Virtual Narrowband NWA calibration from Wideband Error Term Data

data-Based Wideband NWA Calibrations

Presentation Concepts by Stephen Nibblett Jan 16, 2013

(Project Kick-Off Presentation Originally created Sept 17, 2010)

“dBWCAL” CONCEPT

Wide-Band Characterization at

Variable Sample Rates Established

Wide-Band Data Table

Established

(Custom Table per ATE)

Error Terms

800

900

2100 1900

Network Database

Product Test Frequency Ranges and Power Levels Determined

1) Network Setup

• Exact Test Ranges and Accuracies identified and cataloged, kept updated

• Custom Wideband RF Paths Error Term data set instance created (per site)

• Note: sample point densities are intentionally NOT uniform across wideband range, customized as needed!

NWA Measurement Of Cal Standards

(USB Controlled E-Cal)

ATE Rack Characterization

App

Network Database

Save Wide-Band “Custom” Data Table

2) ATE Rack Characterization

• Custom Wideband RF Paths Error Term data set measured at test site

• Characterization Intervals and Times scheduled “Offline” from Production

Product Narrow-Band Test Frequency Ranges

NWA Calibrations

Created at ATE Rack from Custom

Error Terms

- Product Selection

- Station Selection

- NB Error Term Interpolations

Management Test Plan App

(Local or Remote)

Network Database

Saved Wide-Band “Custom” Data

Table

3) ATE Rack Setup and Usage

• Product Line Management plans product(s) to be tested at ATE site for shift

• Custom Interpolated Error Terms created for required site NWA calibrations

• ATE software engine pulls Error Terms and saves local NWA calibrations

•6

“dBW CAL” METHOD

EXISTING CAL Methods: Cal Concept

• What Is Contained In A Cal File? o The Error Corrections!

NWA DUT RF

PATHS

System Errors

• Internally Calculated Corrections:

• NWA based processing

• Based on assumed “known” coeff values:

• “Fitting” errors present

• “Actual” coeffs maybe not correct!

• STNDs re-connected for every cal/path:

• redundant actions, time wasted

• poor repeatability (cable moves)

• Minimum Correction Solution:

• 3 unknowns; 3 standards (S,O,L)

OLD CAL Method: Manual SOL Cal

Short coeff

Open coeff

Load

Thru

Expected

NWA

Meas Short

Meas Open

Meas Load

Meas Thru

Actual

OLD CAL Method: Manual SOL Cal

NWA MEAS RF

PATHS

System Errors

Meas STNDs

STND Coeffs NWA CPU NWA MEM

NWA CPU

Compute System Errors

NWA Cal

Estimated

RAW

Measured

Correction

File

Meas Frequency

Information

• Externally Calculated Corrections:

• Network based processing

• Based on assumed “known” coeff values:

• “Fitting” errors present

• “Actual” coeffs maybe not correct!

• STNDs connected once for every cal/path:

• reduced actions, time saved

• better repeatability (less cable moves)

• Minimum Correction Solution:

• 3 unknowns; 3 standards (S,O,L), STILL OLD TECHNOLOGY!!

OLD CAL Method: ATE “Optimized” Cal

Short coeff

Open coeff

Load

Thru

NWA

Meas Short

Meas Open

Meas Load

Meas Thru

Network

OLD CAL Method: ATE “Optimized” Cal

NWA MEAS RF

PATHS

System Errors

Meas STNDs

STND Coeffs FELIX SQL TABLE

FELIX


NWA Cal

Estimated

RAW

Measured

Correction

File

Meas Frequency

Information

RAW

MEAS

• Internally Calculated Corrections:

• NWA based processing

• Based on known data values (w/ uncertainty):

• “Fitting” errors NOT present

• Broadband “Actual” data is well known

• ECAL re-connected for multi-port (>4) cals:

• redundant actions, time wasted

• ok repeatability (but cables still moved)

• Over-Determined, Data-Based Correction:

• 3 unknowns; 4 imp states (Z1, Z2, Z3, Z4), MORE MODERN!!!

NEWER CAL Method: Agilent E-Cal

Z1 Data

Z2 Data

Z3 Data

Z4 Data

NWA

Meas Z1

Meas Z2

Meas Z3

Meas Z4

ECAL

NEWER CAL Method: Agilent E-Cal

NWA MEAS RF

PATHS

System Errors

Meas STNDs

STND DATA NWA CPU NWA MEM

NWA CPU


NWA Cal

ACTUAL

RAW

Measured

Correction

File

Meas Frequency

Information

dBWCAL Method: Network dB Cals Externally Calculated Corrections:

• Network based SQL table processing, using modern techniques (WLS)

• Broadband Cal Set Data is stored, used “on-demand”

Based on known data values: Data-Based Cal (gathered round-robin):

• “Fitting” errors present, but always corrected

• “Actual” data is acceptably correct, with uncertainty well known

MULTI-PORT CAL SET (w/ECAL) connected once for multi-port (>4) cals:

• NO redundant actions, NO time wasted, and OFFLINE OPERATION!

