Here's another section of my autobiography on STS-37 landing short of the
runway.

Danny Deger

__________________________________________________ _______

At about this time, I became the Entry Training Flow Supervisor. In this
position I was overall responsible for training the crew for the entry phase
of flight. In this context, entry is all the way from the end of the
deorbit burn to landing. Within days of my taking over this job, a shuttle
landed short of the runway. Fortunately, the landing was on the lakebed and
Edwards Air Force Base, where landing short is not a problem. If the
landing had been at Kennedy, we would have lost the orbiter and the crew.
The short landing was a classic case of a chain of errors. In aircraft
accidents (in this case a very close call) there is almost always a chain of
events. If any one link in the chain was not made, the accident would not
have happened. It is very rare that a single event causes an accident. The
short landing of STS-37 started with the weather. The winds at the time of
landing were the strongest the shuttle had every flown in. This was true
for the high altitude winds as well as the winds on the surface. The
original plan was to land on the concrete runway, which has the Microwave
Landing System. A helium filled balloon was launched at Edwards. A radar
on the ground tracked the balloon on its ascent. Based on the motion of the
balloon, winds at altitude were calculated. This data was fed into a
computer simulation of a shuttle entry. Based on this simulation, the
shuttle was predicted to be low on energy as it rolled out on final, but it
will make the concrete runway OK. But then the surface winds pick up and
the concrete runway is out of cross wind limits.
Wayne Hale was the flight director and he made the call to land on a lakebed
runway that is pointed into the wind. Unfortunately, there was not enough
time to run the simulator through the expected winds to make sure the
shuttle will be OK. At the time of the flight there was no rule to do this.
The rule is in place now. If the rule had been in place at the time, the
landing wouldn't have been attempted. A post flight run of the simulation
showed the shuttle being very low on energy as it rolled out on final.
Meanwhile a heavily modified Gulfstream aircraft, called the Shuttle
Training Aircraft - STA, makes an approach to the lakebed runway. The pilot
reports a massive wind shear at 7,000 feet on final. The winds are such
that 20 knots of precious airspeed is lost at 7,000 feet. This important
call was made to the ground and discussed at length in the Mission Control
Center. But for some reason, Steve Nagel, who was the commander of STS-37,
was not told about the wind shear. We now had in place at least two links
in the chain.
Steve and his crew performed the deorbit burn and let the computers fly the
shuttle down to Mach 1. At this point Steve took over. Unfortunately,
Steve had to fly a right hand turn to the lakebed runway. He had been
scheduled for a left hand turn on the concrete runway, so almost all of his
training was with left hand turns. The flight director needles commanded
him to start his turn, so he did. Now Steve made a mistake. He was so
interested in finding the runway early he started to look outside. Because
his seat was on the left, he couldn't see the runway until very late.
Meanwhile, the shuttle flew into a tail wind and commanded Steve into the
maximum bank allowed - 60 degrees. Steve missed the command and maintains
the more normal 45 degrees. Without knowing it, the shuttle was flying wide
and losing energy fast.
Finally Steve picked up the runway. He knew immediately he was in trouble.
He was low energy. He rolled out on final with less airspeed than he should
have. He thought he was OK, and based on the information he had at the time
he was. But then he hit the wind shear and lost 20 knots of his already too
low airspeed. Steve knew immediately he was not going to make the runway.
He drops the nose to get back his airspeed. This put him too low to make
the runway, but it was better than running out of airspeed while still in
the air. He planned to land on speed, 195 knots, but well short of the
runway. As he approached the ground, his velocity vector was telling him he
was going down too fast. He pulled up in response to this indication from
the velocity vector. It turned out the velocity vector had a significant
error because this lakebed runway has no MLS.
Steve ended up landing at 165 knot airspeed 1,600 feet short of the runway.
Everyone thinks he was slow because he was attempting to make the runway. I
thought this for about a year. I finally had a chance to have a one-on-one
interview with Steve where he told me this story and straighten me out on
why he landed so slow.
It is hard to believe we were doing such a bad job of teaching the manual
phase of flying from Mach 1 to rolling out on final, but we were. The pilot
usually takes control at Mach 1 on entry and hand flies the shuttle the rest
of the way. This is about 80,000 feet altitude. We had in the training
flow a single class in the simulator to teach this phase. A big problem was
the flight director needles were turned off during this class and the
student did a 100% manual flying task. This is not the way the shuttle is
flown. The flight director needles are on and used extensively, but the
pilots had zero training on how to use the needles. Even worse, this phase
was not trained as the pilots came out of the pilot pool and were trained to
fly an assigned flight. It was common when I took over for the pilots to
not have taken this class in years.
One of my first duties as Entry Training Flow supervisor was to upgrade
training of this important task. First of all I added a 4 hour class to be
taught every time a crew was selected for a flight. Second I modified the
class to have the flight director needles on during the entire class. If
the crew had needles on the actual entry, they were going to have needles on
during the class.
As I was developing this class, many pilots told me based on landing the
Gulfstream aircraft modified to fly like a shuttle, the needles failed a
lot. I did some research. The Gulfstream can't go to 80,000 feet to do an
entire approach. Typically they go to about 20,000 feet and fly the last
portion of the approach. Many times the needles go "goofy". I worked with
the Gulfstream instructors and finally came up with the answer. The
Gulfstream at 20,000 feet is going much slower than the shuttle would be at
this altitude. The shuttle software "gets lost" and thinks the pilot wants
to make another complete circle before landing. The commands to the pilot
via the flight directory needles become completely unusable. If the pilot
were to follow the needles, the shuttle would crash. Based on this, the
pilots had lost confidence on the needles. I worked with the Gulfstream
instructors and we were able to increase the speed a bit. Safety
considerations would not allow them to fly actual shuttle speeds. But I did
get them to teach the needles often going "goofy" would happen very rarely
in the real shuttle. We have had no more problems in this phase after my
training changes were put into place.