REQ: major electronics analysis project help needed TIA

[JB2]

How often do you need to log each input?

Still to be determined, but preliminary data acquisition will need to
be during major changes in flight profiles, such that sample rates may
only be required when user selects the need and not a continuous
sampling.

Do some inputs need to be logged more often than others?

No, in fact would rather have all samples at the same time for
comparative analysis.

What kind of voltage and current ranges do you need?

Gen output is 40kVA three total & One 1kVA
115v 1 phase 400Hz 650 amps
115v 3phase 400 Hz
220v 2Phase 400Hz
28 vDC 200 amps max. There are two TRs tied together in parallel, EA
max 100 amps. Loss of various equipment will be executed and data
sampled for future capabilities and emergency procedures. The main
concern is the capability of the TRs to handle various loading
configurations but data for all sources is required.

Do you need analog measurements for everything, or is
digital OK for some of the inputs?

Data will be used for reporting purposes, future calculations and
modeling. I will probably end up building a more accurate model then
currently exists (current model is solely analytical, no hard data
available) All the sources are analog, but how those samples are
obtained, I don't think will make a difference. I thought simple
inductive pickups that feed analog to dig conv for recording.

How accurate does each analog reading need to be?

Big question! I would want at most _+2%. I suspect any more
deviation may make modeling too inaccurate.

The 150 channel requirement is a biggie. If you don't need to log each
input that often, it's not as big of a deal, but if you need high data
rates, it may get a little expensive.


Budget?

LOL yea right. in house resources avalablity of equipment., Next
week I can check whats avalable, for now, I'm simply attempting to get
a solid foundation for a POA.

If you have plenty of somebody else's money, and you don't need
to log each channel more than once every several seconds, go buy a data
logger like a Fluke Hydra. This is essentially a good auto-ranging
multimeter with a DP21T switch on the input leads. Some models can only
output the data immediately over a serial port, while some can store data
in an internal memory. The potential problem here is that it costs
US$2,000-$3,000 for 20 channels, and you need to do that seven and a
half times. You might be able to buy one data logger and multiplex the
inputs to it with external silicon, switches, or relays. You need a
40P8T contact arrangement... Outstanding suggestion

Instead of trying to log everything at once on a laptop, it might work
better to select smaller boards with 10 or 20 inputs and let each board
log to local memory, then download all the boards at the end of the
flight. This may also tend to reduce the number of wires strung around
the aircraft, if you can locate the boards near the measurement points.
You need some kind of sync between the boards... they all have to start
their clocks at the same time, or get individually synchronized to GPS
or WWV, or you need to string a wire between them. You may also have to
be careful where you put the boards physically... if they make the
aircraft radios go screwy the pilot will not be happy with you.

I do have GPS signals available

This may be obvious, but it's probably a bad idea for there to be just
one person trying to fly the plane and operate the data logging at the
same time. If one-man operation is required, the interface to the data
logging has to be very simple, or even something that can be started
before takeoff by someone else and stopped after landing.

I have a crew on board, flight crew, trechnicians, engineers (me) 707
A/C

I'm somewhat curious as to what you're doing... most of the people that
want to do something like this probably work for an aircraft
manufacturer and in that case I would think they'd just buy
off-the-shelf stuff. But hey, maybe you're building an experimental
jet in your garage.

I wish LOL no.... DOD trying to avoid using $$$$$ contractors, We
have the resources, but money is limited and this is my 1st project of
this magnitude that I'm heading up.

Once you can answer most of the questions above - in other words, when
you can state your requirements more clearly - you may want to
cross-post or post to sci.electronics.design for more good input.

Matt Roberds

thanks Matt

 

[Don Baker]

I did a similar data acquisition system for the 757 about 3 years ago. If
you email directly with a good email return address then I will gladly share
my experiences and the approach we used.

Don

 

[PeteS]

How often do you need to log each input?

<<Still to be determined, but preliminary data acquisition will need to
be during major changes in flight profiles, such that sample rates may
only be required when user selects the need and not a continuous
sampling.
How many sample points will you need per effective sample? A-Ds have
many subtle errors, and taking multiple samples can help around that.
There are a number of algorithms available to give a more accurate
result, even in the presence of noise. That brings up how fast the
effective sample must be taken, which will help you decide what type of
converter to use.

