[time-nuts] OffsetGenerator

Ulrich Bangert df6jb at ulrich-bangert.de
Sun Dec 25 07:11:24 EST 2011


I have done it just for my own scope but perhaps someone other may have
interest in it too.

Over the last weeks I designed a pcb which holds

a) A clock shaper as designed by Bruce Griffiths based on a ADCMP600 that
should work well to over 100 MHz.

This clock shaper delivers the reference clock for either

b) an ADF4002 pll circuit


c) an AD9852 48-bit dds circuit

In case the reference clock is applied to the dds, then the dds in
conjunction with a second pulse shaper at it's output may deliver the
reference clock for the pll. There are a number of interesting applications
for a circuit like this f.e. as a offset generator for dual mixer systems.
Since you would need 2 isolation/distribution amps in this case the board

d) two AD8007 amps

which by the means of jumpers can be fed by either the dds output or the
vc(x)o of the pll. The serial data streams for both the pll as well as the
dds are generated by an ATMEGA32A microcontroller to which you talk over a
standard rs232 port in a plain text command language. Any bit fiddling is
done by the micro you just have to set parameters of the two devices in
plain text and the control words are assembled by the micro. All settings
are stored in non-volatile memory and loaded into the devices the next time
after power-up.

All rf connectors can either be populated with sma or bnc.

I will not forget to mention that this project is VERY similar in principle
to a project of John Miles published a few years ago. Nevertheless John's
project more aimed at the microwave world of frequencies while mine aimes
more at time nuts applications. The board has a predecessor based on a
AD9850 32-bit dds. The experience with this board has shown that by proper
dithering of the dds's lsb a lot of the negative effects due to the cyclic
phase error phenomenon can be removed. 

I have currently received the first boards and I am going to populate them.
Some parts are missing, will receive them between X-Mas and new years eve.
Keep you informed.

You can download the circuit drawings and two images of the pcb from the
usual place:


Note how the problem with the dds's 'exposed pad' is solved. The exposed pad
NEEDS to by soldered to the ground plane for best cooling.

Best regards and a happy x-mas

Ulrich Bangert
Ortholzer Weg 1
27243 Gross Ippener 

I had nearly forgotten it: The board also holds an AD654 voltage to
frequency converter driven by the pll's loop voltage and connected to one of
the micro's counter/timer inputs. So the board may also be used as an
correct implementation of the tight pll methods as discussed here in depth.

More information about the time-nuts mailing list