DesignCon 2013
Effects of Temperature and
Relative Humidity in
Transmission Systems Using
Differential Signaling
Aldo Morales, Penn State Harrisburg
Email: awm2@psu.edu
Sedig Agili, Penn State Harrisburg
Email: ssa10@psu.edu
Mike Resso. Agilent
Email: mike_resso@agilent.com
Jeff Clark, Penn State Harrisburg
Email: jmc5444@psu.edu
Chris Kocuba, Samtec
Email: Chris.Kocuba@samtec.com
Abstract
Most of the current on-board and cable communication systems are based on differential
signaling. If any asymmetry happens in the differential lines then unwanted RF radiation
will occur. On the other hand, placement of the communication equipment that contains
differential signaling can be located anywhere in the world with different environmental
conditions. To this effect, researchers have explored the effect of the humidity and
temperature on PCB boards but mostly in single mode communication systems. In fact,
environmental issues are very important to the signal integrity (SI) community since they
have a profound effect in the performance of communications systems. In this paper, we
explore relative humidity and temperature effects on PCB boards with differential signaling.
Of particular interest is mode conversion, since this increases unwanted RF radiation.
Key takeaways: We explore relative humidity and temperature effects on PCB boards
with differential signaling. Of particular interest is mode conversion, since this increases
unwanted RF radiation.
Authors Biography
Dr. Aldo Morales has a Ph.D. in electrical and computer engineering from SUNY-Buffalo.
Dr. Morales was the PI for a 3-year Ben Franklin Technology Partners Grant that
established the “Center of Excellence in Signal Integrity” at Penn State Harrisburg. He was
a co-author for the Best poster paper award at the IEEE International Conference on
Consumer Electronics 2007, Las Vegas, Nevada.
Dr. Sedig S. Agili has a Ph.D. in Electrical and Computer Engineering from Marquette
University. He was the Co-PI for a 3-year Ben Franklin Technology Partners Grant that
established Penn State Harrisburg’s “Center of Excellence in Signal Integrity.” He was a
co-author for the Best poster paper award at the IEEE International Conference on
Consumer Electronics 2007, Las Vegas, Nevada.
Mike Resso is the Signal Integrity Application Scientist in the Component Test Division
of Agilent Technologies and has over twenty-five years of experience in the test and
measurement industry. Mike has been awarded one US patent and has twice received the
Agilent “Spark of Insight” Award for his contribution to the company. He received a
Bachelor of Science degree in Electrical and Computer Engineering from University of
California.
Jeff Clark is an Electrical Engineering undergraduate student. He did his undergraduate
internship at Center for Signal integrity at Penn State Harrisburg.
Chris Kocuba was an Electrical Engineering undergraduate student. He did his
undergraduate internship at Center for Signal integrity at Penn State Harrisburg. Chris
now works in Samtec, a connector company in the Harrisburg, PA area.
Introduction
Temperature and humidity effects in the performance of transmission systems have
recently gained interest in the signal integrity community [1-8]. For instance, Hamilton et
al. [1] investigated the temperature and relative humidity impacts on transmission line
loss data and analysis obtained from two Printed Circuit Board (PCB) designs subjected
to a program of drying and moisture absorption were presented. In [2], Miller et. al.
studied the complex permittivity dependence of environmental factors in their impact on
speed of signal propagation, channel loss and impedance matching. In addition, they
examined the impact of temperature variation on eye diagrams operating margin. Sheets
and D’Ambrosia [3] also explored the environmental impact on a XAUI backplane. They
found an increased S-parameter sensitivity in the backplane at different temperatures and
humidity levels, with more pronounced effects at high humidity and high temperatures
(60 C and above). They also found that dielectric constant and tangent loss increase when
temperature. This was essentially due to the increased activation of ionic species in the
resin matrix.
