Pelican's Perch
by John Deakin

Pelican's Perch

How I Learned To Love CANPA

Deakin's prior column criticizing the use of VNAV during non-precision approaches provoked a torrent of dissenting reader responses, much of it from Airbus pilots and other glass-cockpit operators. Never one to take this sort of thing lying down, AVweb's resident iconoclast fires back at these pinball wizards and flying videogame players, then discusses the best way for real pilots to fly various kinds of approaches (LOC, straight-in VOR/NDB, offset VOR/NDB) in various kinds of airplanes (glass cockpit, steam gauge, flib).

John DeakinNo, I haven't changed my mind about the usefulness (and the dangers) of the classic early descent to MDA and flying along level looking for the runway, nor the usefulness (and the dangers) of doing CANPA, the so-called "Constant Angle Non Precision Approach" when appropriate. But my previous column obviously needs a little follow-up. Like the seven blind men, each feeling an appendage of the elephant, everyone seemed to have a different perspective of what I tried to say. So I'll say it again, in a different way, and probably make ‘em all even madder.

(Yes, I know, it's four blind men. I prefer my version.)

There are a wide variety of airplanes, auto-flight systems, pilots and approaches "out there," and what works for one combination is deadly for another. It's a very complex subject. My first column did not emphasize this point, and a lot of nice people jumped on me as a result. But keep those cards and letters coming, folks, I love ‘em!

I really am very pleased at the avalanche of reader mail I got. The column also spawned a very long and healthy "thread" over in CompuServe's AVSIG Forum. Regardless of "right" or "wrong," I think many people learned a lot, and most of all, gave the subject some much-needed thought. A columnist can ask for no more.

Some Terminology

My friend, mentor, and sometime-tormenter Randy Sohn takes violent objection to the common shorthand of "Dive and Drive" for the old-style NPA (Non-Precision Approach). He feels it has a terribly negative connotation, for it is not a "dive," it is a controlled, planned rate of descent to the final MDA, with only minimal configuration changes at the start and end. He and I agree that it's very simple to configure before that descent, perhaps note the power setting, make a simple power change (often to a known value), do the descent, and then reset the previous power at or approaching MDA. Or just do whatever it takes to put the airplane where you want it, but that seems to be a dying skill.

Unfortunately for Randy and other purists, "Dive and Drive" is a part of the aviation lexicon, and like it or not, it's a common shorthand that differentiates one method from another. For some pilots, it's a pejorative word, for others an affectionate shorthand. But out of deference to someone vastly older than I am, I shall avoid the use of "Dive and Drive," and see how "Classic NPA" works.

To further define terms, I use "power" here generically, intending to include "thrust." I do love to irritate the purists who get all bent out of shape over the differences between "throttles," "thrust levers," and "power levers." It drove me nuts when my company insisted that in the DC-8 they were "throttles," and in the 747 and 727, "thrust levers," and then actually got uptight when pilots would slip and say one instead of the other! In rockets, according to NASA, it's "throttle up" and "throttle down," so go figure. All I know is "push to go, pull to slow," which works fine in all the airplanes I fly, prop or jet. Maybe we need a new term, to cover ‘em all. We'll also need some sort of variation for the French, who don't even want the levers to move, no matter what you call them.

Pinball Wizards

737-600panel.jpg (115367 bytes)
737-600panel.jpg (115367 bytes)
737-600panel.jpg (115367 bytes)

Those who screamed the loudest at my "blast" at CANPA seemed not to be the real pilots at all, but people like Airbus drivers who are no more than airborne video game players, disassociated from all reality, at least until the ground rises up and smites them with reality -- and finality. For them, if it can't be done with the push of a button or three, they aren't interested. Worse, they assume the whole world must do it the same way, and that anything else is unsafe.

Why, one fellow I know, condemned to flying the French Fry ... er, flying pinball machine, says he only flies it on the auto-flight system so it won't screw up his real flying skills on real airplanes. Smart man.

When one of these button-pusher-drivers is requested to do a sidestep approach, he whines about re-programming the stupid computer! Heck, during my usual hand-flown ILS, my normal excursions probably overlap the centerline of the other runway, so all I have to do is roll wings level at that point, and I'm all set for the sidestep, and a new set of excursions! No problem.

