Flight Data Recorders

"Peter Clark" wrote in message
...
On Mon, 8 Sep 2008 13:08:56 +0000 (UTC), David Lesher
wrote:

Michael Henry writes:

MAK adds that the quality of the flight-data recorder information
is "unsatisfactory" because of partial thermal damage to its
magnetic tape, caused by the intense fire which followed the crash.

Which made me think: magnetic tape?! Surely we've come far enough that
we can be using solid-state storage in flight data recorders?

Does anyone know why FDRs would still be using what, to me, is archaic
technology?

Co$t is why. And while flash-based systems may offer more survivable
recording; consider the recent Qantas incident, where the
depressurization itself was overwritten by subsequent flight.

They could put in a bigger chip for longer recording. I thought most
DRs were only about 1/2 hr looping?

The original CVRs were 1/2 hour looping, carying three channels (or tracks)
of audio, and would erase the tape when the parking brake was applied at the
destination gate. Part of that was due to a combination of technical
necessity and convenience, and part was the dfproduct of negotiation between
the interested parties. The reusult was robust, in the sense that the
recovery of data did not depent upon any part of the equipment remaining
operable I am not familiar with the newer equipment; but would expect
that it would have been made functionally similar--with the possible
addition of more channels of audio.

The original FDRs contained a single use role of stainless steel foil, which
was marked by stylii. The result was *extremely* robust, but even the
second generation of such recorders had a very limited number of data
channels and required frequent depot level maintenance to replace the foil
rolls. Solid state memory simply does not have similar survivability and
there is virtually no change that it will within the next 25 years.
Therefore, solid state memory, in the event that it is actually in use, is
simply a way to obtain more channels of data at less monetary cost in those
cases where the data can still be recovered.

So long as the purpose of the data is to detect failures which can then be
resonably predicted, then it is probably a good trade.

Paul Saccani wrote in
:

On Mon, 8 Sep 2008 20:46:32 +0000 (UTC), Bertie the Bunyip
wrote:

"Sr20goer" wrote in
:


"GB" wrote in message
...
Michael Henry wrote in
news:00d3a924$0 :
SNIP

It's not all that archaic. Aviation trails the bleeding edge by a
very long margin, and with good reason. The bleeding edge usually
draws blood at some point! I was initially surprised to learn, in
circa 1999, that the B767 didn't use GPS for navigation. I
subsequently learned that they don't need it... GPS isn't really
up to scratch. (I'm not gonna explain that here, that's fodder
for another post!)

GB

GB
You referring to INS?
But they do now use GPS for 'local' navigation (SID, STAR, flextrack
etc). Brian


No, they use GPS to update the INS.

Aren't both sources integrated via a fast Kalman filter? GPS errors
are rapid and errors tend to clump around an accurate position, whilst
INS errors are slow and tend to drift away from an accurate position.
Integrating the systems via a Kalman filter results in a far more
accurate and precise position.


No idea. Works well, lasts a long time.
When we dont use GPS, we use radio update of another type, usually. VOR
DME. Without it, even the best INS can drift appreciably over time. It's
not uncommon to see an error of a couple of miles after a long flight,
and errors of 20 miles are possible after an ocean crossing. Radio
update pretty much eliminates this and I've never seen any map shift
whatsoever on a GPS updated system.

"Paul Saccani" wrote in message
...
Aren't both sources integrated via a fast Kalman filter? GPS errors
are rapid and errors tend to clump around an accurate position, whilst
INS errors are slow and tend to drift away from an accurate position.
Integrating the systems via a Kalman filter results in a far more
accurate and precise position.

Be interested in the technicalities, having been obliged to use a 4th order
Chebishev (sp?) filter for some DA years ago - got a URL ? (Too lazy to
look atm :-)

David Lesher wrote:
... magnetic tape?! Surely we've come far enough that
we can be using solid-state storage in flight data recorders?

Does anyone know why FDRs would still be using what, to me, is archaic
technology?

Co$t is why. And while flash-based systems may offer more survivable
recording; consider the recent Qantas incident, where the
depressurization itself was overwritten by subsequent flight.

I can see how solid state storage is more expensive than tape, but I
would have thought that by the time you go through the whole
certification process the cost of the actual medium is just a small
fraction of the total cost... and if the solid state medium offers any
benefits over tape than that small extra cost would certainly be
justifiable.

The other issue you raise really surprises me: the small amount of
flight time for which recordings are kept. Some other posts have
mentioned half-an-hour loop! Does anyone know what volume of data this
entails?

Regards,

Michael


P.S. A quick mental calculation: I can go to Officeworks (in Australia)
and buy a 16 Gb memory stick for AU$100. OK, it's not going to meet the
same fault-tolerance and quality standards, but let's ignore that for
now For those of us in the iPod generation 16 Gb equates to around
300 hours of MP3 music. Obviously you would encode speech differently
and fit a lot more than that, but let's not worry about that. Of course
you also want to record other parameters (temperatures, pressures,
position of controls, etc). I'm just pulling numbers out of the air but
let's say you want to record 1,000 different parameters each sampled 10
times a second. Furthermore let's say the data for each parameter is
stored as a 32-bit value, which comes out to 40kb of data per second or
just under 150 Mb per hour. That's over 100 hours of parameters in our
16 Gb memory stick. Recording voice and data at the same time and I
figure you can fit about 80 flight hours in a $100 USB stick. Not bad
I'm obviously missing out on some important bits but it has made me
wonder: what would be the market for a really small, light, and cheap
uncertified flight data recorder...

What will all these new filter technologies, you still need to sump your
fuel. Too much water, even though the filter will catch it, will still
plug it up.


