Why 9GA-class fans dominate replacement requests
Sanyo Denki's SAN ACE 9GA type appears in more replacement enquiries than any other fan family — it is specified across servers, storage arrays, telecom rectifiers and industrial controllers, and its combination of high static pressure and compact frames made it the default choice of two decades of equipment designers.
The manufacturer's own product pages describe the 9GA type as a cooling fan series covering frame sizes from 40 mm to 120 mm square, built around a high static pressure impeller design, with the 9GV type covering the 119 x 119 x 32 mm high-airflow class. Those two families earned their installed base for reasons that matter to this guide: they deliver useful pressure at the resistance real electronics enclosures impose, they follow the 4-wire PWM interface that controllers expect, and they carry long-life bearing systems documented at elevated temperatures.
That installed base is exactly why replacement demand is constant. Equipment outlives the original fan's catalogue lifetime or the original's production run; price pressure on the service channel grows; and the engineers who inherited the machine need a fan that behaves like the one the thermal design assumed — without paying a premium for a legacy part number. This guide is the concrete application of the operating point replacement method to 9GA-class requests: what the part number tells you, what it does not, and the checkpoints and validation protocol a credible 9GA replacement must pass.
One warning before the checkpoints. The 9GA type's reputation was earned at the operating point — at pressure. A replacement that matches the frame and the free-air number but not the pressure capability will spin, move air, and quietly under-cool a loaded enclosure. Everything below exists to prevent that specific failure.
What the part number tells you — and what it does not
The 9GA part number is a family code plus frame size, thickness, voltage and option codes. It tells you which datasheet to open. It does not tell you whether a candidate fan matches the duty the original performs in your machine.
Read the original part number first, because it saves real time. The leading code identifies the series and frame family; the following digits give nominal voltage and size; the trailing option codes select the interface — tach output, 4-wire PWM control, locked-rotor sensor — and variants such as moisture-resistant coating. Each option code maps to a different wiring diagram and a different controller expectation, so two fans with the same frame and speed but different option codes are not interchangeable in a machine that monitors its fans.
Then stop trusting the number and open the datasheet. The manufacturer's datasheet for the exact option code is the source of truth for the electrical table, the P-Q curve, the acoustic data and the life curve. Measure the physical fan where the machine is concerned: hole pitch, frame thickness, connector, and the inlet-side geometry, because a high-pressure impeller's shroud and inlet opening are part of its aerodynamic design — a replacement with a different open area adds a restriction in series with the system.
Finally, record the duty the machine imposes — not what the fan is capable of, but what it actually does: airflow at the resistance of the loaded path, inlet temperature at worst case, the control interface actually wired, and the environment. Send that list with the part number when requesting a replacement. A supplier who receives only the part number will match the footprint; a supplier who receives the duty can match the fan.
| Item | Example | Why it selects the candidate |
|---|---|---|
| Original part number and option code | Full code from the fan label, plus the datasheet revision | Fixes frame, voltage, interface and variant — the reference point |
| Original datasheet P-Q curve (or a measured operating point) | Required airflow at the static pressure of the loaded path | The single number a high-pressure replacement must beat |
| Controller interface actually wired | 4-wire PWM at the controller's frequency, or 3-wire tach, or 2-wire | Wrong interface logic logs faults even when cooling works |
| Worst-case environment | Inlet temperature, orientation, dust load, salt fog, altitude | Bearing system, coating and IP treatment follow the environment |
| Noise limit at the operator position | Sound power or sound pressure at a stated distance | A candidate 6 dB(A) louder changes the product |
Mechanical: match the opening, not just the holes
Frame size, thickness and hole pattern are the easy half of the mechanical checkpoint. The hard half is the inlet geometry — on a high-pressure design it is aerodynamic structure, not decoration.