As another side note, didja ever notice airline pilots will scream bloody murder about "safety" unless they have a personal reason for doing something, or there's more money involved? It's really funny how that works. If the captain has a hot blonde waiting, he'll ASK for the sidestep to save a second or three.

As we all know, the real measure of skill for an Airbus pilot is being able to type 60 words per minute on the world's worst word processor. Check your flying skills at the jetway, pal, they're not needed. In fact, they are a detriment. Did you know that when a person is type-rated in any of the Airbii, his pilot certificate is amended to read:


Hey, would I lie?

No, I won't repeat the hoary old joke about the dog, everyone's heard that one by now.

There! All that ought to draw a little fire! Gosh, that felt good!

Seriously, let's do a few real scenarios, starting with the Airbus-type equipment. No, I'm not qualified in any Airbus, and don't want to be. I like flying and computers, but I prefer to have the computers help me, rather than me helping them. Ooops, there I go again, Airbus drivers (never pilots) are such an easy target, and they scream so loudly when wounded.

While I'm at it, let's lump all the 757/767/777 "drivers" in there with ‘em. 747-400, too. They're all a buncha auto-flight freaks, if you ask me.

(In case you hadn't noticed, I thought I'd have a little fun with this column.)

Magic Airplanes


Boeing 737-600
Boeing 737-600

Boeing 777
Boeing 777

Using the Airbus as one example among many, I think we can all agree that there are a number of airplanes flying today that are capable of fixing their position in space with great accuracy, automatically, all the time. (That's great, I love it -- the airplane knows where it is, even if the crew doesn't.) These airplanes usually have some sort of redundant inertial system, updated with constant GPS inputs or double DME inputs. This is all automatic, transparent to the crew -- they don't even have to tune the radios. The system knows where the best stations are and which afford the best geometry, and they auto-tune them without pilot action. All these aircraft have varying degrees of "computing power" that neatly ties in where the airplane is and where the crew wants it to go. In some cases, the accuracy is an astonishing few dozens of feet, or less.

The ideal system is probably inertial updated by GPS. While the inertial system alone is capable of maintaining great accuracy for many minutes (we cross the Pacific on inertial alone), it will eventually drift off by a few miles. Not critical for oceanic flying, but not good enough for terminal operations without updating. With GPS or DME constantly "nudging" it back towards perfection, you can bet your life on it, and we do. Second best is probably inertial updated by double DME. Some airlines and some FAA people prefer the double DME, not trusting GPS quite yet. Eventually, GPS alone may become reliable enough to be used completely stand-alone for all aircraft, but "not yet."

Magic airplane, full ILS

Now, if we take one of these flying videogames and put it on a full working ILS approach, it uses all that "magic" to find the localizer, then the glide slope, where it transitions to the older-style "auto-coupled" approach, often including the landing and rollout. A point worth noting here is that the localizer and glide slope signals converge, and thus get much more sensitive as the aircraft approaches, and much more accurate. The terrain under that slope has been surveyed with a high degree of accuracy, and is protected in many ways by law.

Airplanes of the future will probably all do this, automatically, all the time, without even having the ground transmitters for the localizer and glide slope. I make a lot of fun of it, and some denigrating comments about "drivers" vs. "pilots," but it may even be a better way, I don't know. What I do know is that to me it's not "flying," and I'm glad I served as a pilot during the period I did, and not this new one. I've never flown a HUD (Head-Up Display), but I think I'd like it a lot better. I still do all ILSs hand-flown, unless the weather is so low that only an auto-coupled approach is legal. So I'm a dinosaur -- what else is new?

The above is not the so-called CANPA approach; I mention it only as a starting point, a model to which all other approaches are compared.