In [4], Agili et. al. evaluated the impact of temperature and relative humidity on
microstrips. They also [4] modeled, simulated and predicted the performance of the
microstrips subject to these environmental factors. For instance, they found out, after
extensive analysis of S-parameter data, that insertion loss variations have an exponential
decay. One of their goals was to obtain a diffusion coefficient of the substrate material for
modeling prediction [4]. However, most of the work was done on single ended
transmission system. Considering that most of the high speed on-board and cable
communication systems are based on differential signaling [9-11], engineers try to design
these systems with as much symmetry as possible. If any asymmetry happens in the
microstrips (or cables [11 ]) then unwanted RF radiation will occur and potentially having
the product not to pass FCC standards. On the other hand, placement of the communication
equipment, that contains differential signaling, can be located anywhere in the world with
different environmental conditions.
In this paper, we explore the of humidity and temperature effects on PCB boards with
differential signaling. Of particular interest is S-parameters mode conversion due to the
asymmetry induced by the moisture absorption in the differential lines. Note that, as per
[9], any asymmetry can also convert some of the differential signals into common
signals. This asymmetry can also lead to signal skew that affects bandwidth, degrades
rise time as well as increases electro-magnetic interference (EMI) [10]. The asymmetry
can also affect near and far end crosstalk. This paper is divided into three main sections.
First, a detailed explanation of the experimental set up is provided. Second, a presentation
of the experimental results and insight into these results are presented. Finally,
conclusions are derived and further research is outlined.
Experimental Set up and Procedure
For the experimental set up a similar procedure described in [8] was followed, which is
detailed in a flowchart shown in figure 1. A temperature and humidity chamber, Cincinnati
Sub-Zero Products Model Number: ZHS-32-2-H/AC was utilized to change the
environmental conditions imposed on the DUTs (see Figure 2). For this experiment, the
chamber was set at 55 0C and 95 % relative humidity. A calibrated electronic scale Mettler
AE 100 was used to measure the mass of the DUTs. An Agilent Technologies Performance
Network Analyzer (PNA) E8364B was the chosen measurement instrument, which offers a
wide test frequency range from 10 MHz to 50 GHz. The PNA was calibrated using an
electronic calibration kit N4693A, and short-open-load-through (SOLT) calibration method
was performed. The latest version of Physical Layer Test System (PLTS) available at the
time of measurements was used. Before the measurements, the DUTs were baked in a
temperature chamber at 80 degrees C for 4 days, a condition called “dry state”, to
remove any initial moisture, and then weighed.
Bake all DUTs at 80 oC for 4 days
SOLT calibration
asdfsadfcafacaclicalibration
S-parameter measurements of DUTs
(dry state)
Place DUTs in the environmental chamber
kept at 55 oC and 95% RH
S-parameters measurements of DUTs
No
Every 12 hours for 7 days
complete?
Yes
Results analysis and S-parameter simulation
Fig. 1: A flow chart describing the RH/Temperature procedure
To this effect, three devices under test (DUTs) were used as shown in Fig. 3 and Fig 4,
respectively. One, “golden standard reference board,” provided by a local company,
member of the Center for Signal Integrity at Penn State Harrisburg (Samtec); one DUT
made (“milled board”, without solder mask) by researchers of the above center and a Tyco
XAUI backplane board provided by Agilent. To keep consistency of the measurements
only one operator measured the S-parameters. Since operator’s hands can be a heat source
during the measurement process minimal hands touching time was required. Cable
movement was also restricted after calibration and connector cleaning was performed every
12 hours.
Fig. 2: Cincinnati temperature and humidity chamber
Fig. 3: Devices Under Test used to evaluate temperature and relative humidity effects using
differential signaling.
Fig. 4: Tyco XAUI backplane board provided by Agilent.
