SM 5 BSZ - Preamplifier design.
(June 3 2001)

Antenna noise temperature and receiver noise temperature.

When designing the first amplifier stage of a receiver the antenna noise temperature, the amount of random noise picked up by the antenna and generated by losses in the antenna itself and cables between the antenna and the preamplifier represents the best (lowest) possible noise floor one can possibly obtain. An ideal receiver would not add any noise of its own but real world receivers always add at least some noise and therefore the real S/N is always lower.

This page discusses preamplifiers at 144MHz because that is the frequency where I have practical experience. The discussion is of course valid for any frequency band if the 144MHz noise temperatures are substituted for values typical for other frequencies.

The noise performance of an amplifier is usually expressed as the noise figure of the amplifier. On this page noise is expressed as a noise temperatures because noise temperatures are additive. Noise figures and noise temperature are two different ways expressing the same thing.

The antenna temperature may be as low as 190K on 144MHz (optimum EME conditions). If we allow 5% of the noise at the output to be generated by the amplifier itself we find that the noise temperature has to be as low as 10K.
10 / (190 + 10) = 5% The performance loss compared to an ideal receiver is then 0.2dB which is a small number but not completely insignificant. An amplifier with noise temperature 10K has a noise figure, NF of 0.15dB. With modern devices noise figures like this can be obtained, but only if the losses of the input circuit are extremely small, something that can be achieved in case input selectivity is unimportant. In case very strong out of band signals are present a compromise has to be made.

Saturation and cross modulation

First of all, selecting a transistor that does not saturate easily will make this problem less difficult. Secondly, by use of noiseless feed back (source inductor) it is possible to reduce the gain and thereby make the saturation come at higher input voltages. In case these two measures do not cure the problem, input selectivity is required. see below

The input circuit

A typical preamplifier with a GAS-FET has an input LC circuit that is used to step up the impedance of the antenna to the impedance at which the FET gives optimum noise figure. How much transformation is needed depends strongly on the device. An ATF33143 likes about 150ohms, a MGF1801 about 600 ohms while a MGF1425 needs 5000 ohms for optimum NF on 144MHz.

A typical LC circuit will have a Q in the order of 300 on 144MHz using an air dielectric capacitor and 1mm wire. This means that the losses caused by the LC circuit corresponds to a resistor in parallel with the circuit with a resistence 300 times bigger than the reactive impedance of the capacitor (or inductor). Fig. 1 shows an idealised schematic diagram of the input circuit of a GASFET preamplifier.



Fig.1 Idealised input circuit of a GAS-FET amplifier. The antenna is connected to Ant1 or Ant2. R1 represents the losses of L1, C1 and C2.

In the discussion of the noise contributions from the input circuit, represented by R1 in fig.1 the transistor is assumed to be neutralised. Neutralisation is possible in real world amplifiers and allows stable operation with an output that is matched for maximum power gain. The neutralised GAS-FET is a very small capacitance to ground with negliable losses. The effects of lossless feedback is to reduce the gain at the signal frequency. Noiseless feed back changes the frequency response of the input circuit and it changes the input impedance but it does not change the noise contribution by the input circuit losses.

If the antenna is connected to Ant1, C2 is used to set the transformation from the antenna impedance, usually 50 ohms, to the impedance that will make the transistor noise performance optimal.

In case the antenna is connected to Ant2, there is no need for a C2 capacitor and the impedance transformation is set by the position of the tap on L1.

The input coil L1 is in resonance with C1+C2 at the signal frequency. This means that the reactance in ohms of the capacitor C1+C2 is equal to the reactance in ohms of the inductance L1. The equivalent parallel loss resistor is Q times the reactance in ohms of C1+C2 (or L1). The table below shows typical values for C1, Q and R1 for some typical input circuits at 144MHz. A quarter wave coaxial cable with characteristic impedance Z has an equivalent loss resistance of 1.414 * Z * Q.


    C1(pF)      Q      R1(ohms)     Coil description
     15pF      340       25k     coil from 1mm tinned wire
     15pF      440       32k     coil from 2mm silvered wire
      5pF      450      100k     coil from 1mm tinned wire
      5pF      600      133k     coil from 3mm silvered wire
quarter wave    70     4900      semirigid 2mm (inner = 0.5 mm)
quarter wave   100        7k     RG223 5mm (inner = 0.9 mm)
quarter wave   180       13k     RG8 11mm (inner = 2.2 mm)
quarter wave   400       28k     1/2" Flexwell 16mm (inner = 3.9 mm)

Table 1. Typical Q values and equivalent loss resistances
for input circuits at 144MHz.
The losses originate essentially in the resistance of the coil wire for LC circuits and in the center conductor and the dielectric for coaxial quarter wave resonators.

A thicker wire allows the current to distribute over a larger area with smaller losses and higher Q as a consequence. A smaller capacitor allows a longer wire for the coil. A smaller current is distributed over a larger area so Q increases.

