My autopilot

First let me warn. If you got here by general Internet surfing don't stay. You'll be bored silly. This page is for aficionados and balloon techies. You can get back to my start page by clicking here.

For those still around, I produced a balloon autopilot some 30 years ago, it's still around, and it still works.

Why make an autopilot for a balloon you may ask. Well, at the time I was making a lot of high, long duration, long distance flights over the Alps in Switzerland and Austria. At these heights, 10,000 ft or so, you more or less have to fly on instruments. You are not near enough the ground to sense rate of climb or descent. And let's face it, it's a bit boring having your hand stuck up in the air half the time for a 5 hour flight just to work the burner while you are watching the instruments. Much better to look at the view, and have someone else, or something else, take over the job of turning the burn valve on and off.

To the experts reading this, please forgive my back to basics approach. A reminder of these basics is needed for my autopilot theories. To keep a hot air balloon in level flight you must make periodic blasts with the burner. The burner is far more powerful than needed just to maintain level flight since you must retain the ability to check a descent so in normal flight the burner must be turned on for short periods only. There is a delay of some 10 or 20 seconds before the result of the burn is noticed in the flight characteristic of the balloon, before the "bubble" of hot air reaches the top of the balloon and makes it's effect. The actual delay time is dependent on the size of balloon (in a big balloon, the "bubble" has further to travel, and takes longer) and is also dependent on the power of the burner (the more powerful the burner the quicker the response).

When flown correctly (in "level flight") the balloon will actually fly in a low amplitude sine wave. To maintain this flattened out sine wave type level flight the ideal time to burn is just when the balloon is rounding out to level flight after a climb. You get that feeling, I seem to get it in the stomach, when you're just at the top, and it's time to burn. After the burn, the balloon will round out and start a descent as it cools, and before the recent burn has had it's effect. As the burn takes effect the rate of descent will be reduced and reduced until it takes the positive effect of a climb, which in it's turn will gradually die out since the balloon is continually cooling due to loss of heat mostly by radiation. The balloon again rounds out at the top of it's climb and it's time to burn again and repeat the sequence. (Of course, you're welcome to have other theories, this is mine, and it works).

If the burn was the right length, taking into account the size of the balloon, and the power of the burners, the balloon will now be at the same height as it was at the start of the cycle. If the burn time was too short, you will be lower and if it was too long you will be higher.

The vario, which electronically measures rate of climb and descent, is the ideal instrument for sensing automatically when the balloon is rounding out. I actually used a Ball 400/3 vario which is quite a precise instrument (no longer in production) which was sensitive enough to show a full scale deflection of a very fine 300 fpm as well as a switchable 1500 fpm. With the fine scale in use it is easy to read a rate of climb or descent of only 10 ft per minute. The vario outputs it's information as a varying voltage, and the dial which shows these rates of climb is really nothing more than a voltmeter with a different name and different numbers written on it.

It was relatively easy to use this voltage as a variable input to our own devices to determine electronically just when the balloon was "rounding out". To look at it in terms of the vario dial, after the burn the needle will show a descent for a short while then the rate of descent will level out and then climb back towards zero. When it reaches zero again, with the needle climbing up, that is when the balloon is rounding out at the bottom of the sine wave. The needle will continue to rise as the balloon climbs up and then will fall back as it starts to level out, until the rate of climb drops back to zero. That's the time to burn, rounding out at the top of the climb.

By the way, please take with a pinch of salt any remark that "it was relatively easy ….. electronically". I produced the idea, the design, and the parameters, and supplied the vario, but the electronic genius who put it all into practice was an English balloon pilot Mike Robinson. Without his electronic skills combined with his understanding of my objectives I would not have had an autopilot.

I said the vario dial was really only a voltmeter with different numbers. So, as I said, it is relatively easy electronically to arrange that at this point, which we'll call the threshold, a "burn" signal should be dispatched. Put in a rheostat so that you can manually control the burn length, and a few other bits and bobs to determine when the voltage is dropping to the level when it will show a zero reading on the variometer, and you have an autopilot. It will burn for a set (manually adjustable) length of time at the right time, just when you are rounding out, and, when correctly set, will give you level flight. In practice, I found that this can be set to give a flight where the rate of climb or descent does not exceed 30 fpm, and when left for an hour to it's own devices, the final altitude variation can be less than 500 ft from the start.