Measure instead of decoding: the exact frame dimensions, thickness, mounting hole pattern and pitch, the connector position and keying, and the rotating-direction arrow. A 9GA-class replacement must fit the existing mount without adapters where the machine's service documentation assumes a direct swap; adapters are acceptable only when the mechanical designer owns them deliberately.
The inlet side is where footprint-matching quietly fails. A high-static-pressure impeller works against a shaped inlet — the shroud, the inlet opening diameter and the clearance between impeller and frame. A replacement whose inlet opening or shroud geometry differs changes the effective open area and the tip clearance, and both move the real P-Q curve of the assembly even if the fan alone is stronger. The comparison to make is the free open area of the original's inlet versus the candidate's, measured on the parts, not read from the catalogue.
Finger guards, filters and EMC grilles sit in series with the fan and belong to the system curve, but a thicker replacement frame can change the seal to the enclosure gasket and create a recirculation leak that no datasheet will show. Check the gasket plane and the fastener stack before the sample arrives.
Electrical: the 4-wire interface is a contract with the controller
Most 9GA-class server fans follow the Intel 4-wire PWM specification: 25 kHz nominal control (21–28 kHz allowed), duty cycle commands speed, and the tach line returns 2 pulses per revolution. Matching this interface is a checkpoint, not a formality.
Three things must agree between candidate and controller. First, the PWM frequency range: a controller driving 25 kHz needs a fan specified across that band, because driving a fan outside its specified range distorts the speed response. Second, the duty-to-speed behaviour: manufacturers do not linearise identically, and a controller whose firmware assumes the original's duty curve may flag under-speed at a duty where the replacement is merely shaped differently. Third, the protection behaviour: locked-rotor detection, tach missing, and under-speed thresholds are configured against the original's response, and the replacement must trip and report the same way or the service team inherits false alarms.
Voltage is part of this checkpoint, not a separate one. The 9GA type is offered across the common 12 V and higher-voltage classes; a replacement must hold its operating point at the low end of the machine's supply tolerance, because the curve drops with voltage and the crossing point moves left. Current matters in both directions: running and stall current against the driver budget, and enough current draw behaviour at start-up that a hot-plugged backplane does not brown out.
Where the original is a 3-wire (tach, no PWM) or 2-wire (thermistor-controlled) variant, the same discipline applies with different parameters: the speed control law and the tach electrical levels must match what the board was designed around. Related reading: PWM fan control.
| Signal | What to verify | Failure mode if mismatched |
|---|---|---|
| PWM input | Frequency range covers the controller's 25 kHz band; duty response documented | Speed does not follow command; over-speed at boot |
| Tach output | 2 pulses per revolution, open-collector levels per the original | Controller reads zero speed and raises a fault |
| Locked-rotor / under-speed | Protection and reporting behaviour equivalent to the original option code | Silent stall or false alarms in service |
| Supply window | Operating point held at minimum supply voltage, not just nominal | Curve drops; cooling margin disappears at brown-out |
| Current | Running and stall current within the driver budget | Tripped supply or degraded start-up |
Operating point: the high-pressure claim must survive your system curve
The 9GA type is chosen by designers because of its pressure capability. That is precisely the property a cheaper replacement is most likely to lack — and the only way to prove it is to cross curves.
Worked example (numbers are illustrative of the method, not any specific product). A storage chassis pushes air through a loaded filter bank; the measured system resistance per fan is P = 0.0065 Q² (Pa against m³/h). The original high-pressure fan crosses this system curve at roughly 97 m³/h against 61 Pa, and the thermal budget requires 90 m³/h at that state. Candidate A is a high-pressure design with a shut-off pressure of 95 Pa and 200 m³/h free air; its curve crosses at about 103 m³/h — a 6 % margin over the original, accepted. Candidate B is a general-purpose fan with a flashier free-air rating of 215 m³/h but only 50 Pa of shut-off; against the same loaded resistance its curve crosses at about 81 m³/h — below the requirement at exactly the state that matters. Candidate B has the bigger box number and fails the duty.