• BEST repeatability (cables moved only once); can also incorporate test jig!

Over-Determined Correction Solutions:

• 3 unknowns; 4 (or more) imp states (Z1, Z2, Z3, Z4)

dBWCAL Method: Network dB Cals

NWA

Network Cal Set (dBCAL)

Stored Data Table:

:ECAL (or MANUAL) “Offline”

Stored Data Table:

Meas Data dBCal

:Broadband

:Multi-path

:Multi-pwr

“Online”

USER select

typecodes

Network

generates ONLY

needed corrections:

interpolated / interpreted

from unique instance of

Stored Table: Meas Data dBCAL

Gold Stnd

checks cals

“Offline”

Gold Stnd

checks cals:

no re-cal

if acceptable

“Round Robin”

Data Analysis

1

1a

1b

1b

2a

2b

3a

3b

3c

dBCAL Method: Network dB Cal “offline”

NWA MEAS RF

PATHS

System Errors

Meas STNDs

STND DATA ATE Engine SQL TABLE

SQL TABLE

ACTUAL

BROADBAND

STND DATA

RAW BENCH

BROADBAND

MEAS DATA

Meas Frequency

Information

RAW

MEAS

SAVED FOR DESIGNATED LIMITED LIFETIME OF “VALID” ONLINE USAGE…

INTERPOLATE ERRORS

dBCAL Method: Network dB Cal “online”

Network APP SELECT

TYPECODE

ATE Engine SQL TABLES

Narrowband NWA

Correction Files

Meas

Information

NWA Cal

BENCH

BROADBAND

ERROR DATA

INTERPOLATED

NARROWBAND

CORRECTIONS

TRANSLATE FOR NWA

MEAS

GOLD STND

USER

MEAS

TYPECODE

BENCH

BROADBAND

ERROR DATA

INTERPOLATED

NARROWBAND

ERROR DATA

dBWCAL Phase II: Interpolated Cal Experiment

dBWCAL Method: Projected Goals

• Store Wideband Bench Characterization o Requires “Relatively Flat” Bench Response

o OR Requires Sufficient # Of Data Points

• Retrieve Narrowband Requirement o Define Necessary DUT Measurement Parameters

• Interpolate NB Bench Calibration o NWA & Bench Switch Error Terms Included

• Format & Save NWA CAL FILE

dBWCAL Method: Experiment Goal

• Demonstrate Cal Interpolation Feasibility

• Discover & Define All Critical Requisites

dBWCAL Method: Interpolation Process

• Extract Narrowband Cal Information

From Wideband S2P Cal Files: • Optimally: ENA+Rack Path Characterization

• For Experiment: Calibrate Wideband On ENA

• Interpolate Frequency and Data Points • Optimally: Use Weighted Least Squares

• For Experiment: Use On-Board ENA Interpolation

dBWCAL Method: Experiment “A” Plan

• Calibrate VNA Wideband .1 MHz to 4500MHz o (1601, 6401, 20001) Points (requires 5071C)

o Use Same Cal Kit for Narrow and Wideband

• E-Cal Module is BEST

o Repeat Below (2) Steps For Each Point Setting

• Set Start/Stop to 1830-2000MHz, 401 Points o ENA Automatically Interpolates Calibration: DO NOT RE-CAL!!!

• Gather S2P Files For Narrowband DUT

Then: • Recalibrate ENA Narrowband, Same Parameters

• Gather DUT Narrowband S2P File (Control)

dBWCAL Method: Experiment “A” Reporting

• Using Excel, Plot ALL S2P Data:

o 401 Point ACTUAL NB Calibration “Control”

o 1601 Point WB Cal Interpolated For 401 Points



• Create “Variance” Plots

o Compare Accuracy Of Each Interpolated Cal

o Use “Control” 401 Point Cal Data for Baseline

dBWCAL Method: Experiment “A” Results

• S11 Log Mag Data Correlation: Excellent o Note: Return Loss Differences Past 30dB Is Insignificant

Accuracy

Critical

Don’t Care!


