IEEE 802.3 DTE Power via MDI
Detection Methods - Reliability Analysis
Presented by Yair Darshan, PowerDsine - [email protected]
1
IEEE 802.3af, Nov. 2000.
Topics
! Probability of false positive detection
! Definition of “Absolute” and “Behavioral” methods
for signature detection
! Summary
2
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Probability of False Positive Detection
! Objective
Decreasing the probability of falsely identification of non-PD
as a valid PD
3
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
Probability of False Positive Detection
IEEE 802.3af, Nov. 2000.
Comparison between “AC Coupled Diode” and “Resistor+diode” method
A
R1
2K
R2
75K
A
A
D1
D1
V1
R1
25K
R2
100K
B
C1
1uF
D2
B
B
! Target:
Looking for received/non-received
pulses at specified condition.
R3
1MEG to Open
! Target:
Looking for specific value at
specified conditions.
4
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Probability of False Positive Detection.
! Comparison between “Diode” and “Resistor” methods.
A
R2
75K
R1
2K
D1
R2
100K
C1
1uF
A
A
D1
V1
B
R1
25K
R3
1MEG to Open
B
D2
B
! What is the probability to receive
a: the “correct” result when each of the components changes around
its nominal value
b: When all components value change simultaneously?
! What will be the Envelope size in which “correct” detection is
obtained?
! In order to simplify analysis, we will assume that the Envelope size is
simply a function of the product of signal sensitivity to elements
tolerance.
5
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Expected Positive Signature Detection Envelope
" Description of 3 elements signature
scheme with nominal values of P1, P2
P1
and P3 with ∆P1, ∆P2 and ∆P3
around the nominal value
P2
" ∆Pi= ∆(Expected signal) / ∆(Element-i)
i =1,2,3.
P3
" Probability of “False Positive”
Detection (PFP) is Proportional
to Envelope Size (ENS)
Powered DTE
non-Powered DTE
" PFP ∝ ENS
" ENS∝ ∆P1·∆P2·∆P3
6
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
Probability of False Positive Detection
IEEE 802.3af, Nov. 2000.
! Comparison between “Diode” and “Resistor” Methods
A
R2
75K
R1
2K
A
A
D1
D1
V1
R1
25K
R2
100K
R3
1MEG to Open
B
C1
1uF
D2
B
B
ENS(Resistor) = Fx·Fz·∆P1·∆P2·∆P3
!
ENS(Diode) = Fy· ∆C1·∆R1·∆R2
!
!
Fy=Probability to find AC coupled
diode scheme in non-Powered DTE
!
Fx=Probability to find Resistor+Diode
scheme in non-Powered DTE
!
Fy<1
!
!
!
Fx<1
Fz= Positive detection criteria limits
Practically, V1,R2 tolerance can be
ignored by automatic calibration in PSE
∆Pi=∆Vab/∆(Element-i), i=1,2,3
7
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Probability of False Positive Detection
! Comparison between “Resistor” and “Diode” methods
! Practical values (Refer to Annex B for detailed analysis):
R2
75K
A
A
D1
V1
R1
25K
R3
1MEG to Open
B
D2
B
" ∆Vd= ± 20%
" ∆R1= ± 2%
" DR3=1MΩ to open (Leakage current equivalent)
" Fz =Vab Positive detection criteria limits = ± 2%
" Generating:
" ∆Vab/∆Vd = ±3.2% max
" ∆Vab /∆R1 = ±1.2% max
" ∆Vab /∆R3 = -2% max
!
· · · ·
· · ·
·
ENS =(3.2 2) (1.2 2) abs(-2) (2 2) Fx ≈ 123 Fx
8
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Probability of False Positive Detection
! Comparison between “Resistor” and “Diode” methods
!
Measured values (using Nortel prototype)
A
!
Software Option A:
C1=40nF to Unlimited
R1=190R to 14K
R2= from short to open
R1
2K
D1
R2
100K
C1
1uF
B
!
Software Option A+B:
!
R2 is actually not needed in Nortel’s B.B. (See Annex C for more details)
!
ENS = 750000·Fy·∆R2, Simplifying to ENS =750000·Fy
C1=40nF to 4uF
(∆C1=1000%)
R1=1.35K to 14K, (∆R1=750% )
R2= from 2.7K to open.
9
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Sensitivity Analysis - Changing C1 from 0.5uF to 5uF step 0.5uF
1. 246V
1. 000V
Forward Polarity
0. 500V
A
0V
R1
2K
D1
R2
100K
C1
1uF
- 0. 500V
4 . 8 9 5 2 ms
4 . 9 0 0 0 ms
4 . 9 0 5 0 ms
4 . 9 1 0 0 ms
4 . 9 1 5 0 ms
4 . 9 2 0 0 ms
4 . 9 2 5 0 ms
4 . 9 2 9 6 ms
V( R9 : 2 )
Ti me
5 0 0 mV
B
0V
Reverse Polarity
- 5 0 0 mV
"
- 9 9 0 mV
4 . 7 9 4 3 3 ms 4 . 7 9 6 0 0 ms
4 . 8 0 0 0 0 ms
4 . 8 0 4 0 0 ms
4 . 8 0 8 0 0 ms
Refer to Annex A for
detailed circuit model.
4 . 8 1 2 0 0 ms
V( R2 2 : 2 )
Ti me
10
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Sensitivity Analysis - Changing R1 from 0.5KΩ to 5KΩ step 0.5KΩ
1. 56V
Forward Polarity
1. 00V
A
0V
- 0. 97V
4 . 8 8 9 7 ms 4 . 8 9 2 0 ms
4 . 8 9 6 0 ms
4 . 9 0 0 0 ms
4 . 9 0 4 0 ms
4 . 9 0 8 0 ms
4 . 9 1 2 0 ms
R1
2K
D1
R2
100K
C1
1uF
4 . 9 1 6 0 ms
V( R9 : 2 )
7 2 9 mV
Ti me
5 0 0 mV
B
0V
Reverse Polarity
- 5 0 0 mV
"
- 8 2 3 mV
4 . 7 9 5 0 ms
4 . 8 0 0 0 ms
V( R2 2 : 2 )
4 . 8 0 5 0 ms
4 . 8 1 0 0 ms
4 . 8 1 5 0 ms
4 . 8 2 0 0 ms
Ti me
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
4 . 8 2 5 0 ms
Refer to Annex A for
detailed circuit model.
4 . 8 2 9 6 ms
11
IEEE 802.3af, Nov. 2000.
Sensitivity Analysis - Changing R2 from 10K to 800K step 10K
1. 49V
Forward Polarity
1. 00V
A
0V
R1
2K
D1
R2
100K
C1
1uF
- 1. 00V
4 . 8 9 5 ms
4 . 9 0 0 ms
4 . 9 0 5 ms
4 . 9 1 0 ms
4 . 9 1 5 ms
4 . 9 2 0 ms
4 . 9 2 5 ms
4 . 9 3 0 ms
V( R9 : 2 )
Ti me
1 . 0V
B
0V
Reverse Polarity
- 1. 0V
4 . 79 0 8 ms
"
4 . 79 5 0 ms
4 . 8 0 0 0 ms
4 . 8 0 5 0 ms
4 . 8 1 0 0 ms
4 . 81 5 0 ms
4 . 81 9 1 ms
Refer to Annex A for
detailed circuit model.
V( R2 2 : 2 )
Ti me
12
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Comparison of Expected Positive Signature Envelope
" ENS(Diode)/ENS(Resistor) =(750000/123)·Fy/Fx ≈6100·Fy/Fx
C1
AC coupled diode
software algorithm
R1
Diode method and Resistor method
Hazard Matrixes
Vd
" Ratio ≈6100·(Fy/Fx)
R1
R3
R2
Diode method
" ENS=750000·Fy
Resistor method
" ENS=123·Fx
13
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Comparison of Expected positive signature Envelope
! ENS(Diode)/ENS(Resistor) =6100·Fy/Fx
!
What are the values of Fy, Fx?
!
We know from Nortel’s and Lucent’s tests / Hazard Matrix that Fy<1, Fx<1
i.e. the probability to find AC Coupled Diode scheme or “Resistor +Diode”
scheme in non-PD or in non-compatible 802.3af powered DTE is low
!
Further work is required to determined how low Fy and Fx are. However, we
can practically assume that they are with the same order of magnitude
!
