CPLD Course Draft - System Safety Society

Download Report

Transcript CPLD Course Draft - System Safety Society

Safety Engineers



The Naval Ordnance Safety and Security Activity
(NOSSA) realized the need to educate and
inform the Safety Professionals on CPLDs due to
the numerous Contractors that were starting to
utilize them in Navy systems.
Proper safety analyses were not being
performed.
Booz Allen Hamilton was tasked to develop an
introductory course utilizing both Government
and Industry sources.

In this course, you will learn the basics of
Programmable Logic Devices (PLDs).
 What are PLDs?
 What are the different types?
 What are the positives and negatives of using
PLDs?
 What design and test requirements must be
considered?
 What the Government boards look for?

What is a Programmable Logic Device (PLD)?
 Programmable hardware device with no preset functionality /
configuration
▪ Functionality / configuration programmed via software (VHSIC
hardware description language (VHDL) or schematic capture)
▪ Programmed PLD’s can be used to replace vast amounts of discrete circuitry
 PLD’s considered to be highly reliable by Reliability Engineers
and vendors when implemented by the manufactures’
specifications and used in a controlled environment for a
specified time.
▪ Probability of failure occurrence values of 1x10-20 or lower typically
assigned using hardware only hard fault per standards (IEEE, ANSI,
ASTM, IEC, etc.)
▪ Extremely high reliability values have a great enticement to programs
which wish to meet MIL-STD-882 1x10-6 probability safety
requirements (e.g. Autonomous use)




Static Random Access Memory (SRAM)
Erasable Programmable Read Only Memory
(EEPROM)
Flash
Anti-Fuse
 Static Random Access Memory (SRAM):
▪ Infinitely reprogrammable and is normally configured upon
power-up by another device such as a configuration
Programmable Read Only Memory (PROM)
▪ Programmed information retained only when device is
powered (voltage required to retain data varies by vendor – see
datasheet)
▪ Some new “Hybrid” Field Programmable Gate Arrays (FPGAs)
contain their own internal configuration PROM and are listed
in their datasheets as “Non-volatile, Infinitely Reconfigurable”
 Erasable Programmable Read Only Memory (EPROM) &
FLASH
▪ Are types of memory that use an array of floating-gate
transistors and are programmed using higher voltages than those
normally used in digital circuits.
▪ EPROMs can be erased using either UV light or electricity
(Electrically Erasable PROM (EEPROM)).
▪ Flash is a type of EEPROM.
▪ Have limited re-programmability (varies by vendor – 100 to 1000
times typical)
▪ Programmed information is retained for years (varies by vendor – 10
to 100 years)
▪ The threshold voltage of the Polysilicon Floating Gate is changed when (Vpp
> Vd) Fowler-Nordheim tunneling traps an excess of electrons on the floating
gate causing the transistor to be in the OFF state
e-
e- e-
 Anti-fuse (fuse technology also exists)
▪ One-time programmable device
▪ An anti-fuse is a high impedance contact (Open State) until it has a
relatively high voltage applied to it which turns it into a low-impedance
state (Closed State)
▪ Programmed information retained indefinitely (times vary by vendor)
▪ Once fused it cannot be un-fused, thus anti-fuse FPGAs cannot be
reprogrammed
Type
Reprogrammable
Voloatile
Technology
Radiation
Hardness
Fuse
No
No
Bipolar
Hard
SRAM
Yes, In-Circuit
Programming (ICP)
Yes
Complementary Metal-Oxide
Semiconductor (CMOS)
Soft
Anti-Fuse
No
No
CMOS
Hard
Flash
Yes
No
CMOS
Hard
EPROM
Yes
No
Ultraviolet CMOS (UVCMOS)
Moderate
EEPROM
Yes, ICP
No
Electrically Erasable CMOS (EECMOS)
Moderate
 How can I be sure that each and every device is
programmed as designed?
 Once programmed, is it possible for a PLD’s emulated
circuit design (functionality / configuration) to
inadvertently change?
 What is the reliability of PLD devices from a safety
viewpoint?
▪ If different from the value provided by Reliability Engineering,
why?
 Can they be safely used for autonomous control of
safety critical applications?

