ed

[uncle_bob]

Ok, so the 4040A has at least one more problem or not? If you put it into storage mode and then later bring it out of storage mode you get a "Unit Restart Fault". It has always done this but I thought it was because the previous times I powered it down I removed the power cable to totally shut it down and it has complained about this before. In the manual it is Major Alarm 16 and it just says "Unit Restart" Perhaps this is not really an error and it is just letting you know that it was restarted? I assume these things are intended to be always running and never shut down? Just informational and not indicating that there was some problem?

Hi

The HP5071A does the same thing. When it restarts it flags the fact to you can log the interruption. One assumption is that you ran out of battery power and may want to go buy another battery.

Bob

[hydrogen maser]

monitor3 software is interesting and useful (www.evoria.net/PRS45A/Monitor3.zip), it is windows based and runs just fine on Win7 pro 64bit. It will save all the data in a csv file and has better graphing. Some things don't work in this software: the alarm reset does not work and the download event log also does not work. I assume the event log download does not work because perhaps there is no event log for the model I have - there is no mention of an event log in the 4040 manual or the dos monitor software. It also misidentifies my model as being 4201A rather than the 4040A, I assume it is for later models and they did not even care about models from a different era, perhaps that is why the alarm reset does not work - it needs a slightly different command?.

[edpalmer42]

I forgot to mention Monitor3.  It doesn't work at all on my unit.  The doc file in the zip file confirms that it partially works on 5045A-based units.  I guess if I bypassed my user interface and connected directly to the 5045A that it would be usable on my 4065A, too.

Ed

[hydrogen maser]

Yeah, it says it does not work with the 4065, I also guess that is because the serial port on your 5045 is involved with the rest of your unit while it is not one mine. There are only 6 wires that go from the 4040 into the 5045 and initially I thought that perhaps it was really simple: one power, one ground, and the other 4 wires were just enough to carry out the 4 states that the 4040 displays: initialize, lock, major alarm, minor alarm through via simple open/closed or logic level states, but that appears not to be the case. 2 of the wires are ground, one power and I don't know what the other 3 wires do. I then thought those three wires might be serial data (still seems likely) but as soon as I touched the scope probe to one of the pins for those wires I heard a relay close in the 4040 and decided it was a little risky to be doing that so I left that alone.

[edpalmer42]

Here's my info on the signal level from the Cesium Tube test point.    My thought was that since yours was hopefully undisturbed, I could adjust my level to equal yours. Since I now have your info, I decided to play with mine. 

If I'm reading yours correctly, your baseline is -100 mv, your jumps are plus and minus 50 mv from the baseline, the spikes peak at -250 mv, and the noise level is 200 mv p-p.

First, the sense of the pot is backwards.  Clockwise decreases the level, CCW increases the level.  If I adjust mine for a 100 mv baseline, my jumps are -75 mv, +25, +50, and +75 mv. (!), the spikes peak at -175 mv, and the noise level is only 50 mv p-p.  If I use the 20 MHz bandwidth limit, the noise level drops to 30 mv p-p.  Nothing else changes.  I often have to use the BW Limit because there's a local FM station that regularly shows up on my scope.   

Adjusting the baseline more negative than ~-200 mv causes errors 02, 07, and loss of lock.  Adjusting the level at all changes the Ramsey level which changes everything else.  There's a slow servo (many minutes) that tries to return the Ramsey Level to 2500. 

I had to go old school to get these pictures.  The vertical scale is 50 mv / Div.  Horizontal is 10 ms / Div.  The 20 MHz bandwidth limit is engaged.  The baseline is -100 mv.  One of these days I've got to get me one of them new-fangled digital oscilloscopes.  The problem is that this ornery old HP 1744A analog storage scope just won't die and give me an excuse!  It's over 30 years old for crap sake!  What do I have to do, drive a stake through its heart? 


