Executive Summary:
This report on hydronic cooling attempts to provide context while offering a detailed examination of pump controls, focusing on which pumps operate at variable or constant speed, their corresponding configurations—(i) Primary Only, Constant Volume, (ii) Primary Secondary, and (iii) Variable Primary—and the conditions under which variable–speed pumps adjust and maintain differential pressure for their system. Each configuration is treated as an iteration of the previous one, with increasing emphasis on cost and energy efficiency by minimizing pump usage through speed control (allowable through VFDs or AFDs). The trade–off being that more controllability often comes with greater complexity (more sensors). Note that in a closed system, the pump creates flow by developing pressure head, and the relationship between the two is unique per each pump. Therefore, the pump is tasked not only to maintain the set differential pressure, but the designer must also carefully determine which inputs it responds to, guided by affinity laws, pump curves, and its capacity for optimal performance. This involves an interplay between equipment, including automatic pressure–independent valves, temperature monitors on supply and return lines to ensure appropriate delta T, protective flow meters at chillers or zones, and pressure sensors that measure resistance and drops. Meanwhile, considerations such as area or curve control, configured setpoints (Trim & Respond), optimal operating conditions (e.g., BEP, min/max flow, life–cycle), the addition of multiple pumps or chillers in parallel, and equipment protection measures like staging or rate of change requirements by ASHRAE further influence system variables. Overall, all of these components must work together to adapt and meet a building’s load, requiring comprehensive hydronic balancing in the process. The objective then is to optimize throttling and prevent starvation (on further zones) or over abundance (on closer zones) of flow, as variable–speed pumps alone cannot guarantee issues like low valve authority are solved. ASHRAE’s pursuit is energy efficiency, and the trend has been toward reducing pump speed—likely to advance further with predictive analytics, internet of things, and artificial intelligence. However, the evolution and potential emphasis on sensors in the future will be particularly fascinating, as faster and more precise feedback directly translates to cost savings (aa).
Before proceeding, please note that I am not an expert, just a sleep deprived engineering student that tried his best :)
aa. Footnotes will be used as additional comments, not sourcing.
Introduction:
To preface, this research report requires answering the foundational question: What does the term “Hydronics” actually mean? Fortunately for us, we have that definition readily available, “a system of heating or cooling that involves transfer of heat by a circulating fluid in a closed system of pipes” [1]. In any forced system (ab), a pump facilitates this delivery and maintains the flow of heated or cooled fluid to a conditioned or zoned space [2]. The fluid itself acts as sort of a “conveyer belt,” not necessarily the source of the heating or cooling; it places thermal energy toward a destination [3]. For heating, the liquid absorbs the energy at a heat source (like a boiler or heat pump (ac)) and distributes it to where you want it to be (placing heat via emitters) [3]. For cooling, the liquid absorbs unwanted heat from a space (with fan coils or radiant panels) and carries it away to a heat rejection source (such as a chiller and cooling tower) [3]. Note, most hydronic systems use water due to its high specific heat capacity and its relatively large density value compared to other liquids. To put this into perspective, a given volume of water can contain almost 3,500 times as much heat as the same volume of air for the same temperature change [4]. Additionally, water is a renewable resource, making it both efficient and sustainable. Fluids such as glycol mixtures or brines are also used in specific applications where freezing protection or other properties might be necessary (ad)—careful, these solutions can be expensive and sometimes adding too much will void warranties due to corrosiveness or build up (ae) [2]. That said, setting up such a configuration is no easy task, and this research document will focus primarily on pumps in recirculating chilled water systems rather than heating systems.
ab. The alternative is a gravity system, though this is increasingly becoming less common (if not already gone) [2].
ac. This is not particularly relevant, but modern heat pumps technically can “act” as chillers through reversing valves (change direction of heat movement) [5]. The choice to do so relies on where the system is and what you need.
ad. Depending on the solution, the engineer typically recommends a seal or pH for others to use in their own selection.
ae. In fact, the safest protocol is to abide by the warranty and comply with ASHRAE guidelines.
Hydronic cooling systems, though some configurations more “modern” than the next, have existed in multiple forms for centuries. Take the Ancient Egyptians, for example, who hung reeds in windows and trickled water over them to cool incoming air and absorb heat from sunlight—a type of evaporative cooling [5]. Or consider the Romans, who ran aqueducts through their walls as a rudimentary radiant cooling system [5]. Even the Persians had wind towers that would circulate air down, through, and out of a building [4]. Notice that there are complexities and novelties associated with each variation and era that deserve our respect. However, we now live in a world where many hydronic systems integrate sophisticated controls that continuously respond to dynamic load changes. What are load changes, you may ask? It is simply the amount of heat energy that needs to be added or removed from a space to maintain or receive a “comfortable” temperature (af) [6]. An excess or deficiency in heat transfer leads to optimization problems, which likely translate to cost problems down the line. But the actual load might include factors such as building insulation or window placement; occupancy, since every person acts as a small 98.6 F heat source; equipment usage, as an office full of running computers, appliances, or lighting adds considerable internal heat; and outdoor weather conditions, like the sun striking the west–facing side of a building in the afternoon. Hence, if it is comfort and efficiency we wish to achieve, a hydronic system must repeatedly adjust its rate of heat delivery to the building. Or, at the very least, already be equipped with a response on hand.
af. A building load profile is required (see HM Figure 16), as it provides the most realistic representation of the load. Plus, it is useful for making decisions that account for design day conditions, which is when load is at its worst [37].
Let us relate this back to the pump, whose supplied flow is dictated by the building load, found using what is often called the most important equation in hydronics:
Here, building load is expressed in BTUH, which means British Thermal Units per hour. The flow is denoted as gallons per minute. The number 500 comes from one gallon of water weighing 8.33 lbs., then multiplying that value with 60 minutes in an hour. And lastly, delta T is the sensible heat (ag) absorbed or released by the water—this is typically placed at around 15 (ah) to 20 F for chilled water systems [7][8]. With this, we can start by assuming a basic control model: the temperature is set to a specific cooling point on a thermometer, the system responds to the load by transferring heat elsewhere, and in many cases, a variable–speed pump modulates its flow through VFDs based on its own operating conditions (ai). There are several strategies for controlling the pump itself, which will become clearer as we explore later sections. For now, it is useful to keep in mind some of the main obstacles, like low delta T syndrome and condensation (or falling below dew point (aj)). These can be briefly described next: Suppose “you” (the pump) run a delivery company, and the “trucks” (the chillers) in place are constantly operating at only a third of their “loading capacity” (low delta T). To meet “customer demand” (the required load), you would have to send more trucks, just as a hydronic system must increase flow to compensate, leading to inefficiencies. Not only does this increase wear on equipment, which have their own life cycle costs and rated outputs, but per the pump affinity laws, energy use is proportional to the cube of speed (RPM), which itself is linear to flow (GPM) [2][9]. This means doubling the flow rate (by doubling speed) results in practically octupling energy consumption (this is a lot). Further, when water moves too quickly, the coil surface stays cold since there is no time to effectively absorb heat or properly dehumidify the building. If the coil temperature falls below a room’s dew point, condensation will begin to form. This brings us to our second obstacle, as unmanaged moisture can lead to mold, mildew, and other costly damage [4].
