[time-nuts] Lars GPSDO on EEVblog

Magnus Danielson magnus at rubidium.dyndns.org
Sat Sep 8 16:48:33 EDT 2018

Hi Jim,

On 09/08/2018 10:02 PM, Jim Harman wrote:
> Magnus said,
> It would be interesting to test the linearity of the TIC separately for
> instance.
> I have done some testing of the TIC. It works quite well for the Arduino
> Uno with its 1 V full scale ADC setting, but the exponential shape of the
> RC charging from the 5 V is quite evident if you  use a processor like the
> Micro's 32u4 with a 2.56 V sensitivity. This circuit is also sensitive to
> any noise on the 5 V supply. The linearity is not too important for this
> application as long as it does not affect the loop gain so much that the
> loop becomes unstable.

Agreed. It is expected, but it would be fun to see how well it would do.

> However I have modified the circuit to use a simple 2 ma current source as
> shown in the attached schematic, with significantly improved linearity. H/T
> to Horowitz and Hill's The Art of Electronics for the idea of using an LED
> as the voltage reference for the current source. The diode-connected 2N3904
> has less leakage at a small forward voltage than any small-signal diodes I
> tested, so the output voltage with a very narrow input pulse is close to
> zero.

A more serious current source helps.

It comes as no big surprise that this has been a topic of interest to
several designers, hence there exists a number of patents on it.

Different approaches have been used. The HP5335A for instance uses a
three transistor setup, where the current source, a resistor, is
buffered behind the transistor pair that also switches it in and out.
The HP5335A then acts as pulse extender with x200 and a TTL counter to
complete the ADC process. Modern designs use a proper ADC instead but
the basic problem remains the same.

> I could run some tests on my version by setting the oscillator slightly
> off-frequency and capturing the resulting sloped TIC output, and would be
> happy to share the results if you are interested.

Please do. This is one of the methods and probably the easiest to setup
for most. Another approach is to use a programmable delay generator with
enough resolution, but having one of those around isn't as common as a
RF generator which can be detuned with sufficient resolution.

As I spend the evening refactoring the lab-bench, the RF generators gets
a more prominent placement as well as one of the delay generators. Hope
to get the TICs stacked up nicely for ones.


More information about the time-nuts mailing list