You are here: hb9bbd.ch >> articles >> Noise Figure
Noise Figure
Category: EME Equipment,
Created: 2001-10-22 21:02:20
EME stations using very high transmit power risk to be heard by small stations which they could not hear themselves.
In line with increasing transmit power in the dish feed, special attention has to be devoted to the sensitivity of the receiver system. Its sensitivity is predominantly determined by the line from the antenna to the first stage of the preamplifier and its Noise Figure.
Typical Noise Figures of preamplifiers for the 23cm band are between 0.2 and 0.4dB.
However, more important than the Noise Figure of the preamplifier is the System Noise Figure or the system noise temperature. The line to the Preamplifier, its Noise Figure plus antenna noise temperature determine the system Noise Figure.
A very practical and efficient way to determine system Noise Figure is described here: System Noise Figure and here: Preamplifier Noise Figure Measurment
A widely discussed and debated topic is the accuracy of NF measurments. We have to accept that even the best testequipment is not perfect and various sources of errors may disturb the results. Some of these errors can be corrected, others cannot. The effects of changing ambient temperature e.g. or of mismatch between the noise source and the DUT can be eliminated. Modern NF testequipment automatically correct the temperature effects. Since Low-Noise, narrow-band preamplifiers typically come with poor input match, the error of mismatch has to be compensated. Isolation of the Noise Source to DUT >30dB is needed. Agilent offers a spreadsheet calculator to determine the effects of mismatch, and more:
Noise Figure Uncertainty Calculator
My friend Paul Chominsky, WA6PY has written the following note for further understanding the complex problem of the effects of mismatch:
-– When measuring NF, any change in input return loss of an LNA is directly exchangeable for NF. PHEMT LNAs at 1.3 GHz have a high input Q value. According to the Fano-Bode criterion, we cannot obtain a better input return loss (RL) than a certain value over a specified bandwidth. With such a high Q load, we could design a very narrow band matching circuit and theoretically handle an RL close to infinity, but the question is how to design such a matching circuit and keep its losses to a very low value. The ratio Loaded Q to Unloaded Q of a matching circuit is related to loss and consequently limits NF. We want the lowest possible loss input circuit for best NF. In order to provide a very narrow band for matching satisfying Fano-Bode criterion, we might need a matching circuit that has an unloaded Q value close to infinity. The fact is that in order to achieve very low NF, we do not need very good RL. In ham applications a very good LNA RL is usually not necessary. I do not think that for very low NF using a PHEMT with a gate of 0.2 x 200 microns, we can achieve RL < -4 dB. If we use a device with a broader gate, then the input Q will be lower and we can get a better RL. But the noise temperature (due to the PHEMT) might slightly rise. The problem is that there is no commercial production of such PHEMT devices, because there are no applications other then for amateur radio moonbouncing that need such devices. The radio astronomers cool down their LNAs to 4.2 deg K. Measurement of the RL depends on the RF input power because a PHEMT is a non-linear device. Minimum NF can be obtained only for very low input RF power. NF measurements of high RL devices are very sensitive to the mismatch of the ON and OFF states of the noise source. I never optimize my LNAs alone. I always have all the adapters and isolation relay in place during optimization. The RL of my feed horn is <-26 dB. Thus when I connect my system all together, I can verify by measuring Cold sky/Ground noise that the system works as designed.
Some formula to explain Noise Figure and System Noise Temperature:
- Noise temperature (T) = 290 * (10^(Noise Figure/10)-1) K
- Noise Figure (NF) = 10 * log (Noise factor) dB
Notes:
log must be to base 10. When using calculators and spreadsheets make sure that base 10 is selected. As a test, 10 * log(2) should give an answer of +3 dB.
Noise temperature is measured in units called Kelvin (K) and these are like Celsius (C) temperature degrees but start at zero for absolute zero temperature so
0 K = -273 deg C
273 K = 0 deg C (ice melts)
290 K = 17 deg C (ambient temperature of a cable, for example)
Back to Mainpage |
