SlideShare a Scribd company logo
MemberoftheHelmholtzAssociation
GPR MC/WARR Calibration:
t0-Correction Analysis of Single-channel
vs. Multi-channel Surveys
(Final Update)
Tuesday, August 16th, 2016 | Kevin Fan
MemberoftheHelmholtzAssociation
-Ground Wave is identified** in WARR trace, then cross-
correlated with the corresponding MC trace; the
“MC/WARR time-shift“ is then the time corresponding
to the maximum cross-correlation value 2
Context
-Previous Work:
a) Cross-Correlation Analysis of Baxter‘s July 2015 Selhausen MC/WARR Data*
(“t0-correction Update Pres” in FZJ-GPR)
-Overall Objective: To determine whether 500MHz multichannel GPR data can be calibrated
with CMP data acquired at corresponding locations, via modification of the t0-correction
-Based upon intuitive hypothesis that traces of the 2 surveys, at the same locations, are
similar but just time-shifted version of one another
MemberoftheHelmholtzAssociation
3
Context
 Results:
-Same procedure is iterated over all other channels at the specific location, then
all channels at each and every other location
-A general, channel-specific t0-correction could not be established, due to evident
heterogeneity in the time-shift at different CMP locations along the 120m line
-Mix of time shifts both positive and negative; whereas a systematic shift
between MC/CMP data should be only one or the other?
-Necessitates focus upon only a few locations for subsequent measurements…
MemberoftheHelmholtzAssociation
b) Analysis of July 2016 Selhausen MC/WARR Data, including new MC-calibration
Measurements (Location=5m)
(“MC-Calibration Update” in FZJ-GPR)
-Channel 4 produces very low amplitudes, as well as energies (sum of squared amplitudes)
relative to other channels
-Script MC_calib_processing produces amplitude/energy plots, plots temporal cross-correlation
between measurements taken at different times, and wiggle (as opposed to colour traces)
-Can choose to exclude any combination of undesirable channels/positions from analysis
Context
Apply Channel-specific t0-correction
MemberoftheHelmholtzAssociation
*As determined by manual picking of 1st air-wave arrivals
5
ContextResults:
-Calculated expected air-wave arrival times for every channel/location, based on
theoretical air wave velocity (0.2997m/ns)
-For any particular location on the MC-sled:
t0-Correction = Measured arrival time* - Theoretical arrival time
-Typically, channels 1/2/3/6 have similar t0-correction values, while channels 4/5/7
are also similar, albeit delayed relative to channels 1/2/3/6
-Chose to exclude Location 1 (at 23cm on MC-sled) for all channels, due to difficulties
in manual air-wave picking arising from near-field effects (see next slide)
MemberoftheHelmholtzAssociation
6
Rationale for Dumping Position 1 (of all channels)
-Too early: t0-
correction too
small relative to
other positions
-Too late: t0-
correction too
large relative to
other positions
MemberoftheHelmholtzAssociation
7
Calculation of Cross-Correlation
Coefficient for Quantitative Measure
of Deviation in Traces over Time-
Spans of a few seconds
(at 5m vs. 110m along profile)
-Position 1 removed from analysis
-Generally (except for Channel 4), value is >0.97,
indicating high degree of temporal self-similarity
(over scales of several seconds)
MemberoftheHelmholtzAssociation
8
Current Work: Data Comparison of
July 2016 Selhausen Measurements
Definitions
-Full WARR: Data that, aside from skipped traces, encompasses a continuous
range of offsets
-7 traces WARR (3 forms):
a) MC-sled I: Measurements taken with the MC-sled (termed “MC
-calibration measurements), during which a single-channel WARR is
performed for each and every channel (by varying the Rx-position
over all positions in the MC-sled, for each channel)
b) MC-sled II: Full WARR in MC-sled, that has later (via processing)
been reduced to the subset of 7 traces corresponding to MC positions
c) PE-sled: Same as (b), but with PE-sled
-PE-sled: Measurements taken with the Pulse Ekko single-channel sled
-MC-sled: Measurements taken with the custom multichannel sled
-Channel: A particular transmitter/receiver combination (eg. Tx/Rx4)
-Position/Location: The specific section into which Rx is placed
MemberoftheHelmholtzAssociation
9
Current Work: Data Comparison of
July 2016 Selhausen Measurements
Field Conditions/Measurements
-See spreadsheet Selhausen-04-07-MC in FZJ-GPR for detailed information regarding data files
-Field Measurements taken on 04/07/2016 with Manuela, Anja, Jan; survey design by Manuela
-Weather: Hot, sunny day with no rain
-DVL battery was running low for later measurements (ie. during CMP’s, 110m MC-calibration)
Data Processing Considerations
-Location of Interest: 5m along 120m profile (where 0m is the side closest to the street)
-Only performed all survey methods with Channel 1; Channels 2-6 were only used for MC-
calibration measurements at 5m and 110m, as well as for MOG of the entire 120m profile
-All traces dewowed (calls script dewow_anja_eso.m), ungained, and normalized to
max(abs(‘trace’))=1, to allow for comparison between MC/WARR data
-RPE_Evan.m indicates no skipped traces for MC-calibration data
-Output/Plots generated by script MC_calib_processing.m in main folder
-Semblance Analysis Parameter Space:
a) Ground Wave
- t0 ∈ (0,5) ns
- v ∈ (0.05,0.15*) m/ns
b) Air Wave (newly implemented)
- t0 ∈ (1.5,5.0) ns
- v ∈ (0.25,0.30) m/ns
a)/b) Parameters for both Air/Ground Waves
*Time width = 24ns (arbitrary..?)
MemberoftheHelmholtzAssociation
10
Script Documentation
GPR_Summer_2015_MC - Main script for MOG MC data along profile; must specify data
folder + line (only 1 permitted)
-Accounts for skipped traces by removing them, then compressing the data-set
-Only need to do air-wave picks once; upon completion, will save and use picks
in future
To redo picks, delete t0AirOffsets.mat
Velocity_Analysis_MC_Automated- Called by above - main script for automated ground-
wave picking from MC-data (semblance takes in as input each ‘7-trace WARR’ taken by
the MC system); includes t0-correction and no smoothing
MC_calib_processing – Script to process 7-trace WARR (MC-sled I) data
-Can choose whether to do manual air-wave picks or automated ones (ie. air-
wave semblance); ground-wave picking automated via semblance
-Can exclude any combination of channels/positions from analysis in section
‘Data Parameters’, sub-section ‘MC-calibration Parameters’
Velocity_Analysis_MC_Automated_calib – Similar to Velocity_Analysis_MC_Automated,
but for 7-trace WARR (MC-sled I) data
Velocity_Analysis_MC_Automated_calib_wiggle – Same as above, visualized as
wiggle-traces rather than colour plots
Velocity_Analysis_MC_Automated_calib_wiggle_airground – Same as above, with
automatic air-wave semblance picking instead of manual air-wave picking
MemberoftheHelmholtzAssociation
11
Script Documentation
GPR_Summer_2015_WARR – Main script for WARR data (of form MC_sled II or PE_sled)
-Smooths data (5-point moving average)
Saves air-wave picks; to redo, delete xairpicks or tairpicks (deleting either one
will re-write both)
vel_ana_TE_CO_CMP_Multi_Automated – Called by above; automatic pick of maximum
semblance to determine ground wave
vel_ana_TE_CO_CMP_Multi_Comp – Similar to above, but with manual pick of maximum
semblance (Igor has incorporated into his scripts)
CMPandMC_correlation – Main script used previously for cross-correlation analysis of
Baxter’s July 2015 Selhausen data
PlotAllSelectTraces – Plots the 3 types of 7-trace WARR surveys (see (a)/(b)/(c) in
‘Definitions’) at all offsets
MemberoftheHelmholtzAssociation
12
PE-sled
(Source Data:
Folder 19,
LINE00.DT1)
MC-sled
(Source Data:
Folder 19,
LINE55.DT1)
Full WARR, pre-t0-correction, gained: PE vs. MC Sleds
 Which to pick as Air-wave Arrival: High-frequency or low-frequency content?
MemberoftheHelmholtzAssociation
13
-t0-correction now
applied, based on
direct air-wave pick
(value displayed at
bottom of plots)
-Semblance analysis
used to (attempt to)
identify ground wave