Experimental Results
In this section, we will describe and provide some insight of the results obtained. Mode
conversion for the golden standard is shown in Fig. 5. For comparison purposes,
measurements are shown at dry state (0 hours) and 7 days (it is assumed that saturation has
been reached at that point [1-4]). There is a significant increase in the mode S-parameter
conversion in the 2 to 4 GHz range; giving an indication that design engineers need to be
aware of potential unwanted RF radiation on that range. For the rest of frequency range, it
appears that is not a significant mode conversion. However, the phase shows a wide
variation between dry state and saturation. The phase is shown in Fig. 6, with a zoomed out
in the range of 0.5 to 7 GHz depicted in Fig. 7. In some regions the figures show, for
instance from 4 to 6 GHz, a drastic change.
Golden Standard S21 differential to common mode conversion
0
0 Hrs
7 days
-10
-20
S DC21
-30
-40
-50
-60
-70
0
5
10
15
Frequency GHz
20
25
Fig. 5: Golden Standard mode conversion from “dry state” to saturation state.
Golden Standard phase S21 diff. to common mode
200
0 Hrs
7 days
150
Phase of SDC21
100
50
0
-50
-100
-150
-200
0
5
10
15
Frequency GHz
20
25
Fig. 6: Golden Standard mode conversion phase from “dry state” to saturation state.
Golden Standard phase S21 diff. to common mode
200
0 Hrs
7 days
150
Phase of SDC21
100
50
0
-50
-100
-150
-200
1
2
3
4
5
Frequency GHz
6
7
Fig. 7: Zoomed version of Fig. 6 showing drastic phase change.
The differential SDD21 (insertion loss) parameters show consistent performance
degradation (in some sections at about 4 dB, especially at high frequencies) as shown in Fig.
8.
Golden std S21-parameters differential parameters
0
0 Hrs
7 days
-10
-20
SDD21
-30
-40
-50
-60
-70
0
5
10
15
Frequency GHz
20
25
Fig. 8: Golden Standard differential SDD21 parameters from “dry state” to saturation state.
Golden std phase of S21-parameters diff. parameters
200
0 Hrs
7 days
150
Phase of SDD21
100
50
0
-50
-100
-150
-200
0
5
10
15
Frequency GHz
20
25
Fig. 9: Golden Standard differential SDD21 phase from “dry state” to saturation state.
Golden std phase of S21-parameters diff. parameters
200
0 Hrs
7 days
150
Phase of SDD21
100
50
0
-50
-100
-150
-200
13.5
14
14.5
15
15.5
16
Frequency GHz
16.5
17
Fig. 10: Zoomed version of Fig. 9.
Fig. 9 and Fig. 10, show the golden Standard differential SDD21 phase variation from “dry
state” to saturation state, and zoomed out version, respectively. Both figures show also
consistent change, especially in the 13.5 to 17.5 GHz range as much as 50 degrees phase
change.
For comparisons purpose Figs. 5 and 8 are shown together in Fig. 11, where the reader can
better appreciate the influence of relative humidity and temperature in the golden standard
structure.
Comparisons Golden std S21-parameters differential parameters at 0 Hrs and 7 days
0
SDD21 0 Hrs
SDD21 7 days
-10
SDC21 0 Hrs
SDC21 7 days
-20
dB
-30
-40
-50
-60
-70
0
5
10
15
Frequency GHz
20
25
Fig. 11: Comparison of Fig. 5 and Fig 8.
In the next figures (Fig. 12 to Fig 15), single ended measurements are depicted. The
reason to show single ended is to correlate with previous results depicted in [4] and [8]
obtained, where consistent degradation, due to relative humidity and temperature, in
single ended S-parameters has been reported, especially at high frequency regions.
