Keyboard Shortcuts
ctrl + shift + ? :
Show all keyboard shortcuts
ctrl + g :
Navigate to a group
ctrl + shift + f :
Find
ctrl + / :
Quick actions
esc to dismiss
Likes
Search
How to check the accuracy of an HP 5342A frequency counter
|
Terry Maurice
I have two of these units, one has the oven the other does not.? The two units do not read exactly the same value when fed with a signal from a 10MHz FE-5680A, rubidium frequency standard.? I am monitoring the frequency standard with the two HP units and my Siglent SSA 3021X spectrum analyzer, all units being fed from the rubidium standard through a Downeast Microwave 10-4, 10MHz amplifier and filtered 4-way divider.? The units have been on for about 5 hours before taking the readings below.?
Unit A (with crystal oven) reads 9,999,828 Hz, +/- 2 Hz; unit B reads 9,999,830 Hz, +/-2 Hz and my spectrum analyzer reads 9,999,826 Hz.? Are these values within the expected values for these older frequency counters?? Is the FE-5680A correct? I checked both HP 5342As with their internal references and both show 10MHz.? How can I know, which unit(s) are correct, or do I assume my rubidium frequency standard is off by about 170 Hz? The FE-5680A does generate some heat when left on for prolonged periods, however, It is mounted on an aluminum heatsink and sealed inside a solid metal box, but the metal case is only slightly warm to the touch.? On my oscilloscope, the output of the rubidium standard shows a distorted wave, but once it passes through the DEM 10-4 it is a clean sine curve. ? Any help or comments would be much appreciated. Terry? |
|
Bob Albert
What I do is take the 10 MHz output from the counter time base (rear of unit) and compare to the rubidium 10 MHz signal on an oscilloscope X-Y display.? I adjust the counter time base to get the Lissajous pattern to be as stationary as possible. I suspect the standard is more accurate than the counter but it needs to be checked I suppose.? This is probably where? a GPSDO comes in.? However, 170 Hz is a gross error and you shouldn't have anything near that much.? I would guess a lot less than 1 Hz if all is working properly. Bob
On Sunday, April 26, 2020, 07:35:01 PM PDT, Terry Maurice <ve3xtm@...> wrote:
I have two of these units, one has the oven the other does not.? The two units do not read exactly the same value when fed with a signal from a 10MHz FE-5680A, rubidium frequency standard.? I am monitoring the frequency standard with the two HP units and my Siglent SSA 3021X spectrum analyzer, all units being fed from the rubidium standard through a Downeast Microwave 10-4, 10MHz amplifier and filtered 4-way divider.? The units have been on for about 5 hours before taking the readings below.? Unit A (with crystal oven) reads 9,999,828 Hz, +/- 2 Hz; unit B reads 9,999,830 Hz, +/-2 Hz and my spectrum analyzer reads 9,999,826 Hz.? Are these values within the expected values for these older frequency counters?? Is the FE-5680A correct? I checked both HP 5342As with their internal references and both show 10MHz.? How can I know, which unit(s) are correct, or do I assume my rubidium frequency standard is off by about 170 Hz? The FE-5680A does generate some heat when left on for prolonged periods, however, It is mounted on an aluminum heatsink and sealed inside a solid metal box, but the metal case is only slightly warm to the touch.? On my oscilloscope, the output of the rubidium standard shows a distorted wave, but once it passes through the DEM 10-4 it is a clean sine curve. ? Any help or comments would be much appreciated. Terry? |
|
Hi Terry,
toggle quoted message
Show quoted text
I had to read your question over several times to make sure I understood what you're asking. Please correct me if I'm wrong. It seems like you are measuring the rubidium standard with the counters and the SA while all are also taking an external frequency reference (presumably from the same source?) If that is the case, it would seem to me that either (1) you may have a problem with all three instruments...but the Siglent SA should be new enough that it's performing within specification I'd think...UNLESS (2) the Rb standard is off. Situations like this demand having a primary standard...either a 5071A caesium clock if you are really hardcore or just a good GPSDO for everything else, to compare your Rb standard to. Another really nice tool to have in your arsenal for this sort of work is a 53310A modulation domain analyzer, connected to a source of primary standard timebase. A huge benefit if the 53310A is the only external signal you need for calibration is a decently good 5V DC signal for one of the self-cal steps. It is an effective frequency microscope for aligning other equipment. Additionally, the datasheet for your counter (here's one: ) indicates a timebase error uncertainty at 10 MHz of between 1 and 10 Hz depending on whether the unit is equipped with an oven or is just equipped with the base TCXO, respectively. I would say that aspect appears to be in spec, but the Sean On Sun, Apr 26, 2020 at 07:34 PM, Terry Maurice wrote: I have two of these units, one has the oven the other does not.? The two units do not read exactly the same value when fed with a signal from a 10MHz FE-5680A, rubidium frequency standard.? I am monitoring the frequency standard with the two HP units and my Siglent SSA 3021X spectrum analyzer, all units being fed from the rubidium standard through a Downeast Microwave 10-4, 10MHz amplifier and filtered 4-way divider.? The units have been on for about 5 hours before taking the readings below.? |