(bolded line on plots)
T0-correction +
Corresponding
Semblance
(High-freq.
air picks)
PE-sled
MC-sled
MemberoftheHelmholtzAssociation
14
Analogous
plots for Low-
Freq. air picks
PE-sled
MC-sled
-Difference in t0-
correction when choosing
high vs. Low-freq. Air
wave is ~6ns for both
PE/MC-sleds
-PE-sled pick appears
~correct (characterizing
the ground wave
minimum)
-MC-sled pick in-correct;
ostensibly requires
inclusion of negative
intercept-time in
semblance parameter
space?
MemberoftheHelmholtzAssociation
11:05
15
Full WARR PE-sled, at Different Time, t0-corrected/gained
11:09 12:56
-Compare whether picking high vs. low frequency air-wave has effect upon
temporal variation of t0 (over the course of the day)
High-frequency Picks
MemberoftheHelmholtzAssociation
11:05
16
Full WARR PE-sled, at Different Time, t0-corrected/gained
11:09 12:56
-t0 values larger for low-frequency picks
Low-frequency Picks
MemberoftheHelmholtzAssociation
17
-For low-frequency picks, max
semblance (and associated
values of t0/v) are far more
consistent, than amongst high-
frequency picks
 Implies that minimization of
temporal variability requires
picking of low-frequency air-
waves, even if earlier high-
frequency content is present
Corresponding
Semblance Plots
High-frequency
picks
Low-frequency
picks
11:05
11:05
11:09
11:09 12:56
12:56
MemberoftheHelmholtzAssociation
18
-t0-correction now relies upon (newly implemented) air-wave
semblance, replacing manual picking
General Proviso: Automatic air-wave picking requires fine-
tuning of parameter space!
-Chosen intercept time from air-wave semblance typically
too small (ie. Needs to be later to accurately plick 1st-
break)
-Optimal parameter space here is:
v ∈ (0.25* , 0.30) m/ns
t ∈ (0.18 , 5) ns
-No set of parameters tested thus far has resulted in 100%
accurate picking: typically there is a pattern in which though
most channels are fine, a few channels will have their automatic
pick being too early, for all locations at such channels (for
instance, the plots on Slide 20)
7-trace WARR (MC-sled): Gained vs. Ungained
MemberoftheHelmholtzAssociation
19
Gained
Air Wave,
Traces
Gained
Ground Wav
Traces
Air Wave
Semblance
(of gained data)
Ground Wave
Semblance
(of gained data)
MemberoftheHelmholtzAssociation
20
Ungained
Ground Wave
Traces
Ungained
Air Wave
Traces
Ground Wave
Semblance
(of ungained data)
Air Wave
Semblance
(of ungained data)
MemberoftheHelmholtzAssociation
21
-Gaining of 7-trace WARR (MC-sled) data results in ambiguous
semblance plot (very high-values across parameter space, no
clear maximum), while ungained data contains several peaks
Observations
MemberoftheHelmholtzAssociation
Plotted in ‘PlotAllSelectTraces.m’ 22
Qualitative Comparison of
Different GPR Survey Methods
MemberoftheHelmholtzAssociation
23
Effect of Running the Cables over the
Centre vs. Side of the MC-sled (gained,
normalized, non-t0 corrected)
Centre Side
MemberoftheHelmholtzAssociation
24
MemberoftheHelmholtzAssociation
25
MemberoftheHelmholtzAssociation
26
MemberoftheHelmholtzAssociation
27
MemberoftheHelmholtzAssociation
28
MemberoftheHelmholtzAssociation
29
MemberoftheHelmholtzAssociation
30
Observations
-Effect of cable position appears significant, and may
affect accuracy of both air and ground wave picks
though it appears to influence the presence of early
-Channel 1: Plots for both positions identical upon
visual inspection
-Channels 2-6: 1st-break air wave arrival appears to be
delayed when cables on side, relative to those for the
cables on centre
-Can further investigate with cross-correlation, see if
cable position has influence on semblance results,
etc..
 Require a ‘cable-holder’ that will allow for centre
position to be employed?
MemberoftheHelmholtzAssociation
Discussion: Sources of Uncertainty
a) Skipped Traces: Typically a few skips per CMP – if too many skips, can distort
linearity of air/ground wave, and therefore distort the t0 obtained from semblance
-Current method used in Ekko2Dread2_Update_test removes skipped
traces, then collapses data to remove the holes
-Offset information in header files has been corrected, to account for
skipped traces
-However, the data is not located in the correct position (offset), due to
collapsing approach
-Proposed Solution: Rather than collapsing data, fill in holes via
interpolation (eg. spline)
-See “On Recovering Missing Ground Penetrating Radar Traces by
Statistical Interpolation Methods” (Safont et. al. 2014) for inspiration?
b) Temporal Variation of t0 even for the same Survey Type*: Across multiple
measurements over the same day of the same thing (Full WARR at 5m in PE-
sled), is variable (see next slide)
-Possibilities: Instrumental temporal drift, use of high-frequency air
waves during manual picking
31
MemberoftheHelmholtzAssociation
32
Conclusions
Analysis of Channels in the Multichannel GPR System
-Channel 4 is faulty (very low amplitude/energy), rendering air-wave picking
(and consequently t0-correction) highly imprecise
-Channels 5/7 have sufficient energy, but on average have larger t0-values
than channels 1/2/3/6
-The channel-specific t0-correction may be dependent upon the specific
subsurface conditions at the location (eg. high-conductivity), as implied by the
fact that the 110m calibration measurements showed that even for different
positions of the same channel, t0 was variable; while calibration at 5m
produced much more consistent results
-Attempted CMP measurements with the Multiplexer failed
-Placement of the MC antenna cables on the middle vs. side of the sled
appears to have a significant effect for Channels 2-6, with the side positioning
resulting in a delayed air-wave 1st-break
MemberoftheHelmholtzAssociation
33
Conclusions
Semblance Analysis
-Semblance analysis breaks down (manifested by non-unique maxima,
non-physically reasonable parameters chosen,etc.) under circumstances
such as:
-Wave non-linearity (whether due to skips or otherwise)
-Use of only selected traces (eg. In MC)
-Gaining of data (for 7-trace WARR with the MC-sled)
-Generally, the Semblance Parameter space must be tested for each
data-set
-It is not always clear that 5ns is optimal as the maximum possible
intercept-time for 500MHz data; as max semblance values sometimes
occur at the edge of the parameter space
-Time-width (current default value of 24ns seems arbitrary)
Temporal Variation
-Temporal variation in data on the scale of several seconds is minimal,
as indicated by cross-correlation calculations; regarding diruanl
variations, the t0-correction value is also consistent, but only provided
that the low-frequency air-wave arrivals are chosen
MemberoftheHelmholtzAssociation
34
1. Apply the Channel-1 t0-calibration results to other channels; repeat
semblance, see if results are consistent across channels
2. Apply channel-specific t0-corrections (calculated from
MC_calib_processing.m) to the 120m profile MOG data, then see if
ground-wave semblance values become comparable with WARR data
taken at same positions
3. Perform 7-trace WARR (MC-sled) measurements at locations with a
range of conductivities, so as to better characterize the degree to which
the consistency of the channel-specific t0-correction is affected by
specific subsurface conditions at each location
Future Directions
MemberoftheHelmholtzAssociation
35
4. Implement potential Semblance Solutions, such as:
-Application of Hilbert Transform to 7-trace WARR (PE-sled)
may be an avenue for improvement of semblance and reduction of non-
uniqueness of maxima
-Investigation of the influence of Time-width (‘window’) on max
semblance picks
-Include negative time values in semblance parameter space?
5. Investigate proper treatment of skipped traces, via interpolation
6. Use filtering to improve air-wave picking, such as band-pass, to
remove high-frequency content sometimes seen, as well as low-
frequency waves that occur before the air wave
7. Construct/Upgrade MC-sled such that all antenna cables may be run
through the centre of the sled
Future Directions

