Star Ford

Essays on lots of things since 1989.

Complete and incomplete covers in engineering

on 2017 May 24

I confess I have been irritated my whole life about car dashboard controls for heating and cooling because they are an incomplete cover for the complexity that is going on inside. It has been a rough few decades for user interface enthusiasts!

What is a complete cover? It is a layer or shell over some machine complexity that completely hides it and does not let any of the complexity out. A cover is incomplete if it forces you to understand what is going on underneath, or if it is confusing when you do not understand, or if the cover is insufficient to operate all aspects of the machine. A cover can be thick or thin – the thicker the cover, the more it changes the paradigm of the machine interaction. A cover is optimal when it is complete, regardless of whether it is thick, thin, or absent. Sometimes it is optimal to have no cover.

I will explain this with some of examples, starting with a mechanical mercury thermostat. There are three kinds of people in relation to these devices: (1) Those with a gut fear reaction when they look at dials and numbers; (2) those who understand the two exposed dials – measured temperature and set point – but do not know or care how it works inside; and (3) those who understand that the rotation of the temperature-sensitive coil which is superimposed on the rotation of the set point tips a mercury switch, that the bi-stable 2-lobed shape of the mercury chamber affects the temperature swing, and why mercury is used in the first place. It is a lovely thing, but not really the scope of this paper. I am mainly concerned with the middle category of people who are functional operators of the cover and what kind of cover it is.


The thermostat is a complete cover because you can operate every aspect of the heater with it, without needing to know how it works. It is also a fairly thick cover in the sense that it translates one paradigm to another. The actual heater requires an on/off switch to work, thus the only language it understands is on/off. But the thermostat exposes a set point to the user. It translates the language of on/off to the language of set points. Someone could replace the whole heater and wiring with a different inside paradigm but leave the exposed paradigm there, and the user would not need to know that anything changed, because the operation stays the same. In many systems – especially software systems, the replaceability of layers is an important design point, and complete coverage is one of the factors that makes it possible.

Another example is a machine with a thin cover – a gas lawn mower. The exposed controls on small engines – fuel flow, choke, and the starter – control the machine directly with no paradigm translation. In order to use it effectively, you have to know about the possibility of flooding the engine and you have to know what the knobs are really doing. (The reason small engines do not have a thick cover the way a thermostat does is just because people would not want to pay a lot extra for it.)

I would say that in the case of a small engine, an optimal cover would be thin and complete: one that labels the knobs by what they do inside rather than by pretending anything else. Some engines are labeled with “start/run/off” and “low/high” which is misleading because it suggests that the cover is doing a paradigm translation when it really is not. A user who is misled by the labeling might put it on “low”, and “start”, pull the cord a bunch of times and flood the engine, not knowing that it will not start on “low” and that “start” really means choke. Having that misleading labeling makes the cover thicker but less complete, and therefore worse.

The worst of all possible combinations is a thick, incomplete cover. If you have a machine with a thin or no cover like the lawn mower, you only need to know how the machine works. If you have a machine with a thick complete cover, then you only need to know how the cover works, not the machine. But if you have one with a thick incomplete cover, then you need to need to know both how the machine works and how the cover works; thus you need to understand and think in both paradigms when the incomplete cover is basically trying to confuse you.

Car climate controls are usually thick and incomplete, and this is why I am losing sleep over them. I will break it down into some design options to see why I am saying that. Suppose we wanted a complete thin cover: it would have knobs labeled with what they do and each knob would be connected to one internal part. So a knob might be labeled “intake louver” or “condenser on/off”. To use those controls effectively, you would need to know how the air flows through the pipes and where the heat exchanger and condenser are placed in relation to the louvers and fans. But knowing that, you could get the exact results you want.

The opposite alternative would be a complete thick cover. In that case there would be a paradigm or interface language that the user would think in which would completely describe all possible human variables. These variables are not the machine parts but the variables of interest to the person. Those variables are (probably): Temperature; Air speed; Energy use; Noise; and Air quality (replacing stale air while also avoiding noxious outside air). I am not including defogging the window as a human variable because that is related to visibility, not climate control. Ideally those things should be controlled separately.

I will get into some design points on the cover paradigm below, but basically you would want to be able to control all those variables in a clean intuitive way.

What we have instead in non-fancy cars is a mixture: some controls are thin like the fan and AC – they just turn on one component. But then things get confused with a mode selector and “max” button that combine functions of defrost (which is AC), recirculation, and AC boosting. The temperature dial usually suggests it goes from hot (red) to cold (blue) when in fact it really just controls heat. Few non-engineering types can easily explain what all the controls do or even what the paradigm is.