Return loss and insertion loss are shown in Figs. 12 and 13, respectively, where 3 set of
measurements are depicted, at dry state, at 118 hours and 7 days (saturation). Near end
crosstalk (NEXT) and far end crosstalk (FEXT) are shown in Fig. 14 and 15,
respectively. Notice that in Fig. 14 NEXT is not affected as much since it basically shows
unbalance couplings close to the end of the microstrip line [12] while FEXT is equally
affected by unbalance couplings anywhere along the line [12]. However, since there is
also degradation in the single ended performance, due to relative humidity and
temperature, hence; FEXT is not affected as much, as shown in Fig. 15.
gldstd S11-parameters single ended
0
0 Hrs
118 Hrs
7 days
-5
-10
-15
S11
-20
-25
-30
-35
-40
-45
-50
0
5
10
15
Frequency GHz
20
25
Fig. 12: Return loss, single ended.
gldstd S12-parameters single ended
0
0 Hrs
118 Hrs
7 days
-10
-20
S12
-30
-40
-50
-60
-70
0
5
10
15
Frequency GHz
20
Fig. 13: Insertion loss, single ended.
25
gldstd S31-parameters near end cross talk
0
0 Hrs
118 Hrs
7 days
-10
S31
-20
-30
-40
-50
-60
0
5
10
15
Frequency GHz
20
25
Fig. 14: Golden standard NEXT.
gldstd S41-parameters far end cross talk
0
0 Hrs
118 Hrs
7 days
-10
-20
S41
-30
-40
-50
-60
-70
0
5
10
15
Frequency GHz
20
25
Fig. 15: Golden standard FEXT.
The board built by the Center for Signal Integrity at Penn State Harrisburg, hence called
“milled board”, also shows similar mode conversion performance as the golden standard,
except for an upswing in the 5 to 7 GHz range (see Fig. 16). Differential of SDD21
parameters for the milled board also depict degradation (see Fig 17). Single ended (return
and insertion losses, NEXT and FEXT) show also similar trends as in the golden standard.
Milled Standard S21-parameters differential to common mode conversion
0
0 Hrs
7 days
-10
-20
SDC21
-30
-40
-50
-60
-70
0
5
10
15
Frequency GHz
20
25
Fig. 16: Center milled board mode conversion from “dry state” to saturation state.
Milled standard S21-differential parameters
0
0 Hrs
7 days
-5
-10
-15
SDD21
-20
-25
-30
-35
-40
-45
-50
-55
0
5
10
15
Frequency GHz
20
25
Fig 17: Center milled board differential SDD21 from “dry state” to saturation state.
The XAUI board, on its analyzed range of operation (10 MHz to 6 GHz), shows less mode
conversion performance in most of the range, but with a larger mode conversion on the 2.5
to 3 GHz range as shown in Fig. 18. As expected, the differential parameters show
degradation with respect to relative humidity and temperature, in the order to 2 to 5 dB. The
differential SDD21 parameters are shown in Fig. 19.
XAUI S21-parameters, differential to common mode conversion
0
0 Hrs
7 days
-10
-20
SDC21
-30
-40
-50
-60
-70
-80
-90
0
1
2
3
Frequency GHz
4
5
6
Fig. 18: XAUI board mode conversion from “dry state” to saturation state
XAUI S21-differential parameters
0
0 Hrs
7 days
-10
SDD21
-20
-30
-40
-50
-60
-70
0
1
2
3
Frequency GHz
4
5
6
Fig. 19: XAUI Sd2d1 parameters from “dry state” to saturation state .
NEXT and FEXT for the XAUI show similar behavior as the golden standard, where is
NEXT and FEXT are not affected as much by relative humidity and temperature. NEXT
and FEXT are shown in Figs. 20 and 21, respectively.
XAUI near end crosstalk at 0 Hrs and 7 days
0
NEXT 0 Hrs
NEXT 7 days
-10
-20
-30
dB
-40
-50
-60
-70
-80
-90
0
1
2
3
Frequency GHz
4
5
6
Fig. 20: XAUI board NEXT.
XAUI far end crosstalk at 0 Hrs and 7 days
0
FEXT 0 Hrs
FEXT 7 days
-10
-20
-30
dB
-40
-50
-60
-70
-80
-90
0
1
2
3
Frequency GHz
4
Fig. 21: XAUI board FEXT.
