Three catalogues, one replacement method
After Sanyo Denki, the three catalogues most often open on a replacement bench are Delta, NMB-MAT and SUNON. Each names its fans differently, and each names its bearing technologies differently — which tempts engineers into three different selection shortcuts. The shortcut is the trap; the five-checkpoint method is catalogue-independent.
Delta, NMB and SUNON fans sit in the same machines as the 9GA class: servers, storage shelves, telecom rectifiers, industrial controllers and battery cabinets. Replacement requests start the same way — an expensive or discontinued part number, a service budget, and the hope that a same-size fan will simply drop in. The three manufacturers earned their installed bases with different emphases, and their catalogues reflect it: broad application families with extensive interface and protection variants on one side, precision frame coding with long-life ball bearing platforms on another, and branded bearing technologies on the third.
The naming differences are real, and they matter — but only as maps for finding the right datasheet faster. They do not change what a replacement must prove. Whether the original is a Delta, an NMB or a SUNON, the candidate has to pass the same five checkpoints (mechanical, electrical, operating point, acoustics, life and environment) and finish as a validated sample in the machine. This guide walks the three catalogues with one method, and marks the places where each manufacturer's conventions are most likely to mislead a footprint-first selection. The method itself is detailed in the operating point replacement method and its 9GA-class application.
How each catalogue names its fans — and how to read it
The part number is a map to a datasheet, not a specification. Each manufacturer encodes family, frame and options differently, and each convention has one classic misreading.
Delta organises its DC fans into broad application families by airflow and pressure emphasis, with extensive option variants for interface, speed code and protection features. The classic misreading is family-first selection: choosing within the family that sounds right instead of comparing the candidate's curve against the measured system resistance. The family narrows the candidate list; the operating point picks the fan.
NMB-MAT codes its fans with numeric frame identifiers — the digits carry approximate diameter and thickness information — and pairs them with long-life ball bearing platforms. The classic misreading is treating the numeric code as an exact dimension: the code identifies the series, and the mechanical drawing gives the true frame, hole pattern and thickness. Measure the part, not the number.
SUNON distinguishes series by bearing technology and platform, with its MagLev magnetic levitation bearing as the best-documented example of a technology carried in the series name. The classic misreading is the reverse of the first two: accepting the technology name as the specification. A bearing technology name tells you what to look for in the life curve and the environmental data; it does not replace them.
In all three catalogues the same rule holds: the trailing option codes — interface (tach, PWM), speed class, protection and coating variants — are where machines and controllers actually differ. Two fans that share frame and speed but differ in option code are not interchangeable in a monitored system. Open the exact option's datasheet, and verify the variant list against what the controller expects.
| Catalogue | What the name encodes | Classic misreading | The corrective |
|---|---|---|---|
| Delta | Application family, frame, interface and protection option codes | Family-first selection by description | Compare candidate curves against the measured system resistance |
| NMB-MAT | Numeric frame series code plus bearing platform and option codes | Reading the numeric code as exact dimensions | Measure the physical part; the datasheet drawing is the truth |
| SUNON | Series with bearing technology naming (MagLev is the documented example) | Accepting the technology name as the life specification | Convert the quoted life curve to the real inlet temperature |
Mechanical and electrical: the variants are where machines differ
Frames and connectors look interchangeable across the three catalogues. The option variants behind similar-looking part numbers are not.
Mechanically, the checkpoints are identical to the method guide: frame dimensions, thickness, hole pattern and pitch measured on the part, inlet opening and shroud geometry compared on the parts, and the gasket plane checked against the enclosure seal. The cross-catalogue trap is subtle: a replacement from a different manufacturer with the same nominal frame can have a different inlet opening or impeller shroud, because each high-pressure design shapes its inlet differently. Free open area is a measured number, not a catalogue assumption.
Electrically, the interface contract is the same Intel 4-wire PWM specification across server-class fans from all three manufacturers: 25 kHz nominal control, duty commanded speed, tach output of two pulses per revolution. What differs is the variant matrix around it — which options add locked-rotor detection, which offer moisture-resistant coating, which speed classes exist, and how each manufacturer documents the duty-to-speed curve. A controller tuned to the original's response needs the candidate's response to sit inside the firmware's assumptions; otherwise the machine logs faults that the fan is not responsible for.