RT wrote:
"Paul Saccani" wrote in message
...
Aren't both sources integrated via a fast Kalman filter? GPS errors
are rapid and errors tend to clump around an accurate position, whilst
INS errors are slow and tend to drift away from an accurate position.
Integrating the systems via a Kalman filter results in a far more
accurate and precise position.

Be interested in the technicalities, having been obliged to use a 4th order
Chebishev (sp?) filter for some DA years ago - got a URL ? (Too lazy to
look atm :-)

jeremy wrote:
GB wrote:
Stuff we take for granted now was
unthinkable even 10 years ago. I'm not about to claim that I could
even begin to guess at what computer technology I'll see 25 years from
now.


GB

The famous example of how short sighted we can be is the famous set of
quotes:

"Nobody will ever need more than 640k RAM!" -- Bill Gates, 1981
"Windows 95 needs at least 8 MB RAM." -- Bill Gates, 1996
"Nobody will ever need Windows 95." -- logical conclusion

"You ain't goin' nowhere, son -- you ought to go back to drivin' a
truck" Jim Denny, Sept. 25, 1954 as he fired Elvis Presley after one
performance at the Grand Ole Opry.

"Paul Saccani" wrote in message
...
On Tue, 9 Sep 2008 22:14:49 +1000, "RT"
wrote:


"Paul Saccani" wrote in message
. ..
Aren't both sources integrated via a fast Kalman filter? GPS errors
are rapid and errors tend to clump around an accurate position, whilst
INS errors are slow and tend to drift away from an accurate position.
Integrating the systems via a Kalman filter results in a far more
accurate and precise position.

That really is a fascinating idea - sort of poor man's differential GPS.....
If you kept track of the average position by the GPS and the difference from
there to the inertial position, you'd be able calculate the inertial drift
and thence correct the GPS to inertial accuracy :-)

Be interested in the technicalities, having been obliged to use a 4th
order
Chebishev (sp?) filter for some DA years ago - got a URL ? (Too lazy to
look atm :-)


Chebyshev/Chebychev - ask the missus.

She wasn't available at the time or I woulda :-) Doubt if she ever used
'em herself though, as she was more into control theory (...yuk!).

These books are worthwhile :-

Global Positioning Systems, Inertial Navigation, and Integration, 2nd
Edition

http://www.wiley.com/WileyCDA/WileyT...470041900.html

Kalman Filtering: Theory and Practice Using MATLAB, 2nd Edition

http://www.wiley.com/WileyCDA/WileyT...471392545.html

The wiki write up on Fast Kalman seems reasonable too.

Many thanks - I'll have a prowl...tho me and Matlab never really got on....

Peter Dohm said the following on 7/09/2008 11:55 PM:
This occasionally happens, even though the FDR and CVR have considerable
thermal inertia and are located in a part of the plane that is least
frequently affected by both fire and mechanical damage. Also, the prose
above suggests that there was some data recovered from the cocpit voice
recorder and "some data" is probably more than they would have gotten from
solid state memory.

All the same, there has been a move from magnetic tape and metal foil to
solid state memory in the 20+ years that I have been out of that industry.
However, IMHO, that has been a case of obtaining more detailed data and
information (most of the time) and greatly reduced maintenance cost (the
frequent replacement of tapes and foils was shop maintenance) and accepting
far less robust media in order to accomplish those goals.

Personally, I can see more than one side of the arguments leading to these
changes and I have mixed feelings regarding the wisdom of same.

There's no reason, except cost, why both can't be fitted. Solid state
can potentially record most of a flight and, if it survives provide a
considerable amount of data over an extended timeline. If the solid
state doesn't survive then there's still the legacy FDR and CDR.

A damaged hard disk might offer higher rates of recovery than a damaged
solid state drive. However, I'm not sure how well a hard disk would
tolerate mid air turbulence. Anyway, I think their ability to survive a
fire might be a lot less than that of the older devices.

In article ,
Michael Henry wrote:


The other issue you raise really surprises me: the small amount of
flight time for which recordings are kept. Some other posts have
mentioned half-an-hour loop! Does anyone know what volume of data this
entails?

I don't know the volume of data, but the CVR and FDR have different
amounts of data in terms of time. 1/2 CVRs are old. New ones can
record hours.

btw - for older aircraft, the biggest cost of putting newer FDRs with
(many) more channels is not the FDR itself, but all the sensors that
have to be installed.

--
Bob Noel
(goodness, please trim replies!!!)

"Paul Saccani" wrote in message
...
On Wed, 10 Sep 2008 16:01:32 +1000, "RT"
wrote:
That really is a fascinating idea - sort of poor man's differential
GPS.....

Yes, that order of precision and accuracy for navigation.

If you kept track of the average position by the GPS and the difference
from
there to the inertial position, you'd be able calculate the inertial drift
and thence correct the GPS to inertial accuracy :-)

That's about it. I thought it was used from scratch for aircraft
systems with combined INS and GPS - the cost of the filter is a
fraction of the cost of the system, particularly if their is already
room available on the computing side of things to do it. Its
certainly done with the PPS and SPS on military systems.

However, on reflection, there are probably many aircraft where GPS has
been added to an existing inertial based system and the integration
only goes so far as to allow a position update to correct drift.

Many thanks - I'll have a prowl...tho me and Matlab never really got
on....

Me two...

Heh. Mentioned this to Her Indoors. She was mightily unimpressed. "Of
course we were updating inertial nav - using radar - now do you want a cup
of tea?"

Friggin hell! :-)

Mind you, her field was ICBMs, not IC domestic transport :-)

(For those that tuned in late, my Missus was a fairly senior control systems
engineer in the old USSR)