[Date Prev][Date Next][Thread Prev][Thread Next] - [Date Index][Thread Index][Author Index]

Re: Converter gain and noise.. (Long)

>  Someone recently
>  posted an excellent article on the ins and outs of very low noise receiving
>  systems and the various issues involved, including correct gain structure
>  and clipping in the IF stages (easier done with excessive converter gain).

>  Please don't hesitate to tell us if you find out where it is, I would like 
>  to read it too......


Below is the post. Thanks to Dick, K2RIW and Steve, K1FO for this excellent 
I've also included one other post I got from Ian, G3SEK concerning setup of 
adjustable gain preamps. I use this method with my SSB preamps in both my 
satellite and weak signal work.

Mike, N1JEZ

Why Fight For That Last Few Tenths of a dB in LNA Noise Figure.

    By K2RIW, 03/20/01

-- Some of this information came from the "Do
You Need a LNA Line Driver" presentation by K1FO at the 3/17/01 NEWS
Meeting at Enfield, CT, USA.  In many amplifying devices, there is a
considerable difference between tuning for the best gain, and the best
Noise Figure (NF).  Under certain conditions it is possible to tune a
particular amplifier for best gain and realize a NF of 0.9 dB.  That same
amplifier may deliver 0.3 dB NF, with slightly less gain (maybe 0.5 dB
less), when tuned for best NF.  At first you might say, gee, a 0.6 dB NF
improvement.  Is that small improvement worth fighting for?  The answer
is YES, YES!  In the paragraphs that follow I'll attempt to show you:

  (1) Why it is worth fighting for.

  (2) Why it is hard to detect the improvement with typical laboratory

  (3) What else in the system must be working right to get the full

  (4) The Antenna Impedance problem.

  (5) You'd better have the correct Gain Distribution.

(1) Why is it Worth Fighting For -- You're about
to have a demonstration that dB's of NF are not linear, when dealing with
a cryogenic communication system; and a small NF improvement can make a
big difference in the Signal to Noise Ratio (SNR).  A communication
system becomes cryogenic when you aim the antenna into cold space.

    Two EME System Examples -- Assume I have an
excellent EME antenna that is presently aimed at a high elevation angle,
and it is aimed at a cold direction in the Universe, where the true
Celestial Background is 2.73 degrees Kelvin (the residual of the Big
Bang, by Penzias and Wilson, Nobel Laureates, 1968).  Under those
conditions that excellent EME antenna might have a total Antenna Noise
Temperature (Ta) of 30 degrees Kelvin (they [the experts] call it 30

    -- If your Low Noise Amplifier (LNA)
had a NF of 0.9 dB, that equals an electronic Noise Temperature (NT) of
66.78 Kelvins.  The formula for NT is:

NT = [ALOG(NF/10) - 1] * 290.

Your total possible system NT (Ts) is now the sum of the Ta and the NT.

Ts = Ta + NT.  Ts = 30 + 66.78 = 96.78 Kelvins.

    -- If I was using that LNA when it is
tuned at a NF of 0.3 dB, the LNA's NT would equal 20.74 Kelvins.  Now the
Ts = 30 + 20.74 = 50.74 Kelvins.  The only thing that is always linear
about a cryogenic communication system is that the total Noise Power is
proportional to the Ts in Kelvins.  You have to almost ignore the dB's.

    When I compare a system with a Ts of 96.78K (A) to a system with a Ts of
50.74K (B), that ratio is 2.80 dB.  In other words, that EME signal will
be 2.80 dB further out of the noise when I use the LNA with the better
tuning (I'm ignoring Moon Noise).  That 2.80 dB of difference can easily
make the difference between a QSO and a missed QSO, when I'm listening to
a weak EME signal.

    A Better EME Antenna -- If I was able to make
further improvements to that EME antenna so that the Antenna Temperature
(Ta) was 20 Kelvins (and, that is possible), then that ratio would be
86.78 Kelvins versus 40.74 Kelvins, or 3.28 dB of system improvement in
Signal to Noise Ratio (SNR).  All of this from a 0.6 dB NF improvement
(0.9 dB to 0.3 dB NF) in the LNA!  A ~ 3 dB SNR improvement for a 0.6 dB
NF change, that's a beautiful (apparent) non-linearity!  But, it really
isn't non-linear, it just looks that way (in dB's).

