Professional Documents
Culture Documents
resulting from degraded connectors. Likewise, theoretical challenges include low detectability of faults in
operation conditions due to magnitudes of difference in normal signal voltages, noise voltages, as well as
voltage and impedance variations.
Recent research efforts demonstrated that S parameter measurements provide a method to model and
characterize precursors to partially inserted connector faults where changes in the insertion depth of elec-
trical contacts were captured by measured connector S parameters [Tokgöz et al., 2016a, 2016b; Tokgöz and
Dardona, 2016a, 2016b, 2016c]. Analytical and physics-based models were developed for a connector with
a partially inserted contact, providing an accurate representation for the connector as a linear function of
the insertion depth. The models were capable of representing the connector for multiple insertion depths
from a fully inserted to a barely touching contact by varying only two length parameters besides correctly
reproducing measured resonant frequency shifts and their directions.
It is desired to have a miniaturized device that is capable of replacing the functionality of a network analyzer,
monitoring the health or condition of a connector in real time, and processing measured data. Thus, an UWB
radar module, NVA-R641, from Novelda was used as a source for connector interrogation in the proposed work
[Morawski et al., 2014; Yu et al., 2012]. The radar module has an UWB transceiver with transmitting and receiv-
ing ports. The transceiver is capable of generating Gaussian pulses within 0.85–9.55 GHz frequency range. It
transmits pulses using a Vivaldi antenna connected to the transmitting port. A separate Vivaldi antenna that
is connected to the receiving port receives a pulse reflected from a target, samples it, and communicates the
sampled data to a computer via a USB port for processing. The radar module uses comparators to sample a
received pulse with 64, 128, 256, or 512 samples that are separated by 26, 52, or 280 picoseconds correspond-
ing to the sampling frequencies of approximately 38.46, 19.23, and 3.57 GHz, respectively. The antennas were
removed from the radar module for connector fault measurements. It was initially attempted to use both
transmitting and receiving ports of the radar module to obtain the connector response. However, it was real-
ized that the receiver sensitivity of the radar module was not high enough causing excessive noise and its
receiving port was not matched well to 50 Ω resulting in strong reflections. Therefore, the radar module was
used as an UWB source and a spectrum analyzer was used to gather connector response data as an alternative
to network analyzer measurements.
In this paper, an electrical connector is treated as a two-port network. The transfer function of a two-port
network is derived in terms of its S parameters and port reflection coefficients besides representing it as a
measured spectral ratio. Transfer function representations as a measured spectral ratio and as a function of S
parameters are shown to be in good agreement. It is demonstrated that the connector S parameters extracted
from its transfer function compare well with those measured using a network analyzer at multiple insertion
depths.
This paper is organized as follows: Representations of the transfer function as a measured spectral ratio and
as a function of S parameters are presented in section 2. The proposed method for detection of precursors to
partially inserted connector faults is discussed in section 3. A brief summary of the work done and potential
future work are presented in section 4.
The power spectrum of the input signal, X(f ), is measured in dB using a spectrum analyzer. Then, the power
spectrum of the output signal, Y(f ), is measured in dB after the connector with a single contact is connected
between the input signal and spectrum analyzer as a two-port network for which the transfer function is
predicted. The transfer function, H(f ), is obtained by subtracting X(f ) from Y(f ) in dB over all frequency points
of measurement.
2.2. Transfer Function as a Function of S Parameters
Voltages at both ends of a two-port network, shown in Figure 1, can be expressed as
where V1+ and V1− are respectively the incoming and outgoing waves for the source port and V2+ and V2− are
respectively the incoming and outgoing waves for the load port of the general network shown in Figure 1
such that
V1− = S11 V1+ + S12 V2+ (5)
where Sij for i = 1, 2 and j = 1, 2 are the S parameters of the two-port network. The incoming and outgoing
waves can be related as
V1− = Γ1 V1+ (7)
where ΓL is the load reflection coefficient and Γ1 is the reflection coefficient of the source port of the network.
