With the interest in the high-accuracy Rb frequency references and the upcoming frequency measurement test coming up, I thought I would try to get a sense of the periodic frequency error in the Flex 5000's VFO. I started out trying to use WWV and then average the values but multipath and doppler in the WWV signal makes using it to measure frequency closer than about 1/10 Hz challenging, even with averaging. I was bemoaning the fact until Jerry Flanders pointed out that I don't need an absolute frequency reference, just one that is very stable so I can take relative measurements. What better local reference than the Rb reference oscillator I am using as the reference for the F5K. Even with the antenna disconnected it produces a signal 40dB above the F5K's noise floor that it is just fine for making relative frequency measurements.
For the test I used the Rb oscillator as my reference signal. I used fldigi's 'Freq Analysis' mode to measure the frequency of the incoming signal. It does this by taking the VFO frequency as reported by CAT (to 1Hz) and summing that with the frequency of the audio tone produced by the reference signal. I tuned the F5K to 9,999,000 Hz (1kHz below the reference) and then measured the spectral line at 1kHz. I then stepped the VFO in 1Hz increment until I reached 9,999,050 Hz, a range of 50Hz to see if there were any periodic errors. For this entire range of VFO test frequencies the reported signal frequency varied from 10,000,000.06 Hz to 10,000,000.18 Hz, a range of 120mHz or 0.12Hz. That represents a periodic error of 1.2E-8 or .012 PPM, pretty darned good. Now for the (maybe) bad news. The error is periodic. The pattern is very regular. I am guessing that this comes from the fraction that needs to be loaded into the DDS VFO. If you graph the data as a scatter plot, the pattern of error becomes very obvious. This implies that we could probably add in a very accurate correction factor *if* we took a LOT of data points, or knew the DDS frequency programming algorithm. Regardless, here is my raw data in the form of VFO frequency followed by error value relative to 10,000,000.000 Hz. I have left it in comma-separated-value form so that you can easily cut-and-paste it into a spreadsheet in order to plot the error against frequency. "VFO","Error" 9999000,0.11 9999001,0.07 9999002,0.14 9999003,0.09 9999004,0.16 9999005,0.11 9999006,0.06 9999007,0.13 9999008,0.09 9999009,0.15 9999010,0.11 9999011,0.06 9999012,0.13 9999013,0.08 9999014,0.15 9999015,0.1 9999016,0.17 9999017,0.13 9999018,0.08 9999019,0.15 9999020,0.1 9999021,0.17 9999022,0.12 9999023,0.08 9999024,0.14 9999025,0.1 9999026,0.17 9999027,0.12 9999028,0.08 9999029,0.14 9999030,0.09 9999031,0.16 9999032,0.11 9999033,0.07 9999034,0.14 9999035,0.09 9999036,0.16 9999037,0.11 9999038,0.18 9999039,0.13 9999040,0.08 9999041,0.16 9999042,0.11 9999043,0.17 9999044,0.12 9999045,0.09 9999046,0.16 9999047,0.1 9999048,0.17 9999049,0.12 9999050,0.07 The fact that they all average about 100mHz high suggests that my reference is a skosh above 10MHz but then, I am using the same signal as both the DDS master oscillator and the reference signal. I am not sure if the error is additive or should tend to cancel. Regardless, this test shows the periodic error in the VFO setting. Now before anyone gets worked up with Flex about inaccuracy, remember, we are talking about an error range of +/- .06 Hz at 10MHz if the reference oscillator is right on the money. It seems pretty darned amazing to me that we have a piece of consumer radio equipment with that kind of frequency accuracy. Maybe I will fiddle with the DDS offset to see what effect that has on the graph. -- 73 de Brian, WB6RQN/J79BPL _______________________________________________ Flexedge mailing list [email protected] http://mail.flex-radio.biz/mailman/listinfo/flexedge_flex-radio.biz This is the FlexRadio Systems e-mail Reflector called FlexEdge. It is used for posting topics related to SDR software development and experimentalist who are using alpha and beta versions of the software.
