[time-nuts] Low-Cost Rubidium Performance
magnus at rubidium.dyndns.org
Thu Feb 9 23:34:55 UTC 2012
On 02/10/2012 12:22 AM, Javier Herrero wrote:
> El 10/02/2012 00:13, Javier Herrero escribió:
>> El 09/02/2012 22:28, Magnus Danielson escribió:
>>> Consider that it is de-modulated and then low-pass filtered.
>>> Also, it is the alternating rate and not 1400 Hz difference in DDS
>>> setting which is the key parameter here. The 1400 Hz gives a hint of
>>> the Q-value however, which seems to be lower on these than on any of
>>> my larger rubidiums, but it is maybe to be expected.
>> Yes, you're right... I was thinking on the alternating rate (that in
>> fact I measured, at 416.6666..Hz, but the other number came first ;)
>> ), and in the fact that the FRS, that uses 127Hz as alternating rate,
>> has notorious spurs at 127 and 254Hz (al at a lot of their harmonics),
>> so I was expecting someting similar for the FE5680A at 416.6666Hz and
>> harmonics, but seems not to be there (or the spur forest makes not
>> easy to see these trees :) )
> I answer myself. Perhaps they are there quite notoriously, since in the
> spectra plots that I took when I got mine, now it is clear why there are
> two peaks at around -70/-75dBc at somewher that seems very near
> +/-416.7Hz: http://www.nebulosa.org/images/FE5680A/FE5680A4.jpg Probably
> in the phase noise plot they are masked by all the phase noise floor and
> other spurii that are not so apparent in a quick measurement with the
> spectrum analyzer
All being as expected then.
I think we can focus on tracing the DDS spurs as injected through
another path. Could be lack of isolation of the 60 MHz crystal
oscillator from the rest of the design, or it creeps onto the 10 MHz in
the CPLD, or just plains sneaks in on the output buffer, possibly via
the power lines.
More information about the time-nuts