It is possible to insert (8) into (5) and (6) to yield
S12 S21 ΓL
Γ1 = S11 + (9)
1 − S22 ΓL
After (8) is inserted into (6), the ratio of the load and source voltages is given as
V2 V − (1 + ΓL ) S21 (1 + ΓL )
= 2+ = (10)
V1 V1 (1 + Γ1 ) 1 + S 11 − Γ L (S22 + S11 S22 − S12 S21 )
where (9) is used. The transfer function of the two-port network is determined using two S parameter mea-
surements where the first measurement is conducted without the network as shown in Figure 2 and the
second measurement is performed with the network as shown in Figure 3 as discussed in section 2.1. The
measurement setup without the network can be treated as a series of source impedance with a load as shown
in Figure 2. The S parameters for a series source impedance can be derived as
[ ]
1 1 + ΓS 2(1 − ΓS )
S̄ 1 = (11)
3 − ΓS 2(1 − ΓS ) 1 + ΓS
where
ZS − Z0
ΓS = (12)
ZS + Z0
Figure 2. A schematic diagram for measurement without a connector showing a source connected to a load.
The measurement setup with the two-port network can be defined as a cascade combination of a series
source impedance and the network with a load as shown in Figure 3. Then, the S parameters for the cascade
combination can be determined by multiplying their T matrices and converting the resulting T matrix into S
parameters to obtain
[ (2) (2) ]
S11 S12
S̄ 2 = (2) (14)
S21 S(2)
22
where
(h) (h)
(1 + ΓS )T11 + (1 − 3ΓS )T21
S(2)
11
= (h) (h)
(15)
(3 − ΓS )T11 − (1 + ΓS )T21
( )
(h) (h) (h) (h)
2(1 − ΓS ) T11 T22 − T12 T21
S(2)
12
= (h) (h)
(16)
(3 − ΓS )T11 − (1 + ΓS )T21
2(1 − ΓS )
S(2)
21
= (h) (h)
(17)
(3 − ΓS )T11 − (1 + ΓS )T21
(h) (h)
−(3 − ΓS )T12 + (1 + ΓS )T22
S(2)
22
= (h) (h)
(18)
(3 − ΓS )T11 − (1 + ΓS )T21
Y(f ) S(h)
21
(1 − ΓS )(1 + ΓL )
= ( ) (19)
VS 2 1 − ΓS S(h) − ΓL S(h) + ΓS ΓL Δ h
11 22
where S(h)
ij
for i = 1, 2 and j = 1, 2 are the S parameters of the two-port network, and
Figure 3. A schematic diagram for measurement with a connector showing a cascade combination of a series source
impedance and a two-port network connected to a load.
Finally, the transfer function of the network can be determined by taking the ratio of (19) and (13) to yield
S(h)
21
(1 − ΓS ΓL )
H(f ) = (21)
1 − ΓS S(h)
11
− ΓL S(h)
22
+ ΓS ΓL Δ h
which relates the transfer function to the S parameters and source and load reflection coefficients of the
network.
The source and load are respectively the UWB radar module and spectrum analyzer when the connector with
a single contact is considered as a two-port network for which the transfer function is predicted.
3. Measurement
The UWB radar module from Novelda is shown in Figure 4. It was initially used both as a source that generates
UWB Gaussian pulses and as a receiver for sampling the pulses transmitted through the connector for which
the transfer function is predicted. However, the receiver sensitivity of the radar module is not high enough,
causing excessive noise in the sampled data. The receiving port of the radar module is also not matched well
to 50 Ω resulting in strong reflections. Hence, it was decided to use the radar module only as a source for
UWB signal generation. The low-noise amplifier (LNA) shown in Figure 5 was used to increase the power level
by approximately 17–19 dB. A schematic diagram is shown in Figure 6 for spectrum measurement without a
connector where Agilent 8563E spectrum analyzer was used to collect data and a 6 dB attenuator isolates the
amplifier from the spectrum analyzer.
First, the measurement setup pertaining to Figure 6 was used to collect data without a connector. The
collected data correspond to (10) and serve as baseline for all consequent measurements with a connec-
tor. Then, the measurement setup associated with Figure 6 was modified by placing a partially inserted
Figure 5. The 1–12 GHz LNA, MLA-01122B-C4, from Microwave Technology, Inc.
Figure 6. A schematic diagram of spectrum measurement without the connector for which the transfer function is
predicted.
connector with a single pin-socket contact between the attenuator and spectrum analyzer as shown in
Figure 7. Its transfer function was obtained both as a measured spectral ratio and as a function of the S parame-
ters from direct network analyzer measurements. Comparison between the transfer functions obtained using
the two approaches is shown in Figure 8 for transmission through a connector with a fully inserted contact.
