Re: [time-nuts] FE-5680 frequency jump
I found the cause of the 4 mHz frequency jump. I have an LPRO-101, an FE-5680, power supplies and a Motorola M12T GPS board in a surplus case. When I put the case away to work on another project I piled the hockey puck antenna and lead in the case and it happened to land on the FE-5680. I noticed the antenna stuck firmly to the FE-5680 case when I got the project back out. I found I could get a 6 mHz (0.6 ppb) shift comparing the frequency with residual magnetism from the magnet stuck on the FE-5680 case to the frequency after demagnetizing the FE-5680 case. Unintentional C field adjustment. Dope slap, live and learn. Bob -Original Message- From: Magnus Danielson Sent: Monday, April 01, 2013 5:13 PM To: time-nuts@febo.com Subject: Re: [time-nuts] FE-5680 frequency jump On 04/02/2013 01:12 AM, Attila Kinali wrote: On Mon, 01 Apr 2013 23:57:48 +0200 Magnus Danielson wrote: On 04/01/2013 10:06 PM, Bob Quenelle wrote: I’ve been running an FE-5680 for maybe a total of 50 hours over the last several months. I found that an offset setting of 180 made it track GPS and (previously-set) LPRO-101 10 MHz signals. Even with power cycling, after about 1/2 hour, with an offset setting of 180 the FE-5680 was stable. The last time I turned on the FE-5680, it drifted with a setting of 180 and needed a new setting of –415 to track the other signals. That’s a change of 595 counts and with a resolution of 6.8 uHz per count, a frequency change of 4 mHz (0.004 Hz) and 0.4 ppb. Operation at the new setting is stable for now. The lock signal indicates lock and the power supply voltage is still 15V. I haven’t checked lamp voltage or VCXO voltage as that requires opening the case. How long have it been turned on since last power-up? Let it sit for a day at least. I've found that it is easy to be in too much hurry to judge the situation and trim things efter power-up. The crystal oscillator just doesn't get the time to settle in. That might be indeed the case. Figure 3 in [1] gives "quite high" frequency aging differences after switch on and long run time. Attila Kinali [1] http://www.pi5.uni-stuttgart.de/common/show_file.php/lectures/100/blaetter/The%20Rubidium%20Clock%20and%20Basic%20Research.pdf You are confusing the VCXOs frequency drift with that of the rubidiums (which is the result of the FLL locking of the VCXO to the rubidium "resonance"). If the VCXO still has a fair distance to drift, then false locking can occur while compating the initially quite vigorous drift rate. The only real way to handle that is to sit and wait for it to settle down. Only after that may trimming of the oscillator be done to zeroize the integrator state. A small commercial rubidium doesn't need very long to "get a feel" if it is in good condition or not, but sitting on your hands and let it warm up gives you a fair idea of just how skewed situation it is. That's also true for caesium clocks. So, sit on your hands and let it settle. Better yet, leave on while you do other things. Just recall to put enought cooling on it! Cheers, Magnus ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
[time-nuts] FE-5680 frequency jump
I’ve been running an FE-5680 for maybe a total of 50 hours over the last several months. I found that an offset setting of 180 made it track GPS and (previously-set) LPRO-101 10 MHz signals. Even with power cycling, after about 1/2 hour, with an offset setting of 180 the FE-5680 was stable. The last time I turned on the FE-5680, it drifted with a setting of 180 and needed a new setting of –415 to track the other signals. That’s a change of 595 counts and with a resolution of 6.8 uHz per count, a frequency change of 4 mHz (0.004 Hz) and 0.4 ppb. Operation at the new setting is stable for now. The lock signal indicates lock and the power supply voltage is still 15V. I haven’t checked lamp voltage or VCXO voltage as that requires opening the case. Before I start poking around, thought I’d ask if anyone has seen a similar symptom. Time to open it up? Thanks, Bob ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
[time-nuts] Simple method for comparing 10 MHz signals
I kept putting off buying a nice counter and finally decided to try a phase detector circuit to compare 10 MHz standards. It’s not novel, but I like the results so far. It lets me see things I couldn’t see before. I thought the idea might be useful to some of us who are equipment-limited. The graph shows an LPRO-101 as the white trace and an FE-5680 as the red trace, both compared to a simple GPS standard. The graph is just an example of a data collection run and doesn’t represent any particular level of performance. It does show a lot of common mode change, indicating the GPS is changing during the run. Maybe I should say probably changing. The whole breadboard circuit has 4 IC’s. The blue trace is a measurement of the case temperature of the GPS standard. The circuit uses 1/2 of a 74HC4015 4 bit shift register for each channel. The D input of each 74HC4015 gets the Q-D output inverted by a gate from a 74HC04, forming a divide by 8 “Johnson counter”. At the beginning of a run all 74HC4015’s are simultaneously reset. 74HC86 XOR gates are used as phase detectors. One input of each XOR connects to the Q-A output of the GPS 74HC4015 and the other input connects to the Q-C output of the LPRO-101 or FE5680 74HC4015. Using different taps gets the initial state of the XOR output close to 1/2 scale and known slope. The average value of the XOR goes from 0 to full scale for a phase change of 180 degrees. 180 degrees of the divide by 8 corresponds to 400 nsec, +/- 2 cycles of 10 MHz. I already had a LabJack U6 data acquisition unit, which has several analog inputs and digital I/O. Other similar products are available and inexpensive. LabJack has free data-collection software so you can get a file usable by Excel or whatever without writing any code. For me it was easy and cheaper to convert the phase signal to a voltage and read it. This approach isn’t useful for comparing PPS signals and isn’t as accurate as using a good TIC. I’m looking forward to the TIC design in progress, but this project seems useful for now.<>___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] EIP545A 18GHz counter query
