Eff 'em Tea

[X-Rated]

Naw. I am not in the wrong forum. I want to discuss frequency measurement. Specifically, precision frequency measurement.

ARRLink (http://www.arrl.org/frequency-measuring-test)

I have done actually one of these minor FMT's a few years back. I did pretty well considering I used no computer aids like so many are doing now. But a lot of this frequency measurement is my life. Except for the atmospherics causing phase shift conditions, it is what it is.

Now, the ARRL is doing what has been messed with for a while. They are going to transmit 2 carriers at the same time. One is going to be specified. The other is the one you have to guess.

So, how to do this? Well one way is to use an AM receiver and record the resulting beat frequency and algebraically add the measured tone to the known RF signal. I have not done this stuff, but some use a program that is called TimeLab and that can analyze the audio frequency from your radio.

I can hear you now. "But X-rated, the timebase in my computer is not that good. It may be 10ppm off frequency and that will mess up my measurement."

Let's analyze that error. We have a timebase that is 10ppm off frequency in your computer, so the 2kHz signal you receive into your computer will be 0.02Hz off frequency. Since you are given the frequency of one of the signals, you know what the reference is and all you need is the offset frequency. So if your analysis is 2194.2491Hz, then you add the 14120kHz and get 14122.1941kHz and Voila, you have a frequency to submit. If you did everything perfectly and there were no atmospheric issues, and maintained a 10ppm error in your computer, your result will be about 1.4 parts per billion (0.02Hz) off frequency, good enough to be proud of your result.

Soundcard issues. Obviously if you have a better soundcard, the program will result in a better resolution. If you eyeball your waterfall you may be able to interpolate a good frequency.

There are several programs out there. Honestly, I have not used any of them. These are all frequency based solutions. For the best results, phase based measurements are the best.

[N8YX]

27.02500000

No excuses.

;)

[X-Rated]

Mercy sakes.

[X-Rated]

I checked into our local tech net last night as they were discussing this test. A lot of people have a lot of different ideas. It was hard trying to understand what others are saying sometimes.

The problem with measuring frequency is that you have to account for all excursions. With the atmosphere playing its tricks on us, this can be a challenge. When you look at the phase, as long as your measurement does not move more than a period, missing phase measurements no big deal.

Some people whined that their radio does not do AM. Well, what can I say? The programs can record more than one CW signal and you can still measure the offset between those.

[N8YX]

I'll delve into this later, when I can post from a real computer. Drift will be one of the topics.

[X-Rated]

I'll delve into this later, when I can post from a real computer. Drift will be one of the topics.

I don't get your drift. if you are talking about the drift of the radio, you won't have to worry about this in this test IF you use AM to detect the beatnote. If you use a SSB receiver, the two signals will have the same drift. Then there is the drift of the computer frequency reference. There is not much of that left on the audio frequencies as I have pointed out.

[N8YX]

I'm talking about the drift of the timebase used as a counter reference. You could always go with a rubidium standard such as a Ball-Efratom, or use something which is phase-locked to WWVB.

Spectracom made numerous examples of the latter. One of their offerings actually measured and charted drift as a function of propagation (atmospheric) variance. It output the difference as a proportional voltage which lent itself to automatic compensation schemes.

If a total drift figure could thus be extrapolated based on a frequency/propagation-corrected model then used to calculate an offset for an observed signal, one might be able to obtain still more accuracy when making the measurement - yes?

[X-Rated]

Well, I steer clear of the term "drift" because it has many connotations. Some view it as an offset frequency, some view it as a changing frequency, and some are referring to atmospheric affects.

I own a rubidium. It is pretty cheap.
http://www.ebay.com/itm/10MHzOut-FE-5680A-Rubidium-Atomic-Frequency-Standard-Oscillator-Transceivers-/280905645392?pt=LH_DefaultDomain_0&hash=item416747fd50

I also shot this video of the rubidium signal when compared to an OCXO in the lab. We can see some phase shifting, but what we don't see here is phase pops. That is excellent. Hopefully when you buy a rubidium reference, you will find that it is clean like this one.


http://www.youtube.com/watch?v=dGFVpUsBUQk

I actually used no counter in one FMT.

http://www.k5cm.com/april-15-2009.htm

As you can see there, I was able to get excellent results and no counter was used. I used a signal generator to generate a precise known frequency and I generated a beatnote of 400Hz. I could compare that 400Hz with a known good 400Hz. When the phase matched on the oscilloscope, then I recorded that frequency. At the time I was using a signal generator that had an RF out and an AF out based on the known good GPS reference.

[X-Rated]

I'm talking about the drift of the timebase used as a counter reference. You could always go with a rubidium standard such as a Ball-Efratom, or use something which is phase-locked to WWVB.

Spectracom made numerous examples of the latter. One of their offerings actually measured and charted drift as a function of propagation (atmospheric) variance. It output the difference as a proportional voltage which lent itself to automatic compensation schemes.

