This is not a textbook. It’s a field handbook. Everything in here comes from actual installations — systems that worked, systems that didn’t, and the revisions that brought them back from the edge. If you’re designing a secondary refrigerant loop, or converting one, or troubleshooting one that’s misbehaving, start here.

What you won’t find: theory for theory’s sake. Equations you can look up elsewhere. Hand-waving about “best practices” without telling you what the practice actually is. What you will find: five real cases, the specific problems they presented, and the specific solutions that resolved them.

1. Ammonia Conversion: Replacing a Primary Refrigerant with a Secondary Loop

Ammonia is a brutally effective refrigerant. It’s also toxic, and the regulatory pressure to reduce ammonia charge in occupied buildings has been building for years. The Xinjiang cold storage facility had a direct-expansion ammonia system serving multiple cold rooms. The operator wanted to convert to a secondary loop: keep the ammonia in the machine room, use a secondary refrigerant to distribute cooling to the cold rooms.

The conversion sounds simple. It isn’t.

The first question: which secondary refrigerant? The cold rooms were maintained at -18°C to -25°C. The ammonia evaporator in the machine room was running at -32°C. The secondary refrigerant needed to operate at roughly -30°C with good heat transfer properties. LM-8, water-based and rated to -50°C, was the obvious candidate. Non-flammable, which matters in a food storage facility. Good heat transfer. Proven inhibitor package.

The second question: how much piping? The existing ammonia lines ran through the cold rooms. Converting to a secondary loop meant installing new piping for the coolant distribution. The total pipe run was 380 meters, 3-inch schedule 40 carbon steel. The pipe volume alone was 1,700 liters. Add the heat exchangers, the expansion tank, and the pump housing, and the total system volume was close to 2,500 liters.

The third question, and the one that almost derailed the project: the pump. The existing ammonia system used the compressor’s discharge pressure to circulate refrigerant. A secondary loop needs a pump. The pump needed to move 2,500 liters of LM-8 at -30°C through 380 meters of pipe with a total dynamic head of 28 meters. The pump selection was constrained by the available electrical supply in the machine room — there was no spare capacity for a large motor.

We went with a close-coupled centrifugal pump with a 5.5 kW motor, selected for the viscosity of LM-8 at -30°C. The pump curve was checked at the operating point. The NPSH margin was verified. The motor was within the available electrical capacity. The pump has been running since April without issue.

The conversion took eight days. Four days for piping installation. Two days for flushing and pressure testing. One day for filling and commissioning. One day of margin that wasn’t needed. The system is operational. The ammonia charge in the building has been reduced by 85%. The fire authority is satisfied. The operator is satisfied. The cold rooms are holding temperature.

2. Narrow Delta-T: When 1 to 1.5°C Is All You Have

Most secondary refrigerant loops are designed for a ΔT of 5 to 10°C. The coolant enters the load at temperature T, leaves at T+ΔT, returns to the chiller, and repeats. The flow rate is set by the cooling duty and the ΔT. Q = ṁ × Cp × ΔT. Standard stuff.

The Zhejiang case was different. The process being cooled was a precision manufacturing operation that required the coolant temperature to be held within 1 to 1.5°C across the entire process. The cooling load was 120 kW. With a ΔT of 1.5°C, the required flow rate was enormous.

Let’s do the math. LM-4 has a specific heat capacity of roughly 3.9 kJ/kg·K at the operating temperature of 5°C. For a 120 kW load and a 1.5°C ΔT, the mass flow rate is 120 ÷ (3.9 × 1.5) = 20.5 kg/s. That’s about 74 cubic meters per hour. For comparison, a typical 5°C ΔT loop with the same load would need about 22 m³/h. The narrow ΔT triples the flow rate.

The implications cascade through the entire system design. The pipe size goes up. The pump size goes up. The pump power goes up — by a factor of roughly 9, since pump power scales with the cube of flow rate for a given system curve. The heat exchanger surface area goes up. The expansion tank size goes up. Everything gets bigger, heavier, and more expensive.

The Zhejiang system was designed with 4-inch piping (up from the 2-inch that would have sufficed for a 5°C ΔT). The pump was a 15 kW end-suction centrifugal. The expansion tank was 500 liters. The total installed cost was about 2.5 times what a conventional 5°C ΔT system would have cost. The operator accepted this because the process requirement was non-negotiable.

The system has been running since October 2025. The temperature control is within ±0.3°C of setpoint, which is better than the ±0.5°C specification. The pump is consuming about 13 kW in continuous operation, which is a significant operating cost. The narrow ΔT approach works, but it’s expensive. Don’t design a narrow ΔT system unless you have to.

