[time-nuts] 4 channel 1 Mhz - 25 Mhz distribution amp
wje at quackers.net
Sun Jun 7 16:06:02 UTC 2009
I finally gave up trying to find a 50 ohm distribution amp I liked and
wanted to pay for, so I built a simple 4 channel one that uses either a
Linear LT1365 or an Analog Devices AD8044. It has an ac-coupled input
and transformer-isolated outputs. I took care in the layout to keep
trace lengths exactly the same for all channels, and they're also
striplines. (not that using striplines makes much difference with 0.5"
Warning - it doesn't have perfect unity gain, 0 chan-chan phase error,
infinitely low noise, infiinte isolation, or any of those other
characteristics Time-Nuts expect.
What it does to (using the LT1365 @ 10Mhz) is accept a 10dBm input,
provide 50dB worst-case channel-channel isolation, 68 dB worst-case
output-input reverse isolation, and worst-case 0.2 degree
channel-channel phase error. I'm not quite sure about the noise; it's
down around the noise level of my HP 3588A, which reports -139dBm/rtHz
for the amp. This is also the basic noise figure of the 3588, so the amp
is probably lower. The amp specs say 9nV/rtHz, which works out to
-147dBm referred to 50 ohms, assuming I did my quick calculations correctly.
Interestingly, the phase error between 3 of the 4 channels is down
around 0.02 degrees; just one channel has the 0.2 degree error relative
to all the others. This was measured with my HP 5370 using 10k samples
per reading. The LT1365 datasheet says 0.04 max. Hmm.. maybe I should
check the coax for that channel.
The gain is less than unity at all frequencies. Why? Because I wanted to
be able to accept a 10dBm input without clipping and still get close to
that out. Using the LT1365, this is possible. The AD8044 will start
clipping at about 6dBm because it doesn't have enough current drive.
However, it has a MUCH flatter bandwidth curve, managing 0.2 dB from
<1Mhz to >50Mhz. The LT1365 is significantly worse, but for a
single-frequency distribution amp, this isn't all that important. Using
the LT1365, the gain is -1.5 dB @ 10 Mhz, -0.9 @ 5Mhz. Of course, you
can trim the gain for whatever you want within the voltage and current
limits of the op amp. The opamp is set for a gain of 2 by default (to
drive the series-terminated output transformer).
If anyone is interested, I can put the schematic, pc board layout, and
various plots from my 3588 for the amp on my FTP site.
They said 'Windows or better'
so I used Linux.
More information about the time-nuts