Magic Airplane, LOC approach

Now for the next scenario, let's take that same situation, same magic airplane, but with the ground transmitter for the glide slope inoperative. The flying video game will do the same thing, but this time instead of following an external electronic path (a radiated glide slope from the ground), it will use its internal accuracy, its knowledge of where it is at any given moment, its knowledge of the position and altitude of the beginning of the final descent, and its knowledge of the point in space 50 feet above the threshold of the runway to compute its own electronic glide slope in the cockpit. Some systems don't even need the final descent fix, you simply tell them what degree slope you want, and they compute when to start down. The auto-flight system (or even a mere human) can follow that "glide slope needle" just like a real one. Fly level to the descent point and start down, following the needle the system gives you a nice, steady slope right to the runway, just like an ILS glide slope. Good system, I like it -- on a LOC approach. Again, it is worth noting that this "pseudo glide slope" does not get more sensitive as the airplane nears the landing! On the very close final, the old ILS is still the most accurate system in common use. For this reason and others, even with the "magic," a localizer-only approach is limited to slightly higher minima than the full ILS, because it is still a "Non-Precision Approach." One day, if we ever get one of the long-planned enhancements to GPS (like WAAS or LAAS), we may be making true precision approaches and even auto-landings with it, but that day is not here, yet, for line pilots.

The full ILS is a true "Constant Angle" approach and also a precision approach. The LOC only example above is a "Constant Angle Non-Precision Approach," or "CANPA." The airplane will follow the exact same path in space, regardless of wind, drift angle, or speed as long as the needles are centered. If there's a tailwind, the groundspeed will be higher, and the auto-flight system will increase the rate of descent to track the glide slope, whether the autopilot is flying, or whether the human is just following the needles. If there's a strong headwind, the system will automatically decrease the rate of descent, again maintaining that fixed slope in space. Like the full underlying ILS, the terrain is known, and terrain clearance is assured.

I have no problem with that, in fact I rather like it. From a general safety standpoint, it makes the LOC approach just like the ILS (for the "magic" airplane), albeit with slightly higher minima.

Magic Airplane, VOR (or NDB) approach

Be careful, now, this is going to get complicated. Let's take the same airplane, same "magic," and a VOR approach where the final is aligned with the landing runway. Under those very specific conditions, you can turn off the VOR ground station entirely as far as I'm concerned, the "magic" is far more accurate than the VOR system. (The prudent pilot will keep it on and visible as a backup, of course.) You can tell the "magic" the latitude, longitude and elevation of the point to start the final descent (or let some systems compute it themselves), and the latitude, longitude and elevation of the point 50 feet above the runway threshold, and have at it. Go ahead, make it a CANPA, at least to MDA. But you'd better be aware that from MDA down to the runway, you may NOT have the same terrain clearance you'd have on an ILS system. Unless, of course, it's been specifically surveyed (as it would have been for an ILS), you have no guarantees at all. In fact, there are situations where making a CANPA all the way to the runway will put you below obstructions on the way from MDA to the runway!

What's that? I can hear someone saying "So what, I'm gonna miss at MDA if I don't see the runway, and if I do see the runway, I'll see any obstructions."

Uh, huh. Suuuure you will. Tell me, which is easier to see when the rain is streaming across the windshield: the end of a brightly-lit runway or one of those tall TV towers?

You will probably miss more of these approaches by doing the CANPA than if you did it "the old way" (for reasons explained in my previous column), but the increased safety from not letting Airbus drivers do something unusual may well be "better" than getting in those few extra times. That's an economic decision, and not in my job description.

See how complex this gets?

Magic Airplane, offset VOR (or NDB)

Next step. Take the same "magic airplane" and a VOR (or NDB) approach that is NOT aligned closely to the runway. Remember, a so-called straight-in NPA can be up to 30 degrees off the runway centerline, and does not even have to ever cross it. Yes, there is a movement afoot to limit CANPA approaches to 15 degrees off the runway, but "not yet." Again, the aircraft systems are capable of delivering you to a point in space much more accurately than the underlying system, and I like the idea of using the "lateral nav" functions to do that. If you do a CANPA on this type of approach, descend only to MDA, then do an immediate miss when you don't see the runway, you're safe enough, but you're not doing the job most of us are paid to do. If you do the CANPA and see the runway, you are going to be instantly faced with a whole bunch of variables, things to do, and decisions to make. You are going to have to align with the runway, perhaps making a small correction to the glidepath. You are going to have to keep in mind that you may not have obstruction clearance. And finally, you'll have the usual battle getting all the magic disconnected, when it really doesn't want to go away.

If faced with that last approach (30 degrees off the runway) in an airplane with all the "magic," I'm pretty sure I'd prefer the "classic" early descent to MDA. All on autopilot, of course -- God forbid one of these "drivers" should have to fly.