3. Nitrogen Pressurization: Keeping LM-4 Alive

Same facility, different problem. The LM-4 secondary refrigerant loop was experiencing accelerated inhibitor depletion. The corrosion inhibitor concentration was dropping faster than expected. The pH was drifting downward. The corrosion coupons were showing increased metal loss. Something was consuming the inhibitor.

The culprit: oxygen ingress.

The expansion tank was an open-to-atmosphere design — an older style that’s still common in China. The coolant surface in the tank was in direct contact with air. Oxygen from the air was dissolving into the coolant. The dissolved oxygen was consuming the corrosion inhibitor. The inhibitor was designed to form a passive film on metal surfaces, but with continuous oxygen ingress, the film was being continuously disrupted and the inhibitor was being continuously consumed.

The fix was a nitrogen blanket. The expansion tank was retrofitted with a nitrogen supply line and a pressure regulator. The nitrogen maintains a slight positive pressure — about 0.1 bar — on the coolant surface. This prevents air from entering the tank. The oxygen is excluded. The inhibitor consumption returns to normal.

The nitrogen consumption is minimal — a standard 40-liter cylinder at 150 bar lasts about three months. The cost is negligible. The benefit is substantial: the inhibitor life is extended, the corrosion rate is reduced, and the coolant change interval is doubled. For an LM-4 system with a 2,500-liter charge, doubling the change interval saves about $1,500 per year in coolant cost alone, not counting the labor and downtime.

The nitrogen pressurization was implemented in October 2025. The corrosion coupon results from the following quarter showed a 60% reduction in metal loss rate. The inhibitor concentration has stabilized. The pH is steady. The system is healthier.

4. Piping Upgrade: When the Fluid Change Forces a System Change

This case is covered in detail in the Replacement Guide, but it’s worth repeating here from a system design perspective. The Jiangsu installation was switching from LM-8 (water-based) to LM-11D (water-free) to achieve a lower operating temperature of -55°C. The water-to-water-free transition meant the heat transfer coefficient of the coolant was dropping by roughly 40%. The existing 1.5-inch piping was inadequate for the increased flow rate required to maintain the cooling duty.

The system design lesson: when you change the coolant, you must re-verify the entire hydraulic design. The pipe sizing, the pump selection, the expansion tank sizing, the heat exchanger performance — every component that depends on fluid properties must be checked against the new fluid.

The Jiangsu piping upgrade from 1.5-inch to 2.5-inch resolved the pressure drop issue. The pump was retained. The expansion tank was resized for the higher thermal expansion coefficient of LM-11D. The heat exchanger was checked and found to be adequate — the reduced heat transfer coefficient of LM-11D was offset by a larger ΔT across the exchanger at the new, lower operating temperature.

The cost of the piping upgrade was significant. The benefit was a system that actually works at the new operating conditions. The alternative — switching to LM-11D without the piping upgrade — would have resulted in a system that couldn’t achieve the required cooling duty. The pump would have been operating at the end of its curve, cavitating intermittently, and the flow rate would have been insufficient. The piping upgrade wasn’t optional. It was necessary.

5. Freeze Point Adjustment: 12.42 Tons of Fine-Tuning

Huaqing Ronghao in Beijing operates a secondary refrigerant loop with a 12.42-tonne charge of LM-4. The system was originally designed for a freeze point of -15°C. The process requirements changed, and the operator needed to lower the freeze point to -20°C.

With LM-4, the freeze point is controlled by the concentration of the active ingredient. The standard LM-4 formulation is delivered as a concentrate that is diluted with water on-site. The freeze point is a function of the dilution ratio. To lower the freeze point from -15°C to -20°C, the concentration needs to increase — which means adding concentrate to the existing charge.

The calculation: 12.42 tonnes of LM-4 at -15°C freeze point has a certain concentration. To reach -20°C freeze point, the concentration needs to increase by approximately 5 percentage points. The required amount of concentrate is about 620 kg. The concentrate was added to the expansion tank, circulated for 24 hours to ensure uniform mixing, and the freeze point was checked with a refractometer. The target was achieved on the first attempt.

The adjustment was straightforward, but the scale is worth noting. A 12.42-tonne charge is a significant amount of coolant. The cost of draining and replacing the entire charge would have been substantial. The ability to adjust the freeze point in-situ — by adding concentrate to the existing charge — is a significant operational advantage of concentrated coolant formulations. The alternative, for a pre-mixed coolant, would have been to drain the entire system and refill with the new concentration. That would have cost tens of thousands of dollars in coolant and days of downtime.