Do some inputs need to be logged more often than others?

<<
No, in fact would rather have all samples at the same time for
comparative analysis.
How simultaneous is 'same time' ? Can some systems be treated in
relative isolation, (for instance, can 20 or so [or whatever] of the
inputs be dealt with as a separate / independent set) or must all the
inputs be sampled at the same time?

What kind of voltage and current ranges do you need?

<<
Gen output is 40kVA three total & One 1kVA
115v 1 phase 400Hz 650 amps
115v 3phase 400 Hz
220v 2Phase 400Hz
28 vDC 200 amps max. There are two TRs tied together in parallel, EA
max 100 amps. Loss of various equipment will be executed and data
sampled for future capabilities and emergency procedures. The main
concern is the capability of the TRs to handle various loading
configurations but data for all sources is required.
Do you need RMS measurements only, or must you detect variations in the
cycle? Again, this helps determine the speed of the acquisition, and
the dynamic range it requires.

Do you need analog measurements for everything, or is
digital OK for some of the inputs?

<<
Data will be used for reporting purposes, future calculations and
modeling. I will probably end up building a more accurate model then
currently exists (current model is solely analytical, no hard data
available) All the sources are analog, but how those samples are
obtained, I don't think will make a difference. I thought simple
inductive pickups that feed analog to dig conv for recording.
I would agree that A-D is the best solution here - much easier to
record the data (imo). That said, in an aircraft there are multiple
sources of noise, and the pickups should probably be buffered with
instrumentation amplifiers which have superior common mode rejection.
That may be within the converter, or a separate entity nearer the
source.

How accurate does each analog reading need to be?

<<
Big question! I would want at most _+2%. I suspect any more deviation
may make modeling too inaccurate.2% of full scale? Or 2% of whatever the sample happens to be? If you
expect widely varying samples you might want to get a mu law or A-law
converter (used in digital audio). These have a fairly constant
quantiser error relative to the level of the signal. A linear converter
will have (at best) +/- 0.5LSB error. At small samples, that can be a
large error. (Most converters have larger actual errors than this)

The 150 channel requirement is a biggie. If you don't need to log each input that often, it's not as big of a deal, but if you need high data rates, it may get a little expensive.

<
As noted, if you want to get all the inputs as a snapshot together,
you'll either have to have 150 channels, or determine an acceptable
window for the multiplexer, which will help determine the conversion
time requirements.

There are premade ADC modules you can use (and probably should as it
means you won't have to build them yourself unless you need ultra fast
sampling). There are tradeoffs in terms of accuracy, linearity,
acquisition speed and input signal requirements.

Budget?

LOL yea right. in house resources avalablity of equipment., Nextweek
I can check whats avalable, for now, I'm simply attempting to get a
solid foundation for a POA.

If you have plenty of somebody else's money, and you don't need to log each channel more than once every several seconds, go buy a data logger like a Fluke Hydra. This is essentially a good auto-ranging multimeter with a DP21T switch on the input leads. Some models can only

output the data immediately over a serial port, while some can store
data in an internal memory. The potential problem here is that it
costs US$2,000-$3,000 for 20 channels, and you need to do that seven
and a half times. You might be able to buy one data logger and
multiplex the inputs to it with external silicon, switches, or relays.
You need a 40P8T contact arrangement...

Outstanding suggestionI agree - but see my caveats about acquisition speed above.

Instead of trying to log everything at once on a laptop, it might work better to select smaller boards with 10 or 20 inputs and let each board log to local memory, then download all the boards at the end of the flight. This may also tend to reduce the number of wires strung around the aircraft, if you can locate the boards near the measurement points.

You need some kind of sync between the boards... they all have to start
their clocks at the same time, or get individually synchronized to GPS
or WWV, or you need to string a wire between them. You may also have
to be careful where you put the boards physically... if they make the
aircraft radios go screwy the pilot will not be happy with you.

I do have GPS signals available
If you can separate various measurements (i.e. a snapshot of everything
is not required to be synchronised) this is a good idea. If everything
has to be synchronised to an event, then you'll need some method of
communicating that between measurement nodes. If not, then GPS time is
a good solution.