ag. Sensible relates to change in temperature, while latent relates to humidity—both will be mentioned in this report (check HM Figure 14). Further, heating systems do not mess with humidity unless a humidifier is added, while cooling systems must handle latent as well as sensible loads [5].
ah. According to ANSI/ASHRAE/ISE Standard 90.1 2016 (might be outdated), this is the absolute minimum the delta T should be [7].
ai. This model will be updated later on, but the intent here is to establish the general idea.
aj. Dew point is the temperature at which air is fully saturated and can hold no more water; ASHRAE Guideline 36 2024 considers it a “humidity variable” [30].
Components:
Alright, so there were some equipment–based terms thrown around in the previous sections that might have made the explanation slightly confusing. Luckily for us, ASHRAE simplifies things by categorizing all main hydronic components into two broad categories: thermal (load, source, expansion chamber) and hydraulic (distribution system, pump, expansion chamber) parts [2]. It is best to think of this as a distinction between (Ia) what is involved in exchanging heat and (IIa) what is involved in moving or containing the fluid. Figure 1 shows this in action. In a cooling system, the load initiates the demand, and the resulting inward air heat flow is characterized using the sensible heat equation:
q is the heat transfer rate to or from air, Qa is the air flow rate, ρa is the density of air, and cp is the specific heat capacity of air. But were we not just referring to water in hydronic systems? As it turns out, the same equation applies here too (we just need to change the variables):
The connection between air and water occurs at the terminal (ak), where warm air from the room moves across a chilled coil (at around 42 to 48 F) [7]. Heat naturally transfers toward the colder coil surface by convection, raising the internal water temperature on return (57 F onward (al)) [5][7][10]. This basically describes an air handler or fan coil unit which uses forced convection. In contrast, devices like chilled beams and radiant panels try to attract and remove heat without causing drafts, making use of natural convection or radiation methods instead [2]. These can run into issues with humidity as they do not really address latent heat. Then pumps and distribution components (such as valves or pipes) are responsible for properly supplying and returning water between the source and the load. Now, if it is the load that asks the favor, it is the chiller being a good friend that accepts—a chiller is pretty much a “refrigeration cycle” that continually shifts refrigerant phases to extract heat from the “hydronic loop” (always closed), and reject the majority of it through the “condensing loop” (usually open (am)) [4][11]. That sentence may have been too dense. In other words, it cools the water in the hydronic loop by absorbing the heat and tosses it away. Where does it go? Well, the water from the hydronic and condensing loops never meet (after all, one of the loops is closed), rather the refrigerant acts as a bridge for their heat transfer. Always ready to back up its buddy the chiller, the condensing loop continually brings in water cooled by “outdoor elements (an)” (since it can be open) so the heat has some place to leave. It is this cycle that creates the temperature difference we call delta T. Finally, an expansion tank is needed to manage the volume change of water caused by heating or cooling. Because water is incompressible, an air “cushion” or “bladder” absorbs this change—this can also help maintain positive pressure in the system (ao) [10]. Side note, while not in Figure 1, most systems include an air and dirt separator to remove entrapped air and capture debris circulating around as well.
ak. This report focuses more on water than air, but VAV boxes, like VFDs, bring variability to coil airflow, measured by air sensors (e.g., supply, return, mixed, discharge, and outdoor temp, as well as for CO2 and humidity) [7][30].
al. According to ANSI/ASHRAE/ISE Standard 90.1 2016 (might be outdated), this is the minimum the return temperature should be [7].
am. Chiller 101 (Evaporator (low-pressure refrigerant absorbs Hydronic Loop’s return water heat and becomes superheated) → Compressor (squeeze newly made refrigerant vapor to high-pressure making it hotter) → Condenser (hot refrigerant gas releases its heat to the Condensing Loop, turning back to high-pressure liquid) → Expansion Valve (drops refrigerant’s pressure, causing it to become cold again to restart cycle) [11].
an. The introduction suggested cooling towers, but other selections exist—such as geothermal or lake water cooling.
ao. In fact, since it works in interplay with the pump, the system should hold positive pressure. The whole objective is to control what is happening inside, and part of that comes with preventing everything else from entering.
Configurations and Controls:
Before we get into this section, it is worth noting that there are a plethora of piping and equipment placement choices in hydronic systems; dedicated or manifold, direct return or reverse return, parallel or series, zoning methods, and so on. Although the distinction between constant and variable flow will guide most of our discussions, it is helpful to address the foremost design choice to provide context for the figures to come.
To start, we call pumps primary because they are responsible for moving water through the main hydronic loop. If there is a secondary pump, which one of the sub–sections introduce, it is commonly used to distribute chilled water to the building’s terminal units in their own sub–loops. Anyway, by the end it should become clearer that the real difference between configurations and pump arrangements mostly comes down to whether the pumps themselves are constant or variable speed.
Where does one place the pump relative to the source? Or any of the other equipment, for that matter? Intuition. Next question! If only it were that simple—in his thesis “Point of No Pressure Change” (PONPC (ap)), Gil Carlson shifted the way we think about the initial question, clarifying that we must be focused on “pumping away” from the expansion tank [12][13]. That is, placing the expansion tank to the suction side or before the pump. Why? To answer this, one has to keep in mind that in a closed system, a pump does not create pressure; it creates a pressure difference that makes water flow (aq) [12][14]. Not to mention the expansion tank’s air cushion or bladder keeps the pressure constant at its placement. Think of this as a fixed “reference point” that holds the hydronic system’s static pressure, whether the pump is on or off [4]. With this, the pump has, in a sense, met its match; the expansion tank raises its shield, blocking any and all pressure changes it faces. That is because the pump has no authority to add or remove water from the tank, which is the very principle behind the tank’s defense working (volume change) [4]. Now imagine the issue if the pump, setting off on its noble quest, is blocked and deflected far too early. The pump is going to swing around its “weapon” (the pressure differential) no matter what, it is part of its telos or purpose. The decision then is if this gets added to the system’s static pressure or subtracted from it. Notably, there are serious consequences for its subtraction—it creates a vacuum that invites air and causes cavitation (ar); contrary to our objective of “maintaining positive pressure (as)” [14][15]. Therefore, a pump’s placement relative to the source must be secondary to its placement relative to the expansion tank (or PONPC).