|
Bob Albert
Since the SA and the two counters are not primary standards, someone down the line has adjusted them.? They all agree with one another but I'd vote for trusting the standard.? Align the counters and the SA to the standard. If you can find a way to verify the standard, do that.? But a gross error is unlikely from it. Bob
On Sunday, April 26, 2020, 08:18:57 PM PDT, <[email protected]> wrote:
...whoops hit reply too quick! To continue, the frequency reference appears like it needs to be checked. Bob's suggestion of using your o-scope in X-Y mode is a good place to start. Sean |
|
Bob Albert
One suggestion.? Compare to WWV and you will get your answer immediately. Bob
On Sunday, April 26, 2020, 08:37:34 PM PDT, Bob Albert via groups.io <bob91343@...> wrote:
Since the SA and the two counters are not primary standards, someone down the line has adjusted them.? They all agree with one another but I'd vote for trusting the standard.? Align the counters and the SA to the standard. If you can find a way to verify the standard, do that.? But a gross error is unlikely from it. Bob
On Sunday, April 26, 2020, 08:18:57 PM PDT, <[email protected]> wrote:
...whoops hit reply too quick! To continue, the frequency reference appears like it needs to be checked. Bob's suggestion of using your o-scope in X-Y mode is a good place to start. Sean |
|
On Sun, Apr 26, 2020 at 8:37 PM Bob Albert via <bob91343=[email protected]> wrote:
Unlikely that you could set an OCXO 170 Hz out. The OCXO in my 8566B was 76 Hz out compared to my Trimble GPSDO.? The closest I could get it was 70 Hz out.? I found a NOS OCXO and it was within 0.1 Hz after running for a couple of weeks. ?(The bad OCXO has since died after running it outside the SA.? No 10 MHz at all now.) I'd guess the FE-5680A isn't locked.? They do sit low, then every minute or so, sweep upwards looking for lock.? I'd put the FE-5680A on the SA and watch it from cold.? See if it locks. Orin. |
|
The simple answer is that none of the units are right, and all of them are just fine. You need to study up in the manuals on how the instruments work, what the reference frequencies do, the difference between internal and external references, and how things should be hooked up for various situations.
If you want everything to read the same, choose any one of the available 10 MHz sources (each counter and the SA likely has a 10 MHz ref input and output on the back, and means to select what it's doing), and connect it to all the external reference inputs and signal inputs on every piece of gear. You can buffer it with the amplifier/splitter, or just tee it all together - it will probably work OK this way at 10 MHz. Then think about what's going on. The Rb unit is probably the closest to exactly "right" in frequency, of all these pieces. You may find that the its apparent sine distortion is from not being terminated in 50 ohms, if you just look with a scope. Each independent reference source is slightly different in frequency with respect to the others, but when you use only one for all the references and signal inputs, they all should read the same. The absolute accuracy is a different story. Ed |
|
I should add that the Rb would have to be assumed working properly - if it's a used, questionable one, then you need to make sure it's OK, evidenced, by a "lock" indication a while after powering up - typically a half hour or less. I see upon reading the thread again, that you apparently see similar (but not identical) frequency readings from three different instruments. They read? slightly differently from each other due to their own references, but since they are much closer to each other than the apparent 170 Hz or so difference from the Rb, I'd vote with the others, that there may be something wrong with it. Be sure to look for that lock indication.
Ed |
|
The rubidium should be the most accurate of the bunch, except for the GPSDO. The method I have used is similar to the X-Y lissajous, but I use A, B ch, with B ch trig on a 100+MHz o-scope. Show both traces then adjust A freq source to stop the trace walking, decrease the sweep time down to the nsec. range, adjusting the A freq source for minimum trace walk. The faster sweep time gives you more resolution and allows you to adjust 10 MHz to within .001 Hz. Wait 10-30min. and recheck for A trace movement from the B triggered trace. I have achieved <1E-13 MHz separation from the 10 MHz reference source, using this method.
Don Bitters |
|
I fully agree to Ed Breya. Your Rb standard very probably is not locked, 17ppm off is a typical sign for that.
