Bruce wrote:

Quite a good performer for such a simple circuit.  I found, both in
modeling and on the bench, that there is the usual noise bump at
200-300 MHz and non-monotonic behavior out in the 900 MHz
region.  The latter can be solved by using an MPSH10 for Q1, which
also brings the in-band noise and phase noise down a little.  The
former can be addressed by adding 8-10 pF across R2, at the expense
of lowering the 3 dB point from around 150 MHz to around 80 MHz.  For
use as a 5 or 10 MHz distribution amp, I'd include the cap.

The input impedance stays decently high everywhere the amp has useful
gain -- there should be no problem paralleling 10 of them on a 50 ohm
source.  You can raise R2 just a tad to get back to unity gain, if
needed.  The reverse isolation is about 35 dB.  This can be improved
to around 50 dB by adding an emitter follower at the input, adjusting
R7 and R8 to maintain Q1's base voltage.  The noise increase is negligible.

It is fairly sensitive to power supply noise, so you want a nice
quiet supply.  I used a regulator built with an LM399 and LT1028.

Since the transformer is 1:1, one might be tempted to omit it.  For a
distribution amp that will be connected to a number of different
instruments, however, one is well advised to include it to isolate
the various returns.  6 bifilar turns on a T43-37 toroid core and 14
bifilar turns on a T61-37 both worked fine for me.  If you have 1:1
transformers from a spare Ethernet card, those should, too.

For a Q&D distribution amp, this would be a pretty good candidate.

Best regards,

Charles

On 3/26/2012 8:15 AM, Charles P. Steinmetz wrote:

if one is distributing 10 Mhz, does it really matter what the circuit
does at 300 and 900 Mhz??

73's,
Randy, KI6WAS

Randy wrote:

That depends on what it is feeding and what noise
and other signals are getting to the DA
input.  Some synthesized 10 MHz sources produce energy well above 10 MHz.

I consider nonmonotonic behavior to be a design
flaw in general (except where it is specifically
desired or you are up against the bleeding edge
of technology and it can't be avoided), so I
eliminate it at every opportunity as long as the
fix doesn't cause worse problems.  In this case,
the nonmonotonicity is cured by replacing an 8¢
transistor with a 16¢ transistor, and some other
small benefits are realized at the same time, so
I say it's 8¢ well spent.  Nearly a whole dollar
extra for a 12-output DA.  The noise bump is
cured by restricting the 3 dB bandwidth to ~80
MHz, which does not affect the 10 MHz but may
help the receiving instrument if it is sensitive
to VHF noise (although the magnitude of the bump is not large).

I consider these good prophylactic design
measures.  Practicing them keeps you out of
trouble that you might not even know was
threatening, whether or not it makes a practical
difference WRT a particular design.

Best regards,

Charles

Charles P. Steinmetz wrote:

The stability of Sziklai pair like configurations (in this case a
triple) is enhanced if the input device has a significantly higher ft
than the output device(s).
Stable operation at unity gain is necessary if a feedback capacitor is used.

The above reverse isolation is about 25dB lower than I would expect.
The impedance of the power supply rail at the test frequency should be
low so using bypass caps with low inductance and esr is desirable.
The reverse isolation of an emitter follower is determined by the
current gain at the test frequency.
Thus with an ft of 300MHz it should be around 30dB at 10MHz.

An LM329 has similar noise without the dissipation of the internal
heater in the LM399.
The 10nF cap between the output transistor bases is the primary source
of  this sensistivity.

Bruce

Bruce wrote:

D'oh!  Bruce is right -- I calculated the reverse isolation
incorrectly.  I had only been expecting 40 dB, so I didn't question
the result.  The breadboard actually measured nearly 63 dB.

The Miller capacitance of the output transistors sees to that (with
an even greater phase margin when a faster transistor is used for Q1).

I know.  I just particularly like the 399, and have a pile of
them.  I rationalize using it in this case by noting that the range
of frequencies where phase noise of the DA is important includes sub-
to low-Hz frequencies at which thermal effects could make the
unheated 329 significantly noisier (though if you keep drafts off
both of them, it might not be by a large amount).

Best regards,

Charles

I have seen National LMH6702 current-feedback video amplifier chips in
non-inverting amplifier configuration used to implement a wideband 10 MHz
distribution amplifier.

Joe

From: "Charles P. Steinmetz" charles_steinmetz@lavabit.com
To: Discussion of precise time and frequency measurement
time-nuts@febo.com
Date: 03/27/2012 07:28 PM
Subject: Re: [time-nuts] Distribution amp - Use a video amp unit ?
Sent by: time-nuts-bounces@febo.com

Bruce wrote:

expect.

D'oh!  Bruce is right -- I calculated the reverse isolation
incorrectly.  I had only been expecting 40 dB, so I didn't question
the result.  The breadboard actually measured nearly 63 dB.

used.

The Miller capacitance of the output transistors sees to that (with
an even greater phase margin when a faster transistor is used for Q1).

I know.  I just particularly like the 399, and have a pile of
them.  I rationalize using it in this case by noting that the range
of frequencies where phase noise of the DA is important includes sub-
to low-Hz frequencies at which thermal effects could make the
unheated 329 significantly noisier (though if you keep drafts off
both of them, it might not be by a large amount).

Best regards,

Charles

time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.

The LMH6702 is one of the few current feedback amplifiers that is stable
with a relatively low value (237 ohm) feedback resistor.

Its estimated phase noise floor is around -171 dBc/Hz with +13dBm output
(and input) in the noninverting 2x configuration with a 50 ohm load and
a 50 ohm resistor in series with its output.
Thats about 3dBc/Hz less noisy than the measured 10MHz phase noise floor
of a OPA653 in the same configuration with the same input and output
signal levels.
For offsets below about 100Hz the phase noise of an OPA653 is
indistinguishable from the noise of  measurement setup.

The effect of drafts and other air currents on the phase shift of RF
transformers is also significant for offsets of around 1Hz or less.
Extra thermal mass (encapsulation) and draft shields (even a piece of
paper) can work wonders.

Bruce

Joseph M Gwinn wrote: