September 25 October 6, 2001 notes
Fiber Optic Gyros
It turns out that we did have one major component die in the
crash one of the FOGs is no longer functioning. I should have immediately tested all of them, but I had given
them to Bob to use in fabricating the new mounting bracket for them, so I
didnt find out that one of them was dead until last Saturday. I got it Fed-Exed off to KVH for repair on
Monday, but I havent heard back from them yet.
We are grounded until we get this fixed, which is
unfortunate. I have been looking into
buying another tri-axial FOG unit so we could improve a few things, and leave
the current units as a backup in the future.
While I understand that various micro-machined mechanical
gyros can be calibrated sufficiently to use in high acceleration environments,
the true G insensitivity of laser gyros seems to be a significant benefit to
me. Ring laser gyros offer the greatest
accuracy, but they all seem to be up in the six figure price range, so fiber
optic gyros are the way to go.
Comparing the different FOGs is difficult, because the
figures of merit arent all quoted the same way. In general, the lower the sensed rate range, the more precise
they are. 100 deg/sec is about the
lowest I am comfortable with. You can
easily swivel a gyro faster than this in one hand, but vehicles should be able
to be kept below this rate if you dont need fighter plane quick snap turns.
E-Core 1000 (single axis), $1500 each
E-Core 2100 (single axis), $2500 each
Range: +- 100 deg/second
Scale factor linearity, constant temperature: < 0.5%
temperature: <2% RMS, <1% RMS on digital output versions
Output: 2.5 V +- 2.0 V, digital output is offered at 10 Hz, and
optionally 100 Hz on the 2100
100 Hz instantaneous bandwidth
The difference between the two is the bias stability and
1000: Angle random walk: 20 deg / hr / rt-Hz, 0.33 deg /
2100: Angle random walk: 5 deg / hr / rt-Hz, 0.08 deg /
KVH offers a dual axis mil-spec unit for turret and antenna
pointing, but no three axis units. I
sent a request for information about reduced filtering for higher bandwidth,
but didnt receive a reply. Delivery of
our FOGs took six weeks from the time of order.
Our three E-Core 1000 gyros have worked fine, but a smaller,
more integrated three axis package would be nice, and a digital output would allow
us to get good enough inertial accuracy for navigation as well as stabilization,
and let us continue to use our crappy PC104 A/D system for the less precise
sensors (pressure, temp, throttle, etc).
If we stay with an analog output device, we will need to move to an
external A/D board to get good enough accuracy for navigation.
They only offer FOGs in integrated systems with
accelerometers, and they range from $8500 to $11000, depending on the options.
The most appropriate FOG system, the IMU600CA-201 has the
minor annoyance that it requires a 15-30V power source.
The vertical gyro (VG) systems have additional software to
calculate absolute pitch and roll values from the basic sensors, using the
assumption that planes or vehicles are going to be going mostly level,
allowing the accelerometer to slowly correct long term gyro drift. That isnt of any value to us, because we
could do the same thing in software.
They offer one version designed for automotive use that has a built in 12V
power converter, but doesnt offer the same 10G/100Hz accelerometer output that
we would get from the CA-201.
The Attitude Heading Reference Systems (AHRS) also add a
magnetometer so it can perform similar long-term gyro corrections on the yaw
Range: +- 200 deg/s
Resolution: < 0.025 deg/sec
Scale factor accuracy: < 1%
Non-linearity: < 0.2%
Random walk: < 1.25 deg/sqrt(hr)
Size: 127x152x101mm, 1.6 Kg
They offer analog output, but it is really all digital, with
the analog output being created by a DAC.
I almost purchased one of these last week, but Im afraid
the system latency is a little longer than I would like. It spends about 8ms internally dealing with
the gyros, then it takes another 6ms to pump the 18 byte update packet over a
33Kbaud serial link, which is a noticeable amount of latency on top of the
smoothing already performed by the 100 Hz sensors.
If this took a 12V power source and could output packets
continuously at 115Kbaud, it would be my perfect sensor.
DM 28.000 ($12,795) for just FOGs, DM 33.000 ($15,080) with
Range: +- 200 deg/s
(50 to 800 deg/s optional)
Output: +- 4.75 V
Biad stability: <160 uV
Resolution: < 50 uV (0.002 deg/s with +- 200 deg/s range,
thats 18 bits of resolution!)
Noise (0-100 Hz): <100 uV/sqrt(Hz)
Bandwidth: 0-100 Hz (optional up to 300 Hz)
Linearity error: < 0.2% (compensated)
Power: +5 V in small package, or 9-18 V or 18-34 V in larger
Shock: 200 g, 6 ms
Size: 80x80x80 mm or 80x80x118 mm, 900 grams (light weight
version on request)
This is higher end in all specs than the others, but more
Range: +-100 deg/s
Resolution: 0.05 deg / sec
Bias stability: 360 deg / hour (probably a misprint)
Dimension: 80x80x45 mm single axis, 170x170x137.5 mm
I have requested a quote for these.
Other Reconstruction Notes
Since I was pulling the PC104 computer apart for several
tests, I finally bought one of the little plastic Parvus PC104 board separating
tools, which I highly recommend. No
more bent pins.
I have replaced all the nylon standoffs with metal ones
now. My one gripe with WinSystems is
that they only ship two nylon standoffs with each PC104 board. Very early on, we had some boards pull apart
on a tip-over, so we went to four standoffs per board. We had some metal ones, but they werent
exactly the right size, and they caused some problems for one of the boards in
the stack, so we still had several nylon ones in use. During this crash, one of the boards broke off all four heads of
the nylon standoffs. I finally got the
right size standoffs, so we all metal now.