In order to match the resistor method, the factor 6100·Fy/Fx should be
approximately 1
14
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
“Behavioral” Signature Detection vs. “Absolute” Signature Detection
! Expecting the “Correct Signal” at a specific set of
conditions is “Behavioral” Signature Detection
" Allows for different implementations
" All Implementations will generate “correct results”
" Insensitive to circuit elements value
! Expecting a “Specific” value at a specific set of
conditions is “Absolute” signature detection
" Allows for a single (theoretically) implementation
" Only one value will generate valid result
" Highly sensitive to elements value change
15
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
“Behavioral” Signature vs “Absolute” Signature Detection
" AC Coupled diode scheme is of a Behavioral type
" Wide components value range generates valid identification
" Probability of wrongly powering non-PD may be increased.
" Resistor scheme is of an Absolute type
" Wide components value range will not generates valid identification
" Low probability for powering non-PD
16
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
Demonstration of AC coupled diode behavior
IEEE 802.3af, Nov. 2000.
! Reverse polarity protection used in some DC/DC converters,
exhibits behavior of an AC coupled diode if PD power supply is
accidentally reversed.
" If detection algorithm ignores polarity, it will happen at correct
connection as well.
D1
AC Coupled diode behavior
*Can be detected with Nortel's prototype, software option A.
*Can not be detected with software option B.
R1 is used for fast
discharge of input cap w/o
dissipation power at
normal operation
C1
1U - 4U
Linear
or low power/low cost
DC/DC power
supply(**)
R1
2K - 10K
D1
AC Coupled diode behavior
*Can be detected with Nortel's prototype, software option A.
*Can be detected with software option B.
C1
1U - 4U
Linear
or low power/low cost
DC/DC power
supply(**)
(**) In low power / low cost switching regulator, Bi
Polar transistor is used as the switch element
Test conditions:
! Feeding through data pairs
! PD contains reverse polarity protection circuit
! Connection from PSE to PD is done with straight and with cross-over cable.
17
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Summary
! 802.3af specification should define the detection scheme type required,
Absolute or Behavioral. (I.e. sensitive to component values or not)
! An Absolute type scheme may exhibits higher immunity to False
Positive Detection.
! False positive detection due to AC coupled diode Behavioral nature,
has been demonstrated by using Nortel’s prototype.
" Further work needed to reduce the factor 6100·Fy/Fx.
" It is recommended to reduce the factor “6100” in the AC coupled diode
method.
18
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Summary
! Improvements to AC Coupled Diode software/hardware should be
considered in order to narrow the components tolerance which will
generate positive results.
" Enhancing software option B, by setting min/max range for the higher duty
cycle needed to charge C1 and getting no pulse in either diode polarity
(Enhancements to Nortel’s option B software).
" AC coupled diode detection algorithm can be enhanced towards Absolute
type by implementing the Transformer-less solution of the AC coupled diode
method, as described in PowerDsine presentation Sep./2000
19
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Annex A - Electrical Model for Circuit Elements During Detection
Mode.
Chassis
PARAMETERS:
k_data = 0.998
L_data = 350u
L_data1 = 350u
L_cm = 100u
R1
75
TX_Data_trafo
Rdrive = 12
Vdrive = 3.3
C2
10n
K_Drive = 0.9997
L_Drive = 1.2mH
N_Drive = 1
R2
50
R4
10meg
Drive_Trafo
R8
{Rdrive}
IN1 OUT1
TXCT
0
R25
{R_Line*Length}
2
1
2
L65
{L_Line*Length}
L66
{L_Line*Length}
TXN
R3
10meg
RX_Data_trafo0
RDP
RXP
RDCT
RXCT
RDN
RXN
R26
{R_Line*Length}
TX_Data_trafo
R6
10meg
DTE_N
RX_Data_trafo
aaa
R32
C11
2.7n
IN2 OUT2
bbb
0.1
D2
R10
2k
0
Drive_Trafo
D4
D5
D1N4148
D1N4002 D1N4002
C10
C13
{0.5*C_Line*Length}
R12
D8
DTE_P
{0.5*C_Line*Length}
R13
L1
R11
100k
1k
C4
1U
D1N4002
D6
100uH
0.001
D1N4002
V3
C5
220U
D7
R29
117
D1N4002
0V
DTE_N
D1N4002
R14
50
C6
10n
0
C9
220uF
D9
R31
0.2
R19
50
RX_Data_trafo
R16
10meg
RDP
RXP
RDCT
0
R27
{R_Line*Length}
1
2
RDN
RXN
L67
{L_Line*Length}
1
2
R28
{R_Line*Length}
TX_Data_trafo0
TDP
L68
{L_Line*Length}
RXCT
TDCT
TDN
R30
10MEG
R15
10meg
TXP
0
TXCT
TXN
TX_Data_trafo
R18
10meg
R20
75
DTE_N
RX_Data_trafo
Drive_Trafo0
R21
C1
1n
R9
100
C3
0.01u
V1
TDCT
1
0
D1
D1N4148
TXP
TDN
R7
50
Length = 100 m
R_Line = 0.125 /m (Scaled)
L_Line = 0.3uH /m (Scaled)
C_Line = 15pF /m (Scaled)
TDP
D3
IN1 OUT1
{Rdrive}
V2
C7
1n
D1N4148
C8
0.01u R22
100
C12
2.7nF
Chassis
Chassis
IN2 OUT2
0
R23
10MEG
Drive_Trafo
R24
10MEG
20
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Annex A - Cont.