I found it takes maybe 3 or 4 cycles to get the burn length set pretty close to where you want it but it does not need to be set precisely to operate. Lets go back to the sine wave. You burn at the top with the correct length burn to maintain level flight. The balloon descends x feet until the effect of the burn kicks in. Then it climbs x feet until it rounds out and it's time to burn again. Lets say you need to set a 6 second burn to keep level flight. It takes half that burn time, 3 seconds, to check the descent and bring the balloon level at the bottom of the sine wave. A fraction of a second more and the balloon will commence a slight climb. Now, whatever happens, the balloon will eventually (but sooner rather than later) stop that slight climb and will round out before descending again. At that point it will trigger another burn.

So if you need a 6 second burn to maintain level flight and you set only 3.1 seconds the auto will operate correctly but you will descend in a series of steps each a touch less than x feet in height. Similarly if you set a 9 second burn the balloon will climb in a series of steps each approximately x feet high.

Therefore, if the optimum setting is 6 seconds burn length, the auto will operate at any setting between say 3 seconds and 9 seconds. It will operate, either climbing or descending, but will require a more accurate setting to produce level flight.

So we had the auto working and able to produce satisfactory level flight. But what about setting a rate of climb, or descent? Now I mentioned the Alps. A basic thing of flying in the Alps is that you take off (and land) at the bottom of the valley. But to go anywhere you have to climb around 8,000 ft to get out of the valley. As pilots, are used to maintaining level flight. It is almost second nature. But one thing I found, as a result, we are not good at holding a steady climb of say 500 fpm for 16 minutes. Our inbuilt mental trigger is preconfigured to maintain level flight, and to burn just when we are rounding out for level flight. We can't help it. To climb we then put in a longer burn, and will climb in a series of steps. To get the auto to climb you could do the same thing and increase the burn length to say 9 seconds when only 6 seconds was needed for level flight, but the climb rate would not be as quick as you might like, since after every burn the balloon has to hit level before it burns again. Just as we would tend to do naturally if we were flying manually.

Back to the drawing board. Auto Mk ll.

The ideal way to climb, in my opinion, is to burn for the same length of time as normal, but to instigate the burn at a point before level flight is reached. In this way you would fly in a rising sine wave. For a descent, you would burn after level flight was reached and a descent had started, and therefore fly in a descending sine wave. You can descend by using frequent shorter burns while still in a descent but whilst I will often descend in this style when flying manually, it is difficult to reproduce electronically since it will require adjustment to the burn length. Easier was to produce a control to adjust the fire point, the threshold, so that instead of burning at level, we could set to burn just before, or just after level flight was reached. Using the same burn length, if we set to burn just before level flight, then we would set a climb in a rising sine wave and similarly, for a burn after the level point, a descent. So we now had two controls. The first to set burn length, adjustable between around 2 and 15 seconds and a 2nd control, to set rate of climb.

The 400/3 vario was switchable between 2 scales. The first showed a maximum scale deflection of 300 fpm, the second a maximum of 1500 fpm. The reasonable maximum rate of climb or descent, excluding cold descents, is around 600 fpm and we used that figure as the maximum and minimum settable rates of climb or descent. The vario has a zero setting adjust which allows you to swing the needle between an apparent 200 fpm climb or descent. Using this it was therefore easy to determine the voltage swings for different rates of climb or descent whilst still on the ground! The size of rheostat could be calculated and calibrated using these swings. So we have control number 2, and Auto Mk ll. With a control to set the rate of climb or descent between +/- 600 fpm.