More Related Content

What's hot

Aircraft propulsion non ideal turbofan cycle analysis
Aircraft propulsion   non ideal turbofan cycle analysisAircraft propulsion   non ideal turbofan cycle analysis
Aircraft propulsion non ideal turbofan cycle analysis
Anurak Atthasit
 
Nasa tech briefs ksk 11495, simplified model of duct flow
Nasa tech briefs ksk 11495, simplified model of duct flowNasa tech briefs ksk 11495, simplified model of duct flow
Nasa tech briefs ksk 11495, simplified model of duct flow
Julio Banks
 
P&w tables of compressible flow functions
P&w tables of compressible flow functionsP&w tables of compressible flow functions
P&w tables of compressible flow functions
Julio Banks
 
Default test templates raptor
Default test templates raptorDefault test templates raptor
Default test templates raptor
Eduardo Marchesi
 
Study and Development of an Energy Saving Mechanical System
Study and Development of an Energy Saving Mechanical SystemStudy and Development of an Energy Saving Mechanical System
Study and Development of an Energy Saving Mechanical System
IDES Editor
 
Uc 832 calibration
Uc 832 calibrationUc 832 calibration
Uc 832 calibration
Sascha Schlobinski
 
Aircraft propulsion performance of propellers
Aircraft propulsion   performance of propellersAircraft propulsion   performance of propellers
Aircraft propulsion performance of propellers
Anurak Atthasit
 