ap. The thesis itself is partly frustrating; it claims (i) the pump should be on the supply side of the source, and (ii) the pump should “pump away” from the expansion tank [12]. The problem with (i) is that it is based on exceptionally simple heating configurations common in and before the 1960s–1970s. Nowadays, many modern cooling configurations comfortably place the pump on the return side of the source without issue, as long as (ii) is met.
aq. A centrifugal pump makes differential pressure by having a suction side and a discharge side. High pressure moves toward low pressure, and that is the very principle that begins circulation [14].
ar. Bubbles = cavitation; their implosion causes damage to the pump. This could be made worse considering that most centrifugal pumps are inherently not self–priming [33].
as. I may have presented this as more severe than it truly is, given that atmospheric pressure remains on our side before this happens (check HM Figure 15) [37].
I admit this does not fully resolve our question, though Figure 2 offers insight. In both illustrations, the expansion tank is placed before the pump. However, its position relative to the source varies by type, boiler or chiller, and specific system configuration, which will be explored shortly. It is not necessarily wrong to place a pump before a boiler; many systems do exactly that. In fact, Figures 1 and 3 show the pump located after the source, regardless of it being a boiler or chiller. More so, the idea of placing the expansion tank and air separator after the boiler and before the chiller is guided by solubility principles. The pump’s placement is then influenced by these factors, along with ensuring (Ib) the Net Positive Suction Head available (NPSHa) exceeds the required (NPSHr (at)), which is also affected by pump selection, (IIb) we satisfy the pressure ratings of the sources, and (IIIb) heat introduced by the pump is addressed [2][7]. Dissolved air becomes less soluble in water when temperature increases and pressure decreases (partially Henry’s Law) [2][9][17]. This is easier to visualize in a boiler system. Since the boiler’s supply side is the hottest point in the system, it is the ideal location for an air separator, as this is where gases are most likely to come out of solution [18]. To add on, the expansion tank should be located at the system’s lowest pressure point to define the static pressure and control fluctuations within safe limits [2]. Therefore, placing the pump after the expansion tank and air separator—where it is protected from air bubbles and does not increase pressure within the boiler and risk explosion—aligns with solubility principles, (Ib), and (IIb) [2][14][18]. Likewise, return water entering the chiller is relatively warm and low in pressure, allowing one to apply the same logic [4][18]. This provides the added benefit of protecting the pump and the chiller from cavitation and evaporator coil freezing. Indeed, the pump’s positional priority is then attributed to safeguarding itself and the chiller, corresponding once more to (Ib) and (IIb). While the advantage described in (IIIb) comes with this typical layout, should the pump’s pressure differential raise above the chiller’s rated pressure, one might reconsider selecting it or placing it there. Otherwise, we can enjoy the typical cooling configuration knowing it is well-justified.
at. This helps make sure the pump is not malfunctioning or prone to cavitation; the former being the actual suction–side pressure head, the latter the minimum required for it [2]. Placing the expansion tank before the pump in part “anchors” this pressure in a way.
Primary Only, Constant Volume:
Technically, not all pumps need to adjust directly to the building load, as long as, remember, they are “equipped with a response on hand.” The Primary-Only, Constant Volume configuration exemplifies this perfectly, where a single set of constant–speed pumps distributes water throughout the entire chilled water system [19]. Seeing as this is its only mode of response, it defaults to using it at all times, resulting in the pump running continuously.
At first glance, this may seem counterintuitive—and rightly so (au). A constant–speed pump (the orange triangle) alone is inherently incapable of adapting to the variable demand across spaces. Interestingly, for each coil to receive variable flow while keeping the system’s total flow constant, a three–way control valve is installed per terminal unit (in yellow), along with balancing valves (av) (in green) to “fine–tune” its resistance [20][31]. This configuration gives the unit two options: either route water through the coil (Path A) or bypass it entirely and return it directly back to the system (Path B) [21]. In theory, the pump does not require any adjustment beyond initial startup, as the control valves automatically regulate flow based on demand. When a room needs cooling, Path A is fully opened to the coil. For partial cooling, both Path A and B are partly opened. When no cooling is needed, Path B is fully opened to the return. The main drawback is that, albeit simple from a control standpoint, this method is not very efficient. One can imagine a scenario during off–peak hours when cooling demand is minimal, yet chilled water continues to circulate back to the chiller (low delta T). Plus the pump running every waking moment wastes a lot of energy (pump affinity laws).
Figure 4 has a cleaner presentation than Figure 3—thanks ASHRAE. They both convey the exact same information but explaining with colors is often easier.
au. While not shown, the triple–duty valve is a shutoff (isolate device for service), check (prevent backflow), and balancing valve (strategic throttling of flow) all in one [22]. These are less common now as ASHRAE claims they do not meet energy codes [23]. Here, the likely alternative is a stop–check or butterfly valve besides the pump?
av. Definitely a broader definition—and Bell & Gossett does note the valve should come secondary to other equipment choices—but I will give the balancing valve its due attention in the Hydronic Balancing section [31].
Primary—Secondary (P/S):
Imagine being a mail carrier in a city with only one massive “six-lane highway” (a primary–only loop) connecting the “post office” (the chiller) directly to every single “home” (the loads) (aw). Besides the obvious traffic this would cause, every time a specific home needs a delivery, the entire highway has to be running and pass by all possible homes (ax). If only we could keep the main highway active while deciding when to use the smaller local roads for neighborhood deliveries. Oh, we can? Ahem…without further ado I introduce the Primary–Secondary configuration, which consists of two hydraulically separated loops (ay). The “primary loop” (production) caters to the chillers with constant flow pumps, ensuring consistent circulation to protect them (left–side of Figure 5 (az)),, and the “secondary loop“ (distribution) uses variable– speed pumps and two–way control valves (Figure 6 below; in dark blue) to determine when and how much flow should divert onto the selected “local roads” (right–side of Figure 5) [7][9].
