Users of fan systems would like to experience a steady, continuous flow of air. In this ideal situation, the pressure generated by the fan is constant. A single, instantaneous measurement of the flow rate would be valid for extended periods of time.

Figure 1 (page 7) shows the flow for an ideal system. Figures 2 through 5 (pages 8 and 10) show a variety of conditions for time variation of flow.

Those involved with measurement of flow rates know that ideal flow conditions are not common. Each point of flow measurement is usually time-averaged for 10 sec or more to get an accurate reading. Variations in flow and pressure readings of 10% over short time periods are relatively common.

However, fans that are improperly selected or applied can produce variations far greater than this. Conditions can become so severe that the flow through the fan can oscillate between forward and reverse (flow exiting the inlet) many times per minute (see Figure 4).

Variations in flow and pressure not only make it more difficult to measure the flow, they can create a variety of problems:

  • Dramatic increases in noise;

  • Increases in vibration;

  • Structural fatigue damage to the fan due to continuous loading and unloading of components;

  • Damage to ductwork and other system components; and

  • Fan system interference which prevents the fan from performing its intended service, due to unsteady flow and/or transmitted vibration.

An understanding of the causes of unsteady flow can be helpful in avoiding these problems. Because some of the causes are complex, researchers have shown some interest.

There has not been uniform agreement in the conclusions about what the exact causes are. However, from their research we can learn the conditions that tend to perform normally, and avoid the conditions that don’t.


When elongated objects (such as fan blades) pass through the airstream, they deflect the air. If we change the orientation of the object relative to the flow direction, we can increase or decrease the amount that the air is deflected. If we progressively increase the attack angle of the fan blade, it will increase the amount of deflection.

This change of direction (and relative velocities) allows the fan to generate pressure. If the attack angle becomes too severe, air will no longer follow the blade surface uniformly.

The amount of deflection and the pressure being generated stops increasing and normally will fall off. This is called the stall point.

In a fan, the blades normally rotate at constant velocity. Therefore, to change the angle of attack, the system to which the fan is attached must be changed. Higher flow rates through the inlet increase the attack angle; lower flow rates decrease it.

Therefore, if a fan is operating in stall, it is because the cfm is too low. On a given system, this is caused by selecting a fan that is too large (making the air velocities too low in the fan).

In some fans, the angle of attack is not uniform across the full width of the blade. These are normally not the most efficient fans, although the severity of the stall is often less since only part of the blade is stalling at any one flow rate.

Some people say that radial-bladed centrifugal fans are always in stall, because there is a poor match between the directional velocity of the blade and that of the approaching air. This is essentially true. However, these types of fans can have severe time-varying flows at very low flow rates, since the internal losses are dominated by stall and the pressure falls off at this point.

A fan operating at or near the stall point usually will have severe increases in noise. On some fans it will sound almost as if the impeller is being impacted by a solid object (hammering). Pure stall tends to have a random frequency, but there are special cases where a pure frequency is generated. This will be discussed later.

There is a time-varying nature for the flow of a fan in stall. However, this is normally not the major cause of concern. The increased noise being generated can be a problem, but this too can be dealt with.

The major concern for a fan operating in stall is the potential for mechanical damage. Those who have had a bumpy airplane ride have a feel for how severe aerodynamic shock impacts can be.

A fan continuously operating in stall can sustain structural metal fatigue. This is especially true for axial flow fans having long, slender blades, or blades fabricated from sheet metal.

Centrifugal fans are less prone to damage. Centrifugal fans designed for relatively high pressures but operating at very low pressures (less than 1 in. sp) have been known to operate continuously in stall for many years without damage.

There is another downside to having a fan operate in stall. It means the efficiency of the fan is less than optimum. A smaller size fan costs less and has a lower operating cost. It will also likely outlast a larger fan.

Rotating Stall

This is a special case of stall that normally only occurs in high-efficiency, backwardly inclined centrifugal fans. Most observers also report that inlet vanes are involved.

These fans are encased in a scroll-type housing that helps generate the fan’s pressure. The pressure around the periphery of the fan wheel varies relative to how near it is to the fan outlet (where it is highest). These fans have several blades, typically nine to 12.

We will call the passageway between each blade a cell. The flow through each cell can vary since the pressure around the periphery varies. Near the stall point it becomes possible for most of the cells to have the normal forward flow, while one or two cells have reverse flow.

The air that “squirts” backward through these cells has nowhere to go so it moves to an adjacent cell, deflecting the air which was already traveling through it. This change of attack angle now forces this cell to stall. It then also has reverse flow, passes on its bubble of air, and on and on around the fan wheel.

Most researchers have reported that the frequency of travel of this rotating stall occurs at about two-thirds of the fan rotational rpm. Some have observed two traveling cells at once generating a four-thirds-rpm frequency.