5
6
Conclusions
Our initial results show that boards’ performances, using differential signaling, are affected
by relative humidity and temperature, especially at high frequencies. Differential
parameters, for all board analyzed, show consistent degradation in performance while mode
conversion shows upswing in some regions of the boards’ range of operation. NEXT and
FEXT are also affected in some regions of operation but the change seem to be less than
mode conversion and in differential S-parameters. In general, this experiment seem to
confirm previous work, however more analysis are needed in how moisture absorption and
temperature change the balance lines (for example layer placement) and board weave.
Bibliography
[1]
Paul Hamilton, Gary Brist, Guy Barnes, Jr. & Jason Schrader, Humidity-dependent
Loss in PCB Substrates in
http://members.ipc.org/IPCLogin/IPCMembers/IPC/Route/0507/tech.pdf
[2] J. Miller, Y. Li, K. Hinckley, G. Blando, B. Guenin, I. Novak, A. Dengi, A. Rebelo
and S. McMorrow, “ Temperature and Moisture Dependence of PCB and Package
Traces and the Impact on Signal Performance,” Proceedings of Designcon 12, Santa
Clara, CA, 2012.
[3]
G. Sheets and J. D’Ambrosia “Evaluating Environmental Impact on Channel
Performance,” CommDesign May 2004,
http://www.commsdesign.com/design_corner/showArticle.jhtml?articleID=203005
81.
[4]
S. Agli, A. Morales, J. Li and M. Resso, “Modeling Relative Humidity and
Temperature Effects on Scattering Parameters in Transmission Lines, “ Accepted at
the IEEE Transactions on Advanced Packaging.
[5]
D.I. Amey, S.J. Horowitz, R.L. Keusseyan, “High Frequency Electrical
Characterization of Electronic Packaging Materials: Environmental and Process
Considerations,”. Proceedings of the 4th International Symposium Advanced
Packaging Materials, 1998.
[6]
J-M Heinola, K-P. Lätti, P. Silventoinen, J-P. Ström, and M. Kettunen, “A Method
to Evaluate Effects of Moisture Absorption on Dielectric Constant and Dissipation
Factor of Printed Circuit Board Materials,” Proceedings of the 9th International
Symposium on Advanced Packaging Materials: Processes, Properties and
Interfaces, 2004.
[7]
H. Fremont, et al, “Measurements and FE-Simulations of Moisture Distribution in
FR4-based Printed Circuit Boards”, 7th Int. Conf. On Thermal, Mechanical and
Multiphysics Simulation and Experiments in Micro Electronics and Micro-Systems,
EuroSimE 2006.
[8]
J. Li, M. Minns, A. Morales , S. Agili, M. Resso, “Evaluation of Relative
Humidity and Temperature Effects on Scattering Parameters in Transmission
Systems,” Proceedings of Desigcon10, February 2010, Santa Clara, CA.
[9]
E. Bogatin, Signal integrity Simplified, Prentice Hall, 2003.
[10] J. Nadolny, S. Sercu, “Characterization And Impact Of Skew On Differential
Connector Systems” Proceedings of 18th International Zurich Symposium on
Electromagnetic Compatibility, Zurich, Switzerland, February 20-22, 2001.
[11] J. Nadolny, J. Ferry, and C. Arroyo, “Mode Conversion and EMI Performance of
Shielded Cable Assemblies for 10 Gbps Data Transmission,” Proceedings of
Desigcon08, February 2008, Santa Clara, CA.
[12] C. Di Minico and P. Kish, “Development of Equal Level Far-End Crosstalk
(ELFEXT) and Return Loss Specifications for Gigabit Ethernet Operation On
Category 5 Copper Cabling,” at
www.ieee802.org/3/an/public/material/diminico_IWCS.PDF last accessed
12/01/2012.