Highly resistant low-noise preamplifier for 144MHz S53WW

Robert S53WW is a great designer of VHF equipment. On its website you will also find, for example, a description of the 2m antenna system suitable for contests and a high-quality Javornik 144 / 14MHz transverter. I asked him for permission to translate and publish these articles. Robi sent it to me immediately, for which he deserves a big thank you. We will gradually bring them all to you at www.cq.sk . But if you don't want to wait, visit his site http://lea.hamradio.si/~s53ww/.

Introduction

After a thorough analysis of the various preamplifiers for my new 144/14 The MHz transverter was finally left with the Infeon BF998 MOS-FET. I had a very good experience with the old BF-981 and so I started testing the BF998. This transistor is cheap, easily accessible and has very good characteristics. I didn't use GaAs FET, because they are difficult to obtain and expensive, while the differences are minimal.

It was clear from the analysis of the transverter, that it determines the resulting resistance of the whole assembly preamplifier (if you use a + 17dBm mixer with an oscillator signal of + 20dBm followed by an amplifier with IP3 out > 36dBm and you neglect the resistance of HF TCVR)! OK, commercial HF receivers should not have bad resistance (IP3 in < +15dBm to + 25dBm, AOR-7030 to + 30dBm). In this case, IP3 out should be at least at this level (+30dBm). The noise of HF receivers is at the level of 16dB, at best around 13dB.

If we estimate the loss in the ANT-XVRT coaxial cable at about 0.5dB and we require the resulting noise figure of about 2.2dB, we will find out, that the XVRT gain must be at least 24dB (with noise figure 1dB). It follows that, that IP3 in must be more than + 6dBm (=30dBm-24dB). Meeting this requirement is not easy because of the preamplifier.

Low-noise preamplifier with parallel combination n times BF998

Attempting to install the original preamplifier with the new BF998 did not bring significant improvement. He had a profit 26,5 dB, noise figure 0,8 dB, P1dB 17 dBm a IP3 in 0 dBm. These are not bad values, but they could have been better.

The first picture shows the common connection of a preamplifier with a two-base MOSFET with respect to the maximum gain. Idss is set to the minimum noise level (Idss is given in the range 2 up to 18mA – usually between 10 up to 15mA). The input coil L1 is wound on a diameter of 5mm 1.0mm AgCu wire; has 6 threads with winding length 11mm. It should be placed at least 2mm above the ground and at least 10 mm from the metal parts of the box. Capacitive trimmers must be of good quality. Diodes (Infineon BAR63-04 or BAR64-04 or BAR14-1) serve to protect the MOSFET. They have almost no effect on the noise figure or immunity, their ingestion in real conditions is desirable. The most critical part is the TR1 transformer: it should be wound on a two-hole core of size A7 from material U17 (Epcos, formerly Siemens&Matshushita). Materials from other manufacturers may be used with respect to degraded gain and P1dB. We wind two threads trifilarly – it is always better than autotransformer with 6 threads (=2×3) primary winding and a branch on the second thread from the cold end (in practice, they are wound first 4 threads, by twisting the wires, a branch is created and wound up 2 threads). The TR1 terminals must be shortened to 5 mm and run close to the ground on the printed circuit board, otherwise, oscillations at high frequencies may occur.

While experimenting, I found an interesting involvement of FET (used in NF amplifiers), where two identical FETs connected in parallel have a lower noise than one! So if you choose another BF998 (must have the same Idss, must be selected), you can get up to 0,6dB NF, profit 26,5dB and the same or worse (!?) IP3 in. The problem with IP3 is solved by more appropriate output adjustment – instead of a transformer with a ratio of 3:1 we will use 2:1. In this way we achieve + 28.5dBm IP3 performance (P1dB is + 19dBm). The general connection of the LNA n times to BF998 is shown in the next figure (the achieved values ​​are given in the table).

Of course I also tried the four BF998 (who wouldn't) – the noise figure is reduced to 0,5dB, profit remains 26,5dB (when set to the minimum noise figure) has IP3 out i +34dBm. To reach this IP3 value, it is necessary to change the ratio TR1 to 1,3:1. The L1 branch is in the middle of the winding (on the third thread). In the first attempt, the amplifier oscillated above 3.5GHz, which did not affect the 2m. I tried to place a pair of transistors on each side of the printed circuit board connected by short wires and separate G2 bias circuits. (the placement of components and the blocking capacitor is critical). After this adjustment, the oscillation disappeared, but I would like to note, that all connections are stable only under certain conditions. The layout of the components is critical, in particular it concerns the blocking capacitor 1nF.