(Didja hear about the Airbus that got stuck in a holding pattern for three solid hours, crew couldn't break it out? They had to get a phone patch to Airbus, find someone who spoke English, and get tech support to help them out? Toughest part of it was getting past the "All of our technical support people are busy helping other customers, your call will be answered in the order received if you speak French, otherwise forget it.")

Non-Magic Airplanes

Boeing 747-200
Boeing 747-200


Boeing 707
Boeing 707

But let's quit picking on the Airbus drivers, and switch now to the airplane I fly, a "Classic" 747 (also known as "Rope Start," "Steam Driven," and other pet names). Triple Inertial Nav, triple-channel autopilot, triple ILS/GS. We'll arrive at the terminal area with the inertial systems showing anywhere up to a few miles error, for we do not have automatic updating, and we need the VORs for the arrival. We can tune in a couple VORs, or DMEs, and update the INSs about an hour out, but once we start the descent, we're switching the nav radios from one station to another too quickly to use them for updating. So essentially, the INS is giving us attitude and compass stabilization (slaving), maybe a rough distance to the next fix or the airport, and that's about it.

In the terminal area, we will navigate in the old manner, from VOR to VOR, or more often, use radar vectors to the ILS approach. We rarely use anything else, because we serve only the major international airports. We'll somehow get on the final approach, follow the localizer, then the glide slope just like the Airbus did above, either on auto-flight, or hand-flown. No problem.

No magic, LOC only

Now turn off the ground transmitter for the glide slope again.

Without a glide slope being radiated, and without the "magic," there is absolutely no way we can generate an accurate glide slope, either for display, or for use, either by the auto-flight, or if flying by hand. Nothing, nada, zip. (This will be an alien concept to Airbii pilots.)

Some operators, perhaps having heard about the "magic" and CANPA, have had what they think is really bright idea. They'll do the trigonometry, and figure "Well, if we hit this point here, and our ground speed is x, and we maintain exactly 763 feet per minute descent rate, we'll have a CANPA, too!" Jepp even charts this stuff on many approaches.

Well ... no, you won't have CANPA. Let me point out a couple wee factors, here.

First, you'd better have available a good, accurate TAS (True Air Speed), or you'd better be willing (and able) to calculate one. How many of you still have an E6B, or a Jepp confuser that will even do TAS, for use when the TAS indicator has fizzled? There are a LOT of cockpits without such a backup, and I've heard there may be one or two without TAS indicators!.

Next, you really need a pretty good idea of the wind that will affect that final descent, and if the wind at the FAF is different from the wind at MDA and touchdown, you're going to need to do some pretty good guessing on the fly to come up with a good average. That's really fun when the wind changes direction, as it usually does.

How accurate is your vertical speed indicator? Ever put a stopwatch on it, and tracked the errors you'll find from airplane to airplane? BIG question, how accurate is the auto-flight vertical speed selector? On the ones I fly, set 1,000 fpm, and you might get 1,500, you might get 500.

Given a point in space, say 7.2 DME from the airport, have you ever watched how many variations there are in pilot technique? Some will cheat a few tenths early, some a few tenths late, some will roll that vertical speed selector to -2,000 fpm for a "brisk" nose over until the actual VSI shows 1,000, then they'll match it, some will wimp it down a few hundred feet at a time. All fine on a "classic NPA" where there are all kinds of errors factored in, but all will have an effect on a "simulated CANPA."

How much effect? Remember, all those errors can be accumulative. But let's take a mere 10-knot total error on a 140-knot final. A five-mile final is 30,000 feet long. 140 knots is 233 fps (feet per second), 150 knots is 250 fps. At 140 knots, it'll take 129 seconds to fly that final, at 150 knots, 120 seconds, for a nine-second difference. Big deal, you say. Nine seconds at 150 knots is 2,250 along track, and on a 3-degree glide slope, that's 117 feet above or below the glide slope. For only a ten-knot error, and no other variations that I listed above.

Suppose your actual vertical speed misses the target by only 100 feet per minute. That's not much error. Maybe you can fly a vertical speed within that tolerance, but I can't. In two minutes, you'll be 200 feet high, or 200 feet low.