Huaqing Ronghao’s system has been operating at the new freeze point since May 2025. The adjustment process took less than 48 hours from start to finish. The system was never offline — the concentrate was added while the system was circulating. This is the kind of operational flexibility that concentrated coolant formulations provide.

6. Pipe Sizing Reference

The sizing above is conservative. It assumes a maximum velocity of 2.0-2.8 m/s to keep the pressure drop manageable. For water-free coolants, the pipe size is one nominal size larger than for water-based coolants at the same flow rate. This is a rule of thumb, not a substitute for a proper hydraulic calculation. Every system is different. Calculate the pressure drop. Check the pump curve. Verify the NPSH. Do the work.

7. Expansion Tank Sizing

The expansion tank is the most frequently undersized component in secondary refrigerant loops. The tank must accommodate the thermal expansion of the entire fluid charge from the minimum temperature to the maximum temperature. If the tank is too small, the system pressure will spike on warm-up and the relief valve will lift. If the tank is too large, the pressure will drop on cool-down and the pump may lose suction.

The expansion volume is calculated as: V_exp = V_system × β × ΔT, where β is the volumetric thermal expansion coefficient and ΔT is the temperature range. For water-based coolants (LM-4, LM-8), β is approximately 0.0004 to 0.0006 per °C. For water-free coolants (LM-11 series), β is approximately 0.0008 to 0.0012 per °C — roughly double. The expansion tank for a water-free system needs to be roughly twice as large as for a water-based system of the same volume and temperature range.

The expansion tank pre-charge pressure should be set to the minimum system pressure at the tank location. The tank acceptance volume — the amount of fluid the tank can accept between the pre-charge pressure and the relief valve set pressure — must be at least equal to the expansion volume. If the acceptance volume is less than the expansion volume, the tank is undersized.

This is not complicated. It’s arithmetic. But it’s arithmetic that gets skipped surprisingly often. Don’t skip it.

8. Commissioning Checklist

1. Flush verification. The flush water must be clear, with conductivity within 10% of the supply water. Any residual debris, oil, or old coolant will contaminate the new charge.
2. Pressure test. Test at 1.5× design pressure for a minimum of 30 minutes. No pressure drop. No leaks. Fix leaks before filling — finding a leak after filling means draining the system.
3. Fill with the correct concentration. Mix the coolant with water (if applicable) to the target freeze point. Do not guess. Measure.
4. Circulate and vent. Run the pump at low speed and vent air from all high points. Air in the system reduces heat transfer, causes pump cavitation, and accelerates corrosion.
5. Check the freeze point after 24 hours of circulation. The concentration may not be uniform immediately after filling. Sample from multiple points.
6. Verify the pump operating point. The pump should be operating within its preferred operating region — not at shutoff, not at runout. Check the flow rate, the head, and the power draw.
7. Set the expansion tank pre-charge. With the system cold and the pump off, set the pre-charge to the minimum system pressure at the tank location.
8. Record baseline data. Freeze point, pH, conductivity, pump amps, flow rate, system pressure. These are your reference for future troubleshooting.
9. Run for 48 hours under load. Monitor for leaks, pressure fluctuations, temperature control issues, and unusual noises. Fix problems before they become failures.

9. When to Call Someone

You’ve read this far. You have a system that needs a secondary refrigerant loop. You’ve looked at the pipe sizing table, the expansion tank calculation, the commissioning checklist. You’re thinking: “I can do this.”

Maybe you can. A lot of systems are straightforward. A closed loop with a single chiller, a single pump, and a single heat exchanger. LM-4 or LM-8, water-based, not too cold, not too hot. The standard design works. Follow the checklist. You’ll be fine.

But if your system has any of the following — call someone. Glacier Coolant has an engineering team that does this every day. They’ve seen the failure modes. They know what to check.

Call if: your operating temperature is below -50°C. You’re crossing the water-based to water-free boundary. Your ΔT is less than 3°C. Your process gas is corrosive. Your system is open-loop. Your expansion tank is atmospheric and you can’t retrofit it. Your pump is cavitating and you don’t know why. Your inhibitor is depleting faster than expected. Your corrosion coupons are showing red flags.

This field handbook covers the common cases. The uncommon cases need an engineer on-site. There’s no shame in calling. There’s a lot of shame in flooding a compressor room with coolant because you didn’t check the drain valve. Ask me how I know.