Em ii mannual
Em ii mannualEm ii mannual
Btl control system-lab-manual-10 eel68
Btl control system-lab-manual-10 eel68Btl control system-lab-manual-10 eel68
Btl control system-lab-manual-10 eel68
Gopinath.B.L Naidu
 
Aircraft propulsion non ideal cycle analysis
Aircraft propulsion   non ideal cycle analysisAircraft propulsion   non ideal cycle analysis
Aircraft propulsion non ideal cycle analysis
Anurak Atthasit
 
ENHANCED DATA DRIVEN MODE-FREE ADAPTIVE YAW CONTROL OF UAV HELICOPTER
ENHANCED DATA DRIVEN MODE-FREE ADAPTIVE YAW CONTROL OF UAV HELICOPTERENHANCED DATA DRIVEN MODE-FREE ADAPTIVE YAW CONTROL OF UAV HELICOPTER
ENHANCED DATA DRIVEN MODE-FREE ADAPTIVE YAW CONTROL OF UAV HELICOPTER
ijics
 
A Simplified Speed Control Of Induction Motor based on a Low Cost FPGA
A Simplified Speed Control Of Induction Motor based on a Low Cost FPGA A Simplified Speed Control Of Induction Motor based on a Low Cost FPGA
A Simplified Speed Control Of Induction Motor based on a Low Cost FPGA
IJECEIAES
 
Turbofan Engine Modelling and Control Design using Linear Quadratic Regulator...
Turbofan Engine Modelling and Control Design using Linear Quadratic Regulator...Turbofan Engine Modelling and Control Design using Linear Quadratic Regulator...
Turbofan Engine Modelling and Control Design using Linear Quadratic Regulator...
theijes
 
Mobile instrumentation and_data_processing_system_for_testing
Mobile instrumentation and_data_processing_system_for_testingMobile instrumentation and_data_processing_system_for_testing
Mobile instrumentation and_data_processing_system_for_testing
lkantwani
 
Gps waas class 5 nov10
Gps waas class 5 nov10Gps waas class 5 nov10
Gps waas class 5 nov10
Clark Jay Wilson
 
Automated Testing of C-Band Rotary Field Ferrite Phase Shifter
Automated Testing of C-Band Rotary Field Ferrite Phase ShifterAutomated Testing of C-Band Rotary Field Ferrite Phase Shifter
Automated Testing of C-Band Rotary Field Ferrite Phase Shifter
IJERA Editor
 
Artigo sea
Artigo seaArtigo sea
Artigo sea
alegbrito2
 
space vector
space vectorspace vector
space vector
elsayed soliman
 
JGrass: the Horton machine (FOSS4G2008)
JGrass: the Horton machine (FOSS4G2008)JGrass: the Horton machine (FOSS4G2008)
JGrass: the Horton machine (FOSS4G2008)
Andrea Antonello
 

What's hot (19)

Aircraft propulsion non ideal turbofan cycle analysis
Aircraft propulsion   non ideal turbofan cycle analysisAircraft propulsion   non ideal turbofan cycle analysis
Aircraft propulsion non ideal turbofan cycle analysis
 
Nasa tech briefs ksk 11495, simplified model of duct flow
Nasa tech briefs ksk 11495, simplified model of duct flowNasa tech briefs ksk 11495, simplified model of duct flow
Nasa tech briefs ksk 11495, simplified model of duct flow
 
P&w tables of compressible flow functions
P&w tables of compressible flow functionsP&w tables of compressible flow functions
P&w tables of compressible flow functions
 
Default test templates raptor
Default test templates raptorDefault test templates raptor
Default test templates raptor
 
Study and Development of an Energy Saving Mechanical System
Study and Development of an Energy Saving Mechanical SystemStudy and Development of an Energy Saving Mechanical System
Study and Development of an Energy Saving Mechanical System
 
Uc 832 calibration
Uc 832 calibrationUc 832 calibration
Uc 832 calibration
 
Aircraft propulsion performance of propellers
Aircraft propulsion   performance of propellersAircraft propulsion   performance of propellers
Aircraft propulsion performance of propellers
 
Em ii mannual
Em ii mannualEm ii mannual
Em ii mannual
 
Btl control system-lab-manual-10 eel68
Btl control system-lab-manual-10 eel68Btl control system-lab-manual-10 eel68
Btl control system-lab-manual-10 eel68
 
Aircraft propulsion non ideal cycle analysis
Aircraft propulsion   non ideal cycle analysisAircraft propulsion   non ideal cycle analysis
Aircraft propulsion non ideal cycle analysis
 
ENHANCED DATA DRIVEN MODE-FREE ADAPTIVE YAW CONTROL OF UAV HELICOPTER
ENHANCED DATA DRIVEN MODE-FREE ADAPTIVE YAW CONTROL OF UAV HELICOPTERENHANCED DATA DRIVEN MODE-FREE ADAPTIVE YAW CONTROL OF UAV HELICOPTER
ENHANCED DATA DRIVEN MODE-FREE ADAPTIVE YAW CONTROL OF UAV HELICOPTER
 
A Simplified Speed Control Of Induction Motor based on a Low Cost FPGA
A Simplified Speed Control Of Induction Motor based on a Low Cost FPGA A Simplified Speed Control Of Induction Motor based on a Low Cost FPGA
A Simplified Speed Control Of Induction Motor based on a Low Cost FPGA
 
Turbofan Engine Modelling and Control Design using Linear Quadratic Regulator...
Turbofan Engine Modelling and Control Design using Linear Quadratic Regulator...Turbofan Engine Modelling and Control Design using Linear Quadratic Regulator...
Turbofan Engine Modelling and Control Design using Linear Quadratic Regulator...
 
