On 7/7/18 11:33 PM, Jim Brown wrote:
On 7/7/2018 12:41 PM, Richard (Rick) Karlquist wrote:
Interesting paper, lots of stuff to absorb.
My comments here are about the relationship between noise coming in on
the antenna and circuit noise. FWIW, I know this material well, but find
the way the material is presented difficult to follow. Pity the poor guy
who doesn't already know it. I view this as a teaching failure, not a
technical one.
One important error -- for no degradation of signal to noise ratio by
circuit noise, the margin between circuit noise and noise from the
antenna at the input stage must be 10 dB, not 3 dB.
10dB would mean that the circuit noise is 1/10th of the noise from the
antenna, so the output noise is 1.1 times the antenna noise. So, for a
signal at 10dB SNR (with noiseless circuit), you'd now be at 9.6 dB SNR
(a degradation of 0.4dB)
3dB is circuit noise is half the noise of the antenna, SNR degrades from
10 to 8.2 dB (a degradation of 1.8 dB)
(If your desired signal is at 0dB SNR, then circuit at -10dB makes it
-0.4, circuit noise at -3dB makes it -1.8)
Note that for HF receivers feeding an ADC, where you've hopefully
designed so that the LNA is 10dB below the outside noise, you don't need
another 10 dB over the ADC noise. So we shoot for 3-5 dB LNA noise over
the ADC noise, which puts the ADC noise at 13-15 dB below the input.
The combination of ADC and LNA noise (referred back to the input) is
then around 9 dB below the desired signal.
I certainly agree with your comments about preamps with excessive gain,
and although I'm not a student of preamp design, my distant memory is
that gain is one of the important compromises in achieving a low noise
figure.
These days, unless you're well up into double digit GHz, getting 10-20
dB gain is pretty easy. In "the old days" where you were getting 7-8 dB
of gain, the whole design problem of noise from following stages was
pretty tricky. The second stage noise does contribute significantly.
When it's easy to get 20-30 dB of gain in a single stage, the downstream
stage noise performance isn't as big a deal.
In my experience, the challenge is actually dynamic range - getting
sufficient headroom over your largest signals without consuming lots of
power. For non-battery powered applications it's easier - this leads
you to designs like the LF preamp from Burhans (AMRAD has it) with a low
noise FET on the front end running a LOT of bias current.
For what it's worth, there's a lot of MMIC amplifiers out there with
very good HF noise performance, even though it's not in the data sheets
(typical MMIC data sheets go down the low-in-microwave-terms frequency
of 50MHz or 100 MHz). Off hand, 5-6 dB NF at 5-10 MHz is pretty easy to
get with a single cheap part, but you do need to measure one to see if
it will work for you - the 1/f flicker knee could be pretty high
(several MHz). The GALI-74+ is used in a lot of terrestrial HF radio
telescopes, for instance. +38dBm P1dB, 25+ dB gain at 100 MHz, 2.5 dB NF.
If you want an OpAmp based design the OPA656 is hard to beat - low
current noise and low voltage noise.
One of the issues that must be addressed with low output antennas is,
depending on the noise environment, antenna sensitivity, and feedline
loss, whether the preamp should be at the antenna or can be in the shack.
Preamp at the antenna in most cases - Especially if you're using a
"voltage probe" antenna (length << wavelength).
OK - if you have some big hardline already installed with very low loss
at your frequency (<1dB) then you might not need to.
But if you're going to be doing a new install, I'll bet the cost of
enough gain at the antenna that lets you run good quality but
inexpensive cable TV coax, is cheaper than the low loss coax.
(There is the issue of serviceability.. LNA at the antenna at the top of
a tree that is hard to climb might not be as good an idea, if it fails
from lightning or something).
Another question about the presentation is the validity of the noise
data -- how recent is it? In most of the developed world, noise levels
have increased significantly over the last decade or two with the near
universal use of switch-mode power supplies, solar systems, variable
speed motor controllers, leaky CATV systems (especially the HF signals
used for "backhaul" data from consumer to the system). A few years ago,
radio broadcasters asked for a serious study of the increased noise
level, especially on the AM broadcast band.
This problem exists well up into the low microwave bands. There was a
NPRM from the FCC asking about whether we should be moving to a
"brightness temperature" kind of background noise measurement to reflect
spectral occupancy, rather than trying to add up discrete sources.
That said, HF measurements I've seen that are made from orbit over urban
areas are below the "exceeded 0.5% of the time" curve in the ITU report.
The spacecraft (CASSIOPE) is at a few hundred km, so it sees all the RF
power from a circle about 500-1000km across, depending on the time of
day and frequency.
Over NE US, for instance, the background noise is about -142 dBm/m2/Hz,
which is about 23 dB above the Galactic Background (-165). Over
"quieter areas" (like Hudson Bay), the noise is about -154 dBm/m2/Hz
which is probably the receiver noise.
(kT of 40 and 53 dB, respectively, Galactic noise is about 30dB -
Some measurements made with a well matched vertical monopole and a low
noise receiver in the same general area at the same general time show
about -132dBm/m2/Hz
These are at around 7.5 MHz, around 0800Z (3AM local, in October)
To convert the monopole flux density to received signal power, multiply
by the antenna effective area, which is Ga*lambda^2/4pi = 400 square
meters (26 dB)..
73, Jim K9YC
_______________________________________________
_______________________________________________
TowerTalk mailing list
TowerTalk@contesting.com
http://lists.contesting.com/mailman/listinfo/towertalk
_______________________________________________
_______________________________________________
TowerTalk mailing list
TowerTalk@contesting.com
http://lists.contesting.com/mailman/listinfo/towertalk
|