There are other reports of rotating stall ranging from two-thirds to more than 90% of the operating frequency. This frequency will show up in both sound and vibration measurements, but it is normally found by complaints of noise.


Several years ago there was a report of a grain-drying system that was pressurizing large grain bins. It used a burner section near the inlet of a large centrifugal fan.

Periodically this system would “belch” fire back out the inlet of the burner. This was likely a severe case of system surge.

The sound a surging fan makes is commonly described by observers as “whoosh” or “whoomp.” Several criteria must be met to have surge:

  •  There must be a relatively large volume of air which is pressurized (such as grain bins or a large plenum).

  •  There must be a section of ductwork with relatively high velocities.

  •  The operating point of the system is to the left of the peak pressure (at lower flow rates). In this region the fan curve has a positive slope, such that increasing the flow also increases the fan static pressure.

In concept, a system in surge is like an oscillator. The energy imparted to the air alternates between creating kinetic energy (high velocity in the duct) and potential energy (compressing the air in the plenum). The positive slope on the fan curve allows large amplification of this oscillation to occur.

In extreme conditions, the air can temporarily blow back through the inlet.

In a fixed system, the frequency of surge is constant. Usually the frequency is low enough so that you can count the number of cycles per minute (cpm); it is quite audible. Most severe reports occur at a frequency below 300 cpm. One researcher reported that this effect seems to disappear at frequencies above 450 cpm.


Some people use the term “hunting” to apply to any time-varying flow. However, the appropriate use of this term should apply to an under-damped control circuit.

In variable-volume systems, sensors are used to provide information that controls dampers, vanes, speed controls, or other means of setting the flow rate. If the control system responds too quickly, it will overcorrect and have to readjust in the other direction.

In the extreme condition, a system may continuously hunt back and forth.

Stability and bi-stable flow

Stability refers to the ability of a system, if temporarily displaced, to return to its nondisplaced position. A coin balanced on edge is unstable. A coin at rest on a flat surface is stable.

Some fans are not stable for all flow ranges.

Walking by the inlet (don’t try this!) of a large centrifugal during an air test caused the flow to reduce by over 15%. This fan continued to operate at the lower flow rate until the test was restarted.

We can determine the stability of a fan by performing two air tests. On one test, we start at full flow (free delivery) and measure the flow and pressure as we progressively add resistance. In the second test, we start at shut-off and progressively reduce the resistance.

We now have two flow vs. pressure fan curves. If they do not overlay, we have a region of instability. Since there are only two possible conditions on any system, this is called bi-stable flow.

Although the noise changes between the two flow conditions, neither is particularly objectionable. If the fan is rated in the high-flow condition and it trips to the lower condition, the loss of flow can be a problem.

Bi-stable flow has been observed in backwardly inclined centrifugal fans, usually at performances close to free delivery and almost always at flow rates higher than that where the best efficiency occurs.

Parallel flow

It is relatively common for two or more fans to be operating in parallel. In the two-fan system, each fan is selected for half of the design flow rate.

Fans that have a large dip in the stall region can have another type of problem. Vaneaxial and forward-curve centrifugal fans both can have large dips.

The problem with parallel-flow systems can occur in the starting sequence. If the fans are properly sized and started simultaneously, there is no problem. However, if one fan is started first, the second fan is already exposed to back pressure while it is coming up to speed.

At full speed, a condition can arise where one fan is operating at a flow rate to the right of the peak static pressure point, while the other fan is trapped on the left side of the peak.

It is quite possible to have two identical fans not sharing the load equally. A more severe condition can exist if non-identical fans are operating in parallel.

Some years ago, a complaint regarding a system with two fans in parallel was received from a customer. After installing a second, larger fan in parallel with a smaller fan that had been in operation, the combined flow wasn’t what was expected.

Measurements revealed that the second fan by itself was generating more pressure than the first fan was capable of at any point on its fan curve. The original fan was completely overpowered, and flow was blowing back out of its inlet. The customer was advised to shut off the original fan (saving power) and block solid the duct branch to the original fan (plugging the leak).

There were two lessons learned here:

1. Don’t mix two different fans (or operate two identical fans at different speeds) for parallel operation.

2. If more flow is required on a constant system, boost the pressure capability of the fan or add a second fan in series.


The words stall and surge often “put the fear” into inexperienced contractors applying fans.

We all would like to simply plug a fan into a system and have a continuous, steady flow. It would be nice if system calculations were ultra-precise, making it easy to avoid bad operating points.

However, in the real world, fans are often applied in less than optimum conditions, and many times in conditions where stall is likely. Even then, severe problems are rare.

When problems do occur, there are methods to identify the problem type, and once identified, solutions can be implemented.