Notice, the two–way control valves will outright block flow or increase resistance as the space becomes satisfied. In other words, there is no way to bypass them, you either go through the valve or not (“To valve or not to valve, that is the question” –Jared (probably)) [20]. Because water follows the path of least resistance, and excess flow is no longer required in those zones as they shut, the variable–speed pump (ba) is called to adjust its output [9]. For this, we must refer to another affinity law: the pressure differential (bb) is proportional to the square of the pump speed, and, of course, flow remains linear with speed [24]. However, this, and the pump curve, will only get us so far since they only show specific impeller sizes. What do we do then? Do not forget we have variable–speed pumps. A VFD, by controlling the motor’s rotational speed driving the pump, recasts the pump curve much like a constant–speed pump switching or trimming its impellers, as shown in Figure 7. This was partly neglected, except that the pump affinity laws actually lets us swap “speed (RPM)” for “impeller diameter” in the relations, creating the pump curves below [9]. This means our use of the next affinity law was never in vain.
aw. The common pipe follows the Law of the Tee: “Whatever goes into a tee must come out of that tee” [9]. In other words, what enters must equal what exits, and the flow can either mix or divert depending on the direction.
ax. This is without exception, even homes that do not have packages, as the three–way control valve “diverts” flow, it does not outright “restrict” it [20].
ay. Remember that hydraulic concerns moving or containing fluid. Thus, this sentence is saying that theoretically conditions can be met where the flow of one loop does not affect the flow of the other [2][9].
az. Historically, older chillers and boilers required having constant flow to operate well, making this setup so novel. Suddenly, one could vary the flow and pump speed together without using three–way control valves.
ba. I suppose it could be constant primary and secondary pumps, though the controls for that are similar to Primary–Only, Constant Volume systems.
bb. More specifically the “pump head,” closed systems give us this equivalence. I also noticed that I might be using “pressure differential” and “differential pressure” interchangeably, but nothing is inherently wrong with this.
Primary Only, Constant Volume might not get much attention, but it does give us a decent baseline to work from. With constant–speed pumps, the pump curve remains the same regardless of impeller trimming. This means it is really the system resistance that changes, which makes sense as the load varies, but the pump has fortified its defenses and treats it as no threat. Even so, Figures 7 and 8 show that while the pump curve stays consistent, it can still respond in part to the system curve (perhaps subconsciously) as it becomes steeper or flatter [2][30]. The system curve basically describes the total resistance the pump must be prepared to encounter, and this resistance may either be fixed or variable pressure head [37]. Calculations for that will become better defined in the Hydronic Balancing section, but as the variable portions change, the system curve effectively “rides the pump curve.” In that, the pump’s operating or duty point is always the intersection of the pump and system curves (bc).
For variable–speed pumps, the same principle applies, except, in this case, both curves are continuously changing (it is conscious) (bd). Whether that change is productive comes down to the control strategy, which itself is optimized to the Best Efficiency Point (BEP) (be).
The U-shaped curves in Figure 7, extending out from the pump curves, indicates the region of efficiency, while Figure 9 shows its broader impact and range. Noting that both curves shift, one can now find where the system curve sits and choose the pump curve with the best efficiency (bf). Still, the “choice” aspect depends on the control strategy, and there are two main approaches: Curve or Area control (subsections in HM Figure 9–10). The main difference really comes down to pump control either responding exclusively to calculations and theory (curve control), or drawing on a diverse range of sensor inputs (area control) [32]. The former usually relies on a flow meter or pressure head initial input, sometimes even going sensorless (runs off of motor HP and RPM), which can create problems in more diverse systems (bg) [36][37]. The latter often requires a custom pump controller to interpret the sensor or equipment data and decide how to respond [36].
bc. The energy “added” by the pump should equal the energy “subtracted” by the system → Law of Conservation of Energy [37].
bd. This goes without saying with the preferred operating regions, but avoiding the flat part of the curve is good because DP change with flow may be ambiguous [37].
be. No pump is 100% efficient, but we can operate it within a range that minimizes mechanical stress on the impeller, particularly from “non-symmetrical forces” that manifest as vibrations; it affects maintenance and life cycle [33].
bf. It is recommended that the duty point for constant–speed pumps be set toward the right of the BEP (increase efficiency with more flow), while for variable pumps it is to the left (increase efficiency with decreasing flow) [7].
bg. Diversity in a system means more load changes and components, and this approach can miss giving coils proper flow because it does not capture every single fluctuation in the system [32]. The degree to which it misses depends on the system, but the more diversity present, the more it is likely to miss it (an adaptability problem).
To begin painting the picture, let us craft this scenario. Suppose a valve closes off a zone so that no cooling is demanded, as seen in Figure 8. The system then creates greater resistance for the variable–speed pump to face and overcome. And I reiterate, no matter the conditions, the pump is condemned to clutching or swinging around its cherished “weapon,” the pressure differential. However, the pump is strategic: as valves open and close—changing system resistance and pressure along the way—it adjusts to meet the controlled DP setpoint (bh), correcting these shifts rather than responding with sheer force. As a result, the VFD slows the speed of the pump to reduce flow and address the lack of demand. Then, according to the pump performance curve, the pressure head rises to compensate [29][31]. On the other hand, when valves open and cooling is demanded, once again depicted in Figure 8, the VFD ramps up the pump speed to meet flow requirements, inviting the system resistance to bring the differential pressure back in line [29][31]. Thus, the heroic pump and its faithful sidekick, the control valves, work tirelessly to maintain order and stability in the system (a theme I wish to preserve). Hopefully, it is clear now that the pump varies speed to optimize energy consumption and match demand. Still, a moment comes when the sidekick gathers enough strength and acts as a hero in its own right. Here is the origin story: the pump cannot concern itself with every individual constituent; instead, it strives to serve the entire loop as best it can. Whenever a control valve moves, the system pressure experiences a slight shift, setting off a chain reaction that the pump and remaining valves must react to [23][27]. To counteract this, pressure–independent control valves (PICVs) can be used, minimizing flow variability caused by pressure fluctuations through an integrated regulator (bi) [7][27]. This equips the control valves with a much–needed upgrade to their arsenal, sharpening their ability to manage the heat transfer. For clarity purposes, a valve is considered pressure–dependent when its flow rate varies with differential pressure, and pressure–independent when it remains unaffected.
bh. A setpoint is the desired input a device will adjust to meet, much like how a chiller sends a specific temperature of water out, or the pump attempts to support a differential pressure or flow.
bi. ASHRAE 90.1 2013 (might be outdated) mentions a DP reset strategy where one two–way valve is left nearly wide open to find the absolute minimum pressure required to satisfy the system. However, since PICVs compensate for pressure changes, their valve position does not change with this reset (check HM Figure 9–10) [28][37].