• S11 Log Mag Passband Data Correlation: Excellent o Note: Range Differences for Return Losses Past 30dB Is Insignificant

• Baseline Variation Plot Of (TX) Passband: o Maximum Variation Occurs At Points Past 30dB Return Loss

o Note: Variation Includes Test Cable Connection Repeatability

• S11 Phase Passband Data Correlation: Excellent o Note: MOST Rejection Variation Due to NO AVERAGING & Hi IF BW


• S11 Phase Passband Data Correlation: Excellent o Note: Return Loss Differences Past 30dB Is Insignificant




• S22 Log Mag Data Correlation: Excellent o Note: Range Differences for Return Losses Past 30dB Is Insignificant


• S22 Log Mag Passband Data Correlation: Excellent o Note: Range Differences for Return Losses Past 30dB Is Insignificant




• S22 Phase Passband Data Correlation: Excellent o Note: MOST Rejection Variation Due to NO AVERAGING & Hi IF BW


• S11 Phase Passband Data Correlation: Excellent o Note: Range Differences for Return Losses Past 30dB Is Insignificant




• S21 Log Mag Data Correlation: Excellent o Note: MOST Rejection Variation Due to NO AVERAGING & Hi IF BW


• S21 Log Mag Band Edge Data Correlation: Excellent o Note: MOST Rejection Variation Due to NO AVERAGING & Hi IF BW

o ALL Measurements are Same #Points, IF BW, Power, Freq Span

• Correlation Plot Of Low-Side Attn Slope / Band Edge:


• S21 / S12 Log Mag Slopes Data Correlation: Excellent o Note: MOST Rejection Variation Due to NO AVERAGING & Hi IF BW

o Measurements are Same #Points, IF BW, Power, Freq Span

• Baseline Variation Plot Of Low-Side Attn Slope / Band Edge: o Note: Variation Includes Test Cable Connection Repeatability


• S21 Log Mag Passband Data Correlation: Excellent o ALL Measurements are Same #Points, IF BW, Power, Freq Span

• Baseline Variation Plot Of Passband: o Note: Variation Includes Test Cable Connection Repeatability


• S21 Phase Data Correlation: Excellent o Note: MOST Rejection Variation Due to NO AVERAGING & Hi IF BW


• S21 Phase Passband Data Correlation: Excellent o ALL Measurements are Same #Points, IF BW, Power, Freq Span



• S12 Log Mag Data Correlation: Excellent o Note: MOST Rejection Variation Due to NO AVERAGING & Hi IF BW


• S12 Log Mag Band Edge Data Correlation: Excellent o Note: MOST Rejection Variation Due to NO AVERAGING & Hi IF BW

o ALL Measurements are Same #Points, IF BW, Power, Freq Span

• Correlation Plot Of (TX) Low-Side Attn Slope / Band Edge:



• S21 / S12 Log Mag Slopes Data Correlation: Excellent o Note: MOST Rejection Variation Due to NO AVERAGING & Hi IF BW

o Measurements are Same #Points, IF BW, Power, Freq Span

• Baseline Variation Plot (TX) Low-Side Attn Slope /Band Edge: o Note: Variation Includes Test Cable Connection Repeatability


• S12 Log Mag Passband Data Correlation: Excellent o ALL Measurements are Same #Points, IF BW, Power, Freq Span



• S21 Phase Data Correlation: Excellent o Note: MOST Rejection Variation Due to NO AVERAGING & Hi IF BW


• S12 Phase Passband Data Correlation: Excellent o ALL Measurements are Same #Points, IF BW, Power, Freq Span



dBWCAL Method: Experiment “A” Summary

• 1601 Points o Adequate Accuracy For Most Passband Measurements

• Passband Transmission Log Mag Measurements

• Passband Transmission Phase Measurements

o Adequate Accuracy For Most Attn Slope Measurements • Attenuation Slope Transmission Log Mag Measurements

• 6401 Points o Most Results Very Similar to 1601 Points

• 20001 Points o Most Accurate Tracking For “Sensitive” Measurements

• Transmission Rejection Log Mag Measurements

• Passband Reflection Phase Measurements

• WB-NB Interpolated Calibrations Are

Valid

• Initial Recommendations:

o Perform Wideband Bench Characterizations

o Use 20001 Points For All Characterizations

• Before Recommendations Are Final o Re-Check 1601, 6401 Interpolations, for speed:

• For “Sensitive” Meas Use AVG, Lower IF BW

dBWCAL Method: Experiment “A” Conclusion

•46

“dBW CAL” CONCEPT: To Be Continued?

Engineering

Virtual Narrowband NWA calibration from Wideband Error Term Data