killing the bode plot final - picotest · title: microsoft powerpoint - killing the bode plot...
TRANSCRIPT
TITLE
Killing the Bode Plot
Image
Killing the Bode PlotSteve Sandler (Picotest) ImageSteve Sandler, (Picotest)
Copyright © 2016 Picotest.com, All Rights Reserved
Killing the Bode PlotKilling the Bode PlotSteve Sandler, PicotestSteve Sandler, Picotest
Copyright © 2016 Picotest.com, All Rights Reserved
SPEAKERSSteve Sandler
Managing Director, Picotest@ i | i | @ [email protected] | picotest.com | @stevenmsandler
Steve Sandler has more than 35 years’ experience in the design, analysis, and troubleshooting of power conversion and system level equipment for commercial military and space applicationsequipment for commercial, military, and space applications.
Steve is the CEO of Picotest.com, a company that designs and distributes test equipment accessories designed for test and troubleshooting power systemstroubleshooting power systems.
Steve is also the founder and chief engineer of AEi Systems, a leading analysis and modeling company specializing in worst case analysis of high reliability systemshigh reliability systems.
Steve has authored several books related to power electronics including most recently the book entitled "Power Integrity - Measuring, Optimizing and Troubleshooting Power Systems "and Troubleshooting Power Systems.
Copyright © 2016 Picotest.com, All Rights Reserved
The Bode Plot’s Days are Numbered …..
R.I.P.phase
gainThe worlds most popular stability assessment test
Rest in Peace
1947‐201?
Copyright © 2016 Picotest.com, All Rights Reserved
So What’s Wrong with Bode Plots
Wh ’ The first problem is that the Bode plot is a POORindicator of RELATIVE stability – this is well known and well published It means that Bode
What’s RIGHT known and well published. It means that Bode
Plot results can be inaccurate and misleading.
A th bl i th t th t
RIGHTmight be Another problem is that the measurement
requires us to inject a signal into the loop. Many devices simply don’t provide access to the loop.
might be a better
We’ll show some examples…questionCopyright © 2016 Picotest.com, All Rights Reserved
Stability STILL Matters – A LOT
Closed Loop Performance is directly related to the Stability Performance.
That is why you DO need to assess Stability, maybe even more than ever – it’s essential to optimum, low noise, performance.
bili i ’ i l hStability assessment isn’t going away, only the Bode plot will, fortunately, better methods are currently availablecurrently available……
Copyright © 2016 Picotest.com, All Rights Reserved
Why are Bode Plots Becoming IrrelevantIn other cases we might not have physical1. Poor or No Control Loop Access
3.3V/1.5A LDOIn other cases, we might not have physical space to inject into the loop or their might be a large number of power supplies
p
Fixed regulators/references don’t
6 output power supply
Fixed regulators/references don t allow access to the control loop 8.5mm
Or we might not want to gain access because it i i i
8.5mm More power supplies
requires cutting a trace or a wire
Copyright © 2016 Picotest.com, All Rights Reserved
High Frequency Injection is Impractical
500MHz Bandwidth eGaNLinear Regulator
Copyright © 2016 Picotest.com, All Rights Reserved
Why are Bode Plots Becoming Irrelevant
2. Many new devices have multiple internal loops
All need to be accounted forAll need to be accounted for–Digital Control Loops
Copyright © 2016 Picotest.com, All Rights Reserved
Why are Bode Plots Becoming Irrelevant
3. You can have both good phase margin and good gain margin and yet have poor stabilitypoor stability– There may be multiple crossovers– Control loops can be high order
SM = Stability MarginPM = Phase MarginGM = Gain Margin
p g– Multiple or non‐linear loops
SM provides the most accurate measure of stability
ONLY this one point is UNSTABLEONLY this one point is UNSTABLE
Copyright © 2016 Picotest.com, All Rights Reserved
What is RELATIVE Stability & Why Bode Plots Can Mislead
A STABLEt lcontrol
loop is This is the gain/phase vector
Unity gain negative feedback results in a denominator ofpone that doesn’t
Unity gain negative feedback results in a denominator of zero = unstable
The minimum magnitude of the denominator provides the doesn t oscillate
g pRELATIVEmeasure – also known as the ‘Stability Margin’
Nyquist is a better assessment tool for Stability than Bode
Copyright © 2016 Picotest.com, All Rights Reserved
A DC-DC Converter Example What’s Interesting
• Bode & Nyquist ‐ same data
1 210 100
Bode & Nyquist same data
• Bode didn’t indicate stability
• Scope Owners: load step Bode
Nyquist
0
5
R1/
dB
0
50
TR2
ringing may or may not indicate a stability issue
• Must use a Closed Loop Test Nyquist Impedance etc
Bode
-5
TR
-50
2/° Nyquist, Impedance, etc.
