In message , Jeff writes
On 27/06/2015 17:08, Ian Jackson wrote:
In message , Jeff writes

To me, NF refers to "noise floor".

Lets see him go below that.

GPS received signals are among the lowest "power" signals we
currently
grab. They sit just above the noise floor.


It might to you, but in this context it means either Noise Factor or
Noise Figure.

But you have to be careful, as "noise factor" is a numerical ratio, and
"noise figure" is in dB.

Of course you can go below the Noise Floor, and in some circumstances
and modes the signal is receivable and decodable.

In the analogue cable TV world, the noise figure (in dB) can be looked
at as the amount of noise power that (say) a real-world amplifier
notionally has at its input in excess of that which would be generated
from a perfect resistor as its source impedance.

As a rule-of-thumb, in a 4MHz vision bandwidth, a perfect 75 ohm
resistor generates -59dBmV. [Subtract around 48dB if you want dBmW.]

The output of a noiseless amplifier would be -59dBmV + G, where G is the
gain in DB.

The output of a real-world amplifier would be -59dBmV + NF + G, where N
is the noise figure.

One method of measuring the noise figure is first to feed the amplifier
first from a resistive source, and measure the output noise level. Next,
feed the amplifier from a source containing a known amount of noise, and
note the increase of output noise. The noise figure can then be calculated.

In practice, the noisy source is usually a calibrated noise meter*. The
first reading is taken with the noise meter set at zero additional noise
output, and then the noise output is increased until the amplifier
output level rises by 3dB. This means that the noise meter is now
contributing the same amount of noise as the amplifier, and the noise
figure can be read directly from its output display. [This conveniently
saves having to do any further calculations.]

*Usually, a noise meter has a calibrated output meter or other display,
and this indicates the level of its noise output in a stated bandwidth -
both as an absolute level, and as the equivalent in dB with respect to
the basic minimum absolute level. In the cable TV world, the minimum
would be -59dBmV (probably shown in microvolts) in a 4MHz bandwidth, or
0dB. If, to increase the amplifier output level by 3dB, the noise meter
output had to be turned up to -49dBmV / 10dB, its noise figure would, of
course, be 10dB.



Great way if you have a R&S SKTU!!

Indeed it is. The '3dB rise' method is essentially a good dodge for
engineers to avoid having to do any hard sums.

The normal way these days is the Y-factor method and uses a switchable
noise source with a fixed known and calibrated Excess Noise Ratio
(ENR). The noise power from the device is measured with the source on
and off and the NF calculated from that ratio. That is how Noise figure
test sets normally work.

Which is sort-of what I said in the middle of my ramblings.

It should be relatively easy to conjure-up your own noise measuring
machine by using an old-fashioned, high-gain, rather noisy, wideband
amplifier as the noise source, and follow it with a switched (or
calibrated variable) attenuator. If you know the amplifier noise figure,
and its gain, you know how much output noise it will produce - although
it would help if you can get a friendly guru to check.

For high noise levels and low attenuator settings, the noise is
essentially inversely proportional to the attenuator setting, but if
there's low noise and a lot of attenuation, the noise output becomes
asymptotic to the basic noise floor of the noise generated in the
attenuator itself (ie no matter how much attenuation you switch in, the
noise doesn't get any lower).

--
Ian