(2) Why is it Hard to Detect With Typical Lab Equipment
--  When I make measurements with room temperature
laboratory equipment, everything (the pads, signal generators, FM
Receiver, etc.) is at approximately 290 kelvins.  In that 290K
environment, a change in the LNA's NF from 0.9 dB to 0.3 dB, can create a
change in SNR of 0.6 dB (at best).  Here the dB's are linear.  Unless you
are using some good laboratory equipment (such as WB6KBL's SINAD Meter,
or a good NF Meter), you will probably not detect that 0.6 dB SNR
improvement.  Your ear probably doesn't have enough discrimination to
allow you to hear it, when you hit that "Sweet Spot" in NF tuning.  And,
as section (3), (4), and (5) will show, you will not be able to realize
the benefit of that 0.6 dB of NF improvement, unless the rest of the RCVR
has a NF of nearly 1.0 dB or so, has the proper antenna impedance, and
the proper gain distribution.

(3) What Else in the System Must Be Working Properly to Get
the Full benefit -- If I was using a RCVR that was Gain
Starved, or had a second stage NF of 15 dB, then the tuning of the LNA
takes on an entirely different characteristic.

    High Transceiver NF -- Many of the currently used
Base Station Transceivers have a bare foot NF of 12 to 15 dB.  This
occurs because Japan favors dynamic range over RCVR sensitivity.  Those
Transceivers are front end Gain Starved.  If you lived in a dense
community where there was a Ham Radio Operator living on each street, you
might agree with this approach.

    Add an LNA -- Therefore, almost every American
SSB operator must add a ~ 20 dB gain LNA in front of his Transceiver, if
he desires full sensitivity of his communication system.  If he doesn't
add that LNA, all his SSB friends will eventually call him an Alligator
(he is all mouth), instead of a Rabbit (a guy that is all ears).  During
a terrestrial contest, everybody can hear that Alligator call CQ, and
they answer him, but because of his poor hearing aide (RCVR), he only
hears the locals and only responds to them.  The rest of the contest
operators become frustrated, and learn to ignore him.  Unless some local
explains this to the "Alligator," he will conclude that there wasn't much
activity during the contest.

    It is well known that the total system's cascaded NF (NFs) is equal 

  NFs = NF1 + (NF2-1)/G1 + (NF3-1)/(G1*G2) + ..., where:

  NF1 = NF of the first stage (as a real, anti-LOGed number).

  NF2 = NF of the second stage (as a real number), etc.

  G1  = Gain of the first stage (as a real number).

  G2  = Gain of the second stage (as a real number), etc.

    Sometimes Best Gain = Best NF -- If I was tuning
a communication system's LNA, while it is connected to that bare foot
Transceiver, I would find that the best system NF would be approximately
the LNA tuning with the maximum gain.  Here is an example:

If you experiment with just the first two terms of that Cascaded NFs
formula, and use 15 dB for NF2, 0.3 dB for NF1 and a gain of 10 dB for
G1, you will find that the System Noise Figure (NFs) [in dB's] is equal
to 6.16 dB.  If, now, I use 0.9 dB for NF1 and 10.5 dB for G1, I realize
a NFs of 5.98 dB.

    Notice, that I worsened the LNAs NF by 0.6 dB, while improving the gain
by only 0.5 dB, yet the system NF IMPROVED by 0.18 dB.  What this example
demonstrates is that when your RCVR system is gain starved, and has a
high second stage NF, then the LNAs gain is much more important than it's
NF.  Even if I was using perfect laboratory instrumentation (such as a
perfect NF Meter) while tuning that LNA in that environment, I would end
up tuning it for maximum gain, not best NF.

    The conclusion is that to get the maximum benefit of a Super Low Noise
LNA, you have to put it into the right environment while tuning it, and
using it.  Otherwise, you may be "casting pearls upon swine," you could
be wasting your time and your money.

(4) The Antenna Impedance Problem -- It is well
known that when you are tuning an LNA for the best NF, you are primarily
adjusting the impedance that the front end of the LNA is looking into. 
It would be quite wasteful to carefully adjust an LNA stage while it is
connected to perfect 50 ohm resistive laboratory equipment, and then
connect it to an antenna with a VSWR of 1.41:1.  That 1.41:1 antenna
could be an impedance that consists of a capacitive reactance of 50 ohms
in series with a resistance of 50 ohms.  That antenna would drastically
change the LNA's NF.

    That uncorrected 1.41:1 VSWR antenna could easily raise your LNA's NF
from 0.3 dB to 0.9 db, and hurt your EME RCVR sensitivity by ~ 3 db.  But
on transmit, the 1.41:1 would only cost you 0.127 dB of transmission

    Your corrective choices are either to perfectly impedance match the EME
antenna (with a double stub tuner, for instance) to make it look like a
50 ohm resistive load to the LNA, or do the NF tuning of the LNA while it
is connected to the antenna -- such as by injecting the NF Meter's Noise
Source through a 20 dB Directional Coupler (DC) that is always left in
the antenna line.

