[time-nuts] homebrew 13 dBm distribution amplifier based on NIST design 5 to 100 MHz
dk4xp at hoffmann-hochfrequenz.de
Fri Apr 3 12:25:21 UTC 2009
On Mon, 22 Sep 2008 01:01:31 +0200, you wrote:
>after reading "A low noise 100 MHz distribution amplifier for precision metrology"
>by M. Siccardi, S. Römisch, F. W. Walls, and A. De Marchi (NIST),
>I have implemented a homebrew version of their design.
>Circuits, simulation & measurement data are contained in:
I have done a new revision of the distribution amplifier.
It has now 6 + 1 channels on a double EuroCard (usual VME bus size)
I got first silicon^W^W^W epoxy a few days ago, the board is not yet
completely populated, but channel 1 works already as expected.
I will characterize it after the easter holidays. The circuit is
essentially the LT-spice circuit that I have published already.
There are no transformers / ferrites in the signal path, so the
amplifier should be immune to modulation from transformer or fan
Output power is > 13 dBm.
The 7th channel is for cascading or for feeding that 1pps generator.
There is also a new low noise preamplifier for phase noise measurements
and inspecting noise behaviour of voltage references and such.
Stages are 3 pairs of SSM2210, cascode, AD797, AD797 and an
Opa2132 for *1 Gain and offset regulation.
Gain is *1 for phase detector calibration & 60 & 80 dB.
Can be made smaller if wanted. The *1 output is always
available and DC coupled if needed for VCXO locking.
Bandwidth is >1 MHz completely flat with any gain setting.
lower BW corner is abt 0.5Hz or DC if the offset loop is
defeated(but switching this off makes only sense for lower gains).
Noise is the expected 700 pV/sqrtHz.
The high gain path can absorb upto 20 mV offset as it is
All interesting I/O points are on a 100 mil grid, so the
amplifier can be used as a "macro" on predrilled board
(like Vero Powerplane (r)) if the shielding is to be done
for the whole assembly. All switching is performed with
gettered bistable relays. Monostable relays had too much
power dissipation, messing up the offset when switched on or off.
It's fun when your scope setting reads 200 nV / div and you still
have a meaningful display, at least for small bandwidth.
More information about the time-nuts