All such OCXOs (10544 or 10811) can never be out that far, if they are operating correctly (i.e. getting warm, please check). As both OCXO agree to each other, they seem to be working. Please check the LOCK signal of your Rb clock. After you identified the root cause, further hints how to correctly adjusti either of your clocks can be given. Frank ? |
|
Atomic references are in general, short lived. Cs beam's run
toggle quoted message
Show quoted text
out of vacuum, or out of Cs (some debate about that!). Rb references plate out the Rb metal on the lamp's glass bulb, and get to a point where not enough light can get past the plated out Rb metal film for the detector to operate. The Rb can very often be fixed when the lamp goes dark by using a heat gun to evaporate the Rb on the inside of the bulb where the light is emitted, and let it plate out somewhere else. The biggest problem is gaining access to the Rb bulb. Many of the compact units are essentially assemble only devices. When either a Cs beam standard, or a Rb reference runs out of "physics" to operate, its disciplined VCXO will end up against its rails too high, or too low in frequency (depending on how it was designed). Also, crystal ovens are a reliability problem. They are electrically heated devices, and as such must be protected against thermal runaway situations... lest they catch fire. [I have seen oven failures where the entire oscillator board melted its solder, and parts dropped out.] In many cases, the thermal fuses installed inside of the oven will fail, even though there is nothing otherwise wrong with the oven. But again, they are usually assemble only devices, and replacing the fuse will require serious removal efforts. -Chuck Harris Dr. Frank wrote: I fully agree to Ed Breya. Your Rb standard very probably is not locked, 17ppm off is a typical sign for that. |
|
Terry Maurice
Thanks Chuck for that information.? I suspect that my rubidium reference may be on the old side.? I bought it used over 8 years ago and I used it a fair amount as a frequency reference for my Flex 5000A, when I had that rig operational.
toggle quoted message
Show quoted text
The locked light does come on after a few minutes of operation. And thanks to all for the comments and suggestions.? I will work through them and see what I find. Terry On 2020-04-27 7:47 a.m., Chuck Harris wrote:
Atomic references are in general, short lived. Cs beam's run |
开云体育First, make sure the Rb standard is locked. Second, couple the output of the Rb to a receiver and check it against WWV. ?Adjust the coupling so that the Rb signal is about the same strength as WWV. You should be able to detect errors of a few Hz pretty easily. Even without a recent cal the Rb standard, if locked, should be better than 1Hz at 10 MHz. ?If the standard checks out you can then adjust the ref oscillator in each counter to read exactly 10 MHz on the counter.? However, I'm betting there is a problem with the standard, since all the counters show the same error.? Remember all these Rb standards were pulled because they were pretty worn out. I had one stop locking, and now have two, checked against a GPSDO. ?I can set mine to about .001 Hz against the GPSDO. ? Finally, after using the Rb for a while I realized that none of the internal ref oscillators are as stable as the Rb, so I just feed the counter the Rb as a reference. Since I got my Rb reference GPSDO have gotten a lot less expensive, so if your Rb is misbehaving I'd suggest buying one of those.? Ed. W2EMN On Apr 26, 2020, at 10:34 PM, Terry Maurice <ve3xtm@...> wrote:
|
|
FWIW General Radio in the instructions for one of its early frequency measuring assemblies recommended the "three oscillator" method of measuring differences. Where a source is being compared to WWV the third oscillator can be the BFO in the receiver. The BFO is set for an audio beat and the waxing and waning of the beat from the other two can be heard plainly. It is possible where the signals are steady to detect and measure fractional cycle beats this way. Since your standard runs on a WWV frequency it should be trivial to determine if its on frequency. Since you have three measuring instruments all showing the same approximate error I rather think the standard is off. If its adjustable you can likely get it closer using this method.
