[time-nuts] Pre-Release Docs and Schematics for the Fury Interface Board

xaos at darksmile.net xaos at darksmile.net
Tue Oct 30 14:17:41 EDT 2007


Said,

>
> Hi George,
>
> nice design!

Thanks :)

>
> Here are some thoughts from my side:
>
> The LT1208 has a fairly significant Voffset-versus-temp sensitivity. Maybe  a
> part like the LTC6244HV with about 1/10 the offset drift would work better,
> and reduce the thermal sensitivity of the overall system.
>
> For example, the LT1208 has 25uV per Degree C drift, and with a standard  MTI
> OCXO which has about 40Hz deviation from 0V to 5V EFC this would be about
> 2E-011 drift per degree Celsius worst case just due to the opamp   
> drift, a fairly
>  significant amount.
>
> The LTC6244HV part only allows a power supply range of +-5V so you may  want
> to add low-noise 5V and -5V regulators for the opamps.

Those are some very good points.

What I need then is a device with a +/-12V supply, with better specs  
than the LT1208. Otherwise I can't get the output to swing +/-10V.

+/-10V is what some commercial OCXO's require for EFC. I would like to  
have this flexibility.

I picked the LT1208 because I used it in the past and it will give me  
the output voltage range.

Any ideas here are most welcome. Dual device packaging is not  
important. The right device is.

My simulation, in PSPICE, was only the first step to check on the  
design. In reality, as we all know, things are very different.

This design is far from being cast in stone.

>
>
> The feedback and reference resistors R1, R2, R5, R6 are fairly high  value,
> and thus will add noise to the EFC. Make sure to use metal film   
> types  here. To
> reduce overall noise, you may want to add a simple RC lowpass filter to  the
> output of the opamp, say four X7R 22uF10V caps driven by a 4.7K  metal film
> resistor etc.

Values for R1,R2,R5 and R6 are not really important. It is the ratio  
that matters. So 1% metal film resistors of say, 100, 200 and 400 ohms  
should do fine. I'll check the simulation for current bias points but  
it should be minimal.

The output RC lowpass filter is a very good idea. I'll definitely  
build that in.

>
> Ground loop current coming from the OCXO heater through the OCXO ground pin
> and the ground plane can also cause some EFC error. Make sure to route the
> OCXO ground to the OCXO ground pin on the Fury, and then separate-out the EFC
> ground through a star-type layout. The idea is to prevent the OCXO heater
> ground  current from flowing through the EFC circuitry and SMA   
> cable, in which it
> would  cause voltage drop and thus thermal sensitivity.

Ok, this needs to be done. No question. I will separate the grounds. I  
will go even further and isolate the ground of the OCXO heater from  
everything else.

>
> Also, you may want to allow the OCXO to be powered from the Fury OCXO
> footprints' power pins. The Fury can compensate for thermal   
> sensitivity if this  is
> done. Alternatively, a thermistor may be connected to the OCXO, and driven by
> the Fury's OCXO power pins (thermistor must handle 10.5V!). As long as the
> current varies with temperature on the OCXO power pins, the Fury will learn,
> and  compensate thermal sensitivities in the system. BTW: the thermal tempco
> measurement is disabled via software command by default on the SMA   
> Fury's, and
> can be enabled via SCPI command.

Ok, this one is somewhat difficult. The HP oscillators have +12 and  
+24 V supplies. MTI requires +5 to +15 and Crystek needs +12. So we  
have a wide range here. What does the Fury supply?

A compromise has to be made.

On the Fury, Is there a connector that breaks out the pinout of the  
OCXO or would the individual power pins need to be directly soldered to?

I would rather avoid direct soldering. I want to have the Fury is a  
"untouched" state, if possible.

>
> Hope this helps,
> bye,
> Said

This type of feedback is exactly what I need!

-George





More information about the time-nuts mailing list