Mobile instrumentation and_data_processing_system_for_testing
Mobile instrumentation and_data_processing_system_for_testingMobile instrumentation and_data_processing_system_for_testing
Mobile instrumentation and_data_processing_system_for_testing
 
Gps waas class 5 nov10
Gps waas class 5 nov10Gps waas class 5 nov10
Gps waas class 5 nov10
 
Automated Testing of C-Band Rotary Field Ferrite Phase Shifter
Automated Testing of C-Band Rotary Field Ferrite Phase ShifterAutomated Testing of C-Band Rotary Field Ferrite Phase Shifter
Automated Testing of C-Band Rotary Field Ferrite Phase Shifter
 
Artigo sea
Artigo seaArtigo sea
Artigo sea
 
space vector
space vectorspace vector
space vector
 
JGrass: the Horton machine (FOSS4G2008)
JGrass: the Horton machine (FOSS4G2008)JGrass: the Horton machine (FOSS4G2008)
JGrass: the Horton machine (FOSS4G2008)
 

Similar to MC and WARR Callibration - Final Update

Gsm Cell Planning And Optimization
Gsm Cell Planning And OptimizationGsm Cell Planning And Optimization
Gsm Cell Planning And Optimization
Yasir Azmat
 
Radio Measurements in LTE
Radio Measurements in LTERadio Measurements in LTE
Radio Measurements in LTE
Sofian .
 
Radio measurements in long term evolution
Radio measurements in long term evolutionRadio measurements in long term evolution
Radio measurements in long term evolution
Vamsy Satish
 
Radiomeasurements for lte
Radiomeasurements for lteRadiomeasurements for lte
Radiomeasurements for lte
Rida sheikh
 
Radio Measurements in LTE
Radio Measurements in LTERadio Measurements in LTE
Radio Measurements in LTE
Praveen Kumar
 
LTE Radio Measurements.pptx
LTE Radio Measurements.pptxLTE Radio Measurements.pptx
LTE Radio Measurements.pptx
Ismail734463
 
FR1.T03.2 Zou_IGARSS_2011.ppt
FR1.T03.2 Zou_IGARSS_2011.pptFR1.T03.2 Zou_IGARSS_2011.ppt
FR1.T03.2 Zou_IGARSS_2011.ppt
grssieee
 
An adjustable Comparator for 2-bit/step SAR ADC Configuring with multiple sam...
An adjustable Comparator for 2-bit/step SAR ADC Configuring with multiple sam...An adjustable Comparator for 2-bit/step SAR ADC Configuring with multiple sam...
An adjustable Comparator for 2-bit/step SAR ADC Configuring with multiple sam...
Association of Scientists, Developers and Faculties
 
edas.stamped_e-1569991431
edas.stamped_e-1569991431edas.stamped_e-1569991431
edas.stamped_e-1569991431
Rushabh Mehta
 
G010414852
G010414852G010414852
G010414852
IOSR Journals
 
Monitoring Open Pit
Monitoring Open PitMonitoring Open Pit
Monitoring Open Pit
JOSE ESPEJO VASQUEZ
 
10. Calibration.pptx
10. Calibration.pptx10. Calibration.pptx
Synchronous mac protocol study
Synchronous mac protocol studySynchronous mac protocol study
Synchronous mac protocol study
Mirthful Nahid
 
LTE KPI Optimization - A to Z Abiola.pptx
LTE KPI Optimization - A to Z Abiola.pptxLTE KPI Optimization - A to Z Abiola.pptx
LTE KPI Optimization - A to Z Abiola.pptx
ssuser574918
 
Adige modelling
Adige modellingAdige modelling
Thesis Presentation on Renewal theory based 802.15.6 latest.pptx
Thesis Presentation on Renewal theory based 802.15.6 latest.pptxThesis Presentation on Renewal theory based 802.15.6 latest.pptx
Thesis Presentation on Renewal theory based 802.15.6 latest.pptx
ssuserc02c1f
 
M0088
M0088M0088
Object Detection Capabilities of the Bathymetry Systems Utilised for the 2015...
Object Detection Capabilities of the Bathymetry Systems Utilised for the 2015...Object Detection Capabilities of the Bathymetry Systems Utilised for the 2015...
Object Detection Capabilities of the Bathymetry Systems Utilised for the 2015...
Luke Elliott
 
“Ampacimon” real-time dynamic rating system
“Ampacimon” real-time dynamic rating system“Ampacimon” real-time dynamic rating system
“Ampacimon” real-time dynamic rating system
RTE
 
"Use of PMU data for locating faults and mitigating cascading outage"
"Use of PMU data for locating faults and mitigating cascading outage""Use of PMU data for locating faults and mitigating cascading outage"
"Use of PMU data for locating faults and mitigating cascading outage"
Power System Operation
 

Similar to MC and WARR Callibration - Final Update (20)

Gsm Cell Planning And Optimization
Gsm Cell Planning And OptimizationGsm Cell Planning And Optimization
Gsm Cell Planning And Optimization
 
Radio Measurements in LTE
Radio Measurements in LTERadio Measurements in LTE
Radio Measurements in LTE
 
Radio measurements in long term evolution
Radio measurements in long term evolutionRadio measurements in long term evolution
Radio measurements in long term evolution
 
Radiomeasurements for lte
Radiomeasurements for lteRadiomeasurements for lte
Radiomeasurements for lte
 
Radio Measurements in LTE
Radio Measurements in LTERadio Measurements in LTE
Radio Measurements in LTE
 
LTE Radio Measurements.pptx
LTE Radio Measurements.pptxLTE Radio Measurements.pptx
LTE Radio Measurements.pptx
 
FR1.T03.2 Zou_IGARSS_2011.ppt
FR1.T03.2 Zou_IGARSS_2011.pptFR1.T03.2 Zou_IGARSS_2011.ppt
FR1.T03.2 Zou_IGARSS_2011.ppt
 
An adjustable Comparator for 2-bit/step SAR ADC Configuring with multiple sam...
An adjustable Comparator for 2-bit/step SAR ADC Configuring with multiple sam...An adjustable Comparator for 2-bit/step SAR ADC Configuring with multiple sam...
An adjustable Comparator for 2-bit/step SAR ADC Configuring with multiple sam...
 
