[time-nuts] Frequency divider PCB: Current status on"pre-orders", and pointers to documentation.

[Bruce Griffiths]

David C. Partridge wrote:
> Ooops I did mean to say R4 and R7 - the 2K divider pair.
>
> Regards,
> David Partridge
> Email:david.partridge at perdrix.co.uk
>
> -----Original Message-----
> From: time-nuts-bounces at febo.com [mailto:time-nuts-bounces at febo.com] On
> Behalf Of David C. Partridge
> Sent: 18 March 2010 12:41
> To: 'Discussion of precise time and frequency measurement'
> Subject: Re: [time-nuts] Frequency divider PCB: Current status
> on"pre-orders", and pointers to documentation.
>
> <snip>
>    
>> except possibly to use thin film parts for R4&  R5
>>      
>
The problem with thick film resistors is their flicker when a 
substantial dc (or very low frequency) voltage is present across the 
resistor.

Resistors with a potentially significant dc voltage across them:

R2, R3 if the source has a dc output component.

R5, R6 due to the comparator input bias current.
The dc voltage drop will be around 150mV.
(flicker noise due to the comparator bias current flicker noise will 
also be developed across these resistors)

R4, R7
The dc voltage drop across each is 2.5V.
The flicker noise (and power supply noise) at the junction of R4 and R7 
is a common mode signal to the comparator and its effect on flicker 
phase noise is significantly reduced by the comparator's 70-80dB low 
frequency CMRR.

R8-R23
The dc voltage drop across these resistors depends on the load.
If the load is AC coupled there is no dc component.
However for the lower frequency (1Hz and perhaps 10Hz) outputs there may 
be substantial low frequency current in the output buffer series resistors.

Another, often neglected contribution to output noise is the power 
supply low frequency noise.
Depending on the application the low frequency output noise produced by 
common 3 terminal regulators may be significant.
Worst case one may need to resort to using NiCd batteries (measured by 
NIST to have extremely low noise for load currents of 1mA or less).

However before resorting to this a lower phase noise input circuit 
should be used.
The one used was originally designed to be reasonably rugged with 
respect to input overload whilst operating over a wide frequency and 
amplitude range.

The phase noise contribution of the various coupling capacitors is 
determined by the phase noise characteristics of the dielectric and the 
phase shift due to the capacitor.
If the phase shift is sufficiently small almost any dielectric can be used.
If the low frequency voltage noise across the coupling capacitors is 
large then phase modulation due to the voltage dependence of the 
capacitor value can occur.
In this case its important to use a capacitor with a low voltage 
coefficient dielectric such as NP0/C0G.

If an input transformer is used then if the phase shift due to the 
transformer magnetising inductance significant flicker phase noise can 
be added if the transformer core has a permeability greater than unity. 
Ferrite cores produce more phase noise but the amount depends on the 
particular ferrite used. Powdered iron cores contribute lower phase noise.

If one uses an input filter then a tuned air cored transformer may be 
useful in that it contributes very little phase noise whilst providing 
isolation.

For the fastidious, the fact that a digital divider samples its input 
noise and aliases it down needs to be considered.
This effect can be substantial, bandpass filtering the input signal will 
help but with large division ratios a very narrow bandwidth input filter 
is required.
The very large associated phase shift tempco usually makes this impractical.
A more practical solution is to use a cascade of bandpass (`10%) filter 
+ shaper + divider with low value divisor (<64) modules.

Bruce