[time-nuts] WTB: Oscilloquartz 8607, 8600 or similar OCXO
Bruce Griffiths
bruce.griffiths at xtra.co.nz
Sat Jun 7 18:03:44 EDT 2008
Andreas Tschammer wrote:
> Hi,
>
> I am looking for a high quality reference oscillator (5 MHz or 10 MHz) for a stability tester.
> No success until now on _bay, so I try my luck here..
>
> I intend to copy the system described in http://tycho.usno.navy.mil/ptti/ptti2002/paper32.pdf
> for one measurement channel. I was able to buy a 6160B synthesizer at _bay for a good price for
> the offset generation. After some extended repair sessions on this one I did a test of the phase stability
> of the synthesizer by mixing the 5 MHz reference signal with the 5 MHz output of the synth. After determination
> of the phase sensitivity of the mixer close to 90 deg phase shift I collected 10000 s of phase data with
> a Schlumberger 7150plus. The 1s Allan Deviation is 1E-13 and the 1000s 4E-15 (calculated with Ulrich' s fine plotter
> utility) which seems to be quite promising. The cabling and the ceramic coupling capacitors are pretty sensitive to temperature
> changes in this quick and dirty setup.
>
> Andreas, DL2KCL
>
Don't copy this design slavishly, in the light of more recent
theoretical developments its a bit of a joke.
The zero crossing detector in particular is poorly designed.
To achieve the best performance possible more attention to optimising
the design of the zero crossing detector is required.
Merely using a single active low pass filter followed by an amplifier
and comparator is far from optimum.
A better technique is to use a cascaded series of filter/limiter stages
with increasing gain and bandwidth.
Forget the quaint notion that linear amplification is required for the
first stage or 2.
This is only promoted by those unaware of that the theoretical problems
associated with estimating the noise characteristics of a limiter have
long since been resolved.
When using a set of cascaded filter limiters it should be possible to
use -5V and +5V rails, leading to reduced power dissipation and drift.
The temperature coefficient of the filter components is critical to
minimise the phase shift tempco of the filters.
Details like the mixer IF port termination and drive levels are also
important.
Highest SNR (zero crossing slope to noise ratio) is achieved when both
mixer RF ports are saturated and the IF port is terminated in a capacitor.
In fact its even possible to replace the time interval counter/timestamp
instrument with an ADC and some DSP software.
In this case less zero crossing slope amplification is required than
with a time interval counter.
Other details such as avoiding dc and low frequency ground loops by
isolating the mixer RF input and IF grounds are essential.
Isolation amplifiers driving each mixer input are also necessary to
achieve reliable measurements.
The mixer chosen with common RF, LO and IF grounds is not a good choice.
BNC connectors are not usually conducive to high phase stability.
However using isolation amplifiers with transformer couple inputs can
help isolate the signal source grounds.
The mixer RF to LO isolation isnt specified- this is critical when both
RF and LO ports are saturated.
The minicircuits phase detectors may be a better choice in that the RF
to LO and LO to RF isolation are specified and quite high particularly
at low frequencies.
If a through hole version is selected the RF and IF grounds can easily
be isolated.
Bruce
More information about the time-nuts
mailing list