There are conditions to this “hydraulic separation;” mainly, it only works when the decoupler, or common pipe (shown in light green; “on–off ramp”), experiences no pressure drop (bj) [26]. This pretty much means the pressure is not pushing water into either the primary or secondary loops. In practice, while idealistic, perfect isolation of flow is rarely achievable. This is not necessarily a problem, as it does create a “call to action” for the system to self–correct. But the question still remains: how much pressure drop can the configuration accommodate or tolerate? The metric for this will define the system’s degree of independence [3]. While the answer to this depends on several factors, fundamentally one wants to avoid pumps working against each other and causing equipment damage. If the primary flow outdoes the secondary flow (production is greater than demand), the secondary pumps will reject the excess flow, which then passes through the common pipe and mixes with the return water from the secondary circuit [9][26]. This brings us back to the “low delta T syndrome” discussion, which this configuration does not have the greatest response to (bk) (protected variable–speed pump but at what cost?) [19]. And if the secondary flow outdoes the primary flow (demand is greater than production), the secondary pump will pull all available chilled water from the primary loop and mix it with its own return [9]. The problem this creates is that the “warmer” chilled water supplied to the coils may no longer be cold enough to adequately cool the room (protected chiller but at what cost?). Therefore, there may be a tragic flaw amongst the benefits, though it can be easily mitigated by minimizing pressure drop with a short, large diameter common pipe with no valve restriction [2].
bj. Tertiary systems exist as well, adding a pump in every zone and introducing multiple common pipes for them [9]. These are usually intended for multi–building or other large scale applications.
bk. One has to wonder, if this were to happen, does the constant–primary pump have the capacity to respond? Definitely not, it is not configured to alter its speed, only the variable–secondary pump can do so.
Variable Primary (VPF):
There is a pattern forming with each iteration of the configurations: save energy by introducing periods of reduced pump speed, increase control complexity as more components react, perhaps even resolve the flaws of the previous setup along the way. And although it does not get more complex than this—with one set of variable–speed pumps shouldering the whole system—based on the other configurations, we already have the toolset to understand it quite well. Much like the P/S configurations, two–way valves modulate flow based on cooling demand, and depending on the extent of their closing or throttling to the system resistance, a set of variable–speed pumps equally responds [29]. The main difference here is that this time the variable–speed pumps are positioned elsewhere; once again serving the source rather than distribution, yet still remaining mindful of it. But wait, the constant flow through the chillers had to have served some purpose, right? Did the chillers not prefer constant flow to prevent freezing issues? They do, except, oddly enough chillers have always been capable of handling variable water flow (bl); manufacturers simply did not encourage it initially due to control limitations (the same ones that have since been surpassed) [19][33].
bl. The inclusion of VFDs on chillers only increases this tolerance range. Even so, chillers typically are decided near the end, as coil and tower selections derive system temperature and flow rates [31].
Realistically, as long as (Ic) the chiller’s allowable flow range is maintained, (IIc) the minimum flow is met and bypass flow (bm) is minimized, and (IIIc) ASHRAE’s guidelines for rates of change (generally 2% to 30% design flow per minute (bn)) or staging is followed, there is little concern over some underflow or overflow for the system [2][29][30][31]. (Ic) and (IIc) are inherent in the chiller’s manufacturing requirements, with the bypass ensuring minimum flow for active chillers (also consider reading Footnote cg), while (IIIc) adds a barrier on how quickly the pumps can react. Therefore, a Variable Primary configuration already requires a relationship between the chiller, isolation valves for staging on and off, zone control valves, bypass control valve, balance valves (bo), and the pump—all working together to respond to load changes and protect the system. This is definitely a lot of sidekicks for the pump, but why stop at just one of those? I have said the term “set” here and there for a reason, plus many of the recent figures embody multiple pumps or chillers. Another point, it is not like a system only has to react to other components in “epic battles” (such as the pump versus the expansion tank); sensors exist too, acting as surveillance for any “crime,” often motivating a response before problems have a chance to spread. Before we proceed any further, let us revisit and update the basic control model from the introduction.
bm. The bypass is sized for the largest chiller’s minimum flow requirements and is located close by for quick responses [30]. Its usage is minimized because, while it keeps the chillers within safe limits, it also mixes cold supply water with warm return water.
bn. Depends on manufacturer, but ASHRAE Guideline 36 2024 defaults to a max user–adjustment of 25% per min., plus leaves isolation & bypass valves to be “determined in the field as needed to prevent nuisance trips” [30].
bo. Isolation and balancing valves can be actuated, and balancing valves can also be pressure–independent, meaning hydronic balancing and staging could be made automatic (VFDs need automation)—check HM Figure 7 and 8 [32].
Previously, we noted that a building’s load request starts with a thermostat defining the temperature the system is designed to reach. In reality, however, while an interface may define this “setpoint,” there is a hierarchy of requests to sort through, beginning downstream at the terminal or air handling units and propagating upstream to equipment such as pumps and chillers [30]. These requests must be “reviewed” and “approved” by those above them in order to cause change. Still, before the requests even have the chance to reach the variable–speed pump to alter its motor speed, for context, they must pass through air sensors monitoring outside air (bp), the air–handling unit with its variable airflow, the coil, which can turbulate its flow with a turbulator (bq), then the two–way control valves that open or close to regulate chilled water flow, followed by automatic balance valves (and so on) [10][30][33]. All of these, complemented by many other zones, already have a range within which they can influence heat transfer on their own, thereby making the pump’s burden far lighter. To add on, is it not better for a skilled individual to focus on tasks that make the best use of their abilities rather than being sidetracked on simpler ones? We know that the pump cannot respond to every single request; it relies on its “local heroes” or “sidekicks” (those downstream) to correct and deliver within their own ranges before it steps in (i.e., deadbanding (br)) [30]. Because everyone must train and specialize to hone their skills, they cannot afford to be distracted by overexerting themselves or doing too little. Hence, to determine when to act on the request(s), the system assigns a value to each using an importance multiplier (bs) (greater number means higher value). And the reason a piece of equipment would make a request upstream is when its own setpoints are getting close to exceeding or falling below what it can optimally handle [30].
bp. For instance, lockout temperatures prevent the system from operating if the outdoor temperature is cold enough.
bq. Turbulators create chaotic flow for mixing, while fins increase a coil’s surface area—both great for heat transfer.
br. Deadbanding is a “range of input values in a control or signal processing system where the output remains zero, meaning no action is taken” [34]. The system will provide no cooling actions with these inputs making it “dead.”
bs. The importance multiplier is a value, defaulting at 1, that adjusts the priority of certain spaces’ requests. It might be the case that a small coil is given 0.5, while equipment in critical zones are given 4s [30].
