C
Chris Carlen
- Jan 1, 1970
- 0
Hi:
We have a 4 channel motion control servo system for electrohydraulic
actuation of four engine valves. The basic components in reverse from
the business end are as follows:
1. voice coil motor (to actuate hydraulic valves)
2. PWM H-bridge power amplifier (to drive voice coils, with fuses to
protect coils from pegged amplifiers)
3. TI 6711 DSP with 16 channels each of 100kHz 16-bit A/D and D/A to
digitally implement closed-loop control algorithms
4. LVRT position sensors (for precise control) and proximity (for
monitoring and for a backup check on control)
5. absolute encoder on engine camshaft to inform DSP where valves
should be (position reference data)
6. extra incremental encoder on crankshaft (with one index mark blanked
by a hall sensor on the cam) feeds to DSP to give a backup check on cam
encoder alignment.
Additionally:
1. low voltage power +5V and +/-15V for DSP board and it's analog IO
system.
2. two 48V 600W switchers to power PWM amps.
3. PC connected via USB to DSP to implement GUI. This link need not be
active for the DSP to do its job.
The main safety issue is that we must prevent the valves from
accidentally crashing into the piston. Due to the complexity of the
research engines (with optical access) involved, the cost in repair time
which would result from a failure of the valve control would be huge.
This must be avoided in any fault modes where the mitigation cost/repair
cost ratio is <= 1.
In analyzing possible failure modes of this system which could lead to
this undesirable occurance, I have enumerated the following
possibilities in decreasing order of likelyhood and indicate whether the
problem has been mitigated. If not, in the next paragraph I describe
what is being planned to mitigate an unsolved problem.
POSSIBLE PROBLEMS (feel free to comment on the prioritization order):
1. position sensor failure (mitigated by use of redundant sensors)
2. DSP software or hardware failure (see below)
3. encoder alignment slippage (mitigated by use of redendant encoders)
4. line power failure (mitigated by putting system on a UPS)
5. internal DC power supply failure (see below)
6. power amplifier failure (see below)
FIXES:
re: #2. To deal with the DSP hardware or software going bonkers, I plan
to utilize a watchdog timer (WDT). The DSP board has one built into its
FPGA, but it is of little use because it doesn't have an adjustable
timeout. Rebooting the DSP takes a long time, much too long to wait to
regain control. Thus, an external custom watchdog is being designed.
I plan to use the simplest possible discrete logic to build a WDT using
some one-shots and flip-flops plus a few glue gates. It will detect the
absence of a 20kHz pulse train from the DSP (from the control loop
iteration) and if the pulses stop, will first apply a withdraw pulse to
the valves, then disable the PWM amplifiers. Thus, the WDT must include
4 pairs of analog multiplexers to switch the inputs of the PWM amps
from the DSP to a constant voltage source of appropriate polarity and
magnitude to force the valves to close quickly.
re: #5. To deal with a power failure of one of the DC power supplies
(either high or low power supplies), I am considering using redundant
supplies with a diode network to allow them to be paralleled. This
should mitigate the problem to an extremely low likelyhood of causing a
disaster. Furthermore, the WDT circuit could include power monitoring
comparators to signal the DSP if a supply was malfunctioning.
re: #6. I had considered a means of dealing with power amplifier
failure. Solid state relays could be employed to build a multiplexer
for the voice coils that could switch the coils to a constant withdraw
voltage in the event of the detection of a serious control discrepancy
by the DSP. However, the differential output of the H-bridges combined
with the fact that the motor currents are bidirectional makes this
tricky. I have concluded that the cost of fixing vs. the probablility
of occurence makes this fault mode something we will just have to hope
doesn't happen.
The cost of fixing this might be less than the cost of a failure, but at
present we require that the system be put into service shortly in
prototype configuration, and there isn't enough time to deal with this.
It is still possible to include this capability later when the system
is being finalized.
Would you consider solving the faults in the methods described?
Would you agree with the decision to give up attempting to mitigate the
possibility of amplifier failure while the system is being used for a
few months before packaging and finalizing?
Thanks for input.
--
Good day!
________________________________________
Christopher R. Carlen
Principal Laser&Electronics Technologist
Sandia National Laboratories CA USA
[email protected]
NOTE, delete texts: "RemoveThis" and
"BOGUS" from email address to reply.
