[time-nuts] GPSDO Design

Bruce Griffiths bruce.griffiths at xtra.co.nz
Thu Jan 14 20:54:04 UTC 2010


The time constant for a passive loop filter followed by an amplifier is

tau = Ko*Kd*A/(2+PI*fn)^2

where
Ko is the phase detector gain (radian/volt)
Kd is the VCO gain (rad/sec/volt)
A is the amplifier gain

For a CMOS exclusive OR phase detector with a 5V supply

Ko = 0.79 volt/rad

For a 10811A

Kd~ 0.63 rad/sec/volt (this value will be different for each OCXO and in 
general the EFC curve is nonlinear, the characteristic should be 
measured for you particular OCXO )

For A =2,
fn = 0.01Hz

Tau = 250 sec

With an active filter the time constant (in this case) will be 1/2 this 
or 125 sec.

If you only use a portion (say 10%) of the EFC range then the amplifier 
gain is reduced to 0.2 and the time constant (for a passive filter) 
required becomes 25 sec.
However more manual tuning of the 10811 may be required every few months 
(depends on the 10811A drift) to recentre the EFC voltage.

If the phase comparison is made at 100Hz then the delay of a typical 
HCMOS synchronous divider (used to divide a 10MHz OCXO down to 100Hz) 
will be around 20ns with a tempco of about 80ps/C.
The typical delay of a cascaded HC7490 style divider may be 10x this 
with a tempco of  around 800ps/C.

A change in EFC voltage of 100uV will alter the OCXO frequency by 1 part 
in 1E12, this corresponds to a change in amplifier input bias current of 
100pA (for a 1 megohm filter resistor) or an input offset voltage change 
of 100uV. As the filter time constant increases the required resistor 
value (for obtainable/affordable capacitors) will increase.

Thus the stability of the amplifier bias current and the capacitor 
leakage become critical for longer time constants.

A zero drift amplifier like the LTC1151 may be worth considering as long 
as any (~1000Hz for the LTC1151 ) chopper related noise is filtered out 
with a passive low pass filter at the amplifier output.

Bruce



Bruce Griffiths wrote:
> There are 4 principal sources of noise
>
> 1) The GPS receiver
>
> 2) The 4046 Phase detector
>
> 3) The opamp
>
> 4) The OCXO
>
> In the short term the GPS receiver noise will dominate.
> In the long term the 4046 phase detector noise and drift together with 
> the OCXO noise and drift will dominate.
>
> Unless you make an extremely poor choice of opamp the 4046 phase 
> detector noise and drift will be much larger than that of the opamp.
>
>
> Bruce
>
> John Foege wrote:
>> Hi All,
>>
>> Quick question for the more experienced members here with GPSDO
>> design/operation. Let's assume I'm using a 4096 phase comparator chip
>> followed by some kind of long time constant lowpass loop filter,
>> whether it be analog or digital, is not of concern for the following
>> question.
>>
>> Obviously using a 74HCT4096 would mean that my EFC voltage range would
>> be approx. 0-5V. If I wanted to use an OCXO with say a 0-8V EFC
>> voltage range, then I would be inclined to simply use an op-amp
>> amplifier with a gain of 1.6 to scale the EFC voltage accordingly.
>>
>> But not just any op-amp would do I take it? High-speed would of course
>> be of no concern. Also low-offset would be of little concern, as the
>> PLL would work to correct this, and it therefore seems to be
>> negligible. However, the part that's got me thinking is noise.
>> Obviously any noise at the ouput of the amp would adversely affect the
>> frequency stability of the OCXO.
>>
>> I thought the best way to control this would be to use an extremely
>> low noise op-amp employing a rather large compensation cap to give me
>> a rather small bandwidth, perhaps only a few hundred hertz.
>>
>> Anyone have experience with this? Assuming I have an OXCO with a max.
>> pulling range of 1ppm or 1e-6 over a 10V range, then I effectively can
>> pull 1e-7 per volt. This translates to 1e-10 per millivolt and 1e-13
>> per microvolt. Assuming that is a logical conclusion, then for a good
>> OCXO, in which I can at best hope for 5e-12 stability for tau=1s (e.g.
>> HP10811A), I would strive to to keep the noise at such a level that it
>> is an order of magnitude better than the best short term stability
>> figure. Accordingly, then I should shoot to keep any noise under 1
>> microvolt?
>>
>> I don't have much experience with noise calculations. I know it is
>> specified in nV/sqrt(Hz) generally. Translating this to something
>> practical is basically the assistance I'm looking for here.
>>
>> I would appreciate anyone being able to teach me a bit more about this.
>>
>> Thank you in all in advance.
>>
>> Sincerely,
>>
>> John Foege
>>
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at febo.com
>> To unsubscribe, go to 
>> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>> and follow the instructions there.
>>
>
>
>
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to 
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
>





More information about the time-nuts mailing list