Min. Distance from (1,0) Load Step
103 104-10 -100
f/HzTR1: Mag(Gain) TR2: Phase(Gain)
78 degrees phase margin6.7dB gain margin
0.38
But it Rings Severely
Copyright © 2016 Picotest.com, All Rights Reserved
Output Impedance is as Good as Nyquist12 gain 15 phase PM = 50 Deg, 658 Hz
I d
20.0
40.0
volts
)
90.0
180
gree
s
frequency = 851 hertzphase = 21.3 degrees
BW = 658 hertz, PM = 50.2 degrees Impedance Peak, 860Hz
-20.0
0
gain
in d
B(v
-90.0
0
phas
e in
deg
Plot
2
15frequency = 851 hertzgain = -3.85 dB(volts)
x = 1.09k hertz, GM = 8.70 dB(volts) 500m
1.00
st
10 20 50 100 200 500 1k 2k 5k 10kfrequency in hertz
-40.0-180
12
, ( )
-2.00 -1.00 0 1.00 2.00real
-1.00
-500m
0
imag
Nyq
uis
16 frequency = 851 hertzreal = 598mimag = 232m
Ringing at 860Hz
Distance from (1,0)
200m
600m
1.00
1.40
1.80
dist
ance
Dis
t117
y (min) = 464.318m x = 851.138 hertzFlyback
Converter Flyback Converter Nyquist
851Hz100 200 500 1k 2k 5k 10k
frequency in hertz
Copyright © 2016 Picotest.com, All Rights Reserved
So How Can We Easily and Accurately Assess Stability?
Non‐Invasive Stability Measurement – ‘NISM’
No Control Loop – No ProblempStability Margin can be determined from a single output impedance measurement!
Copyright © 2016 Picotest.com, All Rights Reserved
The NISM Concept is Simple
Yes. You can measure phase
From 1 Impedance Measurement ‐ 3 data points– Impedance Magnitude, Phase, and Group Delay
margin from a single measurement
The NISM software extracts data from output impedance and the ‘Q’ via group delay and allows the ZS and ZL to be mathematically determined so they can be converted tomathematically determined so they can be converted to Stability Margin
Phase margin is determined by setting |Tm| = 1 and solving for phase
Copyright © 2016 Picotest.com, All Rights Reserved
What is NISM and Why It’s a Critical TechnologyNyquist solves our problem but is not always simple to computeNyquist solves our problem, but is not always simple to compute
NISM accomplishes the same thing more simply thru closed loop output impedance
NISM determines control loop margins without needing access to the feedback loop
NISM is based on proven Minor Loop Gain Theory
– Similar to assessing Load Step ‘Q’
The NISM technology is licensed and promoted by Picotest.com
The measurement capability can be found on various VNAs
– OMICRON Lab, Keysight, Copper Mountain
Copyright © 2016 Picotest.com, All Rights Reserved
NISM is Based on Fundamentals
D R D Middl b k l i d th t i fMi L G i T ithDr. R.D. Middlebrook popularized the topic of Minor Loop Gain, Tm with his introduction of the extra element theorem which allowed us to assess the stability of power supplies and input filters Minor loop gain, based on Nyquist criteria, is now one of the most researched electronics topicsMany articles can be found with an internet search of “forbidden regionMany articles can be found with an internet search of forbidden region stability criteria”
Copyright © 2016 Picotest.com, All Rights Reserved
How Do I Make a NISM Measurement
Step 1: Measure ImpedanceStep 2: Set Cursors
Practically speaking, output impedance is measured with a suitable probe in a 1 or 2 port configurationprobe in a 1 or 2 port configuration
The software converts the impedance to group delay and Q
The user positions waveform cursors on the impedance and Q waveforms and the conversion to phase margin is
d h ’read out on the instrument’s screen
Copyright © 2016 Picotest.com, All Rights Reserved
NISM on the OMICRON Lab Bode 100
Copyright © 2016 Picotest.com, All Rights Reserved
NISM on the Keysight E5061B
Copyright © 2016 Picotest.com, All Rights Reserved
NISM – Just as Accurate as a Bode Plot
Accurate to within 1 degreeAccurate even when Bode Plots are notAccurate up to 65 degrees
Lab comparisons of 7 regulators/capacitorsBode plot and the NISM methods were within +0.9 degrees
Noise free simulation results were within +0.3 degrees, attributable to cursor resolution
