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 Rx port.Figure 1 shows the arrangement for insertion loss measurements. The important parts are the circulators. They guarantee that the LNA, an MMIC with 2 dB NF and 40 dB gain always looks into the same impedance (through a filter). That guarantees that the MMIC will always provide the same gain.In front of the circulators there is a section of semirigid cable which has been squeezed to make the impedance very close to 50 ohms on the female SMA connector which constitutes one side of the test position. The MMIC is followed by a high level mixer, a ZFY-11 from Minicircuits which is fed with an LO signal from a HP 8657B which is routed through a MMIC amplifier. The output of the mixer is sent to a SDR-IP through a low pass filter. | |
Figure 1. Insertion loss measurement. Here DUT18 is inserted. | |
The Tx port.The 40 dB power attenuator near the upper left corner of figure 1 is connected to a HP8644A which is set to 1296.06 MHz. The short section of 0.5 inch Heliax with screws allows the impedance at the male SMA connector to be tuned to 50 ohms precisely regardless of what 50 ohm unit is connected at the other side of the 40 dB attenuator. (Anything with better than 20 dB return loss.)Measurement.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 changes in the signal level. That is due to the impedance changes on bending the cables. 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 bent to 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.Stability issues.Figure 2 shows the stability at a bandwidth of 100 Hz. The 8644A has less AM noise than the 8657B as can be seen from a comparison between figures 2 and 3. | |
Figure 2. The stability of the setup shown in figure 1 with the 8657B as LO and 8644A as signal. | |
Figure 3. The stability obtained when 8644A is used for the LO and 8657b for the signal. | |
The LO signal level is set to a level where small variations does not affect the mixer gain. This way the AM noise of the 8657B becomes insignificant. The differences are not large however. As it turns out, the dominating instability is due to variations in the phase noise. By use of a larger bandwidth, 500 Hz instead of 100 Hz, the power of the noise sidebands is largely included in the signal level. At these very high signal levels the receiver noise floor does not contribute despite the wider bandwidth. The stability at 500 Hz bandwidth is shown in figure 4. | |
Figure 4. With 500 Hz bandwidth and 8644A for the signal source with the 8657B as the LO the stability is quite good. Keeping constant room temperature is very important. One essential part of the set-up is a big fan that keeps steady air-flow on the amplifiers which otherwise get unstable temperatures and variable gain. | |
This looked good at first, but after starting to do measurements I found that the signal level would suddenly change by perhaps 0.03 dB and sometimes become unstable changing rapidly by about that amount. It was non-trivial to find the reason because there were many reasons. Level changes happened spontaneously but could also be provoked by hitting the table to introduce small vibrations. The major problems were the following:
| |
Figure 5. Improved insertion loss measurement. | |
Measurements.Table 1 gives the impedances of the female port as measured previously and the insertion losses measured with the setup in figure 5 with links to the raw data.Table 2 gives the impedance of the male port as measured previously together with the insertion losses. | |
Device Z_{re} Z_{im} Loss Stddev (Ohm) (Ohm) (dB) (dB) none 49.99 0.02 0.0000 0.0 DUT1 50.54 0.73 0.0348 0.00055 DUT3 48.39 -1.99 0.0242 0.00027 DUT8 68.56 10.56 0.1702 0.00106 DUT13 53.05 2.12 0.0589 0.00083 DUT18 35.16 0.02 0.1921 0.00041 DUT31 48.98 -2.56 0.0624 0.00054 DUT38 51.30 -19.81 0.2137 0.00053 DUT81 67.99 9.25 0.1941 0.0006 DUT83 69.55 13.90 0.2260 0.00037 DUT138 49.44 20.43 0.2608 0.00025 DUT183 33.83 -1.07 0.2456 0.00044 DUT318 47.18 14.67 0.1834 0.00086 DUT381 50.66 -19.16 0.2390 0.00061 DUT813 65.68 7.00 0.1790 0.00060 DUT831 68.50 14.16 0.2587 0.00032 | |
Table 1. Impedances on the female SMA connector with the different DUTs measured by the 8753E and insertion losses from the setup in figure 1. | |
Device Z_{re} Z_{im} Loss Stddev (Ohm) (Ohm) (dB) (dB) none 49.98 -0.07 0.0000 0.0 DUT1 50.06 -1.08 0.0348 0.00055 DUT3 48.45 -2.20 0.0242 0.00027 DUT8 71.81 1.26 0.1702 0.00106 DUT13 47.42 -2.30 0.0589 0.00083 DUT18 71.30 2.13 0.1921 0.00041 DUT31 52.41 0.33 0.0624 0.00054 DUT38 73.15 4.81 0.2137 0.00053 DUT81 35.42 3.87 0.1941 0.0006 DUT83 44.49 -18.64 0.2260 0.00037 DUT138 73.82 5.71 0.2608 0.00025 DUT183 45.26 -18.94 0.2456 0.00044 DUT318 68.05 -0.47 0.1834 0.00086 DUT381 34.18 2.49 0.2390 0.00061 DUT813 52.42 15.00 0.1790 0.00060 DUT831 57.36 19.41 0.2587 0.00032 | |
Table 2. Impedances on the male SMA connector with the different DUTs measured by the 8753E and insertion losses from the setup in figure 1. | |