Coaxial cables with solid dielectric should not be used at all. Simple LC circuits perform better. One may think of a 50 ohm coaxial cable as 100pF/m or a 33pF capacitor for a quarter wave RG/8 cable. The capacitance is distributed along the inductor of the inner conductor so it corresponds to a single capacitor of 16pF at the top.

If the capacitance is 15pF, the losses correspond to a resistor of about 25 kiloohms from gate to ground for a simple LC circuit or a 1/2" air dielectric 50 ohm cable.

25 kiloohms from gate to ground corresponds to about 2.5% losses for an MGF1801. The losses will attenuate the incoming signals and the sky noise and it will add a noise contribution of its own, 2.5% of the coil temperature. The losses act as an attenuator with an attenuation of 2.5% which is the same as an amplifier with a power gain G = 0.975 and with a noise temperature of ( 1/G - 1 ) * 290K, assuming the amplifier is more or less at room temperature.

The noise temperature seen by the receiver will increase with the noise temperature of the input filter losses. Table 2 shows filter noise temperature for different loss levels.

Assuming the transistor itself has a noise temperature of 15K a lossless input would give an input temperature of 205K while 2.5% losses increases the temp to 212.5K, a S/N loss of 0.15dB. Small but not entirely insignificant. Table 2 gives degradation of 144MHz amplifiers under the assumption that the transistor itself has a noise temp of 15K



 Power   Temp       S/N loss
  loss   (K)        at 205K
  0      0             0dB
  1%     2.9         0.06dB
 2.5%    7.4         0.15dB
  5%     15.2        0.30dB
 10%     32.2        0.63dB
 15%     51.2        0.97dB
 20%     72.5        1.32dB
 25%     96.7        1.68dB


Table 2.
Noise temperature to add to antenna temperature
due to losses in the input circuit of the preamplifier
and influence of input losses on S/N ratio.
The antenna temperature is assumed to be 190K and the
transistor temperature is assumed to be 15K.
These assumptions represent minimum values on 144MHz
in a good location with a ground temperature around 300K
and with very low cable losses.

On higher frequencies where sky temperature is lower, avoiding losses becomes more important. For an antenna temperature of 30K and a transistor temperature of 30K the system noise is 60K without losses. 2.5% losses would lead to a system temperature of 67.4K and a loss of S/N of 0.5dB.

For an MGF1425 25 kiloohms equivalent loss resistance would completely destroy performance. One would have to compromise and run the transistor at a lower than optimum impedance with a degraded performance of the transistor itself. Too small LC ratios is a common reason for poor performance of GAS-FET preamplifiers. Just adding one or a few turns to the coil and reducing the capacitance may improve noise figure significantly while the input bandwidth becomes much larger.

The input circuit in good locations

The output power that can be extracted from a small GAS-FET like MGF1425 is 10mW. Operated in a neutralised amplifier the gain is 30dB at minimum noise figure if the output is matched for maximum power gain. Such an amplifier should not receive more power than -20dBm from the antenna.

To get minimum losses in the input LC circuit the antenna should be connected to Ant2 in fig.1 (C2 absent). C1+C2 can be made about 1pF with stray capacitances only. Fine tuning is made with a screw that changes stray capacitances slightly. With 2mm silvered wire Q will be about 800 and the equivalent loss resistance R1 about 900 kiloohms which gives loss of about 0.5% in the input circuit and an associated loss of S/N of 0.03dB which is surely small enough to be neglected.

When the antenna is connected, the input coil is loaded by the antenna and the impedance becomes about 5 kiloohms at the gate of the FET. The loaded Q of the input circuit then becomes 4.4 only and the 3dB bandwidth is 33MHz.

The MGF1425 example given here is more or less worst case when it comes to poor performance with regards to overload and cross modulation. The noise performance is excellent and I have been using this kind of amplifiers for many years without any problems at all.

For inband signals (caused by other amateurs) serious degradation will occur at signal levels of about -35dBm because of the sideband noise in commercial amateur equipment. There is a 15dB margin until the high gain MGF1425 is overloaded. Only under very special circumstances when the friendly amateur neighbour has excellent transmitter purity (home made equipment) it will be necessary to improve the dynamic range of the input amplifier.

Out of band interference is quite another thing. FM broadcast and television transmitters may produce strong signals at the preamplifier input that would completely saturate my MGF1425 amplifiers. If there are 5 transmitters, all of them producing 0dBm at the antenna connector, the instantaneous power will be +14dBm at the antenna connector. With an input bandwidth of 32MHz, the attenuation in the 88 to 108MHz range is about 10dB causing a +4dBm signal at the transistor saturating it by 24dB!!! It is obvious that the wide band, high gain, low level amplifier can not be used everywhere.

Preamplifiers with high dynamic range

A device like MGF1801 or ATF33143 can deliver about 100mW, +20dBm when optimised for minimum noise. In a neutralised circuit the power gain is then around 30dB so the maximum input signal is about -10dBm. Noiseless feedback is usually implemented by a source inductor that converts the 90 degree feedback via the drain to gate capacitance to a 180 degree feedback due to the 90 degree phase shift introduced between gate voltage and source current.

