Measurements of insertion loss with a signal generator and a receiver.
(March 9 2013)

The DUTs.

This page describes insertion loss on DUTs presented here. Using the HP8712C network analyzer to measure impedances and losses on 1296 MHz. The impedance data from that page is used to split the measured insertion losses into mismatch loss and dissipative losses below.

The Tx port.

The Tx port from which the signal generator sends its signal is the attenuators plus cables combination used as the Rx port in the HP8712C network analyzer measurements. That means that the impedance of the Tx port with or without DUT can be taken from the impedance measurements.

The Rx port.

Figure 1 shows the arrangement for insertion loss measurements. The important parts are the circulators. They guarantee that the LNA, an AD6IW amplifier always looks into the same impedance so it will always provide the same gain.


Figure 1. Insertion loss measurement. Here DUT8 is inserted.


The 40 dB power attenuator near the upper left corner of figure 1 is connected to a HP8657B which is set to 1296.11 MHz. The cable routed to the blue 3 dB attenuator, the attenuator and the adapter were used as the rx port for network analyzer measurements so the impedance seen by the DUT is precisely known in relation to the calibration kit load. The other side of the DUT, DUT8 in figure 1, is connected to an adapter followed by a 90 degree bend with N connectors. The bend is inserted because it happens to bring the impedance near 50 ohms. The two circulators guarantee that the impedance does not change, a 3 dB attenuator is inserted in front of the AD6IW LNA to ensure stability. A second LNA with a PSA04-5043 gives some more gain. The amplified signal is routed to the mixer port of a SBL-1 mixer which is fed by 384.07 MHz, +15 dBm from a HP8657A on the LO port for third overtone mixing. There is a 6 dB attenuator on the mixer to guarantee the impedance. The difference frequency 1296.11-3*384.07=143.9 MHz is obtained on the RF port and routed to a GaAs FET LNA followed by a high level amplifier and a filter for 144 MHz. From that a high level mixer converts the signal to 11 MHz where it is received by a SDR-IP.

The loss measurement is performed by repeatedly insert and remove a DUT. It is essential that none of the RG223 cables that carry 1296 MHz is moved during the process. Moving these cables easily causes several tenths of a dB in the signal level. That is due to the impedance changes on bending the cables. The The circulators must thus not be moved during a measurement sequence so all the necessary movement will be on the black cable that is gently bentto allow the Tx port to be moved axially without any impedance change. The black cable is 1 meter of Flexiform 401 with an extra screen on it.

The input impedance on the Rx port, the SMA female on the circulator is reasonably near the impedance shown in figure 2. The uncertainty comes from measurement through a SMA male to SMA male adapter with unknown characteristics. I do not have a female calibration kit that matches the male one. The electrical length of the adapter is compensated for, but only approximatively.


Figure 2. The impedance of the Rx port. There is an unknown error due to the adapter used for the measurement.


Results with a simple setup.

Before arranging the setup of figure 1 experiments were performed with a RTLSDR USB dongle. Figure 3 shows the signal stability after a long enough warm-up period. It looks like AM noise modulation on the LO causes random fluctuations on the signal level just as phase noise causes random fluctuations on the frequency.

The USB dongle does not allow very accurate level measurements.


Figure 3. The stability of a RTLSDR dongle with the E4000 tuner.


A measurement of the insertion loss of DUT1 with the RTLSDR dongle gives IL=0.030 dB with error limit of +/- 0.0042 dB.

Measurements with the setup of figure 1.

The signal stability is better with this more conventional setup. See figure 3. The stability with the same equipment on 144 MHz is far better which indicates that the much worse sideband noise of the HP generators is the limiting factor on 1296 MHz.


Figure 4. The stability of the setup shown in figure 1.


Table 1 gives the impedances measured previousle and the insertion losses measured with the setup in figure 1 with links to the raw data.