    The DC Line Perturbation -- If, after the NF
tuning of the LNA, you made the mistake of removing the Directional
Coupler, you would be changing the transmission line length, and that
would rotate the antenna impedance to a different place on the Smith
Chart.  This would disturb the LNA tuning.

    Another solution is to add a carefully chosen extra length of line onto
the DC's straight through path, so that the DC plus the extra line is an
exact multiple of a half wavelength (electronically).  Now you could
remove that DC plus extra line, and not effect the antenna's impedance.

(5) You'd Better Have the Correct Gain Distribution
--  To realize the system's best possible
sensitivity requires that you have enough front end gain, and a low
enough second stage NF.  But, this requires a compromise of system NF vs
Dynamic range.  You usually can't have both all at once.  

    The best possible system NF usually requires a lot of front end gain
(sometimes 20 to 30 dB).  But, a system  with that much front end gain
will saturate 20 or 30 dB sooner from strong local signals -- that's the
problem that the Japanese equipment manufacturer's discovered.

    Noise Power Saturation --  Also, bear in mind
that even if you live in the "Out Back," and saturation from local
operator's isn't a problem, there can be another subtle detriment from
the use of super high front end gains -- Noise Power Saturation.  It is
possible that the later stages of your RCVR are being subjected to so
much Noise Power, from all the front end gain, that they are beginning to
saturate on the instantaneous noise peaks.  Even if that saturation is
only a fraction of a dB, it can lower the SNR of a weak signal.

    It is well known that a limiting stage will suppress a weak signal
that's surrounded by noise with, what is called, "Signal-Cross-Noise
Terms."  In other words, it is possible for a super high gain system to
suppress that weak EME signal you're trying to hear, in the later stages
of your own RCVR.  This phenomenon is quite subtle, and not easy to
detect.  But, if the gain in your system is shoving the S Meter above S7
on basic Noise Power, than be wary, it could be happening to you.  The
only quantitative test procedure I know of to detect this condition is
the "Notched Noise Power Fill-In Test," also called the Noise Power Ratio
(NPR) Test.

    IF Filter BW -- It is also possible that your
system is going into and out of Noise Power Saturation, as you change the
bandwidth of the IF filter.  At first, you would think that the broader
IF filter selection would aggravate the problem.  However, it is possible
that the more narrow filter selection allows less noise power into the
final detection stage, and this in turn causes a smaller AGC voltage,
which increases the RCVR's gain, and causes Noise Power Saturation in an
earlier stage.

    Sun Noise Problems -- As your EME system becomes
more refined, and you experience a larger number of dB's of Sun Noise
measurements, it is possible that with the added Sun Noise power, your
RCVR system could be experiencing Noise Power Saturation.  That would
give you a pessimistic Sun Noise measurement.  One simple method of
detecting this problem would be to put a 6 dB pad in various places
(after the LNA), and repeat the Sun Noise measurement.  If you get a
better reading, you may have the problem.

    The best system for high dynamic range is one that has a gain
distribution that's just enough, at each stage, to override the NF of the
next stage.  The best system NF requires considerably more front end gain
than that.  Soon we will all pay more attention to the Noise Power
Saturation characteristics of tunable IF RCVRs.  Then we will
simultaneously have the best system NF and high dynamic range.

--->    I hope this information is helpful.  Please feel free to correct the

    73 es Good VHF/UHF/SHF/EHF/Laser DX,

    Dick, K2RIW.

    Grid: FN30HT84DC27

Here's a method that requires no test equipment at all. It comes from
G4DGU, who designed all the original muTek transverters and outboard
preamps to have adjustable gain. This method uses the sharp threshold
effect of FM detectors at low S/N ratios, and it allows you to optimize
the preamp/transverter gain for your local band noise conditions.

1. Turn the transverter/preamp gain well up. 

2. Find a very weak but steady unmodulated carrier (off-air, not from a
signal generator or a local birdie). Rotate the antenna until you can
just detect the signal in FM mode.

3. Reduce the preamp/transverter gain until you hear the noise increase.
The FM threshold is sensitive to a small fraction of a dB in S/N.

4. Increase the gain just a little,to the point where you can't hear the
quieting improve much. 

5. Switch back to a real DX mode.

Remember that every dB of unnecessary preamp/transverter gain will
probably subtract almost 1dB from your system intermod intercept!

The penalty of adjusting the gain correctly is that you're living just
above the "knee" where S/N will begin to deteriorate rapidly if
something changes. It's worthwhile to repeat this test every few months
- especially just before a contest.
73 from Ian G3SEK          Editor, 'The VHF/UHF DX Book'
                          'In Practice' columnist for RadCom (RSGB)
Via the amsat-bb mailing list at AMSAT.ORG courtesy of AMSAT-NA.
To unsubscribe, send "unsubscribe amsat-bb" to Majordomo@amsat.org