The batteries were secured in the box with epoxied on angle
brackets, which broke loose during the crash, allowing the battery to damage
some other things. I have now cut some
slots in the bottom of the box to allow a big hose clamp to be wrapped completely
around both batteries and through the bottom of the box. They arent going anywhere now unless they
tear the box apart.
We are going to rotate the entire electronics box on the
lander so it is in a more natural position and can be secured with ratchet
straps. The current orientation has the
gyros aligned with the engines, which is convenient, but it isnt a secure
mounting positions. I will have to
correct for the new position in software.
We are going to statically balance the big lander with
ballast so that if the attitude engines are shut down, it should fall straight
back down instead of pitching over.
That is going to require about 30 pounds out by the engines opposite the
High Energy Peroxide Fuels
Michael Carden of X-L Space Systems visited with us last
Friday, and we got some very good historic peroxide documents from him that we
are going over now. So far, the most
interesting point is that some unusual fuels can deliver extremely high Isp and
density-impulse with peroxide:
98% H2O2 and BeH2 has a theoretical vacuum Isp (1000 psi
chamber expanding to 0.2 psi) of 498 (!!!) at a chamber temperature of 3884 deg
98% H2O2 and AlH3 has an Isp of 424 Isp with chamber
temperature of 3923 deg K and a density of 1.58 gm/cm3.
Lox / H2 under these conditions is listed at 470 Isp and a
density of 0.227.
They didn't have a density listed for the BeH2 mixture, and
it seems to be an extremely hazardous material, but the AlH3 sounds
interesting. There was no indication that these combinations had actually
been fired, so it may well be too good to be true, but the possibility of
making a pressure fed SSTO hybrid is sure interesting.
For comparison, an 85% H2O2 / Polyethylene hybrid has a
chamber temperature of 2600 K, so these fuels would have a 50% higher chamber
temperature. Due to the metal content,
they would probably also be much harsher on nozzles.
Long Duration Motor Firing
We fired one of the attitude engines for 60 seconds (three gallons
of peroxide), with everyone a long distance away, because there was still some
concern about the strength of the brass motors if they were allowed to
completely heat soak and reach full chamber temperature.
The motor came through it fine, but the catalyst pack (which
had been used on a few flights already) started tunneling during the last 15
seconds of firing. On opening it up,
there was evidence of both channeling around the outside, and tunneling through
the center. We still need to get better
longevity out of our catalyst packs, because they seem to start wearing out
after about 100 seconds of firing.
There are likely both flow and material issues. We have some pure silver foam coming to try
out, and we are going to try making anti-channeling rings (basically piston
rings) to intersperse with the catalyst discs, and some anti-tunneling buttons
to put in the middle.
Motor Latency Testing
We finally did some accurate timings of the onset and
tailoff of thrust in our attitude engines.
The timings are done at 240 hz, and they are adjusted for the sampling latency
(solenoid signals, which take effect instantly, versus serial A/D, which is
latched and bit serialized).
It takes about 28 msec from the energizing the solenoid for
thrust to begin. I believe that about 5
msec of that is the valve opening time, based on the fact that that is the
minimum pulse time that will let anything out of the solenoid when it is under
pressure. It is possible that the total
time is larger for complete valve movement.
These are already pretty quick valves that draw eight amps of current
and get pretty hot, but that might be possible to juice up a little bit by
changing to an open-sustain split current approach at a higher voltage.
We theorized that some latency could be saved by building a
motor that connected directly on the bottom of the solenoid, and packing
catalyst right up to the top, without a spreading plate, and we tested that
today. This does let a lot of heat
transfer back into the valve, so it would probably require some form of
insulator for production use.
When we first fired the motor on a continuous run for
reference, we got behavior that we have seen at one time or another on all of
our motors, but in an extreme case: it
started off fairly smooth, then abruptly got wildly rough. We attribute this to a compression and
settling of the catalyst pack, which opens up an area that can have things
thrash around in when there isnt a significant pressure drop somewhere
upstream of it. It is possible to have
an unchoked engine run smoothly (in which case the power drops in nearly direct
proportion to the tank pressure), but it is very finicky, and it is almost
always easier to get it running well if it has a significant pressure drop.
We opened it back up and packed a couple more discs in, and
set up a restrictor jet between the solenoid and the feed line. We didnt have any jets in this size, so we
first tried one of the jets from our really old small engine tests. It did run perfectly smooth, but at only ten
pounds of thrust. We then adapted it to
have a 3 fitting in the line ahead of the engine, which is a pretty tight
restriction all by itself. It remained
perfectly smooth, starting at about 50 pounds of thrust, and decaying with the
sqrt a tank pressure, a sure sign we are adequately choking it.
We then ran the pulse timings, which came out fairly
rough. When I first looked at the data, I thought we were getting a
really big improvement in latency, but I was comparing the wrong data. On proper comparison, it seems to be consistently
between 4-8 msec faster. This is
significant, but not enough to bother making new motors and worrying about
We wondered if the roughness was permanent (another cat pack
problem), or if it was related to the pulsing, so we ran another continuous
run, which turned out smooth again. The
intermittent pulsing is definitely causing rough feed flow. The feed hose was vibrating around pretty
visibly during the testing, so something we might want to try next is to test
with the tank rigidly connected to the motor with no flexible link at all.
The flush mount motor initial run, then the run with the 3 fitting
restriction, and the final run (also with the 3) after the pulses tests.
The original engine on top, new engine on bottom.