Simplistic evaluation of data.As a first attempt the data of tables 1 and 2 was fed into the computer program used for evaluation of the previous less accurate study of insertion losses.The result was discouraging. Uncertainties were larger than in the previous study. The reason is that tables 1 and 2 on this page contains data from only 3 DUTs and that most of the DUTs have a high SWR. The dissipative losses of a DUT can be computed from |S21|+|S11| which approaches unity for a low loss DUT. The value has to be slightly less than one and the logarithm times 20 gives the dissipative losses in dB. The problem is that S11 suffers from a reproducibility problem. When impedances are measured once again the values typically differ by a few tenths of an ohm. In cases where SWR is high (all involving DUT8) a few tenths of an ohm is an unacceptable error. The previous study provides better accuracy despite the 10 times less accurate insertion losses because more DUTs with low SWR were measured. The data of table 1 and 2 has to be processed differently. The 85033 calibration kit.The manufacturer specification is that the return loss of the LOAD is 40 dB or better below 2 GHz in the temperature range 15 to 35 degrees Centigrade. That is equivalent to say that the impedance is within a circle with radius 1 ohm in the Smith chart.The impedance depends on the temperature, the male LOAD changes by (-0.039 + j0.033) Ohms for a 10 degree temperature change while the female LOAD changes by (-0.046 + j0.006). It seems reasonable to assume that the calibration kit is within 0.95 ohms from the Smith center at 25 degrees. When the network analyzer is calibrated on the female LOAD and then the female to female adapter of the calibration kit is used to measure the male LOAD, the result is (49.939 - j0.141). Theory.The first case, "none" means that the female connector with impedance (49.99, j0.02) is connected to the male connector with impedance (50.05, -j0.02). The associated voltage reflection coefficient isGAMMA = (Z_{M}-Z_{F}) / (Z_{M}+Z_{F}) Applying the worst combination of points 0.95 ohm away from the measured values gives a reflected power of -34.1 dB at the connection. The reflected power is a loss of 0.0017 dB in transmitted power. The measured insertion loss of DUT1, 0.0348 dB could therefore be anything between 0.0348 and 0.0365 due to the influence of the calibration kit uncertainty on the measuement with no DUT inserted. When sending power from the male connector through DUT1 into the female connector, the impedance seen by the male connector is (50.54 + j 0.73). The true impedance carries the error of the calibration kit male load so when computing the dissipative losses, the influence of the calibration kit error does not grow much. One finds that the dissipative losses of DUT1 is in the range 0.0340 to 0.0361 dB due to the uncertainty of the calibration kit. That is the average of the measurement in both directions. The average is less affected than the individual directions. To this comes the error of the insertion loss measurement with a standard deviation of 0.00055 plus a small error due to reproducibility problems with standard SMA connectors. The calibration kit thus gives an error of +/-0.001dB on DUT1. For DUT8 it is different. Repeating the same math gives a range of 0.0285 to 0.0349 dB for the dissipative loss due to the uncertainties of the calibration kit only. That is +/- 0.006 dB, six times larger uncertainty than for DUT1. This result is approximate, but the underlying physics is simple. The DUT is a two-port. We determine the dissipative losses from the sum of transmitted and reflected power. (If the sum is below one, the rest is dissipative loss.) When the reflected power is very small it is not so critical to know impedances very well but when a significant fraction of the power is reflected, even relatively small errors are important. Evaluation.We set up a set of unknowns:ZERR_FEMALE 2 parameters (The error on the female SMA impedance) ZERR_MALE 2 parameters (The error on the male SMA impedance) S11_DUT1 2 parameters S21_DUT1 2 parameters S22_DUT1 2 parameters S11_DUT3 2 parameters S21_DUT3 2 parameters S22_DUT3 2 parameters S11_DUT8 2 parameters S21_DUT8 2 parameters S22_DUT8 2 parameters In total 22 parameters for DUT1, DUT3 and DUT8. Based on an initial guess we can compute the insertion loss as well as the impedance in 15 cases for each one of table 1 and table 2. That gives 90 equations. We also need |S11|=|S22| which gives three more equations and that the insertion loss of "none" has to be zero. With table 1 and table 2 as input to duteval-1.0.tbz (6627 bytes) the data of table 3 is obtained. The program sets up the equations and solves the least squares problem of finding optimum parameters. One can set different weights to the phases and to the insertion losses. The experimental uncertainties differ by something like a factor of 100 so the errors have to be weighted correspondingly. It is not very critical. | |
Table 3.Results from a least squares fit using duteval. The first 15 lines are from table 1, the next 15 lines are from table 2. | |
The purpose of this investigation was to find the summed loss of DUT1, DUT3 and DUT8 with better precision than obtained previously The old result was 0.0890 dB to compare with 0.0807 dB. The difference is due to different results for DUT8 which has high SWR and is more difficult to measure. What error limits to set on the new result is beyond my skils. The RMS error for the computed insertion losses is 0.0013 dB. The result is based on a large number of measurements and is presumably a bit better. The impedance transformers.Table 4 is a list of the six combinations that all have the same insertion loss, 0.0807 dB (with a small but unknown error.)Four of them have VSWR in the range 1.47 to 1.50 and map the impedance plane reasonably well. The remaining two are significantly closer to the Smith chart center and will not be used. They are marked with an asterisk in table 4. | |
Device Z_{re} Z_{im} Phase Ideal Diff VSWR DUT138 73.82 5.71 11 0 11 1.49257 DUT183 45.26 -18.94 -93 -90 -3 1.50320 DUT318 68.05 -0.47 -1 * 1.36114 DUT381 34.18 2.49 169 180 -11 1.46963 DUT813 52.42 15.00 73 * 1.34407 DUT831 57.36 19.41 59 90 -31 1.46996 | |
Table 4.Impedance transformers. | |