If a total drift figure could thus be extrapolated based on a frequency/propagation-corrected model then used to calculate an offset for an observed signal, one might be able to obtain still more accuracy when making the measurement - yes?

I am so sorry Fred. I am too stupid to know what you mean here.

To me, "drift" is a frequency that changes over some domain. Maybe that is what you mean. Having a reference with no drift is critical in a game like this. I usually refer to this as frequency stability. There is frequency stability over time, over voltage changes and over temperature changes. There are others such as over orientation and G-force changes.

One item that is very cheap on ebay that has excellent frequency stability. LINK TO CMAC (http://www.ebay.com/itm/CMAC-10MHz-ocxo-oscillator-square-wave-sc-cut-/180911738566?pt=LH_DefaultDomain_0&hash=item2a1f2e0ec6) This is a well kept secret. These are tiny oscillators that burn about a quarter watt when stable at room temperature. They use a very good SC cut crystal That as a high-Q. Since it is an SC-cut crystal, the frequency delta over temperature changes is very small. I own 3 of these. I use one at home, one at the office and one for a spare. But I do look at the frequency from time to time and it remains to within 1x10-8 easily over the last 9 months without adjustment. Realistically, you would want to adjust these every so often to make sure they are where you want them to be. These have no mechanical trim adjustments and are totally controlled by an electrical trim input voltage.

When shopping for OCXO's (Oven Controlled Xtal Oscillator) be sure you are getting what you want. I always look for a unit with an electrical trim. If you really want one with a mechanical trim, you can get a regulator and a pot and make your own mechanical trim. You cannot make an electrical trim from a mechanical trim OCXO. Some OCXO's have both options built in.

TCXO's (Temperature Compensated Xtal Oscillators) are not nearly as good as OCXO's. For most amateur work, they are more than adequate. They are typically accurate to within a half ppm (5x10-7) whereas the OCXO's are settable and accurate to within a part per billion (1x10-9). I have designed double oven oscillators with temperature offset to under 1x10-10 from -30C to +70C. Link to ebay knockoff. (http://www.ebay.com/itm/10MHz-OCXO-MORION-Crystal-Oscillator-Double-Oven-ULTRA-PRECISION-MV89A-precision-/280864929676?pt=LH_DefaultDomain_0&hash=item4164dab78c) I have had several designs get stolen and used by other companies. And here I sit broke while they are making all the money.

Wenzel Link. (http://www.ebay.com/itm/Wenzel-Associates-10MHz-OCXO-w-data-sheet-/121075940327?pt=LH_DefaultDomain_0&hash=item1c30b03be7) Wenzel is considered to be the top of the line OCXO. I have never evaluated one. Charlie Wenzel was an engineer at Austron and quit there to start his own company. He had many breakthroughs in low noise and is recognized by the National Institute of Standards and Technology as being one of the best for low phase noise designs. In this link, let's examine the auction.

The one picture includes the specification.

+7dBm into 50 ohms. (Kinda small, but you will need to know if you can live with that without amplifying. It will be sinewave)
-165dBc @10kHz Phase noise. (Pretty good. You will expect at least this from a Wenzel. Many will be -155dBc at 10kHz. Will you see the difference? Probably not)
+/-1x10-7 @0 to 70C. (Good enough for amateur work. The shack will probably be 22+/-2C.
1x10-6 per year aging. (Look at the tuning range... 1x10-5 which is +/- 5x10-6. If this aging number is accurate, the oscillator can be set on frequency for 5 years and then it is aged out of range. Realistically, this aging number is way over inflated. Probably more like 5x10-9 per year.)
Now look at the pin descriptions. No electrical trim.

There are oscillators with sine outputs and ones with CMOS outputs. Use the ones you want. Some people believe that CMOS outputs have more noise than the sine outputs. This is not necessarily true. They have more harmonics. This is by definition. Whatever you are running your OCXO into, you will need to bias it properly anyway. If your input requires a CMOS input, you can use a CMOS oscillator directly. If you have a CMOS oscillator and need a sinewave signal, you can probably use a 0.01uF cap in series with the output to get a 0V average signal.

Finally, the oven control circuit is important on the OCXO. This is information beyond any specification and is something only you can test after you have purchased the damned thing.

9023

This is the measurement of the current at startup. Yeah, it is a little effed up, but just look at the right hand side of the trace. You can see the current start high and then drop to a lower value. This is pretty much ideal. It is actually slightly underdamped. You can see the current drop and then go below the nominal current, then it comes back and settles in. This is really important because if you are concerned about frequency accuracy over temperature, this shows that it will correct quickly to temperature changes and maintain your nominal frequency. What is bad is if it is overdamped and the current slowly drops from it max and never dips below the nominal current or it is way underdamped and rings all day long.

Anyway, this is what I have on OCXO's for today. There will be a test next week.