<<

This may be obvious, but it's probably a bad idea for there to be just one person trying to fly the plane and operate the data logging at the same time. If one-man operation is required, the interface to the data logging has to be very simple, or even something that can be started before takeoff by someone else and stopped after landing.

I have a crew on board, flight crew, trechnicians, engineers (me) 707
A/C

I'm somewhat curious as to what you're doing... most of the people that want to do something like this probably work for an aircraft manufacturer and in that case I would think they'd just buy off-the-shelf stuff. But hey, maybe you're building an experimental jet in your garage.

I wish LOL no.... DOD trying to avoid using $$$$$ contractors, We
have the resources, but money is limited and this is my 1st project of
this magnitude that I'm heading up.

Once you can answer most of the questions above - in other words, when you can state your requirements more clearly - you may want to cross-post or post to sci.electronics.design for more good input.

Matt Roberds

thanks Matt

The basic questions I would have before suggesting a specific solution
are noted above, summarised here:

1. Do all the inputs have to have synchronised samping?
2. Can a window be established for samples to enable sample
multiplexing?
3. Should you oversample to overcome errors?
4. How fast must the samples be when an 'event' occurs?
5. You may need to buffer the source transducers to minimise common
mode noise.
6. What sampling error is acceptable and relative to what?

Not noted: Are any of the sampled sources outside of the cabin? If so,
you'll have to be careful about temperature depencies on the transducer
/ amp / connectors.

Cheers

PeteS

 

[JB2]

RGR that Pete, thanks!!!! I will be working on answering theses
questions next week. When I come up with a more comprehensive battle
plan, I'll post some further defined criteria, but what's been brought
up thus far give me some good starting points and things to think
about.
(nothing outside the cabin, but I will be doing an unpressurized test
to conclude the systems do not fail at altitude in the event of
uncommanded depressurization)

JB2 (Jeff)

How often do you need to log each input?

<<Still to be determined, but preliminary data acquisition will need to
be during major changes in flight profiles, such that sample rates may
only be required when user selects the need and not a continuous
sampling.
How many sample points will you need per effective sample? A-Ds have
many subtle errors, and taking multiple samples can help around that.
There are a number of algorithms available to give a more accurate
result, even in the presence of noise. That brings up how fast the
effective sample must be taken, which will help you decide what type of
converter to use.

Do some inputs need to be logged more often than others?

<<
No, in fact would rather have all samples at the same time for
comparative analysis.
How simultaneous is 'same time' ? Can some systems be treated in
relative isolation, (for instance, can 20 or so [or whatever] of the
inputs be dealt with as a separate / independent set) or must all the
inputs be sampled at the same time?

What kind of voltage and current ranges do you need?

<<
Gen output is 40kVA three total & One 1kVA
115v 1 phase 400Hz 650 amps
115v 3phase 400 Hz
220v 2Phase 400Hz
28 vDC 200 amps max. There are two TRs tied together in parallel, EA
max 100 amps. Loss of various equipment will be executed and data
sampled for future capabilities and emergency procedures. The main
concern is the capability of the TRs to handle various loading
configurations but data for all sources is required.
Do you need RMS measurements only, or must you detect variations in the
cycle? Again, this helps determine the speed of the acquisition, and
the dynamic range it requires.

Do you need analog measurements for everything, or is
digital OK for some of the inputs?

<<
Data will be used for reporting purposes, future calculations and
modeling. I will probably end up building a more accurate model then
currently exists (current model is solely analytical, no hard data
available) All the sources are analog, but how those samples are
obtained, I don't think will make a difference. I thought simple
inductive pickups that feed analog to dig conv for recording.
I would agree that A-D is the best solution here - much easier to
record the data (imo). That said, in an aircraft there are multiple
sources of noise, and the pickups should probably be buffered with
instrumentation amplifiers which have superior common mode rejection.
That may be within the converter, or a separate entity nearer the
source.

How accurate does each analog reading need to be?