This is the entire reason the 9GA class is hard to replace casually: high-pressure impellers give up free-air number to gain pressure, so their free-air ratings often look worse than general-purpose fans of the same size. A selection made on free-air numbers will systematically choose the wrong fan in high-resistance paths. The comparison that decides is the crossing point with the measured system curve, at the loaded filter, the hot inlet and the minimum supply voltage — the worst-case conditions from the method guide.
Where the original datasheet curve is unavailable, reconstruct the duty from the machine: anemometer traverse at the vent corrected for outlet area, plus the differential pressure across the filter bank. It is approximate, so add margin discipline — under 10 % airflow margin at the worst-case point is unproven unless the manufacturer states an aerodynamic tolerance and the comparison was made at the bottom of the band.
Acoustics: the tone that the overall dB(A) hides
High-pressure impellers are efficient but tonal. A replacement can match the original's overall sound power and still sound wrong to an operator standing next to the rack.
The blade-passing frequency of a high-pressure impeller sits higher than a general-purpose fan's at the same speed, because more blades pass the inlet each revolution. Overall sound power can be identical while the character changes from broadband rush to a perceptible tone — and perceptibility is what field engineers and their customers actually notice. The comparison must therefore be made in the real housing, not in a datasheet table, and it must include a listening check at the operator position and at the rear of the rack where tones reflect.
PWM operation adds a second acoustic trap. At low duty the fan runs near its minimum speed, and some designs produce a low-frequency tonal hum or a periodic speed-hunting behaviour when the controller's loop interacts with the fan's response. The original's behaviour at idle duty was tuned into the product's acoustic signature; the replacement's idle behaviour needs the same 30-second check across the duty sweep that the thermal test already runs.
The arithmetic stays the same as in the method guide: +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. Related reading: fan noise and acoustics.
Life and environment: convert the L10 claim to your inlet temperature
Long life is one of the reasons the 9GA class was specified in the first place. A replacement's life claim is only comparable after both numbers are converted to the temperature the machine actually imposes.
Life curves are quoted at a reference inlet temperature — commonly 40 C or 60 C in this class. The machine's fan inlet at worst case is often well above both, especially in a stacked chassis where each fan breathes the preheated exhaust of the row below. The Arrhenius rule of thumb from the method guide applies: every 10 C rise in bearing lubricant temperature roughly halves the L10 life. Convert the original's quoted life and the candidate's quoted life to the real inlet temperature, then compare — the ranking sometimes reverses, and the reversal is always cheaper to find on paper than in the field.
Bearing system follows the environment. The 9GA class is offered with long-life ball bearing systems; a replacement must match or exceed the bearing class, not just the hours at reference temperature. For telecom outdoor and coastal cabinets, add the corrosion dimension: moisture-resistant coating variants and salt fog severity classes tested to the IEC 60068-2-52 method, with the severity matched to the exposure the cabinet is rated for. Altitude derating, dust load and orientation complete the environmental row of the checkpoint table.
Validation: five steps in the machine that ships
A 9GA-class replacement is proven in the assembly, with the controller it must obey and the acoustic environment it must live in. Bench results are pre-work, not validation.
Step one, mechanical and electrical handshake: the sample fits the mount, seats the connector, spins up, and the controller reports correct tach and no faults across a full power cycle. Step two, control sweep: command the full duty range and record speed response, protection behaviour and idle-duty acoustics against what the original did. Step three, thermal worst case: run the machine at maximum load with the filter loaded and the inlet at worst case, and compare component temperatures and airflow-derived readings against the original's baseline.
Step four, acoustic acceptance: the listening check at operator position and rack rear across duty sweep, plus a measurement if the product has a noise specification. Step five, endurance and re-check: run the agreed hours — long enough to see bearing break-in effects — and repeat steps one to four, because a fan that passes cold and fails after 500 hours is the failure mode service teams actually meet. The sample validation in the real enclosure guide details this protocol; the point that matters for 9GA-class replacements is that steps two and four are where footprint-matched candidates usually die, and they are cheap to run.