And the reason a piece of equipment would make a request upstream is when its own setpoints are getting close to exceeding or falling below what it can optimally handle (bt) [30]. A setpoint must therefore be (Id) adjustable (bu), (IId) define the equipment’s operating range, and (IIId) include conditions for when it can be changed [7][30]. This is made easier by saying that all setpoints can either be reset–(Id) or fixed–(IId), with requests acting as signals for resets to happen. Consider this, a control valve opening from 80% to 90% will likely send multiple reset requests to the pump to increase its speed, allowing the valve to return to an optimal state. The pump can only ignore so many of these requests before it has to kick in and reset its current speed setpoint to place the valve back to what it can manage. Now, the amount of requests become a new metric altogether, which is where Trim & Respond logic (bv) (IIId) comes into play. If we think of a setpoint as a medium, we can continuously shape and sculpt it (trim) until it feels perfected, or until others begin to critique the beauty of our work (downstream requests exceed ignorable threshold), at which point we step back (the setpoint is raised) [30]. This process avoids wasted energy by searching for the best possible setpoint for the moment and establishing a clear reason to intervene.
bt. HM Figure 3 is practically the same as Figure 12.
bu. ASHRAE requires that any and all setpoints are user adjustable even when not explicitly stated [30].
bv. Ah, a classic counter: rogue zones are problematic areas that continue making requests, potentially keeping the Trim & Respond logic to its maximum setpoint, while other zones may be perfectly fine and trying to close their valves to accommodate and avoid overcooling [21][30]. This usually implies the zone needs service done to it, and might be ignored in the meantime.
On the topic of intervention, it would be unwise to claim the pumps lack a mediator. Naturally, a controller like the VFD or operating system fills this role. What is interesting is that the system actively engages in fault detection, sets maintenance alarms (bw), and documents equipment runtime, almost like tracking skills or attributes in a video game (bx) [30]. It then decides whether the pump is ready for the next level or quest. However, much like a game where a player swaps characters, premature rotation wastes potential, since the character still has much to contribute on the battlefield—a concept known as hotswapping (by). Similarly, no single character should bear all the responsibility in a boss fight; uneven wear reduces the lifespan of that component, which goes on to hurt the reliability of the system [30][33]. Accordingly, a strategy must be in place, and ASHRAE largely bases this on Lead/Lag or Lead/Standby logic combined with staging time (hours since last reset) and lifetime (hours since startup) [30]. All Lead/Lag and Lead/Standby logic really does is figure out which pump to stage next so the one already running does not end up doing all the work. When a leading pump is getting close to full capacity, the lagging pumps kick in to help out. If the lead’s runtime far surpasses the others, a schedule calls for it, or the operator interferes, it is gradually swapped out, taking on the role of a lagging or standby pump [30][39]. The pump on standby is meant to be redundant, existing purely as a backup, ready to take over the lead at any moment [39]. The whole point of ASHRAE’s fault detection is to monitor operational states and anticipate when things might go wrong. To do this, the system considers the time needed to respond to commands, like a reset request. And when a failure occurs, depending on the severity, alarms are slightly delayed to see whether it is a recurring issue [30]. To prevent confusion from multiple alarms, hierarchical suppression prioritizes faults for upstream equipment, suppressing others so the most critical components are addressed first [30]. The system may also consider other actions, such as placing the affected equipment later in the staging sequence as a result.
bw. As with requests, alarms are assigned value based on priority in equipment and classification: Level 1: Life–Safety message, Level 2: Critical equipment message, Level 3: Urgent message, and Level 4 Normal message [30].
bx. Much of this is due to their Automatic Fault Detection and Diagnostic (AFDD) system, which prevents or switches the gears on operation under suboptimal conditions that could incur unnecessary costs [30].
by. Hotswapping occurs when a piece of equipment, such as a pump, has accumulated more runtime than a currently idle pump and is immediately swapped in to balance wear and tear. This can put stress on the system, while a more seamless approach is to rely on staging (when they are already tapping out) [30].
Pressure drops naturally occur throughout the system due to the cumulative resistance of all flow–path components, including friction head loss (straight length (bz)), pipe fittings (equivalent length), and cooling branches (equipment) [31][37]. Everyone draws from the pump, a consequence of its heroic reputation and chiseled design, which is why the balancing valve acts as a mediator, ensuring no zone is starved of its desired flow [6]. The point is, the pump acts like a rechargeable battery, supplying the “energy” (pressure drop) that the system “consumes” (e.g., throttling). Whether through pressure or flow sensors related to speed via the pump affinity laws, calculations based on the pump curve and operating conditions, or changes in differential pressure, temperature, or valve positions (ca), all signals reach the pump’s controller [26][31][32][36]. The controller then decides when and how to respond to these inputs and, with enough requests in mind, adjusts the pump motor to reach the new setpoint. Next, the pump’s operating range must be defined in the field. ASHRAE specifies 0% speed as 0 Hz and 100% speed as the VFD’s maximum (cb) [30]. The minimum is set by observing the frequency at which the pump just begins to rotate, while the maximum is found by gradually lowering the frequency until downstream equipment can safely handle the load [30]. Therefore, the pump has always had more at its disposal than I initially conveyed; there was always fully fledged reasoning to its madness (cc). Faced with greater or fewer “enemies” (cooling demand), the pump, acting justly, has its own “objectives” (setpoint) changed to meet the challenge fairly (cd). In regard to its “surveillance” (the sensors), Figures 6, 10, 11, and 12 have the DP sensors positioned off the side of the chiller, near the farthest side of the system, or scattered across zones. As long as they address the critical load, where the most pressure drop and hydraulic loss happen, they function effectively [7][37]. While often assumed, placing a DP sensor at the most remote circuit is not necessarily the optimal design choice for this reason (a balancer must do the math). Notice that the flow meters are also located near the chillers, likely to provide feedback to the pumps, since chillers are highly sensitive to flow rates and pressure (ce) [31]. And, at the very least, temperature sensors should be on both the supply and return lines to monitor system performance [7]. The decoupler in a P/S configuration is another suitable location, as the mixing that occurs there helps determine when to increase or decrease production [9][35]. The same applies to the VPF, where, even when bypass usage is limited, knowing the mixture to maintain minimum flow to the chiller is beneficial [31]. Nevertheless, temperature control largely depends on the interaction between valves and coils, making it so placing them at each zone is favorable (check HM Figure 18 and 19). After all, this is how we assess the system’s success in meeting its cooling objective—by monitoring the overall delta T and responding to its variations (cf).
bz. According to Bell & Gossett, the industry standard is 4.5’ of pressure drop per 100’ equivalent straight length; we can expect ~2–4 ft/sec for copper pipe, ~5–8 ft/sec for steel pipe, and ~10–13 ft/sec for stainless steel pipe too [37].
ca. Bell & Gossett’s PDH cautioned that the connection governing valve position is prone to loosening, which can possibly throw things off [37].