Device    Zre      Zim      Loss     Stddev     Equ
         (Ohm)    (Ohm)    (dB)     (dB)
none     52.02    0.21     0.0000     0        1
DUT1     50.34    0.69     0.0320   0.0008     2
DUT3     49.09   -2.56     0.0195   0.0008     3
DUT4     50.45    0.16     0.0503   0.0012     4
DUT7     51.84   -2.46     0.0600   0.0035     5
DUT13    54.04    2.24     0.0614   0.0004     6
DUT43    52.97    3.10     0.0836   0.0040     7
DUT73    55.71    0.68     0.0853   0.0008     8
DUT37    46.71   -2.12     0.0806   0.0012     9
DUT34    49.16   -0.95     0.0730   0.0011    10
DUT31    49.67   -1.00     0.0513   0.0017    11
DUT8     71.81    0.51     0.1652   0.0010    12
DUT83    75.04    5.42     0.2364   0.0019    13
DUT81    74.35    0.20     0.2284   0.0009    14
DUT813   69.27   -2.47     0.1969   0.0015    15
DUT318   55.75   14.82     0.1972   0.0008    16
DUT381   44.59  -18.52     0.2484   0.0014    17
DUT831   74.80    2.61     0.2560   0.0018    18
DUT183   34.89    3.87     0.2513   0.0006    19
DUT138   59.53   19.49     0.2665   0.0011    20
DUT6     53.04    1.89     0.1102   0.0012     -
DUT18    36.77    4.60     0.1886   0.0015    21
DUT38    45.70  -17.41     0.1847   0.0021    22
DUT48    36.75    0.00     0.2002   0.0014    23
DUT84    73.83    1.24     0.2445   0.0035    24
DUT14    51.73   -0.41     0.0844   0.0016    25
DUT41    51.89   -0.11     0.0840   0.0011    26
DUT5     51.43   -0.19     0.0673   0.0013     -

Table 1. Impedances from the HP8712C network analyzer and insertion losses from the setup in figure 1.

Evaluation of unknowns.

The impedances in table 1 all contain the unknown calibration error from the calibration kit.

The insertion loss should be the sum of the mismatch loss and the sum of the dissipative losses for the DUT. Disregarding DUT5 and DUT6 which are not measured together with other DUTs we have this set of unknowns:

x(1)=The dissipative loss of DUT1
x(2)=The dissipative loss of DUT3
x(3)=The dissipative loss of DUT4
x(4)=The dissipative loss of DUT7
x(5)=The dissipative loss of DUT8
x(6)=The calibration error real part
x(7)=The calibration error imaginary part
x(8)=The Tx port impedance error real part
x(9)=The Tx port impedance error imaginary part
There are 26 equations that connect these 9 unknown variables. For each line in table 1, the loss should be the sum of the dissipative losses plus the insertion loss. The insertion loss can be computed from this formula:

IL = 4 * Rt * Rr / [ (Rt + Rr)2 + (Xt + Xr)2 ]

Rt is the real part of the Tx port impedance.
Xt is the imaginary part of the Tx port impedance.
Rr is the real part of the Rx port impedance.
Xr is the imaginary part of the Rx port impedance.

From the equation it is obvious that we can not determine Xt and Xr separately, only their sum. Also Rt and Rr are strongly coupled. This means that we have 26 equations and 7 independent variables. The RMS deviation produced by a least squares fit will give a good insight in the measurement errors.

A simple Fortran program with a Makefile for Linux can be downloaded here: losscalc100.tbz (5536 bytes) Running it with table 1 as input produces the listing displayed in table 2.