We have a 4 channel motion control servo system for electrohydraulic
actuation of four engine valves. The basic components in reverse from
the business end are as follows:
1. voice coil motor (to actuate hydraulic valves)
2. PWM H-bridge power amplifier (to drive voice coils, with fuses to
protect coils from pegged amplifiers)
3. TI 6711 DSP with 16 channels each of 100kHz 16-bit A/D and D/A to
digitally implement closed-loop control algorithms
4. LVRT position sensors (for precise control) and proximity (for
monitoring and for a backup check on control)
5. absolute encoder on engine camshaft to inform DSP where valves
should be (position reference data)
6. extra incremental encoder on crankshaft (with one index mark blanked
by a hall sensor on the cam) feeds to DSP to give a backup check on cam
encoder alignment.
Additionally:
1. low voltage power +5V and +/-15V for DSP board and it's analog IO
system.
2. two 48V 600W switchers to power PWM amps.
3. PC connected via USB to DSP to implement GUI. This link need not be
active for the DSP to do its job.
The main safety issue is that we must prevent the valves from
accidentally crashing into the piston. Due to the complexity of the
research engines (with optical access) involved, the cost in repair time
which would result from a failure of the valve control would be huge.
This must be avoided in any fault modes where the mitigation cost/repair
cost ratio is <= 1.
In analyzing possible failure modes of this system which could lead to
this undesirable occurance, I have enumerated the following
possibilities in decreasing order of likelyhood and indicate whether the
problem has been mitigated. If not, in the next paragraph I describe
what is being planned to mitigate an unsolved problem.
POSSIBLE PROBLEMS (feel free to comment on the prioritization order):
1. position sensor failure (mitigated by use of redundant sensors)
2. DSP software or hardware failure (see below)
3. encoder alignment slippage (mitigated by use of redendant encoders)
4. line power failure (mitigated by putting system on a UPS)
5. internal DC power supply failure (see below)
6. power amplifier failure (see below)
FIXES:
re: #2. To deal with the DSP hardware or software going bonkers, I plan
to utilize a watchdog timer (WDT). The DSP board has one built into its
FPGA, but it is of little use because it doesn't have an adjustable
timeout. Rebooting the DSP takes a long time, much too long to wait to
regain control. Thus, an external custom watchdog is being designed.
I plan to use the simplest possible discrete logic to build a WDT using
some one-shots and flip-flops plus a few glue gates. It will detect the
absence of a 20kHz pulse train from the DSP (from the control loop
iteration) and if the pulses stop, will first apply a withdraw pulse to
the valves, then disable the PWM amplifiers. Thus, the WDT must include
4 pairs of analog multiplexers to switch the inputs of the PWM amps
from the DSP to a constant voltage source of appropriate polarity and
magnitude to force the valves to close quickly.
re: #5. To deal with a power failure of one of the DC power supplies
(either high or low power supplies), I am considering using redundant
supplies with a diode network to allow them to be paralleled. This
should mitigate the problem to an extremely low likelyhood of causing a
disaster. Furthermore, the WDT circuit could include power monitoring
comparators to signal the DSP if a supply was malfunctioning.
re: #6. I had considered a means of dealing with power amplifier
failure. Solid state relays could be employed to build a multiplexer
for the voice coils that could switch the coils to a constant withdraw
voltage in the event of the detection of a serious control discrepancy
by the DSP. However, the differential output of the H-bridges combined
with the fact that the motor currents are bidirectional makes this
tricky. I have concluded that the cost of fixing vs. the probablility
of occurence makes this fault mode something we will just have to hope
doesn't happen.
The cost of fixing this might be less than the cost of a failure, but at
present we require that the system be put into service shortly in
prototype configuration, and there isn't enough time to deal with this.
It is still possible to include this capability later when the system
is being finalized.
Would you consider solving the faults in the methods described?
Would you agree with the decision to give up attempting to mitigate the
possibility of amplifier failure while the system is being used for a
few months before packaging and finalizing?
Thanks for input.
--
Good day!
________________________________________
Christopher R. Carlen
Principal Laser&Electronics Technologist
Sandia National Laboratories CA USA
[email protected]
NOTE, delete texts: "RemoveThis" and
"BOGUS" from email address to reply.