(June 27 2013)

About this page.

The purpose of this page is to demonstrate principles that can be used for optimizing the system performance of microwave statations such as amateur radio EME stations on 1296 MHz.

Sergey Zhutyaev, RW3BP has demonstrated one elegant way of optimizing the system from the antenna to the low noise amplifier by use of a horn antenna pointed into the sky. Follow these links:

Interesting graphs but text in Russian: 1296 MHz Small EME Station with Good Capability (Part 2)
Accurate Noise Figure measurements on 1296 MHz. Accurate Noise Figure measurements on 1296 MHz.
LNA Optimization. 1296 MHz Small EME Station with Good Capability (Part 4)
Antenna Switching. 1296 MHz Small EME Station with Good Capability (Part 5)

This page and sub-pages show in detail experiments done in order to find simpler ways to achieve the same goal.

What is losses?

When we read the word loss or attenuation in a text about radio technology it is nerly always implicitly understood that the word means insertion loss, the influence some building block has on the signal level when the block is inserted in a 50 ohm system.

Traditionally one measures the insertion loss with a network analyzer. It is directly related to the scatter parameter S12. There are in total four such parameters S11, S12, S21 and S22 and together they specify the two port device under test (DUT). A relay like CX 520 D which is specified to have an insertion loss of 0.2 dB on 1 GHz might be a better choice than a HP 8761B which is typically below 0.1 dB according to the Agilent specification. The CX 520 might cause some mis-match while the 8761 is extremely well matched but might have dissipative losses on 1 GHz because at such a low frequency the gold plating is too thin.

By measuring all the S-parameters one can evaluate to what extent insertion loss is caused by mismatch and to what extent it is caused by heat generation within the DUT. A small mis-match is harmless but a small disipative loss will degrade the noise floor because the DUT is very hot in relation to the cold sky.

It is far from trivial to perform S parameter measurements with good enough accuracy when the insertion loss is in the order of 0.1 dB but we can find the dissipative losses by use of S/N measurements or relative NF measurements.


On this page and associated sub-pages DUT (Device Under Test) refers to a low loss two port device with a male and a female connector of the same kind. The DUT could be a short cable with a male and a female connector or a combination of two adapters like a BNC to N adaptor plus a N to BNC adaptor. The DUT could also be a relay plus a female to female adaptor.

The DUT can be inserted in a system between a female and a male connector of the same kind without additional adaptors and this way the effect of a DUT does not contain any uncertainty caused by additional adaptors.

Demonstration of principles.

As a first step many measurements were performed on 144 MHz to verify methods and to gain experience from studying well known things in new ways. The insertion loss of a DUT is the sum of the mismatch loss and the dissipative loss. We can measure insertion loss and mismatch loss accurately with a network analyzer although it is not easy.

The network analyzer gives the impedance change with high accuracy when a DUT is inserted. That is because the contact gender does not change. That gives the data required to compute the mismatch loss with high accuracy provided that the impedance of the contact with the opposite gender is known accurately. That can not be measured directly, but by use of several DUTs that transform the impedance in different directions it can be evaluated. The insertion loss is better measured with a good radio than with a network analyzer due to the higher dynamic range, but the netvork analyzer is needed to measure the impedance transformation caused by the different DUTs.

The dissipative losses can be measured as the change of the S11 parameter for open and short from the calibration kit when a DUT is inserted. This should be accurate for high efficiency DUTs according to this paper Comparison of Adapter Characterization Methods but the method is useless with the instruments at my disposal for a DUT that is not perfectly matched to 50 ohms. My guess is that the method is only accurate when both mismatch and dissipative losses are small.

Dissipative losses do affect the NF as expected and measurements by use of NF changes seems to be the best method to use for radio amateurs. An interestingg observation is that the shallowness of the NF vs impedance surface gives the absolute NF within about a factor of two. That gives a far better absolute NF than the instrument uncertainty (noise head calibration) for ultra low noise amplifiers. By use of circulators we can measure relative noise figures with a very high accuracy. For all the details look here: Demonstration of principles on 144 MHz

Measurements on 1296 MHz.

Several low loss DUTs with SMA connectors have been studied with a HP 8712C network analyzer. See figure 1. The purpose of measuring short sections of cable, one of which giving a significant mis-match is to determine the impedance of the Tx port of the 8712C. When inserting a DUT in the signal path between the Tx and Rx port of the 8712C one needs to know the impedance at both sides to compute the mismatch losses both with and without the DUT. Measurements on 1296 are more difficult than measurements on 144 MHz because the impedance variations caused by moving cables is about one order of magnitude larger. The HP8712C gives a mediocre accuracy on the insertion loss of a DUT. For that reason the insertion loss is measured with a signal generator and a receiver. Attenuators are used to make sure that impedances are accurately known relative to each other. The calibration error due to errors in the (home made) calibration kit is of course not known. For details look here: Using the HP8712C network analyzer to measure impedances and losses on 1296 MHz. and here Measuring insertion loss with a signal generator and a receiver on 1296 MHz.