Current and voltage checkpoints also translate directly: running and stall current against the driver budget, the operating point held at minimum supply voltage, and start-up behaviour that the backplane tolerates. Related reading: PWM fan control.
| Item | What to compare | Source |
|---|---|---|
| PWM band and duty response | Frequency range covers the controller's 25 kHz; duty-to-speed behaviour inside firmware assumptions | Candidate datasheet plus the controller's tuning notes |
| Tach and protection | 2 pulses per revolution, locked-rotor and under-speed reporting equivalent to the original option code | Both option-code datasheets |
| Voltage window | Operating point held at minimum supply voltage, not nominal | Candidate P-Q data at the real voltage |
| Current | Running and stall current within the driver budget | Electrical tables of both fans |
| Variants | Coating, IP treatment and sensor options matched to the environment | Option code lists and test reports |
Operating point: comparing curves across brands
Cross-brand replacement is where free-air comparisons fail loudest. The three catalogues publish curves in the same P-Q form precisely so that this comparison can be made honestly.
Worked example (numbers are illustrative of the method, not any specific product). A telecom rectifier shelf cools through a loaded filter bank; the measured system resistance per fan is P = 0.0024 Q² (Pa against m³/h). The original fan crosses at roughly 142 m³/h against 48 Pa, and the shelf's thermal budget requires 130 m³/h at that state. Candidate A from a different manufacturer — a higher-pressure design with 95 Pa shut-off — crosses at about 148 m³/h, a 13 % margin over the requirement: accepted. Candidate B, a general-purpose type with a bigger free-air rating of 225 m³/h but only 45 Pa of shut-off, crosses at about 117 m³/h — below the requirement at the loaded state that decides the selection.
The comparison needs both curves at the same supply voltage and the same temperature basis, and it needs the system curve from the machine — measured if the original is installed, reconstructed from vent velocity and filter pressure drop if it is not. The margin discipline is the same as in the method guide: under 10 % airflow margin at the worst-case point is unproven unless the manufacturer states an aerodynamic tolerance and the comparison used the bottom of the band.
One cross-brand courtesy saves a sample cycle: ask each candidate manufacturer to confirm the crossing point against your system curve in writing. Reputable manufacturers will do the arithmetic; a supplier who answers with the free-air number has told you something about their selection process.
Bearing technologies: the name is not the life data
NMB's long-life ball bearing platforms, SUNON's MagLev magnetic levitation bearing and Delta's standard ball bearing options are real engineering differences. They become comparable only when every life claim is converted to the temperature your machine imposes.
SUNON's MagLev bearing removes contact between rotating and stationary parts, which the manufacturer documents as its approach to long life and dust tolerance. NMB's platforms are built around long-life ball bearing systems with published life curves at reference temperatures. Delta offers standard and long-life ball bearing options across its families. None of these claims is comparable as printed, because each is quoted at its own reference inlet temperature.
The conversion rule from the method guide applies across all three: every 10 C rise in bearing lubricant temperature roughly halves L10 life. Convert each candidate's quoted life to the machine's real worst-case inlet temperature — in a stacked shelf or a battery cabinet, the fan inlet is often far above the reference — and then compare. The ranking the catalogues suggest sometimes reverses, and finding that on paper costs nothing.
Where the environment adds dust, humidity or corrosion, the bearing conversation extends to sealing and coating variants. Each catalogue marks coated boards and sealed variants in its option codes; match the variant class to the exposure and ask for the environmental test report that covers it — salt fog severity to the IEC 60068-2-52 method where the exposure is coastal or industrial.
Acoustics across brands: verify the sweep, not the datasheet
The three manufacturers document sound power at stated operating points. The acoustic character that field engineers notice shows up across the duty sweep, in the housing.
High-pressure designs from any of the three catalogues carry a blade-passing tone; general-purpose designs of the same frame sound different at the same overall level. A cross-brand replacement can match the datasheet sound power and still change the machine's acoustic signature, because tonality and frequency content are not captured by a single number. The comparison is made in the real housing, at the operator position, across the duty sweep.
PWM behaviour completes the check. Each manufacturer's duty-to-speed response differs in shape; at idle duty, some designs hum or hunt where others run silently. The controller's sweep that the thermal validation already runs is the right moment for the listening check — 30 seconds across the duty range, operator position and rack rear, before and after the endurance interval.
The arithmetic is unchanged: +3 dB(A) in sound power is a doubling of acoustic energy, and a candidate 6 dB(A) louder changes the product rather than the fan.
Validation: the five steps, with cross-brand additions
The validation protocol does not change per brand. What changes is the attention the handshake and sweep steps deserve, because a cross-brand replacement meets a controller tuned to a different manufacturer's response.