K K4
K_Linear
COUPLING = {k_data}
K K1
K_Linear
COUPLING = {K_Drive}
R31
K K5
K K6
K_Linear
K_Linear
COUPLING = {k_data} COUPLING = {k_data}
R32
TXP
TDP
0.5
C2
L10
{L_data1}
0.5
L12
{L_data}
L11
{L_data}
R33
0.1
L13
{L_cm}
R34
0.1
TDCT
R1
20PF
R2
IN1
L16
{L_data}
OUT1
1
1
{L_Drive}
L2
R36
{L_Drive*N_Drive*N_Drive}
IN2
R35
0.1
TXN
0.5
0.5
OUT2
C1
L15
{L_data1}
R37
TDN
L3
L14
{L_cm}
L17
{L_data}
R38
Tx_Data_Trafo
TXCT
0.1
20PF
K K2
K_Linear
COUPLING = {k_data}
Drive_Trafo
K K3
K_Linear
COUPLING = {k_data}
R25
RDP
R26
RXP
0.5
L4
{L_data}
0.5
L5
{L_data}
L6
{L_cm}
R27
RXCT
L7
{L_cm}
L8
{L_data}
0.1
L9
{L_data}
R28
R29
RXN
RDN
0.5
0.5
R30
RDCT
0.1
Rx_Data_Trafo
21
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Annex B - Detailed Sensitivity Analysis for Resistor Method
!
!
Vab = (Vtest/R2 + 2Vd/R1) / (1/R1+1/R2+1/R3)
For the following values:
(Eq-1)
" R1=25K ±2%, R3=1MΩ to Open, Vd=0.75V ±20% (±5 tolerance & ±15% change
due to ±50°C around 25°C ambient temperature)
" Vtest=24V±2%, R2=75K ±2% are considered to be locked on their nominal value
by internal PSE calibration.
!
!
Under the above conditions Eq-1 can be written as follows:
Vab = (0.00032 + 2Vd/R1) / (0.0000133 +1/R1+1/R3)
(Eq-2)
! ∆Vab/Vd = ±3.2 % max.
! ∆Vab /∆VR1 = ±1.2 % max.
! ∆Vab /∆R3 = -2% max.
!
Combined Worst case analysis = Vab +4.3%, -6.1%
22
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000
IEEE 802.3af, Nov. 2000.
Annex C - Why R2 is not needed in AC coupled diode scheme
! R2 is not needed in AC coupled diode scheme due to the fact that
the capacitor C1 (1uF) is discharged through R1 (2K).
DTE
Power source
output
capacitor, C
D1
D1N4148
PSE
V2_OUT
V1_IN
C1
1uF
LINK
T1
V1
DTE
D1
D1N4148
0
V1_OUT
R1
2K
R1
2K
V3
V2_IN
0Vdc
C1
1uF
T2
V2
Ceqv=
C1xC/(C1+C)
D1
D1N4148
R1
2K
0
Simplified detection elements model during detection mode, with out R2.
23
Reliability Analysis of Detection Options. Yair Darshan, PowerDsine. Rev-000