This control can also be used as a sort of wake up call. If you set the auto to a descent of say 300 fpm then fly level manually the auto will do nothing. However should you be distracted and omit a burn (and this happens to all of us!) then when the descent reaches 300 fpm the auto will fire. This will do a couple of things. First, it will check the descent, and second, it will wake you up……

But then, I wondered, what about downdraughts? If you have set for level flight and are flying along nice and
level, with maximum rates of climb or descent appearing at say a nice low 30 fpm, and you catch a downdraught? When the auto is set finely like this, a downdraught to knock you out of equilibrium, when the preset burn time is insufficient to ever bring you back to level flight. And, unless you can round out at level flight, the burn sequence is not triggered again. You'll just go down faster and faster. It is quite rare to find such a downdraught at the comparatively high altitudes the auto is designed to fly at, in normal ballooning conditions, but it can happen, and certainly did happen when I was testing the auto in level flight at only 100 ft altitude where you will frequently find downdraughts. Of course, one would not normally expect to use the auto to fly at only 100 ft. A similar effect can be noticed if you set the auto for a high rate of climb, and the burn length is not set long enough to enable you to initially reach this climb rate. If this happened when you were flying manually, you would simply put another burn in.

But here comes Mk lll.

This uses a search system. This searches if a burn has been made within a certain (adjustable) period after the previous burn. If a set time has elapsed since the previous burn, and the rate of climb has not reached the previously set burn point, or threshold, this system assumes something has gone wrong, and throws in an extra burn. If that doesn't trigger the primary system, then after a lesser delay, it throws in another. When eventually you are back to passing through the threshold it cuts back to normal operation.

This active search system ceases when the rate of climb goes up through the threshold. Since, if it has gone up through the threshold it is a reasonable assumption that at some time it will come back down through the threshold! To make it simple to spot at a casual glance the system on lamp shows green when it is above the threshold and red when the search system is active. When green, the system will trigger normally, when red it will trigger if the threshold is not reached in a set time. This system is also useful when setting to climb at a high rate, the timer will put in the extra burns if they are needed to maintain the climb rate. The timer for this backup control can be adjusted without affecting the main burn length timer.

The setting of this backup timer control again takes only 3 or 4 burner cycles, and, like the first control, once it has been set for one balloon, it will be very close to the correct setting for subsequent flights with the same balloon in the same conditions, and often neither control will need more than minor adjustment. Since this is only a backup system it's setting is neither critical nor urgent. If it fires earlier than it should you just extend the period a touch. If it fires too late then when it does fire the automatic firing of the burner will alert you that there is a problem, time to think about putting in an extra manual burn and reducing the delay period.

We put in a manual fire button so that you could fly manually (without using the autopilot) just using the electric
valve to save arm fatigue. We put in a red light to show when the electric valve was open. Not needed in flight of course but handy when you are testing or demonstrating the system.

We thought about using an altitude hold. But the 400/3 vario that I was using did not have a altitude sensor, and the later varios, the various models 500 and 600, which did have an altitude sensor, were not so sensitive in normal operation. Mike did modify the circuit to work with a model 500 - Dick Wirth had the idea of selling an auto as part of a Thunder package - but I don't think it worked as well as mine, based on the 400/3. Also, my auto was designed to be used on long, but not extraordinarily long, flights, when the pilot was awake and able to monitor what was happening, as is the case with an aeroplane autopilot. For flights of 24 hours or more an altitude hold would be needed, with either audible warning type or auto hold and you would have to put up with less accurate operation of the auto.

Richard Ball even made me a vario with an altitude output. To add an altitude hold would have been really easy to do given the search systems already running. (If altitude drops below a certain level, chuck in an extra burn). But I found that for the use I wanted, and the accuracy and search system we already had, that it was unnecessary. I think I would add it in though if I was contemplating a solo RTW!

There was another control I had forgotten about because it never had cause to come into action. It frightened me out of my wits when ground testing a few months ago. It was a siren that sounded if after 5 re-burns the threshold had not been reached. Normally you would notice when still hurtling towards the ground at a grand rate of knots after say 5 closely space burns had had no effect, but Mike fitted the siren anyway as a precaution, just in case. So when it went off in my study whilst I was poring over the circuitry I nearly went through the roof. I didn't know you could get so much noise from such a tiny speaker! If you were dozing it would certainly wake you up.

I have flown the system for some 50 flights and 200 hours. It works fine. Apart from the thought of an altitude hold I can't think of any worthwhile reason to plan a Mk IV.

Operating instructions

Click here to go back to the start page