<<
Big question! I would want at most _+2%. I suspect any more deviation
may make modeling too inaccurate.2% of full scale? Or 2% of whatever the sample happens to be? If you
expect widely varying samples you might want to get a mu law or A-law
converter (used in digital audio). These have a fairly constant
quantiser error relative to the level of the signal. A linear converter
will have (at best) +/- 0.5LSB error. At small samples, that can be a
large error. (Most converters have larger actual errors than this)

The 150 channel requirement is a biggie. If you don't need to log each input that often, it's not as big of a deal, but if you need high data rates, it may get a little expensive.

<
As noted, if you want to get all the inputs as a snapshot together,
you'll either have to have 150 channels, or determine an acceptable
window for the multiplexer, which will help determine the conversion
time requirements.

There are premade ADC modules you can use (and probably should as it
means you won't have to build them yourself unless you need ultra fast
sampling). There are tradeoffs in terms of accuracy, linearity,
acquisition speed and input signal requirements.

Budget?

LOL yea right. in house resources avalablity of equipment., Nextweek
I can check whats avalable, for now, I'm simply attempting to get a
solid foundation for a POA.

If you have plenty of somebody else's money, and you don't need to log each channel more than once every several seconds, go buy a data logger like a Fluke Hydra. This is essentially a good auto-ranging multimeter with a DP21T switch on the input leads. Some models can only

output the data immediately over a serial port, while some can store
data in an internal memory. The potential problem here is that it
costs US$2,000-$3,000 for 20 channels, and you need to do that seven
and a half times. You might be able to buy one data logger and
multiplex the inputs to it with external silicon, switches, or relays.
You need a 40P8T contact arrangement...

Outstanding suggestionI agree - but see my caveats about acquisition speed above.

Instead of trying to log everything at once on a laptop, it might work better to select smaller boards with 10 or 20 inputs and let each board log to local memory, then download all the boards at the end of the flight. This may also tend to reduce the number of wires strung around the aircraft, if you can locate the boards near the measurement points.

You need some kind of sync between the boards... they all have to start
their clocks at the same time, or get individually synchronized to GPS
or WWV, or you need to string a wire between them. You may also have
to be careful where you put the boards physically... if they make the
aircraft radios go screwy the pilot will not be happy with you.

I do have GPS signals available
If you can separate various measurements (i.e. a snapshot of everything
is not required to be synchronised) this is a good idea. If everything
has to be synchronised to an event, then you'll need some method of
communicating that between measurement nodes. If not, then GPS time is
a good solution.

<<

This may be obvious, but it's probably a bad idea for there to be just one person trying to fly the plane and operate the data logging at the same time. If one-man operation is required, the interface to the data logging has to be very simple, or even something that can be started before takeoff by someone else and stopped after landing.

I have a crew on board, flight crew, trechnicians, engineers (me) 707
A/C

I'm somewhat curious as to what you're doing... most of the people that want to do something like this probably work for an aircraft manufacturer and in that case I would think they'd just buy off-the-shelf stuff. But hey, maybe you're building an experimental jet in your garage.

I wish LOL no.... DOD trying to avoid using $$$$$ contractors, We
have the resources, but money is limited and this is my 1st project of
this magnitude that I'm heading up.

Once you can answer most of the questions above - in other words, when you can state your requirements more clearly - you may want to cross-post or post to sci.electronics.design for more good input.

Matt Roberds

thanks Matt

The basic questions I would have before suggesting a specific solution
are noted above, summarised here:

1. Do all the inputs have to have synchronised samping?
2. Can a window be established for samples to enable sample
multiplexing?
3. Should you oversample to overcome errors?
4. How fast must the samples be when an 'event' occurs?
5. You may need to buffer the source transducers to minimise common
mode noise.
6. What sampling error is acceptable and relative to what?

Not noted: Are any of the sampled sources outside of the cabin? If so,
you'll have to be careful about temperature depencies on the transducer
/ amp / connectors.

Cheers

PeteS

 

[John Larkin]

Still to be determined, but preliminary data acquisition will need to
be during major changes in flight profiles, such that sample rates may
only be required when user selects the need and not a continuous
sampling.

No, in fact would rather have all samples at the same time for
comparative analysis.