edas.stamped_e-1569991431
edas.stamped_e-1569991431edas.stamped_e-1569991431
edas.stamped_e-1569991431
 
G010414852
G010414852G010414852
G010414852
 
Monitoring Open Pit
Monitoring Open PitMonitoring Open Pit
Monitoring Open Pit
 
10. Calibration.pptx
10. Calibration.pptx10. Calibration.pptx
10. Calibration.pptx
 
Synchronous mac protocol study
Synchronous mac protocol studySynchronous mac protocol study
Synchronous mac protocol study
 
LTE KPI Optimization - A to Z Abiola.pptx
LTE KPI Optimization - A to Z Abiola.pptxLTE KPI Optimization - A to Z Abiola.pptx
LTE KPI Optimization - A to Z Abiola.pptx
 
Adige modelling
Adige modellingAdige modelling
Adige modelling
 
Thesis Presentation on Renewal theory based 802.15.6 latest.pptx
Thesis Presentation on Renewal theory based 802.15.6 latest.pptxThesis Presentation on Renewal theory based 802.15.6 latest.pptx
Thesis Presentation on Renewal theory based 802.15.6 latest.pptx
 
M0088
M0088M0088
M0088
 
Object Detection Capabilities of the Bathymetry Systems Utilised for the 2015...
Object Detection Capabilities of the Bathymetry Systems Utilised for the 2015...Object Detection Capabilities of the Bathymetry Systems Utilised for the 2015...
Object Detection Capabilities of the Bathymetry Systems Utilised for the 2015...
 
“Ampacimon” real-time dynamic rating system
“Ampacimon” real-time dynamic rating system“Ampacimon” real-time dynamic rating system
“Ampacimon” real-time dynamic rating system
 
"Use of PMU data for locating faults and mitigating cascading outage"
"Use of PMU data for locating faults and mitigating cascading outage""Use of PMU data for locating faults and mitigating cascading outage"
"Use of PMU data for locating faults and mitigating cascading outage"
 

