Helium Drive Data Recovery

Full In-House Helium Recovery

Why does in-house helium capability matter?

Some labs advertise helium drive recovery but open the drive on a standard bench without helium refill capability. In atmospheric air, the heads experience incorrect aerodynamic lift, causing head-platter contact that strips the magnetic coating from the platters. A failed recovery attempt on a helium drive does not just fail; it destroys any chance of a second attempt.

Other labs that lack the equipment send the drive to a third party, adding cost, transit time, and a middleman between you and the engineer doing the work. You lose visibility into the process and cannot communicate directly with the technician handling your drive.

We keep the entire process under one roof. From the initial PC-3000 diagnosis through the head swap, helium refill, and final imaging, your drive stays at our Austin lab. You speak directly with the technician working on your case.

What are the failure modes of helium drives?

Helium Leak and Permeation

Helium atoms are small enough to permeate through micro-imperfections in the welded seal over years of operation. As helium leaks, atmospheric air enters and changes the gas density inside the chamber. The heads begin flying at incorrect height, causing intermittent read errors that escalate to full head-platter contact. Symptoms include increasing acoustic noise, degrading read performance, and SMART attribute 22 reporting declining helium levels. By the time the drive fails, the internal atmosphere is a helium-air mixture that won't support stable head flight. Recovery requires a head swap with fresh helium refill.

Preamp Failure in High-Platter Drives

Helium drives with 8–10 platters can have up to 20 individual read/write heads. Each head connects to a preamp IC mounted on the head stack assembly. Thermal stress from continuous operation in enterprise or NAS enclosures degrades preamp components over time. When a preamp channel fails, the associated head stops reading. If multiple channels fail, the drive reports as inaccessible. PC-3000 head maps identify which heads are functional, and we image from working heads first to capture maximum data before swapping to donor heads for the failed channels.

Firmware Corruption

HGST and WD helium platforms are susceptible to translator table corruption from sudden power loss. The translator maps logical block addresses to physical platter locations; if this table is damaged, the drive may report 0 bytes capacity or fail to initialize entirely. Seagate Exos models can experience Service Area module corruption that prevents the drive from completing its startup sequence. Both failure types are repairable through PC-3000 terminal access without breaking the hermetic seal.

Motor Bearing Seizure

Spindle motor failure in a helium drive is the hardest recovery scenario. With 8–10 tightly packed platters, a motor swap requires removing the entire platter stack, transferring it to a donor chassis, refilling with helium, and imaging. If the seized motor caused the platters to shift rotationally, alignment errors compound across every platter surface. The platter damage tier ($4,000–$5,000) covers this scenario, plus helium cost and donor.

Why does helium recovery cost more than standard HDD recovery?

Three factors push helium recovery pricing above standard hard drive data recovery rates: helium gas cost, donor drive cost, and the complexity of working with high-platter-count head stack assemblies.

Helium cost: $400-$800 additional for head swap and surface damage tiers. This covers the helium refill required after opening the sealed chamber. The donor drive is also consumed in the process and can't be reused. A head swap on a 10-platter drive with 20 heads takes longer and carries more risk than swapping 4 heads on a 2-platter consumer drive.

Firmware-level recoveries ($900–$1,200) don't require helium or a donor, because the hermetic seal stays intact. If your helium drive is inaccessible but not clicking, the repair cost is closer to standard HDD pricing. We diagnose the failure type for free and quote the exact tier before any work begins. A +$100 rush fee to move to the front of the queue is available if you need priority service.

Technical Methodologies for Helium Drive Recovery

The sub-sections below describe the bench procedures used on helium drives at our Austin lab. Each procedure is the same regardless of drive vendor; the parameters differ by family.

Sealed-Chamber Refill: Helium Purity and Partial-Refill Drift

We refill with industrial helium at 99.999% purity (grade 5.0). Lower-purity helium carries trace nitrogen and moisture that raise the average gas density inside the chamber above the 4 g/mol target the heads were designed for. The air-bearing pressure under the slider shifts, and the head lands or crashes outside its calibrated fly-height window during seek operations.