Gen output is 40kVA three total & One 1kVA
115v 1 phase 400Hz 650 amps
115v 3phase 400 Hz
220v 2Phase 400Hz
28 vDC 200 amps max. There are two TRs tied together in parallel, EA
max 100 amps. Loss of various equipment will be executed and data
sampled for future capabilities and emergency procedures. The main
concern is the capability of the TRs to handle various loading
configurations but data for all sources is required.

Data will be used for reporting purposes, future calculations and
modeling. I will probably end up building a more accurate model then
currently exists (current model is solely analytical, no hard data
available) All the sources are analog, but how those samples are
obtained, I don't think will make a difference. I thought simple
inductive pickups that feed analog to dig conv for recording.


Big question! I would want at most _+2%. I suspect any more
deviation may make modeling too inaccurate.

LOL yea right. in house resources avalablity of equipment., Next
week I can check whats avalable, for now, I'm simply attempting to get
a solid foundation for a POA.

I do have GPS signals available
I have a crew on board, flight crew, trechnicians, engineers (me) 707
A/C


I wish LOL no.... DOD trying to avoid using $$$$$ contractors, We
have the resources, but money is limited and this is my 1st project of
this magnitude that I'm heading up.

thanks Matt

My customers (biggish aerospace folks) tend to use VME systems for
this sort of thing. Your choices are

1. Log the raw data (waveforms) with ADC boards. This is simple but
would produce huge heaps of data that would have to be reduced into
things like RMS voltages, currents, RPMs, etc. This may require signal
conditioning ahead of the ADC channels. At high channel counts,
acquiring raw data can run out of system bandwidth.

2. Use signal conditioner boards that measure RPM, voltage, current,
power, whatever, and log their output. More expensive per channel, but
much less data and data reduction, and little or no outboard signal
conditioning.

Both require an infrastructure (VME crate, controller or embedded CPU)
and acquisition boards. Figure $4k or so for the baseline stuff, plus
$100-$500 per channel, roughly, depending on how you do it.

My company makes VME boards that acquire AC power, RPMs, load cells,
thermocouples, stuff like that. In something like jet engine testing,
it's common for hundreds of channels to be acquired at hit rates down
to a few milliseconds. See VITA.COM for most of what's available in
VME.


John

 

[Ban]

I would suggest the LHXX40 series of transducers
http://www.lem.com/inet/products.ns...04FC59E/$File/td_lh40series_internet_engl.pdf
, which have a RMS or an instantaneous analog output and can be connected to
any multichannel datalogger. You can probably rent those transducers from
LEM, if you need them just for a short time. You need of course still power
supply lines and coax output cable. the clamps are super easy to install and
are very precise and have even an autorange feature to increase accuracy for
lowish currents. For a big project like this I would get an experienced
analog specialist to help me setup and execute the data acqusition, so you
get really valid data.

 

[JB2]

thanks John, I'm simply doing an electrical load analysis of the
aircraft. All the sources and how they handle the various loads. ( I
know, sounds easy) that's probably why I was given the project and no
one else has touched it.

 

[John Larkin]

thanks John, I'm simply doing an electrical load analysis of the
aircraft. All the sources and how they handle the various loads. ( I
know, sounds easy) that's probably why I was given the project and no
one else has touched it.

AC voltage/current/power acquisition is shockingly neglected. When I
went to engineering school last century, the freshmen took classes
together but later on the class split up into electronics and power
factions, and we barely socialized after that. The result is that 60
Hz is the most neglected frequency in electronics.

Here's one of our AC acquisition modules...

http://www.highlandtechnology.com/DSS/V180DS.html

and we're planning a couple more, hence my "Hilbert" post. Several
outfits already use this one for analysis of aircraft power systems.

John

 

[Mike Monett]

John Larkin wrote:

[...]

Here's one of our AC acquisition modules...

http://www.highlandtechnology.com/DSS/V180DS.html

and we're planning a couple more, hence my "Hilbert" post. Several
outfits already use this one for analysis of aircraft power systems.

John

Just out of curiosity, how are you measuring power factor?

Mike Monett

 

[Fred Bloggs]

Here's one of our AC acquisition modules...

http://www.highlandtechnology.com/DSS/V180DS.html

And just exactly what do you have in the way of providing for protection
from catastrophic sensor failure on those HV circuits, dopey?