 DUT100  ( 50.340   0.690)   il= 0.0320   Dissipat.= 0.0323  err= 0.0001
 DUT300  ( 49.090  -2.560)   il= 0.0195   Dissipat.= 0.0209  err= 0.0014
 DUT400  ( 50.450   0.160)   il= 0.0503   Dissipat.= 0.0534  err= 0.0022
 DUT700  ( 51.840  -2.460)   il= 0.0600   Dissipat.= 0.0564  err=-0.0039
 DUT130  ( 54.040   2.240)   il= 0.0614   Dissipat.= 0.0532  err=-0.0003
 DUT430  ( 52.970   3.100)   il= 0.0836   Dissipat.= 0.0743  err=-0.0011
 DUT730  ( 55.710   0.680)   il= 0.0853   Dissipat.= 0.0773  err= 0.0020
 DUT370  ( 46.710  -2.120)   il= 0.0806   Dissipat.= 0.0773  err= 0.0020
 DUT340  ( 49.160  -0.950)   il= 0.0730   Dissipat.= 0.0743  err= 0.0004
 DUT310  ( 49.670  -1.000)   il= 0.0513   Dissipat.= 0.0532  err= 0.0005
 DUT800  ( 71.810   0.510)   il= 0.1652   Dissipat.= 0.0358  err= 0.0032
 DUT830  ( 75.040   5.420)   il= 0.2364   Dissipat.= 0.0567  err=-0.0008
 DUT810  ( 74.350   0.200)   il= 0.2284   Dissipat.= 0.0681  err=-0.0005
 DUT813  ( 69.270  -2.470)   il= 0.1969   Dissipat.= 0.0890  err=-0.0015
 DUT318  ( 55.750  14.820)   il= 0.1972   Dissipat.= 0.0890  err=-0.0023
 DUT381  ( 44.590 -18.520)   il= 0.2484   Dissipat.= 0.0890  err=-0.0027
 DUT831  ( 74.800   2.610)   il= 0.2560   Dissipat.= 0.0890  err= 0.0013
 DUT183  ( 34.890   3.870)   il= 0.2513   Dissipat.= 0.0890  err=-0.0017
 DUT138  ( 59.530  19.490)   il= 0.2665   Dissipat.= 0.0890  err= 0.0002
 DUT600  ( 53.040   1.890)   il= 0.1102   Dissipat.= 0.1055  err=-0.0000
 DUT180  ( 36.770   4.600)   il= 0.1886   Dissipat.= 0.0681  err= 0.0053
 DUT380  ( 45.700 -17.410)   il= 0.1847   Dissipat.= 0.0567  err= 0.0027
 DUT480  ( 36.750   0.000)   il= 0.2002   Dissipat.= 0.0892  err=-0.0030
 DUT840  ( 73.830   1.240)   il= 0.2445   Dissipat.= 0.0892  err=-0.0001
 DUT140  ( 51.730  -0.410)   il= 0.0844   Dissipat.= 0.0857  err= 0.0005
 DUT410  ( 51.890  -0.110)   il= 0.0840   Dissipat.= 0.0857  err= 0.0013
 DUT500  ( 51.430  -0.190)   il= 0.0673   Dissipat.= 0.0683  err= 0.0000
 RMS error=  0.00199  alfa= 0.000000  step= 0.000008
 1  DUT1  Dissipative loss=0.0323
 2  DUT3  Dissipative loss=0.0209
 3  DUT4  Dissipative loss=0.0534
 4  DUT7  Dissipative loss=0.0564
 5  DUT8  Dissipative loss=0.0358
 6  DUT6  Dissipative loss=0.1055
 7  DUT5  Dissipative loss=0.0683
 8  Source impedance real part= 52.090
 9  Load impedance real part= 50.576
10  Sum of complex parts=  1.177

Table 2. Results from the evaluation of dissipative losses from insertion losses measured with the setup of figure 1.



The RMS error in the evaluation of the equations is 0.002 dB. That is 10 times better than the result obtained with the insertion loss measured with the HP8712C network analyzer. The size of the RMS error is as expected, the RMS value of the 26 estimated standard deviations in table 1 is 0.0017. With proper crystal oscillators the accuracy would probably have been another order of magnitude better.

What error limits to set for the individual DUTs is beyond my skils. In total 8 unknowns are determined from 26 equations.....