Power Meter

Interdigital Filter

Power Meter

Fig 3—System for measuring sun noise.

If the dishes are pointed at the horizon for terrestrial operation, the situation is much different. At least half of each antenna pattern is illuminating warm earth, so we should expect the noise temperature to be at least half of 290 K, or about 150 K. Adding 1 dB of loss increases the noise temperature to 179 K, a 1 dB increase. At higher noise temperatures, losses do not have a dramatic effect on signal-to-noise ratio. In practice, the antenna temperature on the horizon may be even higher since the upper half of the pattern must take a much longer path through the warm atmosphere, which adds noise just like any other loss.

Most receiving systems use at least one frequency-converting mixer that has two responses, the desired frequency and an image frequency on the other side of the local oscillator. If the image response is not filtered out, it will add additional noise to the mixer output. Since most preamps are broadband enough to have significant gain (and thus, noise output) at the image frequency, the filter must be placed between the preamp and the mixer. The total NF including image response is calculated:

G, image

G, r desired any filtering, G

image

Fig 4—A noise source built on an SMA connector.

Eq 10

assuming equal noise bandwidth for desired and image responses. Without noise figure, which is the same as adding 3 dB. Thus, without any image rejection, the overall noise figure is at least 3 dB regardless of the NF of the preamp. For the image to add less than 0.1 dB to the overall NF, a quick calculation shows that the gain at the image frequency must be at least 16 dB lower than at the operating frequency. A filter is a convenient way to reduce gain at the image frequency, but it must be between the preamp and mixer.

So far we have discussed the sources of noise and a figure of merit for evaluating the receiving system's response to noise. How can we measure an actual receiver?

The noise figure of a receiver is determined by measuring its output with two different noise levels, T[10t and T"cold> applied to its input. The ratio of the two output levels is referred to as the Y-factor. Usually, the ratio is determined from the difference in dB between the two output levels, YdB:

Then the receiver Te may be calculated using Y, ratio) :

and converted to noise figure:

Eq 12

1desired s0

NF= 10 log

Eq 13

The two different noise levels may be generated separately, for instance by connecting resistors at two different temperatures. Alternatively, we could use a device that can generate a calibrated amount of noise when it is turned on. When such a device is turned off, it still generates noise from its internal resistance at Tc0\d, the ambient temperature (290 KJ; usually this resistance is 50 £2, to properly terminate the transmission line that connects it to the receiver. When the noise generator is turned on, it produces excess noise equivalent to a resistor at some higher temperature at 2\ot. The noise produced by a noise source may be specified as the Excess Noise Ratio (ENRdB), the dB difference between the cold and the equivalent hot temperature, or as the equivalent temperature of the excess noise, Tex, which is used in place of Thot in Eq 12. If the ENR is specified, then the calculation is:

Eq 14

The terms T"ex and ENR are used rather loosely; assume that a noise source specified in dB refers to ENRdg, while a specification in "degrees" or kelvins refers to Tcx.

An automatic noise-figure meter, sometimes called a PANFI (precision automatic noise-figure meter), turns the noise source on and off at a rate of about 400 Hz and performs the above calculation electronically.4 A wide bandwidth is required to detect enough noise to operate at this rate; a manual measurement using a narrowband communications receiver would require the switching rate to be less than 1 Hz, with some kind of electronic integration to properly average the Gaussian noise.

Noise-figure meters seem to be fairly common surplus items. The only one in current production, the HP 8970, measures both noise figure and gain but commands a stiff price.

AIL (later AILTECH or Eaton) made several models; the model 2075 measures both NF and gain, while other models are NF only. The model 75 (a whole series whose model numbers start with 75) shows up frequently for anywhere from $7 to $400, typically

Use minimum lead length

Use minimum lead length

Noise Output

Noise Output

Fig 4—A noise source built on an SMA connector.

Fig 5—Schematic diagram of a noise source built on an SMA connector. Q1—Tiny silicon NPN RF transistor R2—Select to set current (see text), such as NEC 68119.

Was this article helpful?

## Post a comment