Not a big deal? Go try it, and see how bad it looks if you're 100 feet high or low on a real GS, at DH or MDA! A real glide slope puts you within a foot or two of where you're accustomed to being, 100 feet high is downright scary in a big jet!

Of course, the folks who love CANPA love simulators, where I'm guessing the procedure was developed. We don't have real pilots running simulators anymore; they are often just "simulator instructors" with very limited real-world experience. The simulator uses an utterly precise wind model, known in advance, the VSI selector gives exactly 700 fpm when you set 700 fpm, and it all just works beautifully. Folks, I don't give a rat's posterior how well it works in the box, it doesn't work all that well out on the line, in real weather, even assuming perfect technique! Yes, you may have done it for real. You might do it several times for real, and see it work out. But let a few of those errors all affect things in the same direction, and you're not going to be very happy with where you wind up, either high or low when you break out and see the runway.

You just might have been better off to do the "classic NPA."

No magic, VOR/NDB approach

Now it gets really interesting. Remember, with no "magic," there are errors in the signal radiated from the VOR, there are additional errors in the aircraft receivers, and terrain, ice and weather can have additional effects, not to mention a little sloppy flying. No, of course not, you never fly sloppy, but I sure do. Auto-flight systems don't even track VORs very well, and none that I know of will track on the ADF at all (remember, no "magic" here). Try that CANPA approach on these, and you'll be totally unable to predict where you're going to break out, high, low, left, right, early, or late. If you break out at the last moment, you will have a whole bunch of things to put together, and it can be a real scramble, probably not a good idea in a big jet.

Oh, and remember, there's that TV tower that is sticking up into the slope you're hoping to maintain to the runway when you can see it, and it's raining, rough, and dark, with your alternate going down the tubes. Or perhaps this is your alternate, and it would be really, really nice to get in, the first time? Yeah, right, I know, your alternates never go down the tubes, and you've always got a nice, above-minimums alternate. Riiiiiight.

A Real Non-Precision Approach

Nope, for my money, if you're going to do NPAs and you don't have the "magic," you're simply better off, more likely to make it in, and probably safer doing it the classic way. Hit that final fix in the minimum-flap landing configuration, checklists complete, and note the approximate power (thrust) setting. Start a positive descent somewhere between 1,000 and 1,500 fpm, just before reaching MDA reset the previous level-flight power, and hold MDA or slightly above, while glancing outside, looking for the runway. Use any reasonable means to get a better shot at the runway, like "Mr. Brown's Barn" if you know the approach that well (if you don't, fly it as charted). Once visual, as you see the proper glide slope developing (from below), start easing it down to intercept it, all the while watching for that TV tower.

If you can't do that, then don't do them at all.

Bonanzas and NPAs

Most of the same comments apply, except small airplanes are far more maneuverable, and can tolerate being high or low, left or right, far more readily than the big jets. In my Bonanza, on a 10,000 foot runway, I can be over the end at 500 feet, and still land safely. You can't do that with a big jet.

Remember, the NPA was designed from the outset not to put you in a landing position, but to simply deliver you to the general area of the airport in weather suitable for maneuvering to land. Some are better than others at doing that, and some will even deliver you pretty close to the end of the runway. But you can't make ice cream out of equine waste, and you can't make a precision approach out of a non-precision approach.

On A Personal Note...

It is Christmas Eve as I finish this column, past deadline, as usual. If the powers at AVweb actually consent to posting this, the 25th Pelican's Perch, it should appear on or about January 3, 2000. Assuming most of you are Y2K compliant and have made it through that overblown barrier, please allow me to wish you all the very best in the coming year. I'm very fortunate, have perfect health, few problems, far too many things to do, and above all, some wonderful friends, many of whom I've never met in person. I refer to you, my readers, as well as others. You know who you are.

Be careful up there!

...John Deakin

John DeakinJohn Deakin is a 34,000-hour pilot who worked his way up the aviation food chain via charter, corporate, and cargo flying; spent five years in Southeast Asia with Air America; and joined Japan Airlines 31 years ago, where he is a 747 captain ... oops, we mean copilot. He also flies his own V35 Bonanza (N1BE) and is very active in the warbird and vintage aircraft scene, serving as an instructor in several aircraft (including the Lockheed Constellation) and as an FAA Examiner on the Curtiss-Wright C-46, DC-3, and Martin 404.