MC and WARR Callibration - Final Update

  • 1. MemberoftheHelmholtzAssociation GPR MC/WARR Calibration: t0-Correction Analysis of Single-channel vs. Multi-channel Surveys (Final Update) Tuesday, August 16th, 2016 | Kevin Fan
  • 2. MemberoftheHelmholtzAssociation -Ground Wave is identified** in WARR trace, then cross- correlated with the corresponding MC trace; the “MC/WARR time-shift“ is then the time corresponding to the maximum cross-correlation value 2 Context -Previous Work: a) Cross-Correlation Analysis of Baxter‘s July 2015 Selhausen MC/WARR Data* (“t0-correction Update Pres” in FZJ-GPR) -Overall Objective: To determine whether 500MHz multichannel GPR data can be calibrated with CMP data acquired at corresponding locations, via modification of the t0-correction -Based upon intuitive hypothesis that traces of the 2 surveys, at the same locations, are similar but just time-shifted version of one another
  • 3. MemberoftheHelmholtzAssociation 3 Context  Results: -Same procedure is iterated over all other channels at the specific location, then all channels at each and every other location -A general, channel-specific t0-correction could not be established, due to evident heterogeneity in the time-shift at different CMP locations along the 120m line -Mix of time shifts both positive and negative; whereas a systematic shift between MC/CMP data should be only one or the other? -Necessitates focus upon only a few locations for subsequent measurements…
  • 4. MemberoftheHelmholtzAssociation b) Analysis of July 2016 Selhausen MC/WARR Data, including new MC-calibration Measurements (Location=5m) (“MC-Calibration Update” in FZJ-GPR) -Channel 4 produces very low amplitudes, as well as energies (sum of squared amplitudes) relative to other channels -Script MC_calib_processing produces amplitude/energy plots, plots temporal cross-correlation between measurements taken at different times, and wiggle (as opposed to colour traces) -Can choose to exclude any combination of undesirable channels/positions from analysis Context Apply Channel-specific t0-correction
  • 5. MemberoftheHelmholtzAssociation *As determined by manual picking of 1st air-wave arrivals 5 ContextResults: -Calculated expected air-wave arrival times for every channel/location, based on theoretical air wave velocity (0.2997m/ns) -For any particular location on the MC-sled: t0-Correction = Measured arrival time* - Theoretical arrival time -Typically, channels 1/2/3/6 have similar t0-correction values, while channels 4/5/7 are also similar, albeit delayed relative to channels 1/2/3/6 -Chose to exclude Location 1 (at 23cm on MC-sled) for all channels, due to difficulties in manual air-wave picking arising from near-field effects (see next slide)
  • 6. MemberoftheHelmholtzAssociation 6 Rationale for Dumping Position 1 (of all channels) -Too early: t0- correction too small relative to other positions -Too late: t0- correction too large relative to other positions
  • 7. MemberoftheHelmholtzAssociation 7 Calculation of Cross-Correlation Coefficient for Quantitative Measure of Deviation in Traces over Time- Spans of a few seconds (at 5m vs. 110m along profile) -Position 1 removed from analysis -Generally (except for Channel 4), value is >0.97, indicating high degree of temporal self-similarity (over scales of several seconds)
  • 8. MemberoftheHelmholtzAssociation 8 Current Work: Data Comparison of July 2016 Selhausen Measurements Definitions -Full WARR: Data that, aside from skipped traces, encompasses a continuous range of offsets -7 traces WARR (3 forms): a) MC-sled I: Measurements taken with the MC-sled (termed “MC -calibration measurements), during which a single-channel WARR is performed for each and every channel (by varying the Rx-position over all positions in the MC-sled, for each channel) b) MC-sled II: Full WARR in MC-sled, that has later (via processing) been reduced to the subset of 7 traces corresponding to MC positions c) PE-sled: Same as (b), but with PE-sled -PE-sled: Measurements taken with the Pulse Ekko single-channel sled -MC-sled: Measurements taken with the custom multichannel sled -Channel: A particular transmitter/receiver combination (eg. Tx/Rx4) -Position/Location: The specific section into which Rx is placed
  • 9. MemberoftheHelmholtzAssociation 9 Current Work: Data Comparison of July 2016 Selhausen Measurements Field Conditions/Measurements -See spreadsheet Selhausen-04-07-MC in FZJ-GPR for detailed information regarding data files -Field Measurements taken on 04/07/2016 with Manuela, Anja, Jan; survey design by Manuela -Weather: Hot, sunny day with no rain -DVL battery was running low for later measurements (ie. during CMP’s, 110m MC-calibration) Data Processing Considerations -Location of Interest: 5m along 120m profile (where 0m is the side closest to the street) -Only performed all survey methods with Channel 1; Channels 2-6 were only used for MC- calibration measurements at 5m and 110m, as well as for MOG of the entire 120m profile -All traces dewowed (calls script dewow_anja_eso.m), ungained, and normalized to max(abs(‘trace’))=1, to allow for comparison between MC/WARR data -RPE_Evan.m indicates no skipped traces for MC-calibration data -Output/Plots generated by script MC_calib_processing.m in main folder -Semblance Analysis Parameter Space: a) Ground Wave - t0 ∈ (0,5) ns - v ∈ (0.05,0.15*) m/ns b) Air Wave (newly implemented) - t0 ∈ (1.5,5.0) ns - v ∈ (0.25,0.30) m/ns a)/b) Parameters for both Air/Ground Waves *Time width = 24ns (arbitrary..?)
  • 10. MemberoftheHelmholtzAssociation 10 Script Documentation GPR_Summer_2015_MC - Main script for MOG MC data along profile; must specify data folder + line (only 1 permitted) -Accounts for skipped traces by removing them, then compressing the data-set -Only need to do air-wave picks once; upon completion, will save and use picks in future To redo picks, delete t0AirOffsets.mat Velocity_Analysis_MC_Automated- Called by above - main script for automated ground- wave picking from MC-data (semblance takes in as input each ‘7-trace WARR’ taken by the MC system); includes t0-correction and no smoothing MC_calib_processing – Script to process 7-trace WARR (MC-sled I) data -Can choose whether to do manual air-wave picks or automated ones (ie. air- wave semblance); ground-wave picking automated via semblance -Can exclude any combination of channels/positions from analysis in section ‘Data Parameters’, sub-section ‘MC-calibration Parameters’ Velocity_Analysis_MC_Automated_calib – Similar to Velocity_Analysis_MC_Automated, but for 7-trace WARR (MC-sled I) data Velocity_Analysis_MC_Automated_calib_wiggle – Same as above, visualized as wiggle-traces rather than colour plots Velocity_Analysis_MC_Automated_calib_wiggle_airground – Same as above, with automatic air-wave semblance picking instead of manual air-wave picking
  • 11. MemberoftheHelmholtzAssociation 11 Script Documentation GPR_Summer_2015_WARR – Main script for WARR data (of form MC_sled II or PE_sled) -Smooths data (5-point moving average) Saves air-wave picks; to redo, delete xairpicks or tairpicks (deleting either one will re-write both) vel_ana_TE_CO_CMP_Multi_Automated – Called by above; automatic pick of maximum semblance to determine ground wave vel_ana_TE_CO_CMP_Multi_Comp – Similar to above, but with manual pick of maximum semblance (Igor has incorporated into his scripts) CMPandMC_correlation – Main script used previously for cross-correlation analysis of Baxter’s July 2015 Selhausen data PlotAllSelectTraces – Plots the 3 types of 7-trace WARR surveys (see (a)/(b)/(c) in ‘Definitions’) at all offsets
  • 12. MemberoftheHelmholtzAssociation 12 PE-sled (Source Data: Folder 19, LINE00.DT1) MC-sled (Source Data: Folder 19, LINE55.DT1) Full WARR, pre-t0-correction, gained: PE vs. MC Sleds  Which to pick as Air-wave Arrival: High-frequency or low-frequency content?
