|
Hello, all, here comes some background information on DCF77, for those of you interested in such. I want to expand on what Martin has written. How precise is a DCF77 clock? It depends... To summarize how most DCF77 clocks work: At the start of each second, the transmitting power is reduced to a quarter of the normal amplitude. That power reduction lasts for either 100 ms or 200 ms. The time information is coded via these duration differences. For the record: Changing transmitter power is also called "amplitude modulation" or, in short, AM. That amplitude reduction is not instantaneous. Amplitude reduction starts precisely at the start of the second. From there, it takes about 800 µs until the low 25% level is reached. What we need is the moment these 800 µs start. Subtract the time the radio waves need to get from the transmitter to your location, then you have the start of the second. So, to know "How precise is my DCF77 clock?", you need to know "how precise can it determine the start of that 800 µs delay?" (provided it is an AM receiving clock - most are). And you need to know how far the radio waves travel. Unfortunately, that varies a bit, e.g., between day and night. Typical variations are below 50 µs for most places in Germany. The PTB resources mentioned below have some detail information on that. I won't get into it. This is radio. Radio means you have signal, but you also have noise. A thunderstorm can put out quite some signals down there in the long wave range where DCF77 operates. But you have noise, man-made and natural, even in the absence of thunderstorms. What can you do to separate signal from noise? Let me use an analogy. Imagine you have a lab at a sea shore and want to measure tides (your signal), but you also have waves (your noise). What would you do to separate the two? One obvious thing you could try: You could put a pipe in the seawater and fill it more or less loosely with cloth. So water can sip through the pipe, but not move quickly. At the other end of the pipe, you get the (slow) tidal signal, but you see (almost) nothing of those faster changes caused by individual waves. At the downside, you also get a delay. That is not so bad in itself. Why - you could simply add a fudge to cancel the delay. But, unfortunately, the delay is not constant. If you have a strong tide, you get a smaller delay, if you have a weak tide, you get a longer delay. This is similar on how they build radios. Where we were using a pipe with cloth in our tide lab, in radio stuff, they reduce bandwidth. What they get is improved signal-to-noise ratio, as did we. The price to pay is a delay - not so bad. But also, more importantly, they get slower signal rise and fall of the signal. (Here, the analogy breaks slightly.) That slowness in turn produces uncertainty regarding the precise moment of the start of amplitude reduction, that marks the start of the second. The PTB says: > To guarantee a reception free from interferences, many of the radio-controlled clocks on the market work with bandwidths around 10 Hz and uncertainties of approx. 0.1 s, which is regarded as sufficient for this purpose. This "10 Hz bandwidth" is small, leading to slowly rising and falling signals. But this kind of clock will pick up the signal far away from the transmitter. For your Joe Average Consumer, it does not matter if the clock is some many dozen ms imprecise, but it does matter whether it works only in certain parts of Germany as opposed to a sizable part of Europe. The same PTB paper also mentioned: With elaborate signal processing and higher bandwidth, > uncertainties between 50 and 100 μs have been achieved for distances in which the ground wave clearly prevails. (where "the ground wave clearly prevails" translates to about "closer than 250 km from the transmitter", depending on the season and time of day). Better resilience against noise for those that want high time precision can also be had. Besides the AM explained above, the same time information is also impressed onto the DCF77 signal using "phase modulation". "Phase modulation", abbreviated as PM, is basically the same thing as frequency modulation, FM. Phase modulation means they steer the transmitter a bit higher in frequency and a bit lower in frequency in rapid succession, according to a precise schedule. There is a paper describing details, and also describing (in rather abstract terms) a radio clock that was build to study this when it was introduced. Whereas DCF77 consumer clocks use a bandwidth of 10 Hz, that clock uses a bandwidth of 2584 Hz. I know only a little about signal processing. As far as I understand, 10 Hz is the lowest feasible bandwidth for AM reception of the DCF77 signal. Going lower than that means you see less than you want of the amplitude change. But with AM, you can pick your own compromise: You can go to a higher bandwidth. This means you'll need a better signal to be able to decode, but if your signal is good enough so you _are_ able to decode, you'll get a more precise time. My understanding also is: For DCF77 PM reception, those 2600 Hz is the lowest bandwidth that can be used. But if you are willing to use such a broad bandwidth, (which means you need a good DCF77 signal to receive anything at all), PM reception will give you better precision, compared with AM reception using that same bandwidth. So far my explanation. Official resources: The German PTB, the institution that runs DCF77, makes available comprehensive DCF77 information at http://www.ptb.de/cms/fachabteilungen/abt4/fb-44/44-literatur.html#c7142 For the English-speaking among you, I'd recommend http://www.ptb.de/cms/fileadmin/internet/fachabteilungen/abteilung_4/4.4_zeit_und_frequenz/pdf/PTBM_50a_DCF77_engl.pdf For those able to read German, try http://www.ptb.de/cms/fileadmin/internet/fachabteilungen/abteilung_4/4.4_zeit_und_frequenz/pdf/2009_Bauch_PTBM__DCF77.pdfDetails about the phase modulation and how it could be received are described in http://www.ptb.de/cms/fileadmin/internet/fachabteilungen/abteilung_4/4.4_zeit_und_frequenz/pdf/5_1988_Hetzel_-_Proc_EFTF_88.pdf Greetings, Andreas Martin Burnicki sent at 07.09.2012 21:31: All of the above is correct, and in addition the time derived from the DCF77 standard AM modulation is often not very accurate. Usually there are limited antenna and filter bandwidths to suppress electrical noise, and these filters are causing a signal delay which results in a time offset depending on the filter bandwith. |
_______________________________________________ pool mailing list [email protected] http://lists.ntp.org/listinfo/pool
