[time-nuts] GPS lock of the FE5680. Current experiment and question

David davidwhess at gmail.com
Sat Feb 11 20:19:23 UTC 2012


There are no temperature coefficient effects or calibration drift with
the current sources and ramp circuits in this case.  Between 1 PPS
measurements, there is more than enough time to gate a sample waveform
from the convenient 10 MHz source to the time to voltage converter to
calibrate it.

One interesting thing I just noticed with the Tektronix 2440 design is
that the trigger starts two different time to voltage converters.  One
measures to the positive sample clock edge and the other to the
negative sample clock edge so if one goes metastable, the other can be
used.

On Sat, 11 Feb 2012 10:54:40 -0500, paul swed <paulswedb at gmail.com>
wrote:

>Getting very interesting.
>Bob had mentioned just sample the 10 MC sine wave. What I used to do on
>homebrew Loran C.
>
>Thats easier to do because today its nothing to buffer that 10 MC signal to
>drive a fast sample and hold. This eliminates the ramp circuitry and
>constant current sources used in the ramp and tempco effects.
>
>This all seems to work out reasonably because the 5680s are in general
>pretty darn stable. (Boy is that a relative term in time-nuts land)
>
>Now to dig through the ole junk box for a sample and hold chips. Most
>likely older and useless. Go hunting at mouser or digikey for modern stuff.
>Hate to have to go to discrete pulsed diodes.....
>Regards
>
>On Sat, Feb 11, 2012 at 10:33 AM, Azelio Boriani
><azelio.boriani at screen.it>wrote:
>
>> This is the simplest part if a microprocessor can be used: by the serial
>> port you get the sawtooth correction in nS to be applied to the sampled
>> data. The sampled data must be converted to nS or the sawtooth correction
>> must  be converted in a suitable sampled data correction. It is possible
>> even to hardware correct the PPS with a delay line before using it  (see
>> the already mentioned gpstime.com/files/tow-time2011.pdf by Tom Clarck and
>> Rick Hambly).
>>
>> On Sat, Feb 11, 2012 at 1:45 PM, Bob Camp <lists at rtty.us> wrote:
>>
>> > Another way to build an analog phase detector...
>> >
>> > Next layer on the onion is how to get the sawtooth correction out of the
>> > GPS and into your loop.
>> >
>> > On Feb 11, 2012, at 12:05 AM, Chris Albertson <albertson.chris at gmail.com
>> >
>> > wrote:
>> >
>> > > All these different suggestions build down to one thing, the precision
>> > > with which you measure the phase when you sample it each second.  The
>> > > single flip flop will tell you which half cycle. a simple two bit
>> > > counter made with two '74 FFs tells you which half cycle and with
>> > > direction.
>> > >
>> > > The "best" maybe  is if you let the PPS set a FF and the 10MHz reset
>> > > it.  The FF's output gates a constant current to a capacitor and
>> > > charges it to some voltage.  Then you measure that with a 10-bit ADC.
>> > >  This measures the phase to maybe 1%, gives you direction and is
>> > > pretty cheap to build
>> > >
>> > > Let's see if I have the numbers right?  If you check a 10MHz signal
>> > > once per second with just the FF then you have 1E-7.  You would need
>> > > 1000 seconds for 1E-10.   But if you measure phase to 1/10th of a
>> > > cycle you get to 1E-10   ten faster.  Right?
>



More information about the time-nuts mailing list