Joint Electron Device Engineering Council
(JEDEC) identified hardware failure modes which Potential Mitigations:
are tested by all vendors (See JEDEC or vendor
Use manufacturer data
websites for more information):
 Time-dependent dielectric breakdown
▪ A conductive path in the dielectric which shorts the
polysilicon’s anode to the cathode resulting in a loss of
charge on the polysilicon
 Surface inversion
▪ An accumulation of drifted ions which increase the
mechanical stress on a device causing loss of charge on the
polysilicon floating gate
 Electromigration
▪ Accumulation of aluminum ions which may cause fractures
to occur within the device leading to internal electrical
failures
 Corrosion
▪ Devices with pure aluminum subjected to moisture
 Temperature cycling / thermal shock
▪ Fatigue failure leading to lifted bonds, fractured / broken
bond wires, solder fatigue, cracked die, or a lifted die
integrity specification and a
robust data integrity
checking scheme.
Use proper grounding
techniques
Use appropriate vendor/part
Use proper encapsulation
Adhere to manufacturer
specification
For further details on potential mitigations, consult your CHENG.
Acronym
SEU
SED
Definition
Single Event Upset
Single Event Disturb
SET
Single Event Transient
SEDR
SEGR
Single Event Dielectric
Rupture
Single Event Gate Rupture
SEL
Single Event Latchup
SES
Single Event Snapback
MBU
Multiple Bit Upset
SEFI
Single Event Functional
Interrupt
Description
Change of information stored
Momentary disturb of information
stored in memory bit
Current transient induced by
passage of a particle, can
propagate to cause output error in
combinational logic
Essentially antifuse rupture
Rupture of gate dielectric caused
by a high current flow
High current regenerative state
induced in 4-layer device (latchup)
High current regenerative state
induced in NMOS device
(snapback)
Several memory bits upset by
passage of the same particle
Corruption of control path by an
upset
• Safety Interlock
– For this discussion, a safety interlock can be anything downstream that
prevents a PLD devices outputs from causing a safety mishap due to a
worst-case failure
• EXAMPLE: Which devices can be considered as interlocks?
These are the only interlocks!
Battery Power
Note: Worst-case failure of either SW or MOSFET
will result in loss of this single interlock!
Manual Safety Switch
Microprocessor
EEPROM Based FW Device
Safety Critical
Serial Message
(32 bits long)
Decode
Activate Safety
Function Signal
Hardware Timer
Interlock Function
I/O 53
Explosive
Squib
Using An Alternate Interlock Technology
Battery Power
- Interlocks for FPGA include CPLD, or inline chip,
or MOSFET associated with CPLD.
FPGA
EXAMPLE ONLY:
- Optocoupler Chip
- Comparator Chip
- Differential Line Receiver Chip
- Buffer Chip
CPLD
EXAMPLE ONLY:
- Optocoupler Chip
- Comparator Chip
- Differential Line Receiver Chip
- Buffer Chip
NOTE: No Common Mode Failures exist!
Explosive Squib
•
There are 23 mandatory PLD Safety Requirements to
consider when developing a project where PLDs are
used.
• 20 Design Requirements
• 1 Analysis Requirement
• 2 Test Requirements
•
•
There are also four recommended requirements that
should be done to further reduce risk caused by PLDs.
These requirements are not all encompassing. Project
variables could lead to additional requirements.




Programmable Logic Devices shall be programmed to
power-up and power-down in a known safe state.
Signal input rise times into safety critical PLDs shall meet
vendor datasheet requirements.
• Failure of meeting input rise time requirements may result
in the PLD becoming upset at any time.
PLDs considered being either a High, Serious, or Medium
safety risk that utilize sequenced logic to activate safety
critical functions shall ensure that out-of-sequence safety
critical inputs do not result in a safety critical output.
Safety critical circuitry (SC) with identified SC PLDs should
be tested utilizing the same equipment which will be used in
the end-design.

Extremely high reliability values given to PLD devices do
NOT reflect their susceptibility to Single Event Upset
(SEU) failure mechanisms or human error in device
configuration / set-up
 System Safety must assume a much higher reliability value
(probability of occurrence) for these parts (e.g. no lower than
1x10-4 for proven stable parts or 1x10-2 for unstable PLDs)
 PLD stability and interlocks external to PLDs are critical to show
safety boards your system is safe

EEPROM & SRAM based PLD devices are very susceptible
to SEUs and therefore are unsuitable for autonomous
control over safety critical functions

Anti-fuse based PLD devices are slightly susceptible to
SEUs but provide better mitigation against most PLD
failure modes
 May be used for autonomous control over some lower risk
safety critical functions

Match each of the definitions with the appropriate type of PLD technologies:

1. Infinitely reprogrammable and is normally configured upon power-up by
another device such as a configuration Programmable Read Only Memory
(PROM).