cb. Plus a deceleration rate of 1 Hz per second when turned off so that other equipment remains stable (remember that chillers are sensitive and can only take so much flow variation) [30].
cc. The most common method is using differential pressure, whose natural relationship to the pump provides an excellent introduction to this research. ASHRAE 90.1 2010/13 (might be outdated) also requires that controls be based on either desired flow or differential pressure [36].
cd. I will put this more explicitly, the Building Management System (BMS), Building Automation System (BAS), or, more broadly the “pump controller” (gray boxes in Figure 12) will dictate the operational strategy, setpoints, and staging of all pumps in relation to other HVAC equipment [2][29].
ce. This has not been mentioned, but too much flow can cause vibration and erosion eventually damaging the evaporator tubes [31]. Conversely, as already covered, too little flow can lead to freezing.
cf. I recognize that the other models could have benefited from a detailed review like this, but I prefer presenting them as iterations to build understanding gradually. Once this is read and the previous configurations are grasped, the intention is that the same logic can be applied to those models.
Multiplicity of Pumps:
In moments of triumph, it is natural to call upon allies. In a similar vein, while the pump affinity laws describe only a single pump or a comparison of two, most choose the same pump in a multiple setup for predictability and simplicity. Not only does selecting a higher–head pump alongside a smaller pump introduce a risk of deadheading (cg), yet the performance curve for identical pumps allows you to increase output in either flow or pressure (check Figure 13) [37]. But the choice remains: should we configure our pumps in parallel or series?
Leaving aside that series pumps have never appeared in any of the figures, justification is difficult because of the benefits of having backup pumps for switching and the extra selection criteria for suction and discharge pressure (ch) [30][33]. Besides, a single pump usually provides enough pressure head for the system, and with variable speed drives, adding more through series is often unnecessary. Notice that when pumps are arranged in parallel, the pumps operate at the same pressure and together deliver the total system flow, with each contributing its share to meet the demand [2]. That is, assuming the system curve intersects both pump curves and adapt when staging on pump on or off (ci). The pressure differential offered by both pumps exceeds what a single pump can deliver, as shown in Figure 13. This makes sense, since when more flow is required, as indicated by control valves opening, system resistance drops and the pumps must compensate. The decision to stage pumps on or off then depends on the factors discussed earlier and typically occurs when they have been operating near maximum or minimum speed for a set period of time; once again, ASHRAE’s way of ensuring an action is done when absolutely necessary [30]. This is not to say that simply running pumps in parallel guarantees energy savings—the mountain of Sisyphus does not get any smaller just because more people are helping him push the boulder (a misapplication of the pump affinity laws) [7]. More so, it helps prevent an oversized pump from running inefficiently under part–load conditions [7][30].
cg. Deadheading is when a pump operates with little or no flow, which can potentially damage it as it struggles to discharge fluid. If not careful, a larger pump can leave a smaller pump in this type of situation [31]. I prioritized minimum flow for chillers as they seem more delicate, but if the load is satisfied and all control valves close, a P/S setup can struggle to handle deadheading, whereas a VPF can just open its bypass [29][31].
ch. Additional pumps in series must be able to withstand the increased suction pressure produced by the preceding pump, and if one pump fails, the entire series is affected.
ci. While this may appear intuitive, it is important to stress that pumps should be compatible and the system curve must intersect all pump curves, otherwise the benefits are not there and one of the pumps is not doing anything [37].
Hydronic Balancing (kept brief):
Hydronic balancing involves more than just calibrating equipment to work together; it ensures each branch receives its design flow and optimizes throttling throughout the system (since excessive throttling often indicates oversizing (cj) ) [6]. Here, ASHRAE uses the term proportional balancing, where each branch receives a percentage of the provided flow; no more or no less is taken, everything is fairly distributed. To achieve this, the easier or closer circuits must be made as resistant as the harder ones [6]. This requires balancing valves, which, as mentioned before when attributing it to fine–tuning, also restricts anything more than design flow to enter each branch [37]. Control valves also have an authority aspect, referring to the ratio between the valve’s pressure drop and the total branch pressure drop (valve authority) [31]. Higher authority provides better control, making the actuator’s valve movements more accurate and predictable (with around 50% considered optimal) [31][37]. Lower authority means the valve must move significantly before having any noticeable effect on flow [31]. The balancing valve does have an impact on this, as if it is throttled too much, it increases the branch resistance, which in turn reduces the authority of the control valve. We discussed the system curve earlier, which is derived from the Darcy Weisbach equation for variable head, along with constant static head calculated from straight and equivalent pipe lengths multiplied by the friction loss rate—all to determine the critical load [6][37]. In practice, having more variable heads provides the system with greater versatility and increases the pump’s control authority and interaction with the components.
cj. ASHRAE’s definition for a "balanced system” requires 97% of design heat transfer at design flow with a 10% flow tolerance [6]. To add on, a balancing report must be created and kept on record.
Fin~
Honorable mentions (Appendix):
(Check downloadable Google document!)
References (IEEE):
[1] Merriam-Webster, “Hydronic,” Merriam-Webster.com Dictionary. [Online]. Available: https://www.merriam-webster.com/dictionary/hydronic
[2] ASHRAE, 2012 ASHRAE Handbook—HVAC Systems and Equipment (SI), ch. 13, “Hydronic Heating and Cooling.” Atlanta, GA, USA: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc., 2012.
[3] J. Siegenthaler, Idronics #12: Fundamentals of Hydronic Systems, 12th ed. Milwaukee, WI, USA: Caleffi North America, Inc., 2011.
[4] Caleffi, “Contemporary Hydronic Cooling in Commercial Buildings,” YouTube, [Online]. Available: https://youtu.be/LPOMMPkThPg.
[5] M. Olson, Idronics #13: Hydronic Cooling, 13th ed. Milwaukee, WI, USA: Caleffi North America, Inc., 2013.
[6] C. Edmondson, “Hydronic Balancing Series,” JMPCO Blog. [Online]. Available: https://jmpcoblog.com/hydronic-balancing-series.
[7] Trane, Comprehensive Chilled-Water System Design: System Catalog, Rep. APP-PRC006A-EN, 2021.
[8] C. Edmondson, “Federal Buildings To Comply with ASHRAE 90.1‑2010 – Just Like the Rest of Us!,” JMPCO Blog, Jul. 30, 2013. [Online]. Available: https://jmpcoblog.com/hvac-blog/federal-buildings-to-comply-with-ashrae-901-2010‑just‑like‑the‑rest‑of‑us?rq=ashrae.