Not that it's likely to help you, but a signature analyzer uses a shift register with feedback to generate a 4 hex character "signature" from a serial data stream. It would only help if the troubleshooting tree includes a list a bad signatures for specific failures. http://www.prc68.com/I/HP5004.shtml Bob -Original Message- From: Chuck Harris Sent: Tuesday, November 27, 2012 7:45 AM To: Chris Wilson ; Discussion of precise time and frequency measurement Subject: Re: [time-nuts] EIP545A 18GHz counter query This one is usually an easy fix. The EPROMS on the controller card are using tin plated sockets, and they become tin-whisker cities. The counter will usually have enough oomph to blow any transient whiskers away if it is left running, but if it sits the whiskers will grow quickly and prevent the CPU from passing its power on self-test, and you get the "--" display. Take a high pressure air gun and blow under, around and through the EPROM sockets from all directions and angles. When you plug the board back in, it should start to work again... for a while. A more permanent fix involves removing the sockets and replacing them with gold plated sockets with machined pins. Also, on many of the 545A counters there is a design mistake on the power supply board where the wire tie that holds an electrolytic capacitor passes through the board. The holes the wire tie passes through are plated, and come through alongside of the +9V unregulated traces... bringing ground and +9V together. Drill or file the plating out of the holes to prevent spurious blowing of the mains fuse. -Chuck Harris Chris Wilson wrote: 27/11/2012 14:18 I bought a 18GHz EIP545A counter which the vendor said was working fine the day before and when on overnight soak test, and also when last used some months agao. But when he checked it on the morning of my coming over to see it he found it had developed a fault... I bought it off him cheaply, as seen, hoping it might be fixable. Here's the basic tale: It was acquired displaying just dashes. I checked the PSU voltages and found the PSU section outputting well away from the manual figures. I corrected these and checked supplies for ripple and they were all good. I then decided to remove any board that was not essential to operation. It came with a GPIB option board, so I pulled it. Luckily this did some good and a working display appeared. There were three tantalum caps on this board, as a matter of interest I pulled a leg on each and tested them, all were OK. Refitting it killed the display back to "dashes" again, so I set it aside as having an "unknown fault", and continued. I can get it to accurately display up to EXACTLY 188. MHz. Inputting 190. MHz gives zeros. It has three frequency bands. Band one works perfectly right up and well beyond its 10Hz to 100MHz range. Band two often doesn't work at all, and just diplays zeros, it's a 50 ohm input, compared to the 1meg 20pF Band 1 input. Sometimes by ramping the frequency up slowly from 100 MHz I can get a seemingly random reading. Band 3, the GHz range, doesn't work at all. Again, only zeros are displayed. The machine has quite a good range of self diagnosis tests. Tests are via the key pad. The first test checks the VCO and other stuff, and should display an accurate 200. MHz. It displays well over, always in the range 253.5 and the display isn't stable, it hovers over several KHz. There's a tree menu to see where this issue could lie. One limb suggests using a known good counter on the output of the VCO. I did this and the output is miles high in frequency, about double, and unstable. The tree menus goes on to suggest a phase lock circuitry fault. Further tests seem to depend in part upon using something called a Signature Analyser, which I have never even heard of They suggest an HP5004A, which is apparently pretty ancient. Is there a cheap way of acquiring something to do this sort of testing? I believe given a few pointers my scope, multi meter, sig gen and my other (working) Racal counter could take me further. It's a nice old thing, and I would quite like a means of counting into the higher frequencies it offers, but don't want to spend too much time or money on it. It works a damned sight more than when it first landed, which has kind of given me incentive to push a bit further, given I have a .pdf copy of the repair manual. Here's the page of the schematic I think is relevant, if it is a phase locked loop problem. This morning I realised there's a chip missing from a socket on the board in question. It's U6, a flip flop, part of the pre scaler I think? I assumed it was for some option, not fitted, but I am not so sure now. A Chinese Ebay seller is breaking one of these machines and he lists all the boards seperately, with decent photos. His board has this socket populated, and looking at the schematic, (linked at the bottom of this post), I think it's probably a vital componen