The magnitude of S11 of a connector with a single contact obtained from network analyzer measurements
is shown in Figure 9 at multiple insertion depths from a fully inserted (100%) to a barely touching (∼0%)
contact. Since the connector was not designed to operate at high frequencies of interest in Figure 9, multiple
reflections of signals are expected at these frequencies due to impedance discontinuities internal to the con-
nector. These reflections cause nulls and peaks in the magnitude of S11 depending on whether they interact
destructively or constructively, respectively, as shown in Figure 9. It is desired to extract the magnitude of S11
of the connector from its transfer function instead of conducting network analyzer measurements to detect
the resonant frequency shifts in the magnitude of S11 as the insertion depth is varied.
When |ΓS | and |ΓL | go to zero, the transfer function in (21) reduces to
lim |H(f )| = |S(h)
21
| (22)
ΓS →0,ΓL →0
Since the connector of interest is almost lossless, and |ΓS | and |ΓL | are very small for the measurement setup
shown in Figure 7, |S(h)
11
| can be approximated from (23) for a connector with a single contact at multiple
insertion depths as shown in Figure 10.
It is observed by comparing Figures 9 and 10 that the magnitude of S11 of the connector extracted from its
transfer function agrees reasonably well with the magnitude of S11 obtained from network analyzer mea-
surements. Discrepancies between them can be attributed to the fact that the approximation in (23) is used
assuming that |ΓS | and |ΓL | are very small, and the connector is lossless. In addition, the attenuator in Figure 7,
which lowers |ΓS |, reduces the signal-to-noise ratio contributing to discrepancies. These sources of error cause
degradations in accuracy when the locations and depths of the nulls at the resonant frequencies are deter-
mined from the magnitude of S11 extracted from the transfer function. However, the resonant frequency shift
around 3 GHz as the insertion depth is varied and its direction can be reproduced reasonably well when it is
extracted from the transfer function. This resonant frequency shift is sufficient to characterize the connector
for multiple insertion depths. The output port of the UWB radar module can be better matched for improved
accuracy. A minispectrum analyzer with a USB interface can be used to collect data and send it to a computer
for postprocessing toward the development of a miniaturized device for detecting precursors to connector
faults [DeLisle, 2014].
Figure 7. A schematic diagram of spectrum measurement for transmission through a connector with a single contact.
Figure 8. Comparison between transfer functions obtained as a measured spectral ratio and as a function of S
parameters for transmission through a connector with a fully inserted contact.
In addition, the measurement setup corresponding to Figure 6 was also modified by inserting a T adapter
with a 50 Ω load at its third port instead of the connector as shown in Figure 11. Comparison between the
transfer functions of the T adapter with a 50 Ω load obtained as a measured spectral ratio and as a function
of S parameters is shown in Figure 12 for transmission through a T adapter.
Moreover, the measurement setup associated with Figure 6 was modified by inserting a T adapter between
the attenuator and spectrum analyzer as shown in Figure 13. The pin side of a connector with a single contact
was connected to the third port of the T adapter, whereas its socket side is terminated with a 50 Ω load as
shown in Figure 13. The transfer function of a combination of the T adapter, connector, and 50 Ω load was
found both as a measured spectral ratio and as a function of S parameters. Comparison between the transfer
functions obtained using the two approaches is shown in Figure 14 for the combination of the T adapter,
connector, and 50 Ω load.
The S parameters for the combination of the T adapter, connector, and 50 Ω load shown in Figure 13 can be
expressed in terms of the S parameters of the T adapter and connector as
[ (r) (r) ] [ (t) (t) ] [ ]
S11 S12 S11 S12 S(c) 𝛿11 𝛿12
̄Sr = = + 11
(24)
S(r) S(r)
21 22
S(t) S(t)
21 22 1 − S(c) S(t) 𝛿21 𝛿22
11 33
Figure 9. Magnitude of S11 based on network analyzer measurements for a connector with a single contact at multiple
insertion depths from a fully inserted (100%) to a barely touching (∼0%) contact. The resonant frequency around 3 GHz
shows a pronounced shift as the insertion depth changes.