Copyright © 2016 Picotest.com, All Rights Reserved
LT1086 Linear RegulatorVout
Min = 25mANom = 500mA
X2RH1086
1 80V
R1249
ADJ
C10.1uF
C20.1uF
Vin5
Vout
CoutCinT409G226M015
+/-0.5V
I125m
Vin
3.28%45.24% 45.24%
Max = 1AIN OUT
ADJUST I2AC = 1
f 1 Vf req1
1.80V
557mV
5.00V
NISM Bode Plot PM
Measured 59 deg 56 deg
R2110
Cadj68n
3.28% 5.87%B1Currentmag(f req)/(2*3.14159265)
R31
f req1 Vf req10V
12010
10-1
100
TR1 TR
2
1
0
50
100
0
100
200
R1/
dB
TR2/
f/Hz TR1 TR2 PM: Cursor 1 52.580k 325.128m 504.129m 58.904 °Cursor 2 40.777k 302.517m 597.409m
C2-C1 -11.803k -22.611m 93.280m
-110
10-3
10-2
102 103 104 105 106 107
T 2
f/H
f/Hz TR1/dB TR2/°Cursor 1 52.743k 0.000 55.467
-100
-50
-200
-100
102 103 104 105 106 107TR
/°
f/Hzf/HzTR1: |Mag(Gain)| TR2: |QTg(Gain)|
f/HzTR1: Mag(Gain) TR2: Phase(Gain)
Copyright © 2016 Picotest.com, All Rights Reserved
LP2998 DDR Termination Regulator
29.43 deg
28.9 deg
Copyright © 2016 Picotest.com, All Rights Reserved
TPS7A4501 Linear Regulator 5mA
NISM Bode Plot
Measured 51 deg 53 deg
A low‐noise, fast‐transient‐response 1.5‐A low‐dropout (LDO) voltage regulator
htt // ti /lit/ / l 259 / l 259 df
100 40
60
150
http://www.ti.com/lit/ug/slvu259a/slvu259a.pdf
010
10-2
10-1
100
TR1 TR
2
40
-20
0
20
50
100
TR1/
dB
TR2/°
f/Hz TR1 TR2 PM: Cursor 1 6.165k 1.264 811.220m 51.214 °Cursor 2 5.782k 1.259 835.934m
C2-C1 -382.160 -5.214m 24.714m-110
10-3
10
102 103 104 105 106
f/Hz5mA : |Mag(Gain)| 5mA : |QTg(Gain)|
f/Hz TR1/dB TR2/°Cursor 1 7.153k 45.204m 53.384-60
-40
0102 103 104 105 106
f/Hz5mA (2.37k) : Mag(Gain) 5mA (2.37k) : Phase(Gain)
Copyright © 2016 Picotest.com, All Rights Reserved
VRG8666 (RH3080) Linear RegulatorX1VRG8666
NISM Bode Plot Vin
Vout
SET
VcBodeInj
2
43
1
VRG8666
V15
C2330u C3
330uI125m
R219
1 2001210
Measured 20 deg 21 deg R1250k
C11u
1
20
40
60
80
0
50
100
150
R1/
dB
TR2/°-30
-20
-10
0
10-1
100
TR1/
dB TR2
f/Hz TR1/dB TR2/°Cursor 1 14.330k 0.000 21.222-40
-20
0
-200
-150
-100
-50
102 103 104 105 106 107
T
°
f/Hz
f/Hz TR1/dB TR2 PM: Cursor 1 14.818k -21.472 2.546 20.497 °Cursor 2 15.304k -21.510 2.696
C2-C1 485.650 -38.053m 150.007m-60
-50
-40
10-3
10-2
102 103 104 105 106 107
T
f/Hz f/HzTR1: Mag(Gain) TR2: Phase(Gain)
f/HzTR1: Mag(Gain) TR2: |QTg(Gain)|
Copyright © 2016 Picotest.com, All Rights Reserved
High Speed OpampUsing the Keysight E5061B VNA with the NISM softwareUsing the Keysight E5061B VNA, with the NISM software, we were able to test the stability of this 245 MHz op-ampThe big WOW is that we obtained the (very poor) phase margin from the impedance measurement using NISM (just about 2 degrees)g )This is a great capability; to be able to accurately assess stability at 100's of MHz or higher without lifting any wires (which would interfere with the measurement)
T tTest Point
NISM Bode Plot
Measured 2 deg N/A
Copyright © 2016 Picotest.com, All Rights Reserved
Applications for NISM
• Op-amps• Voltage Regulators – including LDOs, POLs, VRMs…. g g g , ,• Voltage References• Audio Amplifiers – including switching typesp g g yp• Input Filter Stability• System Level Box-to-Box Stability• Current Regulator Stability (electronic load, LED, etc.)• Almost any control loop
Copyright © 2016 Picotest.com, All Rights Reserved
A New Path To Stability Testing is Here
Technology, Integration, Shrinking Form Factors Have Forced the IssueReduced the practical application of Bode Plot measurement
The load (impedance and current) are critical to an accurate stability
assessment
What’s needed is an in-circuit measurement
NISM – Non-Invasive Stability Measurement– The final nail in the Bode Plot coffin is the availability of an alternative assessment
– Available on popular VNAs
I l d d ith th I t t il bl ft dd– Included with the Instrument or available as a software add-on
Copyright © 2016 Picotest.com, All Rights Reserved
Thank you!---
QUESTIONS?
Copyright © 2016 Picotest.com, All Rights Reserved