The five steps are the method guide's: mechanical and electrical handshake, control sweep, thermal worst case, acoustic acceptance, endurance re-check. For a cross-brand replacement, two of them carry extra weight. The handshake must confirm that tach levels, locked-rotor reporting and boot behaviour sit inside what the controller's firmware expects from the original's option code. The control sweep must compare duty-to-speed behaviour across the whole range, because the firmware's assumptions about response shape were learned from the original brand's curve.
The sample validation in the real enclosure guide details the protocol. The cross-brand discipline adds one administrative step: record which option code, coating variant and speed class the sample carries, and keep it with the validation report — service teams order spares by option code, and a report that says only the frame size will produce the wrong spare.
- Handshake: fit, connector, tach report, fault-free boot cycles against the controller's expectations.
- Control sweep: duty-to-speed response across the full range, protection behaviour, idle acoustics.
- Thermal worst case: loaded filter, hot inlet, component temperatures against the original's baseline.
- Acoustic acceptance: listening check in the housing, across the sweep, at operator position and rack rear.
- Endurance re-check: agreed hours, then repeat the sweep and thermal comparison; record the exact option code with the report.
What MEGA Tech does with Delta, NMB and SUNON replacement requests
Replacing an established fan from any of the three catalogues with equal duty performance at lower cost is the core of our business. The workflow is the same as for the 9GA class, and so are the boundaries.
What we do: send us the original part number and datasheet — or better, the measured duty from your machine — and we propose candidates whose curves match the operating point, produce samples with the mechanical and electrical interface your machine expects, provide measured P-Q and acoustic data for the sample, and support the five-step validation protocol in your assembly. We work across all three catalogues daily, and we will tell you when the original's duty is best served by a candidate from a different family than the one the footprint suggests.
What we need from you: the original part number and option code (or the physical fan), the controller's PWM frequency and protection thresholds, the worst-case environment, and any noise or life specification the product carries. A measured operating point — airflow at the loaded resistance, inlet temperature — moves the selection from pre-footprint to pre-validated.
Where our boundary is: we do not claim equivalence by footprint, and we do not ask you to accept it — equivalence is decided by the operating-point comparison and the validation protocol in your machine. We do not supply liquid cooling loops, heat pipes, cabinet safety design or fan controllers; our scope is the fan and the data that lets you verify it.
Engineer FAQ
The questions engineers actually ask about cross-brand replacements, with the short engineering answers.
Q: Can I replace a Delta fan with an NMB, or a SUNON with a Delta? A: Yes, if all five checkpoints pass — brand is not a checkpoint. The operating point, interface, acoustics and life data decide, followed by a validated sample in the machine.
Q: The numeric code on the original matches the candidate exactly. Are they interchangeable? A: No. Each manufacturer's code encodes its own series logic. Match the measured frame, hole pattern, connector and option code, then compare the datasheets.
Q: The candidate quotes a longer life at 40 C. Is it more durable? A: Not necessarily. Convert both life claims to the machine's real worst-case inlet temperature; every 10 C rise roughly halves L10 life, and the ranking can reverse.
Q: Do MagLev and ball bearing fans sound different? A: Bearing technology affects tonal character and lifetime behaviour, but the datasheet sound power at the operating point plus a listening check in the housing is the only comparison that predicts what the operator will notice.
Q: We monitor fan speed. What is the first thing that breaks in a cross-brand swap? A: Usually nothing mechanical — the controller logs under-speed or missing tach because the candidate's duty-to-speed shape or protection reporting differs from the original option code. The control sweep finds it before the machine does.
Q: Can one supplier really cover all three catalogues? A: The method is catalogue-independent, and so is the sampling workflow. What you should verify is the same as with any supplier: measured curve data, the exact option code on the sample, and validation support in your machine.
Questions for a design review
Take these to the review with the original part number, the three candidate datasheets and the measured system resistance on the table.
A cross-brand replacement is a small design review with a tight scope. These questions close the gaps that produce field surprises, and every one can be answered before samples are ordered.
- Which option code does the original carry, and does each candidate match interface, sensor and coating variants — not just frame and speed?
- Has every candidate's curve been crossed against the measured system curve at minimum supply voltage, with the filter loaded?
- Have all three life claims been converted to the machine's real worst-case inlet temperature before ranking?
- Does the controller's firmware assume a duty-to-speed shape that any candidate violates?
- Has the acoustic comparison been made in the housing across the duty sweep, including the idle-duty behaviour?
- What aerodynamic tolerance does each manufacturer state, and was the crossing comparison made at the bottom of the band?
- Does the validation report record the exact option code, coating variant and speed class of the sample that passed?