I'm not sure what this tells us, if anything.  Is your tube on its last legs?  Is there a subtle fault in your unit causing higher noise levels?  Yours is serial #02350, mine is #01067.  Mine has that mezzanine board, but yours likely has those circuit changes already rolled into the main board.  Even though yours has more noise, you're getting better StdDev values than I am.  Do I have a noisy circuit in my unit?  You're seeing numbers in the 130 - 140 range while I'm getting 170 - 180 although it seems like that number changes everytime I run the unit.  Now that I've got the level set I'll have to see if it stabilizes.  Maybe there's a sweet spot for that level adjustment.


Ed

[hydrogen maser]

Hm, your traces look quite different from mine, perhaps some of it is due to the "nosier" nature (appearance anyway) of digital scopes or perhaps it is something different in my unit as we do not have identical main boards.

[edpalmer42]

Does your Rigol have a 20 MHz Bandwidth Limit?  If so, what does your signal look like with it engaged?  I have no idea what bandwidth the buffer uses.  I can't find any data on the buffer chip which, I think, is U27 - a Burr-Brown 49522 in my unit.  Since the important features seem to be the DC levels rather than the slopes, I don't think it's a high bandwidth circuit.  The slew rate on the spikes is only about 0V050 per ms, i.e. 50 volts per second.

Ed

[hydrogen maser]

I will take a look at it with bandwidth limiting after I look into something here. Remember those 4 caps that I was (as it turned  out unnecessarily) worried about because of the low impedance on them? Looks like one of them might not be working. Because the noise on the 5v rail was higher than all the other rails I thought it was worth looking into that. I went back to check the voltages on all my caps vs what your unit shows and am seeing some strange results. It might be that it is just because it has not locked yet, I will take a look at it after it locks. 

[hydrogen maser]

Well this looks very wrong. Cap CI reads -1.28v on the POS lead and zero on the neg referenced to ground or just itself floating. CI is in parallel with CH - CH reads +4.89 on the POS lead and zero on the neg. What the heck is going on? And no, I did not put CI in backwards, unless the silkscreen is reversed from what it should be on both sides of the board. I had trouble with the positive pads on both sides of the board for CI, they lifted off because there are not any top or bottom traces for it, just on the internal layers and some of those old caps were a little stubborn about coming out.

[edpalmer42]

Looks like the plus lead on CI isn't connected.  Did you check for continuity?

Remember that your values won't be quite the same as mine.  I'm going to modify my guesses on what the voltages are for:

+ 5 volts:  CA/CB, CH, CI
+15 volts: C20/CF
-15 volts:  CD/CE
+24 volts:  C19/CC
System Voltage:  C3/C4
Cs Oven:  CH/CI
CG/CJ minus lead to gnd:  Mass Spec voltage
CG/CJ :  Ionizer voltage
C18: Unknown

For the Mass Spec and Ionizer voltages, the D*1 command reports the readings, the D*2 command reports what they should be.

Ed

[hydrogen maser]

The problem with electronics repair is that humans are not well equipped for it because you often need three hands. Think I see what might be the issue.

[hydrogen maser]

Yep, my lifted pad via repair skills suck. I was in hurry because I got excited, it was looking like this thing might be fixable and I did it the very easy way and just stuffed as many lead wires as would fit in the via - test for continuity and shorts - good - soldered it - test for continuity and shorts - good. But it did not take because I think I did not flow enough solder into the joint and it just had a mechanical connection. Put more flux and more solder on the joint and it tests good now, but given that it looks like this device works I think it is time to step back - remove that cap - and do a proper via repair (a new skill to learn ). Alternatively I could just remove the cap. The old cap was a 220uf in parallel with another 220 and all the 220's got replaced with 470uf so a single cap should work fine in this case as long as the multiple internal traces (it looks like there are two) that go to the positive side of that cap don't mind being disconnected from each other (if they are) and the 5v rail at that point.   

[edpalmer42]

It's just a power supply capacitor.  If you'd used 220uf capacitors, I'd suggest adding a jumper wire between CH and CI.  Since you doubled the size from 220 to 470 I'd just leave it alone since it *is* working.  The more soldering you do on an old board, the more traces and pads will suffer and the more likely that something will go horribly wrong!