[9] C. Edmondson, “Understanding Primary-Secondary Pumping Series,” JMPCO Blog. [Online]. Available: https://jmpcoblog.com/understanding-primary-secondary-pumping-series.
[10] Taco Comfort Solutions, “Chiller Plant Equipment and Piping Design,” YouTube, video, Jul. 30, 2024. [Online]. Available: https://youtu.be/NOwX3rY6sSk.
[11] Thermal Care, “How Does a Chiller Work?,” Thermal Care, 2023. [Online]. Available: https://www.thermalcare.com/how-does-a-chiller-work-2/.
[12] G. R. Carey, Pumping Away… What Does It Really Mean? FIA Inc., 2009. [Online]. Available: Internal company publication (uploaded).
[13] G. R. Carey, “Pumping Away,” Indoor Comfort Marketing, Boiler Facts column, pp. 12–18, Jan./Feb. 2024.
[14] G. R. Carey, “The Point of No Pressure Change,” Indoor Comfort Marketing, Nov. 1, 2018. [Online]. Available: https://www.indoorcomfortmarketing.com/the-point-of-no-pressure-change/.
[15] T. Holder, “Best Practices for Hydronic Systems Part 3: Where to Locate the Expansion Tank,” JMPCO Blog, Feb. 12, 2021. [Online]. Available: https://jmpcoblog.com/hvac-blog/best-practices-for-hydronic-systems-part-3-where-to-locate-the-expansion-tank.
[16] Guangzhou Tofee Electro‑Mechanical Equipment Co., Ltd., “Air Dirt Separator.” [Online]. Available: https://tofee.com.cn/air-dirt-separator/.
[17] “13.4: Effects of Temperature and Pressure on Solubility,” in General Chemistry: Principles, Patterns, and Applications (Averill), LibreTexts, 2022. [Online]. Available: https://chem.libretexts.org/Bookshelves/General_Chemistry/Book%3A_General_Chemistry%3A_Principles_Patterns_and_Applications_(Averill)/13%3A_Solutions/13.04%3A_Effects_of_Temperature_and_Pressure_on_Solubility.
[18] Caleffi S.p.a., "Where should an air separator be located?," Caleffi, Jun. 15, 2015. [Online]. Available: https://www.caleffi.com/en-us/blog/where-should-air-separator-be-located.
[19] W. P. Bahnfleth and E. Peyer, Variable Primary Flow Chilled Water Systems: Potential Benefits and Application Issues, vol. 1, ARTI-21CR/611-20070-01, Air-Conditioning and Refrigeration Technology Institute, Mar. 2004.
[20] Deppmann, "Three-way & Two-way Control Valve Basics in Hydronic Systems," Deppmann, 5 Feb. 2024. [Online]. Available: https://www.deppmann.com/blog/monday-morning-minutes/three-way-two-way-control-valve-basics-hydronic-systems/.
[21] J. Siegenthaler, Idronics #5: Zoning Hydronic Systems, 1st ed. Milwaukee, WI, USA: Caleffi North America, Inc., 2009.
[22] L. Eck, “What is a Triple Duty Valve?”, Liberty Supply HVAC News, Mar. 21, 2022. [Online]. Available: https://libertysupply.com/blogs/hvac-news/what-is-a-triple-duty-valve.
[23] JMP Equipment Company, “Hydronic Balance Series,” YouTube, Sep. 2015. [Online]. Available: https://youtu.be/yjTfXCsOc2s.
[25] C. Edmondson, “Pump affinity laws are key to saving money in a hydronic system,” JMP Equipment Company Blog, Jul. 10, 2013.
[26] A. Bhatia, "HVAC Chilled Water Distribution," PDHonline | PDH Center, Fairfax, VA, Course M328, 2020. [Online]. Available: https://pdhonline.com/courses/m328/m328content.pdf.
[27] E. Hipolito, The Benefits of Pressure Independent Control Valves in Building HVAC Systems, Pumps & Systems, Aug. 15, 2023. Available: https://www.pumpsandsystems.com/benefits-pressure-independent-control-valves-building-hvac-systems.
[28] T. Holder, “Why PIC Valves Aren’t the Best Choice for Area Control with ASHRAE Recommended DP Reset,” JMP Equipment Company HVAC Blog, Mar. 8, 2018. [Online]. Available: https://jmpcoblog.com/hvac-blog/why-pic-valves-arent-the-best-choice-for-area-control-with-ashrae-recommended-dp-reset.
[29] C. Edmondson, “Variable Primary Chilled Water Series,” JMPCO Blog, Jan. 19 2021. [Online]. Available: https://jmpcoblog.com/variable-primary-chilled-water-series.
[30] ASHRAE, Guideline 36-2024: High-Performance Sequences of Operation for HVAC Systems, ASHRAE, Atlanta, GA, 2024.
[31] Bell & Gossett, Variable Primary Flow Systems, Bulletin No. TEH-910, ITT Corporation, Morton Grove, IL, USA, Dec. 2010. Accessed: Aug. 10, 2025. [Online]. Available: https://www.bellgossett.com.
[32] C. Edmondson, “Variable Speed Control Series,” JMPCO Blog, Aug. 2, 2023. [Online]. Available: https://jmpcoblog.com/variable-speed-control-series-1.
[33] A. Bhatia, "Design Considerations for Hydronic Pump System Design," PDHonline | PDH Center, Fairfax, VA, Course M150, 2020. [Online]. Available: https://pdhonline.com/courses/m150/m150content.pdf.
[34] Wikipedia contributors, "Deadband," Wikipedia, The Free Encyclopedia. [Online]. Available: https://en.wikipedia.org/wiki/Deadband.
[35] Primary Secondary Chilled Water System Working | BMS Training 2021, YouTube video, 4 years, 6 months ago. [Online]. Available: https://www.youtube.com/watch?v=MqSFcSxxdfw.
[36] K. DelPiano, "Variable-Speed Pumping Control Strategies," presented at the Dayton ASHRAE Chapter, 2018.
[37] M. T. Licastro, "Hydronic Systems Design Seminar," presented at the Little Red Schoolhouse, Aug. 19, 2025.
[38] M. Bingham, "A Quick Look at Smart Pump Curves," JMP Equipment Company Blog, Jul. 29, 2025. [Online]. Available: https://jmpcoblog.com/hvac-blog/jg5tuf81qzmf1h7tgbzyb7x0ag7yif.
[39] "Commissioning Lead/Lag Pumping Operations: Part I, Identifying a Designer's Intent," ACHR News, Jan. 09, 2023. [Online]. Available: https://www.achrnews.com/articles/162893-commissioning-lead-lag-pumping-operations-part-i-identifying-a-designers-intent.