2. This is a 1 time programmable device and programmed information is retained
indefinitely.

3. This PLD type is a non-volatile memory chip made of floating-gate transistors
that retains programmed information for years. It is reprogrammed using UV
light.
a. Static Random Access Memory (SRAM)
b. Erasable Programmable Read Only Memory (EPROM)
c. Flash
d. Anti-Fuse

MIL-STD-1901A


4.8 Electronic Logic Functions. Any electronic logic related to safety functions
performed by the Ignition System (IS) or Ignition Safety Device (ISD) shall be embedded
as firmware or hardware. Firmware devices shall not be erasable or alterable by credible
environments which the IS or ISD would otherwise survive.
FESWG PLD Guidelines

2.1.
To minimize the subversion of Safety Functions (SFs) due to unintentional
and/or unrecognized modes of operation, including failure modes, each SF
implemented with logic shall use the least complex logic device that can practically
perform the required functionality.
 2.2.
To avoid degradation of a safety feature or provision for arming delay, any
logic device used in the implementation of that feature or provision for arming delay:
▪ 2.2.a.
Shall not be re-programmable or corruptible by intentional or
unintentional means. Shall not unsafely degrade the SF. Properly implemented
fixed-in-structure devices are acceptable to the Services.
▪ 2.2.b.
Shall not have the SF logic configuration reside on volatile memory.
▪ 2.2.c.
Should be rated by the manufacturer to meet or exceed the lifecycle
environments of the system. Otherwise, provide engineering rationale and
associated risk(s) for logic devices not rated to meet or exceed the lifecycle of the
system.
Ask your CHENG for a copy of these and other standards and guidelines.
Note 1: This PLD safety process does NOT apply to the design of Electronic Safety Arm
Devices (ESADs) or Arm-Fire Devices (AFDs)
Note 2: This PLD safety process was developed with the inputs of various USN / USAF safety
board members and NASA’s Office of Logic Design

What is the PLD Safety Process?
 It is a process designed to integrate with the PLD development
process into the overall system safety assessment.
 It provides a step-by-step top-level view of how PLD safety can be
performed through the various stages of a development program.
 It can also be used to determine if a mature system’s use of a PLD is
being done in a safe manner.

Why is a PLD Safety Process needed?
 Modern circuit designs are starting to use PLD devices, such as FPGAs
and CPLDs, to control safety critical functions on an increasing basis
 Designers are typically not aware of PLD devices susceptibility to
Single Event Upset (SEU), therefore they consider PLD to be very
reliable
▪ Those that are aware of SEUs typically assume that basic SEU mitigation
techniques provided by vendors are sufficient to prevent a safety mishap
 No other process exists on how to analyze or perform PLD safety
Congressional policy and Cyber security requirements will soon be released.

What this PLD Safety Process will or will not do
for you
 You will not become a PLD Expert (e.g. VHDL




programming, testing, etc..).
You will not become a PLD failure expert.
You will become familiar with the PLD safety
requirements that Government safety boards look for.
You will be aware of the process a PLD must go
through to safely integrate it into a safety critical
system.
You will be able to select appropriate resources to do
the safety tasks of the PLD safety process.





Why is having a PLD Safety Process
important?
When should the safety process be
implemented?
Are all the steps in the process important?
Who are the people involved in the safety
process?
What are some drawbacks for not having a
safety process?

The PLD Safety Process is broken down into
Six Steps. The results of each step corresponds
to particular sections of the PLD Safety
Worksheet.
 Step 1: Identify all PLD devices, Tag all
safety critical (SC) PLD devices, Apply PLD
safety requirements, and identify all insystem interlocks
 Step 2: Determine PLD HRI and PLD
suitability
 Step 3: Perform worst-case PLD safety
analysis
▪ Per NASA’s Office of Logic Design,
PLD is best analyzed using a “WorstCase” approach.
 Step 4: Complete PLD safety worksheets
and perform PLD testing
 Step 5: Continuous monitoring of CM and
creating of PLD Safety Analysis Report
 Step 6: Perform a PLD change analysis
after V&V testing is complete

For Firmware with High Criticality, an
Antifuse and SRAM technology combination
will result in a Medium Level of Rigor.

A PLD SME or Safety Engineer is required to
do Step 2: Determine PLD HRI and PLD
suitability.

PLD Safety Process provides a structured method that
any Safety Engineer, regardless of technical
background, can follow to keep a system safe

This PLD safety process was developed to help Safety
Engineers understand the minimum level of effort
required for the various combinations of technology
to satisfy Government Safety Boards

PLD devices can be used safely within safety critical
systems only if proper PLD/HW technology
combinations are used and appropriate safety
interlocks exist downstream of the PLD devices
The end-result of this new PLD safety process is the introduction of a safe product for
customer utilization!