Ed

[hydrogen maser]

I put it all back together and that cap measures OK now - same as its twin. Everything looks good since it has locked. No significant differences in this run since the last one. Ed: I looked at that waveform with 20mhz bandwidth limiting and it does not look that much different than it did before on my scope, it certainly does not look like the waveform images you posted. Don't know what to make of that.   

[edpalmer42]

I wouldn't worry about it.  It's locking and you're getting good StdDev readings so, for now, I think you're good to go.  Keep some notes and logs showing the current status and watch for variations and trends.  Power it up for an hour or two every six months to let the ion pump keep the tube healthy and hopefully, it will provide good service for many years.

As Dave would say, "Winner, winner, chicken dinner!"

Ed

[hydrogen maser]

All the voltage rails look better now, yet to be seen what happens with deviation. 

[edpalmer42]

Hi HM,

I just stumbled across a piece of information on your Cs tube that I thought you'd like to have, if you haven't found it already.  I got it from the Wayback Machine searching for www.ftsdatum.com .  The rows of periods are part of the quote.  I didn't cut anything out.

Model FTS 7100 Series . . .

. . .provides standard cesium performance with an expected life of between eight to eleven years. Used in our own instruments, these tubes also fit Hewlett Packard HP5061A or B Cesium Instruments and FEI Cesium Instruments. In either case, the tube is a drop in replacement. Contact your local representative or FTS directly for more details or a quotation.

Model FTS 7500 Series. . .

. . . provides standard cesium performance with an expected life of between ten to thirteen years. Used in our own instruments, these tubes also fit Hewlett Packard HP5061A or B Cesium Instruments and FEI Cesium Instruments. In either case, the tube is a drop in replacement. Contact your local representative or FTS directly for more details or a quotation.

Model FTS 7600 Series. . .

. . . . . .available in two basic configuration. The first configuration is a high performance cesium tube with an expected life of between four to five years, and the second is an extended life tube providing standard performance with a life expectancy of fifteen to twenty-two years. Used in our own instruments, these tubes also fit Hewlett Packard HP5061A or B Cesium Instruments and FEI Cesium Instruments. In either case, the tube is a drop in replacement. Contact your local representative or FTS directly for more details or a quotation.

My tube is model #7504A.  Yours is #7614-1 ( -10?  Not sure.).  So now, is that the high performance / short life version or the standard performance / long life version??    Not sure, but the fact that your unit doesn't say anything about having the high performance option is promising.

I tried Monitor3 and it works on mine when I connect directly to the 5045A module.  I'm going to see if I can hack the program to run when hooked to the rear connector, i.e. to the 4065A port rather than the 5045A port.  Here's a screen with a few of the more interesting parameters.  I have no idea what caused the spike in the Clock & Zeeman signals.  But it had almost no effect on the variance or the output.

Ed

[hydrogen maser]

Perhaps a rs-232 splitter? My tube is the FTS-7614-12, can't find anything on any of the other part numbers yet, still looking. I'm hoping mine is the long life one, seeing as how mine is one of the simplest 4000 series and has no options installed in it, it does not look like anything special, but who knows. Found some interesting docs in the wayback on the 5045A - the pdfs are still there and accessible.

https://web.archive.org/web/19990221231232/http://www.datum.com/res_technical.html

Right now I am mapping out all the wiring that comes out of the 5045A because I think the 4040A chassis is really unnecessary and I want to see if the 5045A will run without it and just a lab power supply but I need to figure out how the run/storage signals are being sent into it, or I guess I could just take the 5045A out, power it up and see what happens on the serial port. Would not surprise me if it was happy with that and went right into initialize and run. Taking a while because I have not had a lot of time to play with it.   

   

[edpalmer42]

An rs-232 splitter might work, but I'd rather just connect to the existing port on the back.  I want to compare the query/response chatter between the 5045A and the computer vs. the 4065A and the computer.  I also have to check that I'm not having yet more rs-232 connection problems.  It also looks like Monitor3 will accept COM ports higher than 9, it just won't work with them! 

