[time-nuts] WWVB Measurements
lenny at Codematic.com
Sat Mar 26 23:31:14 UTC 2011
My design isn't really that sophisticated hardware wise. Its an 8051
variant, and the CMMR-60P which can be obtained from Sparkfun.com. I do
wind my own antenna, which is just a quad, made from PVC and a metal
electrical box. The circuit sits in the center. It's done with a 2.5 inch
square, 4 layer board from pcbexpress.com.
The software samples the bit signal coming from the CMMR, every 10 ms
(essentially), and uses a 100 bit array to qualify it using a weighting
algorithm. The start of the bit train, and with it the sequence of 10ms
samples, is essentially polled very fast to catch the beginning of the
second as soon as possible. Most people disqualify the signal if it is
+-50ms of its target, such as <750 > 850ms for mark. I don't, if i can
determine its value, from the others then i let the Frame State machine
process it. My assumption is that the signal may change over time due to
atmospheric conditions, but slowly and equally across all the bits. So if a
mark is degraded, then they should all be equally. Finally, qualifying a
frame of data, where there are no "check" bits isn't as bad as it seems
because there is structure.. for example there must be a Mark every 10 bits,
the middle bit is always zero, the DST bits have a pattern, numbers should
increment, have min / max values...etc.
I am still unsure on how accurate i can get this, especially during the
daylight hours. That was the root of my question. I need a way to evaluate
how good it actually isn't. :) My initial breadboard was able to detect and
decode a signal for several days before i stopped it, so i am hoping its
I am also hoping i can get a realistic assessment from the people here about
what can best be expected from this signal. It is of course not as good as
GPS, or those (simply awesome) cesium atomic clocks here. I am not a math
guy really, but am an embedded software engineer, makes it great for system,
and hardware hacking... not so good for "allen deviation" stuff... or
whatever that is. :(
Whats the point of doing a project like this if you Already know how to do
it ? :)
Feel free to email if you have other questions.... maybe we can advance your
schedule a bit... get it in a couple of weeks rather than months. :)
On Sat, Mar 26, 2011 at 6:54 PM, Chris Albertson
<albertson.chris at gmail.com>wrote:
> You are ahead of me by many months. I'm building a WWVB receiver
> also. Actually I expect I will need to build several before I get
> 24x7 coverage. My breadboard works only at night in the So.
> California area. My plan is to place the entire receiver, antenna
> and all on a mast far from the house and use an RS422 serial line to
> send the data back to a computer indoors.
> Do you intend to publish your work? I'd be most interested in how
> you decode the signal. I'm conflicted between two approaches (1) To
> declare the signal "invalid" if there is any error at all or (2) to
> try and extract as much signal out of the noise as I can. I may do
> the latter and then have some kind of quality indicator. The WWV
> audio decoder built into the NTP reference implementation can extract
> time code from what sounds like white noise and static to the human
> ear using sophisticated DSP. My first receiver will use #1.
> About measuring the PPS.If you had a nice HP Universal counter with a
> computer interface that would be best. You put the PPS from a good
> GPS on one channel and the PPS from WWVB on the other. Lacking that
> and if you only need to get down to uS level you can use two serial
> orts in a Linux box and use PPS line disciplin on each oert the kernal
> will time stamp the PPS when they happen and software can read and log
> the time stamps. Use the command "ldattach pps <device>" for each
> serial port. Good to about 1 uSwhich for WWVB might be enough
> On Sat, Mar 26, 2011 at 1:26 PM, Lenny Story <lenny at codematic.com> wrote:
> > Greetings All,
> > This is my first post to this board.
> > I've completed the first run of a WWVB receiver board and Antenna (custom
> > wound quad). Its an 8051 microcontroller, with a serial port really, but
> > can decode the signal accurately pretty much all day long. (I'm just
> > of boston, MA).
> > I'm wanting to evaluate its performance, my guess is i'll have to produce
> > plot of its PPS. In reading the LeapSecond.com site (awesome btw), the
> > "Allen Deviation" is used. As this is my first technical, experience in
> > this area, is there a resource or method that is preferred by those who
> > this technology ?
> > The code reports the time delta between each detected second. If i log
> > PPS deltas for an entire day (or week) of detected signal, is that enough
> > data to start figuring out how to do the "Allen Deviation" calculations ?
> > Any resources can you recommend to figuring out the graphs i need to
> > ?
> > Thanks in advance for your help !
> > -Lenny Story
> > _______________________________________________
> > time-nuts mailing list -- time-nuts at febo.com
> > To unsubscribe, go to
> > and follow the instructions there.
> Chris Albertson
> Redondo Beach, California
More information about the time-nuts