toggle quoted message
Show quoted text
?? My understanding from long ago is that a rubidium oscillator is used mainly as a filter to eliminate phase noise from a cesium clock. On 4/27/2020 8:22 AM, Edward Newman via groups.io wrote:
First, make sure the Rb standard is locked. Second, couple the output of the Rb to a receiver and check it against WWV. ?Adjust the coupling so that the Rb signal is about the same strength as WWV. You should be able to detect errors of a few Hz pretty easily. Even without a recent cal the Rb standard, if locked, should be better than 1Hz at 10 MHz. ?If the standard checks out you can then adjust the ref oscillator in each counter to read exactly 10 MHz on the counter. --
Richard Knoppow dickburk@... WB6KBL |
|
The “three oscillator” approach is a practice that has been used by NIST (NBS) for decades for comparison of their frequency & time standards.? It also provides a level of redundancy should there be a failure of one of the standards.? But now with improved methods of frequency/time generation (the current cesium fountain clock) it becomes more of a moot point as to what is accurate in their domain. ? The principle is simple: ? Compare source A against source B. Note frequency difference. But which source is off-frequency? Compare source a against source B against source C. Most likely you will see two sources that closely agree. More are better but there can be limits. ? Greg |
|
Bob Albert
Complicating the issue is the presence of a frequency adjust trimmer on each unit, which presents an unknown as to whether it has been adjusted and by whom.? It's a strong possibility that this voting process is flawed, and that the one renegade reading could be the correct one. However with an arror of over 100 Hz at 10 MHz, comparison with WWV is easy and desirable. We are not agonizing over one part in 10 ^-11 here, but a sloppy 10 ^-5. Bob
On Monday, April 27, 2020, 10:14:54 AM PDT, Greg Muir via groups.io <big_sky_explorer@...> wrote:
The “three oscillator” approach is a practice that has been used by NIST (NBS) for decades for comparison of their frequency & time standards.? It also provides a level of redundancy should there be a failure of one of the standards.? But now with improved methods of frequency/time generation (the current cesium fountain clock) it becomes more of a moot point as to what is accurate in their domain. ? The principle is simple: ? Compare source A against source B. Note frequency difference. But which source is off-frequency? Compare source a against source B against source C. Most likely you will see two sources that closely agree. More are better but there can be limits. ? Greg |
|
Another thing that can happen to an Rb source is that its built-in OCXO may age/drift too far out of its normal range, for the correction action to properly lock it to the Rb signal. However, the lock indication should not come on in this case, so there is probably something else going on.
I'm not familiar with the Rb in question here, but I know for instance, that old ones like my Efratom M-100s, have provisions for adjusting the OCXO back to a proper tuning range to acquire lock. One of my units had a part failure in its PLL, and was stuck way off in frequency until that was fixed, and the OCXO readjusted. It also properly indicated the failure by never indicating a lock, until after it was fixed. It's odd that you get a lock, with it apparently so far off. If you have a detailed manual for the unit, you should study it to see the internal frequency plan. A definite, constant offset from the desired frequency could be a result of the control loop locking to a wrong harmonic from the internal synthesizer, due to a part failure or misadjustment. That's the only way I can picture getting an actual, valid lock indication, yet wrong frequency. Most Rbs synthesize a very particular frequency from the 10 MHz OCXO reference, then multiply it up to the ~6.8 GHz Rb resonance, then use phase detection techniques to steer the OCXO in the right direction and place to get and hold a particular difference between the Rb resonance and the synthesized harmonic. There are various frequency plans used, but this difference frequency is often set up to be in the 100 Hz to 300 Hz range. Your 170 Hz error falls in this range, so I can picture maybe the unit has slipped a cog, so to speak, and is locked to a wrong harmonic, or on the wrong peak of the phase detector signal. If you can find a coarse adjustment on the Rb's internal OCXO, you can try tweaking it toward the right frequency, as measured on your other gear - assuming now they are correct and the Rb is wrong. If the Rb isn't busted, you may find as you tweak, that the lock indicator will go out, then magically pop back on when the output frequency reads near 10.000000... MHz. If this happens, then you have reset it to the right cog, and it should be OK operationally, and likely more accurate than the other gear. Then you can look into fine adjustments and perfecting the setup. Good luck. Ed |
|
I agree with Bob. I've used beat frequency method WWV with my HP counter and RF generator to set the internal references in each. Once I have used the signal strength meter on my Radio Shack HF receiver DX-160 tuned to WWV to move imperceptibly in no-fade conditions, I will use a guitar tuner also and offset my RF generator from WWV? highest broadcast frequency (20MHZ if possible) and offset by +- 440HZ (20,000,440 HZ . That is the "A" note setting on the guitar tuner, you can use a lower octave for more accuracy, (220 hz offset), but your receiver may filter that audio freq out and you won't know why you can't hear the 'beat whistle', and/or your guitar tuner may filter more than mine does, and not allow you to see the first octave (x2).? Then have the tuner mic near the SW radio speaker, Adjust the RF level out on the RF generator so it mixes with WWV and not overpower it, and you center steady up the tuner needle by adjusting the source, or whatever needs adjusting. Stay away from 5MHZ WWV, noisy and better accuracy during daytime up higher. At least 15 MHZ. I've done it at 25 MHZ also when it was broadcast too.
I would also verify the 10 mhz distribution amp by feeding it the source, and placing 2 of the amps outputs into counter ch A and ch B and setting it up for ratio, and the ratio should always be one. Just to make sure there are no "slips". NielsenTelecom |
|
Even calibration via WWV can be a real chore if you are located some distance from the transmitters and can only receive the sky wave instead of the ground wave signal.? The erratic movement of the received signal phase gets to a point where it becomes nearly impossible to determine where you are with the DUT setting unless you can do some time variant analysis.
Greg |
to navigate to use esc to dismiss