Noiseless feedback lowers the gain at 144MHz, but it does that by lowering the Q of the input circuit (loading the gate by a noiseless resistor) so although it improves IM3 and saturation close to 144MHz it has practically no effect at 88 to 108MHz.

Using a "cavity", an air dielectric 50 ohm transmission line gives a loaded Q of 9 if a MGF1801 is connected at the top and the line is loaded by the antenna to give 600 ohms at the transistor. The suppression of FM broadcast transmitters will be about 15dB which is adequate in most cases. To avoid noise figure degradation the unloaded Q of the transmission line should be 900 (1% losses, see table 2) which requires a center conductor of about 8mm (see table 1).

In really extreme situations one would like the preamplifier to tolerate +15dBm (1W) in the 88 to 108MHz range. (6 transmitters, all at 0 dBm). This would require an attenuation of 25dB in the input circuit which implies a loaded Q of about 40. Using a big cavity, for example constructed from 2" air dielectric cable with the transistor halfway along the cavity the unloaded Q may be 1800 with losses around 2.2%.

Rather than using a single huge high Q resonator to suppress interference one may use two coupled resonators. This is much easier and gives better performance. The reason is that the effective Q multiplies while losses add.

Two LC circuits with loaded Q = 6.5 and an unloaded Q of 500 (2mm Cu wire, 5pF, see table 1) will give 1.3% losses each while providing a common Q of 40.

A properly designed 2 resonator preamp is not difficult to tune for optimum NF. It has a flat response over 10 to 20 MHz with steep falloff's at both sides. Amplifiers of this kind are not commonly used and I am not aware of any detailed descriptions. There are other solutions that require no instruments at all to be set up properly.

Using stubs to remove out of band signals

In case problems due to FM broadcast stations occur, some 0.75 wavelength shorted stubs is a very good idea. Such a stub has a very high, purely resistive impedance at 144MHz. The amount of power lost in each stub is the same as when the signal travels through 1.5 wavelengths of the same cable, 0.075dB = 1.8% for 0.5" air or foam dielectric 50 ohm cable. It is a good idea to make the stub from 75 ohm cable because losses are then reduced by the impedance ratio.

A 0.75 wl stub at 144MHz is a 0.5wl stub at 96MHz. At this frequency the attenuation is extremely high because a shorted 0.5wl stub is very close to zero ohms.

A stub will work well only if it is placed at a point along the transmission line where the impedance for the interfering signal is high and reasonably resistive. In case the stub is placed at a point along the transmission line where the impedance is low and reactive, the stub may amplify the interference because a few MHz away from 96MHz it acts as a high quality inductor or capacitor and at some frequency it will improve the matching between the antenna and the preamplifier, both of which are far away from 50 ohms in the 88 to 108 MHz band.

By placing three stubs after each other along the transmission line with a separation of a quarter wave at 96MHz one can asssure that the center stub looks into a very high impedance in both directions. Three 75 ohm 0.75wl stubs from 0.5" air or foam cable will give losses at 144MHz of 3.6% with an associated S/N loss of 0.22dB. The attenuation from 91 to 101MHz is in excess of 30dB. Using one inch 50 ohm cable for the three stubs one can expect more than 30dB attenuation from 89 to 104MHz with a S/N loss of 0.15dB at 144MHz.

It is a very good idea to mount the stubs between the relay and the antenna. It will make sure your transmitter does not emit any spurs in the FM broadcasting band.

In case there is a power divider very close to the relay, place the first stub at the power divider, then as many second and third stubs as you have lines going away from the power divider to avoid lengthening the feed cable. Using N stubs at the other side of a N-way power divider will not increase losses since each stub will only see 1/N of the power.

There is of course no reason to use expensive connectors. The stubs can be soldered like the rest of a stacking cable system. Just beware of water.....

Output filter and gain reduction

Usually GAS-FET preamplifiers are not designed for power matching at the output. Some resistors are normally used to ensure the transistor does not see impedances that will cause oscillations. Some gain as well as available output power is lost this way. Resistive loading is equivalent to putting an attenuator at the output of a power matched amplifier. If the gain loss is permissible, it is better arranged by noiseless feedback on a power matched amplifier because noise figure and gain will be the same but available output power will be a few dB higher which means that IP3 is a few dB better.

Since the IM3 properties near 144MHz rarely is of any significance at all, resistive non-matched output circuits are nearly always used.

Source resistor noise.

The source leads of the GAS-FET are normally decoupled by capacitors to allow a DC voltage on the source. Do NOT connect the DC voltage through a resistor!!! The source lead and the decoupling capacitor form an inductor to ground and there will be some RF voltage at the point where the DC voltage is connected.

Always connect the DC voltage through an inductor that is decoupled at the other end where resistors or whatever the bias circuitry wants is connected. The source impedance should be purely reactive, any resistive component will cause noise figure degradation. This may be a large effect, in the order of 0.5dB for NF.


To SM 5 BSZ Main Page