  • 13. MemberoftheHelmholtzAssociation 13 -t0-correction now applied, based on direct air-wave pick (value displayed at bottom of plots) -Semblance analysis used to (attempt to) identify ground wave (bolded line on plots) T0-correction + Corresponding Semblance (High-freq. air picks) PE-sled MC-sled
  • 14. MemberoftheHelmholtzAssociation 14 Analogous plots for Low- Freq. air picks PE-sled MC-sled -Difference in t0- correction when choosing high vs. Low-freq. Air wave is ~6ns for both PE/MC-sleds -PE-sled pick appears ~correct (characterizing the ground wave minimum) -MC-sled pick in-correct; ostensibly requires inclusion of negative intercept-time in semblance parameter space?
  • 15. MemberoftheHelmholtzAssociation 11:05 15 Full WARR PE-sled, at Different Time, t0-corrected/gained 11:09 12:56 -Compare whether picking high vs. low frequency air-wave has effect upon temporal variation of t0 (over the course of the day) High-frequency Picks
  • 16. MemberoftheHelmholtzAssociation 11:05 16 Full WARR PE-sled, at Different Time, t0-corrected/gained 11:09 12:56 -t0 values larger for low-frequency picks Low-frequency Picks
  • 17. MemberoftheHelmholtzAssociation 17 -For low-frequency picks, max semblance (and associated values of t0/v) are far more consistent, than amongst high- frequency picks  Implies that minimization of temporal variability requires picking of low-frequency air- waves, even if earlier high- frequency content is present Corresponding Semblance Plots High-frequency picks Low-frequency picks 11:05 11:05 11:09 11:09 12:56 12:56
  • 18. MemberoftheHelmholtzAssociation 18 -t0-correction now relies upon (newly implemented) air-wave semblance, replacing manual picking General Proviso: Automatic air-wave picking requires fine- tuning of parameter space! -Chosen intercept time from air-wave semblance typically too small (ie. Needs to be later to accurately plick 1st- break) -Optimal parameter space here is: v ∈ (0.25* , 0.30) m/ns t ∈ (0.18 , 5) ns -No set of parameters tested thus far has resulted in 100% accurate picking: typically there is a pattern in which though most channels are fine, a few channels will have their automatic pick being too early, for all locations at such channels (for instance, the plots on Slide 20) 7-trace WARR (MC-sled): Gained vs. Ungained
  • 19. MemberoftheHelmholtzAssociation 19 Gained Air Wave, Traces Gained Ground Wav Traces Air Wave Semblance (of gained data) Ground Wave Semblance (of gained data)
  • 20. MemberoftheHelmholtzAssociation 20 Ungained Ground Wave Traces Ungained Air Wave Traces Ground Wave Semblance (of ungained data) Air Wave Semblance (of ungained data)
  • 21. MemberoftheHelmholtzAssociation 21 -Gaining of 7-trace WARR (MC-sled) data results in ambiguous semblance plot (very high-values across parameter space, no clear maximum), while ungained data contains several peaks Observations
  • 22. MemberoftheHelmholtzAssociation Plotted in ‘PlotAllSelectTraces.m’ 22 Qualitative Comparison of Different GPR Survey Methods
  • 23. MemberoftheHelmholtzAssociation 23 Effect of Running the Cables over the Centre vs. Side of the MC-sled (gained, normalized, non-t0 corrected) Centre Side
  • 30. MemberoftheHelmholtzAssociation 30 Observations -Effect of cable position appears significant, and may affect accuracy of both air and ground wave picks though it appears to influence the presence of early -Channel 1: Plots for both positions identical upon visual inspection -Channels 2-6: 1st-break air wave arrival appears to be delayed when cables on side, relative to those for the cables on centre -Can further investigate with cross-correlation, see if cable position has influence on semblance results, etc..  Require a ‘cable-holder’ that will allow for centre position to be employed?
  • 31. MemberoftheHelmholtzAssociation Discussion: Sources of Uncertainty a) Skipped Traces: Typically a few skips per CMP – if too many skips, can distort linearity of air/ground wave, and therefore distort the t0 obtained from semblance -Current method used in Ekko2Dread2_Update_test removes skipped traces, then collapses data to remove the holes -Offset information in header files has been corrected, to account for skipped traces -However, the data is not located in the correct position (offset), due to collapsing approach -Proposed Solution: Rather than collapsing data, fill in holes via interpolation (eg. spline) -See “On Recovering Missing Ground Penetrating Radar Traces by Statistical Interpolation Methods” (Safont et. al. 2014) for inspiration? b) Temporal Variation of t0 even for the same Survey Type*: Across multiple measurements over the same day of the same thing (Full WARR at 5m in PE- sled), is variable (see next slide) -Possibilities: Instrumental temporal drift, use of high-frequency air waves during manual picking 31
  • 32. MemberoftheHelmholtzAssociation 32 Conclusions Analysis of Channels in the Multichannel GPR System -Channel 4 is faulty (very low amplitude/energy), rendering air-wave picking (and consequently t0-correction) highly imprecise -Channels 5/7 have sufficient energy, but on average have larger t0-values than channels 1/2/3/6 -The channel-specific t0-correction may be dependent upon the specific subsurface conditions at the location (eg. high-conductivity), as implied by the fact that the 110m calibration measurements showed that even for different positions of the same channel, t0 was variable; while calibration at 5m produced much more consistent results -Attempted CMP measurements with the Multiplexer failed -Placement of the MC antenna cables on the middle vs. side of the sled appears to have a significant effect for Channels 2-6, with the side positioning resulting in a delayed air-wave 1st-break
  • 33. MemberoftheHelmholtzAssociation 33 Conclusions Semblance Analysis -Semblance analysis breaks down (manifested by non-unique maxima, non-physically reasonable parameters chosen,etc.) under circumstances such as: -Wave non-linearity (whether due to skips or otherwise) -Use of only selected traces (eg. In MC) -Gaining of data (for 7-trace WARR with the MC-sled) -Generally, the Semblance Parameter space must be tested for each data-set -It is not always clear that 5ns is optimal as the maximum possible intercept-time for 500MHz data; as max semblance values sometimes occur at the edge of the parameter space -Time-width (current default value of 24ns seems arbitrary) Temporal Variation -Temporal variation in data on the scale of several seconds is minimal, as indicated by cross-correlation calculations; regarding diruanl variations, the t0-correction value is also consistent, but only provided that the low-frequency air-wave arrivals are chosen
  • 34. MemberoftheHelmholtzAssociation 34 1. Apply the Channel-1 t0-calibration results to other channels; repeat semblance, see if results are consistent across channels 2. Apply channel-specific t0-corrections (calculated from MC_calib_processing.m) to the 120m profile MOG data, then see if ground-wave semblance values become comparable with WARR data taken at same positions 3. Perform 7-trace WARR (MC-sled) measurements at locations with a range of conductivities, so as to better characterize the degree to which the consistency of the channel-specific t0-correction is affected by specific subsurface conditions at each location Future Directions
  • 35. MemberoftheHelmholtzAssociation 35 4. Implement potential Semblance Solutions, such as: -Application of Hilbert Transform to 7-trace WARR (PE-sled) may be an avenue for improvement of semblance and reduction of non- uniqueness of maxima -Investigation of the influence of Time-width (‘window’) on max semblance picks -Include negative time values in semblance parameter space? 5. Investigate proper treatment of skipped traces, via interpolation 6. Use filtering to improve air-wave picking, such as band-pass, to remove high-frequency content sometimes seen, as well as low- frequency waves that occur before the air wave 7. Construct/Upgrade MC-sled such that all antenna cables may be run through the centre of the sled Future Directions