Design (A) below is the typical low end design, but “MAX” is not really defined and the center knob does not control cooling despite its blue color.

Design (B) has sixteen buttons – I do not know where to even start with that.

actual controls

Design (C) in the picture is a bizarre thing with seven controls, the word “A/C” is shown three times, and “MAX” is shown twice. What is the difference between the cracked window icon on the left and the triple wave icon on the right?

Fancier cars with thermostats do not always fix the problem. For example in the Prius (pictured below) you can select an automatic set point, but it also allows you to change the fan speed and other variables after you have chosen automatic, which makes it ambiguous. It also does the irrational thing of heating and cooling to the same set point. For example on a winter morning if you want to heat to 72, and then later in the day the car happens to get to 73, that does not mean you want to turn on the cooler, but the Prius turns it on. You are probably OK up to 78 or so because the day warmed up, and people have a range of acceptable temperatures based on how warmly they are dressed, not an exact target temperature.


If you are a careful operator you have to deal with several controls to get the desired effect – fan, heat, AC, mode, and recirculate. Or you just haphazardly turn on or off whatever seems relevant without looking at the whole system. Because who wants to do all that paradigm translation just to drive?

I think we get to weird places like this because there is no profession for cover design. Instead we have engineers saying “we’ll make whatever you want” and product designers pushing for certain “features” that have marketing value. But the product designers think about it as baseline (the way it was always done) plus new features rather than rethinking the cover to have a single paradigm.

So here is my suggested replacement – a complete cover with a really understandable paradigm and only three controls instead of the usual five or six.

complete covers

Here is how it works. If you push all three sliders to off, you close off the outside world, and every system is off. If you want it cooler you slide the cold slider up. If you want it hotter you slide the hot slider up. Since you cannot have it both ways, the hot and cold sliders are connected by a cable that pulls one down when you slide the other up. You can have them both off, but you cannot have them competing against each other. If you want more fresh air, you push the fresh slider up. If the temperature is fine but it is too breezy, you slide the fresh slider down. The reason for the diagonals is to map visually that the hot and cold are mutually exclusive, but at the same time, up means active or on, while down means off.

The beauty of this paradigm is that it handles everything with three controls, people can understand it, and that you can set it and forget. The center point on each slider can have a sticky spot where it clicks into place, where most people will find it comfortable: a mixture of fresh and recirculated air, cooled to 78 and heated to 72.

Here is the way the cover translate to the machine. As you move the cold slider up, it changes the set point on a display between 85 (towards the bottom) to about 65 (the top). The fan and operation of the cooler are adjusted to achieve that temperature, and most of the air comes out the front panel, so you get added wind chill. As the actual temperature gets close to the set point, the air speed lowers.

As you move the heat slider up, it changes the set point on a separate display between 55 (towards the bottom) to 85 (the top). The fan and louvers and adjusted to achieve that temperature. Most of the air comes out near the floor. The cable (or other device) ensures that the heat set point is always at least two degrees below the cooling set point.

The fresh slider has a bottom position that closes off outside air, a middle region that gives some air with a mix, and an upper position that gives the most air, all from the outside. The actual air speed is based on the maximum need of the fresh slider and the temperature set point that is in effect.

The electric fan can be used to boost actual air speed, but it only has to run when there is not sufficient ram air flow. The point is not to control the devices directly, but to control the three variables of interest to people. Actually I listed five variables above; three of them (heat, cool, and wind) are controlled by the sliders. The other two (noise and energy use) and limited by turning all the sliders down or off.

I have not forgotten the defrost control, but it should not be part of the paradigm of climate control (even though it uses some of the same components in the machine). The defrost control would be up near the window and it would be an on/off switch with a light. If it is on, it does have to run the AC and fan to get the defrost effect, but it does not change the set point and therefore should not affect what you are doing with climate control.

I have tried to show that designing a complete cover allows the cover to translate paradigms, which meets two needs: one, the need for a clear human operator paradigm that is unencumbered by the needs of the machine, and two, the need for engineering flexibility in achieving the optimal underlying system. The pattern of complete covers allows both needs to be met well without having to compromise for each other. The pattern extends to any kind of machine or household appliance, and software systems as well. It is not an established field as far as I can tell, and its application is pretty hit-or-miss. Some toasters utterly fail, while others do fine. Some web sites have it and some do not. Some web sites have it and then undergo and “upgrade” and completely lose any rational cover. It seems to depend on specific people involved and their intuition, and not on any established principles or anything you learn in college. The opportunities are enormous for a change in the way things are designed by using and teaching this principle.


One response to “Complete and incomplete covers in engineering

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: