[time-nuts] VC-OCXO EFC stability.

Charles Steinmetz csteinmetz at yandex.com
Tue Jan 6 23:29:00 EST 2015

Luis wrote:

>With a multiturn pot and a 78L05 I can get a 0 to 5V EFC to tune
>an OCXO on desired freq. But... maybe other voltage regulators
>or other scheme have better temp stability than the old 78L05.
>Before I crawl lost in new'ish fancy regulator land does anybody
>know the killer solution/IC for this job ?

First, think hard about whether you really want the full 0-5v EFC 
range.  Even with a 25-turn pot, it can be difficult to adjust the 
frequency in very fine increments with the full EFC range across the 
pot, so it is often beneficial to restrict the range to get finer 
adjustment.  Of course, the voltage that sets the oscillator exactly 
on frequency should be somewhere near the middle of the pot's 
adjustment range.  (If you know which way the oscillator is drifting, 
you can leave more range in the direction that it will [most likely] 
need to be adjusted in the future, and less in the other direction.)

One circuit that has very low noise, first-order temperature 
compensation, and gives a limited tuning range (for fine adjustment) 
is posted at ko4bb.com.  Go to <http://www.ko4bb.com/manuals/> and 
search for the file named:

      "HP 10544 10811 EFC fine adjustment.pdf"

[searching for "10811" will find it]

That circuit is designed to give a symmetrical EFC range of +/- 1.2v 
maximum, which can be reduced further by using fixed resistors at 
either end of the potentiometer.  With slight modification, it could 
easily provide an EFC voltage from about 0.8v to about 3.2v, and by 
adding another LED and temperature-compensating diode the range could 
be increased to run from about 0.8v to about 4.4v.

There are lots of other solutions, and I'm sure a number will be 
suggested.  Do remember that most "ultra-low noise" voltage 
references are not really all that quiet ("ultra" is a relative 
term...), so be prepared to do some heavy noise filtering if you go that route.

One other possibility is to use an LM329, LM399, or LTZ1000 (in order 
of increasing performance) as a reference with a scaling 
amp.  Because those references are all ~6.9v, it's best to use an 
inverting scaling amp to realize 5v or less -- but that is actually 
good (aside from the bother of needing a negative supply to bias the 
reference), because you can configure the amp as a multiple-feedback 
filter and the gain will continue to fall below unity as frequency 
increases.  You need to use a very low noise op-amp (e.g., LT1128), 
and the feedback resistors must have low values -- which tends to 
make the required capacitors impractically large if you want 
significant noise attenuation below 10Hz (also, make sure the op amp 
has sufficient output current capability to drive the low-resistance 
feedback loop).  It's brute force, but it can work well.

Best regards,


More information about the time-nuts mailing list