Partial refill is not an option. If the chamber is filled to anything less than the original helium volume, the residual atmospheric air mixes with the helium and the composite gas density sits between the two reference points. The heads then drift across the platter surface during seek operations because the air-bearing equilibrium oscillates with rotational position. We measure refill volume against the chamber spec for each drive family before reseal and PC-3000 imaging. Helium cost: $400-$800 additional for head swap and surface damage tiers. This covers the helium refill required after opening the sealed chamber.

Helium Donor Matching: Preamp Revision and HSA Family

Our standard donor-matching process for helium drives extends beyond model number and firmware. The head stack assembly carries a preamp IC whose channel-to-head wiring must match the target drive's preamp revision; an Ultrastar HC520 preamp wired for a 14TB platter map will not address heads correctly when transplanted into an HC530 chassis even though both share the Ultrastar form factor. Micro-jog (track offset) calibration values are also stored against the original HSA, so a mismatched donor head reads off-track until PC-3000 recalibrates the micro-jog table on the imaged volume.

• ✓Seagate Exos X-series: match site code, head map (HM), and preamp revision against the target's Service Area identifier. Same-family donors (Exos X10 to X10) preserve micro-jog tolerances; cross-family substitutions require firmware-level re-mapping. See the Seagate Exos 10TB failure profile for the most common Exos failure modes we see.
• ✓WD Ultrastar DC HC5xx: match DCM field, country code, microcode revision, and HSA generation. HC520, HC530, HC550, and HC560 are not interchangeable at the HSA level even when capacities overlap.
• ✓Toshiba MG08: match HDD code, head count, and firmware revision. MG08 helium platforms utilize TDMR heads and tighter track pitches than the previous MG07 helium line; donor stacks from MG07 chassis cannot be transplanted into MG08 drives. The Toshiba MG08 failure profile documents the HSA and firmware-side failure patterns.

Cross-Generation Donor Matching: Exos, Ultrastar, and MG Helium Families

Helium platform generations are not backward-compatible at the head stack level even inside the same product line. Areal density, track pitch, and TDMR head geometry change between generations, which means the donor must come from the same generation as the patient drive. The matrix below summarizes the boundaries we work within when sourcing donors at our Austin lab.

Capacity alone is not a matching criterion. A 16TB Exos X16 and a 16TB Exos X18 share a marketing label but use different head stacks, different servo formats, and different adaptive parameters in the Service Area. The HSA from one will not address heads correctly when wired into the other. PC-3000 reads the patient's firmware family, DCM, and microcode revision through the diagnostic terminal before any donor is opened for transplant.

Why Air-Filled Donor Heads Fail in Helium Chassis

A common failed-recovery pattern we see from outside labs is an attempt to substitute an air-filled HSA into a helium chassis when a same-family helium donor cannot be sourced. The slider air-bearing surface (ABS) on an air-filled head is etched for the aerodynamic lift profile of atmospheric air at 28.96 g/mol average density. Helium sits at roughly 4 g/mol; lift force at the slider drops by about a factor of seven. An air-filled head transplanted into a helium chassis flies too low, contacts the platter surface within seconds of spinup, and strips the magnetic substrate.

The reverse case is also unrecoverable: a helium-tuned HSA placed into an air-filled chassis flies too high to read the servo bursts, the read channel reports continuous loss-of-sync, and the drive enters a busy state without imaging a single sector. The slider geometry is fixed at the wafer-fab step and cannot be re-calibrated in the field. Helium and air HSAs are physically interchangeable mechanically but aerodynamically incompatible. Donor matching for helium drives starts from a sealed helium platform, never from a same-capacity air-filled donor.