Figure 1. DUTs with SMA connectors.
1) About 30 mm of RG 400 cable. Electrical length 0.284 wavelengths.
2) Modified connectors. Electrical length 0.243 wavelengths.
3) Modified connectors. Electrical length 0.129 wavelengths.
4) A pair of adapters. Electrical length 0.252 wavelengths.
5) Relay HP8761B with adapters. Electrical length 0.679 wavelengths.
6) Relay CX-520-D with adapters. Electrical length 0.0.275 wavelengths.
7) Male to male adapter and two female chassis connectors. Electrical length 0.275 wavelengths.
8) Modified connectors. Part with high impedance. Electrical length 0.338 wavelengths.

The sum of the losses of DUT1, DUT3 and DUT 8 was determined to 0.142 dB from insertion losses measured with the HP8712C and to 0.089 dB with 10 times better accuracy from measurements with signal generators. To try to improve accuracy I have made measurements with a HP8753E on DUT1, DUT3 and DUT8. This is a more modern network analyzer borrowed from Mart, SM0ERR. Unfortunately I was unable to process the data correctly, but the data should be good and allow a reasonably accurate set of all the two port parameters for the three devices with a good picture of the statistical errors.

Since DUT1, DUT3 and DUT8 are intended to be used to map the NF vs impedance surface it is important to know the summed dissipative losses in these thre DUTs. Having failed (due to lack of theoretical skil) to measure directly on the the HP8753E I made new studies with a conceptually simpler approach. Using the HP8753E network analyzer to set and measure impedances on 1296 MHz. and here Measuring insertion loss with a signal generator and an improved receiver on 1296 MHz. The result is that the impedance-transforming networks formed from different combinations of DUT1, DUT3 and DUT8 have dissipative losses of 0.0807 dB.

An amplifier with optimum NF for a source impedance near 50 ohm is required to allow accurate measurement of small losses. Such measurements are important in the design of low noise front ends on 1296 MHz.

Manual use of a HP8970A with circulators for evaluation of the optimum input impedance of low noise amplifiers shows that we can compare amplifiers to within about 0.2 K (0.03 dB in NF) by use of standard equipment with the addition of circulators.

Using a HP8970A with circulators under computer control for highly accurate NF comparisons. When the HP8970A is under computer control one gets one more decimal on the noise temperature and one can use the computer to average the averages while updating the temperatures for each measurement. The reproducibility is very good. It is possible to evaluate how the ENR of the noise head varies with the temperatur and introduce a correction for that. This correction includes the variation of the losses through the circulators attenuators and cables. The noise head seems to be the part that is most sensitive to temperature changes.

Comparing L LNAs uses the HP8970A with circulators under computer control to compare several L LNA amplifiers from AD6IW. The amplifiers I have tested are all very similar, the largest uncertainty seems to be due to the poor accuracy of the DC supply voltage (a simple analog meter). The NF depends on the temperature which in turn depends on the supply voltage. The outcome of this test is that the zero point is about 17 K on the NF scale. This is corrected by changing the loss before the DUT from 4.6 to 4.8 dB in the VEE program.

NF measurements at the Orebro EME meeting May 2013. gives NF results with the 1296 MHz NF measurement system described on the subpages of this page. The purpose of the measurements was to try to establish an accurate relative NF scale for 1296 MHz and to try to find a better value for the zero point.

The L LNA by AD6IW has its lowest noise figure for a feed impedance near 50 ohms. Just measuring how S/N is degraded will give a good estimate on the dissipative losses of whatever DUT that is inserted immediately in front of the L LNA as long as the DUT does not make a large impedance change. Rather than measuring S/N, SINAD of FM quieting one can use a standard NF meter provided that the noise source is equipped with circulators as described on the sub-pages of this page.

For precision measurements it is necessary to know how the noise figure changes with the impedance. That was evaluated by use of DUT1, DUT3 and DUT8 here: evaluation of the optimum input impedance

By use of three more DUTs, see figure 2, the NF vs source impedance surface has been investigated again by use of the computer controlled HP8970A. evaluation of dissipative losses from NF measurements.

Figure 2. More DUTs with SMA connectors.
A) A 50 ohm transmission line with an electrical length of 0.25 wavelengths on 1296 MHz.
B) A transmission line with a high impedance section. Provides transformation to 70 ohms.
C) A 50 ohm transmission line with an electrical length of 0.125 wavelengths on 1296 MHz.
D) A pair of adapters. The male to male adapter is identical to the Male to male adapter in DUT7.

The result is shown in table 1.

Device    Loss(NF)      Loss(S11)       Diff
                 (dB)          (dB)           (dB)
DUT1      0.0338         0.0355           -0.0017 
DUT3      0.0221         0.0245           -0.0024
DUT4      0.0547           -
DUT5      0.0606           -  
DUT6      0.1107           -
DUT7      0.0632           - 
DUT8      0.0324         0.0340           -0.0016 
DUTA      0.0359         0.0348            0.0011
DUTB      0.0308         0.0333           -0.0025
DUTC      0.0319         0.0330           -0.0011
DUTD      0.0720           -

Table 3. Output from nfdut.

Table 3 is the answer to the question about losses in amateur stations for EME and radio astronomy. Losses can be measured fairly easily if one is satisfied with an accuracy of 0.01 dB.

SMA connectors typically add less than 0.03 dB but they can add more as is shown by DUTD which has higher losses than DUT7. The very small SMAf-SMAf has higher losses than the combination of two chassis-mount female connectors.

The CX-520-D relay has much higher losses than the HP8761B relay. To me that was unexpected - maybe the CX relay is not silver plated inside?