Figure 10. Magnitude of S11 as predicted from the transfer function of a connector with a single contact at multiple
insertion depths from a fully inserted (100%) to a barely touching (∼0%) contact. The pronounced shift in the resonant
frequency around 3 GHz with the insertion depth is reproduced.
where S(t)
ij
for i = 1, 2, 3 and j = 1, 2, 3 are the S parameters of the T adapter, S(c)
11
is the reflection coefficient of
the pin side of the connector when its socket side is terminated with a 50 Ω load, and
where
S(t)
21
(1 − ΓS ΓL )
Ht = (30)
1 − ΓS S(t)
11
− ΓL S(t)
22
+ ΓS ΓL Δ t
Figure 11. A schematic diagram of spectrum measurement for transmission through a T adapter with a 50 Ω load at its
third port.
Figure 12. Comparison between transfer functions obtained as a measured spectral ratio and as a function of S
parameters for transmission through a T adapter with a 50 Ω load at its third port.
S(r)
21
(1 − ΓS ΓL )
Hr = (31)
1 − ΓS S(r)
11
− ΓL S(r)
22
+ ΓS ΓL Δ r
with
Δt = S(t) S(t) − S(t)
11 22
S(t)
12 21
(32)
and
𝛿Δ = S(t) 𝛿 + S(t)
11 22
𝛿 − S(t)
22 11
𝛿 − S(t)
12 21
𝛿
21 12
(34)
Figure 13. A schematic diagram of spectrum measurement for the combination of the T adapter, connector, and 50 Ω
load.
Figure 14. Comparison between transfer functions obtained as a measured spectral ratio and as a function of S
parameters for the combination of the T adapter, connector, and 50 Ω load.
4. Conclusion
It was demonstrated that a miniaturized UWB source and a minispectrum analyzer can be used to detect
precursors to electrical connector faults related to a partially inserted pin-socket contact as an alternative to
S parameter measurements using a network analyzer. A pin-socket contact of a connector was treated as a
two-port network. The transfer function of a two-port network was successfully derived and related to its S
parameters as well as the reflection coefficients at its ports. The transfer function data obtained using the
derivation agreed well with its representation as a measured spectral ratio. The measured resonant frequen-
cies of the magnitude of S11 of the connector shift in opposite directions as the insertion depth changes, which
was captured in the magnitude of S11 extracted from its transfer function. Deviations between the magni-
tude of S11 extracted from the transfer function and the values measured directly using a network analyzer
are attributed to the reflection coefficients at the connector ports, losses within the connector, and noise.
The presented work increases the technology readiness level of the techniques developed for the detection
of precursors to partially inserted connector faults. The proposed method will be helpful for field implemen-
tation using miniaturized RF electronics. Future activities should focus on modifying the receiver circuit of the
UWB radar module for higher number of samples, higher sensitivity, and better impedance matching. Devel-
opment of onboard processing electronics and algorithms for real-time identification of fault precursors is
also desired by the avionics industry.
Acknowledgments References
The authors gratefully acknowledge
the NASA Ames Research Center Auzanneau, F. (2013), Wire troubleshooting and diagnosis: Review and perspectives, Prog. Electromagn. Res. B, 49, 253–279.
for their support under contract Cerri, G., R. De Leo, L. Della Nebbia, S. Pennesi, V. M. Primiani, and P. Russo (2005), Fault location on shielded cables: Electromagnetic
NNA12AB91C. The authors would modelling and improved measurement data processing, IEEE Proc. Sci. Meas. Technol., 152(5), 217–226.
like to thank Kevin R. Wheeler, Stefan DeLisle, J.-J. (2014), PC-driven test instruments pick up speed, Microwaves and RF, 53(12), 38.
Schuet, and Doğan Timuçin, NASA Deutsch, A., C. W. Surovic, J. S. Campbell, P. W. Coteus, A. P. Lanzetta, J. T. Holton, and A. D. Knight (1997), Electrical characteristics of
Ames Research Center, for their use- high-performance pin-in-socket and pad-on-pad connectors, IEEE Trans. Compon. Packag. Manuf. Technol., 20(1), 64–77.
ful comments and suggestions. Sameh Furse, C., Y. C. Chung, R. Dangol, M. Nielsen, G. Mabey, and R. Woodward (2003), Frequency-domain reflectometry for on-board testing of
Dardona (dardonsi@utrc.utc.com) aging aircraft wiring, IEEE Trans. Electromagn. Compat., 45(2), 306–315.
can be contacted to obtain the Kwon, D., M. H. Azarian, and M. Pecht (2011), Nondestructive sensing of interconnect failure mechanisms using time-domain reflectometry,
presented data. IEEE Sensors J., 11(5), 1236–1241.