- Fit and handshake: mount, connector, tach report, fault-free power cycles.
- Control sweep: duty-to-speed response and protection behaviour against the original's.
- Thermal worst case: loaded filter, hot inlet, component temperatures versus baseline.
- Acoustic acceptance: operator position and rack rear, across the duty sweep.
- Endurance re-check: agreed hours, then repeat the sweep and thermal comparison.
What MEGA Tech does with a 9GA-class request
We are a custom cooling fan manufacturer, and replacing an established series at lower cost with equal duty performance is the core of our business. Here is exactly how the workflow runs, what we need from you, and where our limits are.
What we do: given the original part number and datasheet — or better, the measured duty from your machine — we propose candidates whose curves match the operating point, produce samples with the mechanical and electrical interface your machine expects (frame, connector, PWM and tach behaviour, coating variants where the environment demands them), provide measured P-Q and acoustic data for the sample rather than only catalogue figures, and support the five-step validation protocol in your assembly.
What we need from you: the original part number and its option code (or the physical fan), the datasheet if you have it, the controller's PWM frequency and protection thresholds, the worst-case environment, and any noise or life specification the product carries. If you can share a measured operating point — airflow at the loaded resistance, inlet temperature — the candidate filtering starts pre-validated instead of pre-footprint.
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 above, executed in your machine, on your controller. 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. Contact us with the part number and the duty, and the first question back will be about the operating point, not the price list.
Engineer FAQ
The questions engineers actually ask about 9GA-class replacements, with the short engineering answers.
Q: Is a 9GA replacement valid if it only comes in a different frame size? A: Yes, if the mounting can be adapted and all five checkpoints pass — the footprint is not the specification, the duty is.
Q: The replacement's free-air CFM is higher than the 9GA's. Does that mean it is stronger? A: No. High-pressure designs give up free-air number to gain pressure. In a high-resistance path the low-pressure fan with the bigger box number crosses the system curve lower and under-cools the machine.
Q: Can I keep the controller untouched and just swap the fan? A: Only if the PWM frequency range, duty-to-speed behaviour, tach levels and protection reporting match what the firmware expects. Otherwise the controller logs faults or misjudges speed even when cooling works.
Q: Both datasheets quote L10 life at 60 C. Are they comparable as printed? A: Only after conversion to the real inlet temperature. Every 10 C rise roughly halves bearing life, and a stacked chassis can put the fan inlet far above 60 C at worst case.
Q: We only need one replacement unit for service, not a redesign. Is the process the same? A: The checkpoints are the same and the sample validation shrinks to the steps that matter for one unit: handshake, control sweep, thermal worst case, listening check. Skipping the control sweep is how fault logs appear after the swap.
Q: What if the original datasheet curve is not available? A: Reconstruct the duty from the machine — anemometer traverse plus filter differential pressure — add margin discipline, and let the sample's measured curve decide. Treat catalogue free-air numbers as marketing until proven.
Questions for a design review
Take these to the review with the original part number, its datasheet and the measured system resistance on the table.
A 9GA-class replacement request 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 the candidate match the interface, sensor and coating options — not just frame and voltage?
- What airflow does the original deliver at the loaded system pressure, and does the candidate's curve cross at or beyond it at minimum supply voltage?
- Have both L10 quotes been converted to the real worst-case inlet temperature of the machine?
- Does the controller's firmware assumption about duty-to-speed behaviour still hold with the candidate's response?
- Has the acoustic comparison been made in the real housing, including the blade-pass tone and the idle-duty behaviour?
- What aerodynamic tolerance does the candidate's manufacturer state, and was the crossing comparison made at the bottom of the band?
- Does the validation plan include the endurance re-check, and who owns the thermal budget decision if a checkpoint lands slightly below?
