An amplifier optimized for 50 ohms.Based on the results for the G4DDK design I made an experimental amplifier for proof of concept. The input has a low loss quarterwave section with two tuning screws intended to allow the impedance to be adjusted to around 75 ohms at the point where a coil similar to the one used in the DDK design connects to the FET.
Figure 1 shows a photo of the entire box and figure 2 shows the details around the FET.
|Figure 1. Experimental LNA optimized for 50 ohm source impedance.|
|Figure 2. Details of the experimental LNA optimized for 50 ohm source impedance.|
Refer to figure 3 for the schematic diagram.
|Figure 3. Schematic diagram for the experimental LNA optimized for 50 ohm source impedance.|
L1 is a low loss inductor for DC-grounding on the gate side.
C1 and C2 are screws at both ends of the transmission line which is a silvered bar from 5x5 mm brass.
L2 is tuned with the gate capacitance just like in the G4DDK design.
C6 is the source decoupling capacitor. This capacitor must have extremely low losses. It is actually 15 capacitors in parallel on the edges of a 20 x 20 mm PCB. A 3 mm wide strip of 0.05 mm copper is soldered to PCB and the transistor is soldered to that strip. See figure 2. This way the source inductance is extremely low
R1 is the source resistor and L4 is a ferrite tube that ensures that the low-impedance noise source constituted by R1 can not inject much current into C6. (R1 and L4 should be interchanged in figure 3.) R1 is actually a fixed resistor plus a trimmer.
L3 is 2 turns of very thin wire wound on a 50 ohm resistor R2.
C5 is a 1 pf capacitor to make the drain essentially see R2 at higher microwave frequencies.
T1 is a tuned transformer, red and blue wires in figure 3. It provides nearly resonance with C5 and it provides a signal with 180 degree phase shift at the decoupled end of the red wire.
C3 is the neutralization capacitor, the wire just above the transistor in figure 2.
C4 is decoupling the cold end of T1.
L5 and C7 provide better isolation to the DC supply which could help in case a second stage would be added.
C6 is a 1nF coupling capacitor which is uncritical.
L6 (no label in figure 3) provides matching to 50 ohms.
The idea with this design is to avoid anything that might be lossy and to provide very high gain by use of neutralization. High gain in the first stage makes it easier to avoid a noise contribution from the second stage.
The S-parameters are shown in figure 4.
|Figure 4. The S-parameters of the experimental LNA optimized for 50 ohm source impedance. Note that the network analyzer is uncalibrated.|
It is obvious from figure 4 that the amplifier is not unconditionally stable. At about 1.5 GHz the phase shift of the signal feeding the neutralization capacitor is far from 180 degrees. A better network on the drain would solve this problem in case an amplifier for actual usage would be designed according to the ideas presented here. The amplifier is however perfectly stable with wideband 50 ohm terminations and the neutralization is fine on 1296 MHz with a backwards isolation of 46.6 dB.
This amplifier was compared to several other amplifiers at the Orebro 2013 EME meeting Optimum NF was obtained with R1=0 and a total current of 17 mA to the unit.
Modification for current.The unit was modified to allow a positive bias on the gate. A forward biased 1N4148 provides about 0.6V to the gate. A low loss capacitor is inserted between C2 and L2 with a low loss inductor to a decoupling capacitor to which the bias is connected. It was found that the noise temperature Te goes down by 1.7 K when the current is increased from 17 to 23 mA.
Losses in L2.The original coil L2 was made from 0.35 mm enameld copper wire. The diameter of the copper was 0.31 mm and the length 40 mm. After replacing the L2 wire with the inner conductor from 2.2 mm semirigid cable, re-tuning gave an improvement by 0.45 K. The diameter of the new silvered wire is 0.50 mm so the RF resistance should have changed by a factor 0.62*0.96 = 0.596 where 0.96 is the square root of the resistivity ratio. Eliminating the remaining loss of L2 should improve Te by 0.65 K but doubling the diameter to 1 mm silvered wire should improve by only 0.3 K ( 0.004 dB on the NF ).
Optimised NF.Before modifications this amplifier had a Te that was 3.03 K higher than the best amplifier at the Orebro EME meeting 2013.
After the modifications it is only 0.9 K worse than the RW3BP-13 amplifier. Presumably the individual MGF4919G transistors differ slightly. The true NF should be around 0.16 dB after modifications.
The S-parameters of the modified unit are shown in figure 5.
|Figure 5. The S-parameters of the experimental LNA after modifications. Here the network analyzer is properly calibrated. The higher current has increased the gain by nearly 1 dB.|