According to the data sheet for the 5045A, all it needs is 18 - 64 volts at 30/55 watts (operating/warmup).  But I don't have any detailed info on the connections.  To check the storage mode, look for leads into the 5045A that change state when you plug in the AC cord while switched to storage mode.  Power for the ion pump has to get in there somehow.

Ed

[hydrogen maser]

Figured out how the run/store works. The DB15 that goes into the power supply section of the 5045A on my unit is wired as such: Only 4 pins populated in the wiring connector - Pins 1 and 5 are ground, pin 8 has 29V applied when in run mode and zero volts in storage, pin 10 has zero volts in run mode and 29v in storage. Pretty simple and it makes sense that there are only power and ground wires going into the power supply board for 5045A. There are only two other wires connected from the 5045A to the 4040 chassis. The other DB15 that comes off the main board on the 5045A has only two pins populated in the connector attached to it, pins 14 and 15. These almost have to be serial data and it would make sense that serial data would come from the main board. Spent a good amount of time untangling all the 4040 wiring and getting it separated into different bundles grouped by purpose. It was worth doing because there are not that many different wire colors and it was really confusing about what went where. Making progress.

[edpalmer42]

On my unit, the 5045A has a DB-25 connector with 3 wires for the serial data (xmit,rcv,gnd), then a DB-15 with 2 wires for something, then a DB-15 with 4 wires to the power supply.

It doesn't look like I'll be able to hack Monitor3 to work with my unit.  I hoped it was something like a signature that was different between the 5045A and the 4065A, but it turns out that the 4065A puts out more data so a simple hack isn't possible.  Wiring out the 5045A isn't an option because if I do that, I'll lose the front panel display.  I think I'll throw together a program that polls the unit through the rear panel, parses the data, and puts out a file in the same format as Monitor3.  That will allow me to use Monitor3 or any other plot program to display the data.

Ed

[hydrogen maser]

Ed, you were saying:  "The ion pump won't run in storage mode." in an earlier post. You might check that you have voltage/no-voltage swapped between pins 8 and 10 of the 5045A PS DB15 when in storage mode vs run mode. If there is no voltage (or not enough voltage+current) on pin 10 when in storage mode that might explain why the ion pump was not running in storage mode.

[hydrogen maser]

My long term plan for this device is that I would like to have a display on the front with the time (7 segment LED's?) + a touchscreen display running monitor3 and/or whatever software on windows 7). Should be able to get thin versions of both of those things and bolt on the front, serial to the touchpad, perhaps arduino or raspberry to hook into the main serial or this other mystery two wire for 7 segment or other stuff. Not going to be a quick project, but I don't see why it is not doable. Can also use NTP on the touchpad via wifi to get approximate current time for sync, not entirely all thought out yet. Should be fun.

[edpalmer42]

My long term plan for this device is that I would like to have a display on the front with the time (7 segment LED's?) + a touchscreen display running monitor3 and/or whatever software on windows 7). Should be able to get thin versions of both of those things and bolt on the front, serial to the touchpad, perhaps arduino or raspberry to hook into the main serial or this other mystery two wire for 7 segment or other stuff. Not going to be a quick project, but I don't see why it is not doable. Can also use NTP on the touchpad via wifi to get approximate current time for sync, not entirely all thought out yet. Should be fun.

Sounds like some interesting upgrades, but I wouldn't do that much work for a Cesium.  I use an Efratom FRT Rb standard as my house frequency standard.  It runs 24/7 and I distribute the signal to all of my test equipment that can accept a 10 MHz input as well as using it for a precision 10 MHz source.  I suggest that a unit like that would be a better candidate for the kind of upgrades you're considering rather than a Cesium that will only occasionally be turned on.

I think some models of Cs standard have an 'OCXO mode' where the Cesium tube is powered down, but the OCXO is hot and online.  If you hacked the 5045A so that was possible, then your upgrades would provide more value.

Ed

P.S.  I see that this message thread has over 4600 views.  I'm assuming that at least some of you lurkers have a 5045A-based Cesium unit.  This isn't a private conversation, you know.  If you have any questions or info to add to the conversation, please contribute!