 

[Fred Bloggs]

Here's one of our AC acquisition modules...

http://www.highlandtechnology.com/DSS/V180DS.html

And from which competitor did you illegally pirate the algorithms?

 

[John Larkin]

John Larkin wrote:

[...]

Here's one of our AC acquisition modules...

http://www.highlandtechnology.com/DSS/V180DS.html

and we're planning a couple more, hence my "Hilbert" post. Several
outfits already use this one for analysis of aircraft power systems.

John

Just out of curiosity, how are you measuring power factor?

PF = TruePower / (RMSamps * RMSvolts)

This loses the leading/lagging angle info, which is one reason to do a
next-generation product, all dsp-ish and stuff.

John

 

[John Larkin]

And just exactly what do you have in the way of providing for protection
from catastrophic sensor failure on those HV circuits, dopey?

For voltage, we have an external potential acquisition box that has
resistor strings and transzorbs and stuff. For current, we use a
"current sensor" (which is a CT with a built-in burden resistor) or a
regular 5-amp CT with an external 0.05 ohm load resistor. Either way,
no serious power gets to the VME card.

John

 

[John Larkin]

And from which competitor did you illegally pirate the algorithms?

I don't know of any competitor for this module - do you? - much less
one that publishes his algorithms. I just sat down and wrote the code.
I've been doing AC power meters for almost 20 years, and things have
just sort of evolved. I did a few thousand end-use survey meters (back
when that sort of thing was popular+funded) and over 1700 apartment
submeters for Battery Park City, right across the street from the
ex-World Trade Center buildings.

Well, you did ask.

John

 

[JB2]

Had not gotton that far. Most of the emphasis is on the DC
capabilites, as they are tjhe ones in doubt. The AC sources have no
prolems, previous loading calculations for teh AC did not even
approach 30% of rated maximum. Although, PF is definantly a
consideration and will be looked at. I can compare no load to loaded
AC and calculated PF overall, but not system by system. I wqas
planning on looking at the AC voltage and current rms values.

John Larkin wrote:

[...]

Here's one of our AC acquisition modules...

http://www.highlandtechnology.com/DSS/V180DS.html

and we're planning a couple more, hence my "Hilbert" post. Several
outfits already use this one for analysis of aircraft power systems.

John

Just out of curiosity, how are you measuring power factor?

Mike Monett

 

[JB2]

Guys, our equipment research department is supposed to have a ton of
equipment. this week I'll be spending some time "shopping" there and
attempting to find appropriate test gear. The budget is almost non
existent in terms of contract money. This is going to be an in-house
procedure. So, before everyone gets off on a rant about their
equipment, I just want to put together the plan 1st. I'm going with
the KISS method of testing. no frills....

 

[John Larkin]

Guys, our equipment research department is supposed to have a ton of
equipment. this week I'll be spending some time "shopping" there and
attempting to find appropriate test gear. The budget is almost non
existent in terms of contract money. This is going to be an in-house
procedure. So, before everyone gets off on a rant about their
equipment, I just want to put together the plan 1st. I'm going with
the KISS method of testing. no frills....

Well, this is a discussion group.

John

 

[Fred Bloggs]

I don't know of any competitor for this module - do you? - much less
one that publishes his algorithms. I just sat down and wrote the code.
I've been doing AC power meters for almost 20 years, and things have
just sort of evolved. I did a few thousand end-use survey meters (back
when that sort of thing was popular+funded) and over 1700 apartment
submeters for Battery Park City, right across the street from the
ex-World Trade Center buildings.

Well, you did ask.

John

All those applications have been converted to smart networked nodes-

 

[John Larkin]

All those applications have been converted to smart networked nodes-

Yeah, Ethernet - especially PoE - is the ultimate expansion bus. I
can't see why anybody would want to use IEEE-488 or RS-232 for
instrumentation any more.

John

 

[Mike Monett]

PF = TruePower / (RMSamps * RMSvolts)

This loses the leading/lagging angle info, which is one reason to do a
next-generation product, all dsp-ish and stuff.

John

Neat cheat! Thanks.

Mike Monett