Wiring diagram of preamplifier n times BF998 (similar to one BF998). The values ​​of the variable components can be found in the table. The adjustment of input and output is also different (see text). The material of the transformer is the same as in the previous case, when swallowing more BF998 I recommend A4 size. Preamplifiers can be tuned without measuring technology with a small trick to the minimum noise figure: we tune them to maximum profit on 136 MHz (the noise figure hardly changes over a wide range).

Here you can download the PCB drawing ftp://lea.hamradio.si/pub/vhfct/lnaa2.zip.

	n1	n2	R1	NF [dB]	T [K]	G [dB]	IP3in [dBm]	P1db [dBm]
1 x BF998	4	2	150	0,8	61	26,5	0	17
2 x BF998	2	2	100	0,6	44	26,5	2	19
4 x BF998	1	3	10	0,5	36	26,5	7	23

Push-pull LNA s BF998

Push-pulse amplifier is equal in gain to a two-stage single amplifier and IP3 is about 3dB better than in the case of a parallel design. So I tried to connect a pair of BF998 to a push-pull. The first results were excellent: IP3 inbolo +6 dBm (P1dB +23 dBm), but the noise figure was at least 1.3dB! After changing the input circuit, I reached the noise figure 0,9-1,0dB, IP3 out 32,5dBm at profit 26,5dB (setting to minimum noise figure). The resulting circuit is a combination of two single-stage amplifiers according to the first figure.

Push-pulse connection of two BF998. The input coil L1 is wound on a diameter of 11mm wire with 1.0mm AgCu; has 4 threads and winding length is 7mm. The binding coil has 1 thread on the same diameter as L1 enameled Cu wire so, that it cannot short-circuit L1. Transformers TR1 and TR2 are on the same material as a single LNA (size A7, material U17, Epcos). The output transformer TR2 has two bifilar turns.

Then I tried to push-pull the connection of the four BF998. The intention was to get IP3 in around + 9dBm at a noise figure of about 0.8dB. Although I achieved IP3 in as I imagined, I did not reach the desired noise figure. It was still higher than 1.3dB, which seemed inappropriate to XVRT. However, this value may be suitable for other purposes, it still requires experimentation.

Warning:
These preamplifiers are not suitable for direct connection in front of a receiver or transverter because they have a very high gain. Their effective use (although I do not recommend using preamplifiers) it is possible with a damping element at their output, which will reduce the profit. In most cases, a 6-10dB gain is sufficient for a preamplifier, if it really has to be used. With the -20dB attenuator, the total noise figure deteriorates from 0.5 / 0.6 / 0.8dB to 1.3 / 1.4 / 1.5dB. With the -16dB attenuator, the total noise figure deteriorates from 0.5 / 0.6 / 0.8dB to 0.8 / 0.9 / 1.1dB. So I recommend you to use a -16dB attenuator behind the preamplifier (the pi-cell values ​​are 68 ohms to ground and 150 ohms between them).

High quality VHF and higher preamplifiers can be found on the Dragana YU1AW website: www.qsl.net/yu1aw/low_noise.htm.

Do septembra 2002 I built six preamplifiers with a noise figure of 0.5-0.6dB (BF998 Philips, marked Mop) and three with a noise figure of 0.8-0.9dB (BF998 Infeon, marked Mos). The bad news is, that Epcos has stopped distributing U17 material and it is therefore very difficult to find U17 / A4 two-hole cores. Their replacement is not yet known, but cores of similar size of material with a permeability between 20-30 are applicable…

After measuring nine preamplifiers, I recommend a simple way to set the 4xBF998 LNA to minimal noise: tune trimmer C to maximum profit, then increase its capacity, while profit does not decrease by 1dB! Tuning is not critical, the input circuit is wideband when setting the minimum NF (about 10 MHz).

Does, S53WW
http://lea.hamradio.si/~s53ww/
Translated and edited with the permission of the author by William, OM3-0122