Maul, C., J. W. McBride, and J. Swingler (2001), Intermittency phenomena in electrical connectors, IEEE Trans. Compon. Packag. Technol.,
24(3), 370–377.
Morawski, R. Z., Y. Yashchyshyn, R. Brzyski, F. Jacobsen, and W. Winiecki (2014), On applicability of impulse-radar sensors for monitoring
of human movements, in 20th IMEKO TC4 International Symposium and 18th International Workshop on ADC Modelling and Testing,
pp. 754–759, Res. on Electric and Electron. Meas. for the Econ. Upturn, Benevento, Italy.
Murrell, S. R., and S. L. McCarthy (1997), Intermittence detection in fretting corrosion studies of electrical contacts, in Proceedings of the 43rd
IEEE Holm Conference on Electrical Contacts, pp. 1–6, IEEE, Philadelphia, Pa.
Pecht, M., and V. Ramappan (1992), Are components still the major problem: A review of electronic system and device field failure returns,
IEEE Trans. Compon. Hybrids Manuf. Technol., 15(6), 1160–1164.
Schneider, L., K. Howard, S. Glover, T. Lockner, and M. Dinallo (2005), A new capability to detect and locate insulation defects in complex
wiring systems, IEEE Elect. Insul. Mag., 21(4), 14–20.
Schuet, S., D. Timuçin, and K. Wheeler (2011), A model-based probabilistic inversion framework for characterizing wire fault detection using
TDR, IEEE Trans. Instrum. Meas., 60(5), 1654–1663.
Schuet, S. R., D. A. Timuçin, and K. R. Wheeler (2012), Shielded-twisted-pair cable model for chafe fault detection via time-domain
reflectometry, Tech. Rep. TM2012216001, NASA Ames Research Center, Moffett Field, Calif.
Sharma, C., R. Harrison, and C. Furse (2007), Low-power STDR CMOS sensor for locating faults in aging aircraft wiring, IEEE Sensors J., 7(1),
43–50.
Song, E., Y.-J. Shin, P. E. Stone, J. Wang, T.-S. Choe, J.-G. Yook, and J. B. Park (2009), Detection and location of multiple wiring faults via
time-frequency-domain reflectometry, IEEE Trans. Electromagn. Compat., 51(1), 131–138.
Sun, A. C., H. K. Moffat, D. G. Enos, and C. S. George (2007), Pore corrosion model for gold-plated copper contacts, IEEE Trans. Compon.
Packag. Technol., 30(4), 796–804.
Tokgöz, Ç., and S. Dardona (2016a), Physics based RF/microwave characterization of wave interactions within electrical connectors with
partial insertion faults, Radio Sci., 51, 1489–1502, doi:10.1002/2016RS006101.
Tokgöz, Ç., and S. Dardona (2016b), Physics based modeling of connectors faults, in International Symposium on Antennas and Propagation
(APSURSI), pp. 785-786, IEEE, Fajardo, P. R.
Tokgöz, Ç., and S. Dardona (2016c), Connector fault interrogation using UWB sources, in International Symposium on Antennas and
Propagation (APSURSI), pp. 2119-2120, IEEE, Fajardo, P. R.
Tokgöz, Ç., S. Dardona, N. C. Soldner, and K. R. Wheeler (2016a), Modeling and characterization of partially inserted electrical connector
faults, J. Appl. Phys., 119, 104501, doi:10.1063/1.4943178.
Tokgöz, Ç., S. Dardona, and N. C. Soldner (2016b), Modeling of partially inserted connector faults, in International Symposium on Antennas
and Propagation (APSURSI), pp. 2077–2078, IEEE, Fajardo, P. R.
Wheeler, K. R., D. A. Timuçin, I. X. Twombly, K. F. Goebel, and P. F. Wysocki (2007), Aging aircraft wiring fault detection survey, Tech. Rep. 1342,
NASA Ames Research Center, Moffett Field, Calif.
Yu, Y., J. Yang, T. McKelvey, and B. Stoewe (2012), A compact UWB indoor and through-wall radar with precise ranging and tracking, Int. J.
Antenn. Propag., 2012, 678590, doi:10.1155/2012/678590.
Erratum
In the originally published version of this paper, an author’s name was misspelled as "Çağata Tokgöz." The
author’s name should read "Çağatay Tokgöz." This error has since been corrected and this version may be
considered the version of record.