Helium Loss Diagnosis: SMART Attribute 22 and Hermetic Seal Integrity

SMART attribute 22 (Current Helium Level) reports the residual helium concentration inside the sealed chamber as a normalized value. The factory baseline is 100. The manufacturer-defined fail threshold sits at 25 on most enterprise helium platforms; a value at or below the threshold indicates the drive can no longer maintain calibrated fly height and is at risk of head-platter contact. The raw value field on some platforms exposes the underlying pressure or concentration measurement; on others the raw field is reserved.

We read attribute 22 through PC-3000 SMART access rather than through the host operating system because a helium drive that has lost too much gas often refuses to complete IDENTIFY DEVICE through the SATA controller. Reading SMART through PC-3000 bypasses the host-side timeout and surfaces the attribute value even when the drive will not enumerate normally. A drive reporting a normalized value above the manufacturer threshold but trending downward is imaged immediately under our multi-pass strategy; once the value reaches or falls below the threshold, the drive is opened on the clean bench, the heads are inspected for surface damage, and the chamber is refilled before any further imaging is attempted.

Helium Refill and Reseal Verification

Helium head swaps are performed on our 0.02 micron ULPA-filtered clean bench. After the failed head stack assembly is replaced, the chamber is refilled with helium before the platters are spun for imaging. The refill step matters because the head slider is calibrated for helium density, not atmospheric air.

After refill, the lid is resealed for the imaging window and the drive is connected to PC-3000 Express or DeepSpar Disk Imager. We image by head map, skip damaged zones on the first pass, and return to weak sectors only after the stable heads have been copied. The donor heads, helium, and reseal materials are all consumables on this tier; pricing for head swap is $3,000–$4,500. Helium cost: $400-$800 additional for head swap and surface damage tiers. This covers the helium refill required after opening the sealed chamber. Helium donor drives must be an exact match. Typical donor cost: $200–$600 depending on model and availability, plus helium refill cost ($400–$800) required after opening the sealed chamber. A +$100 rush fee to move to the front of the queue is available when the customer needs priority on the queue.

PC-3000 Portable III Service Area Workflow on Helium Firmware

Helium firmware repair runs on PC-3000 Portable III or PC-3000 Express with the vendor module loaded for the target family. The Service Area on a sealed helium drive is read through the PCB and the diagnostic terminal; the chamber stays sealed for any firmware-tier recovery. Pricing for this tier is $900–$1,200 because no donor and no helium refill are required.

1. LDR microcode load. When the resident firmware fails to boot the drive into Ready state, PC-3000 issues a vendor-specific loader command (LDR on Seagate, equivalent kernel-mode entry on WD/HGST and Toshiba) that pushes a known-good microcode image into RAM and brings the drive up in safe mode. From safe mode the SA modules can be read, edited, and rewritten.
2. Translator rebuild on Exos. Sudden power loss commonly corrupts the Exos translator and defect tables. PC-3000 reads the surviving SA modules (SysFile 28 static translator, SysFile 35 G-list, SysFile 1B P-list), regenerates the translator against the current defect map, and writes the corrected modules back to the Service Area. The drive then reports its actual capacity through the standard ATA interface.
3. WD/HGST module handling. WD and HGST helium firmware is module-based; PC-3000 reads each module, validates CRC, and replaces damaged modules from a same-family ROM image. Common repair targets include the translator, the SMART log, the defect lists, and the adaptive parameter modules. The hermetic seal stays intact for this entire workflow.
4. ROM extraction and transplant. When the on-PCB ROM has lost its adaptive parameters, the chip is desoldered, read on a programmer, edited to inject the original drive's adaptives, and reflashed. The PCB is then reinstalled on the original sealed chamber.

SMART Differential Diagnosis: Helium Loss Event vs Head and Preamp Failure Event

Attribute 22 in isolation tells us whether the hermetic seal is compromised. Read together with attributes 1, 5, 187, 188, 197, 198, and 240, it separates a slow helium leak from an acute head stack or preamplifier failure before the chassis is ever opened. The table below contrasts the canonical telemetry signatures we evaluate on Seagate Exos, WD Ultrastar HC5xx, and Toshiba MG08/MG09 patients during pre-bench triage.

The vendor sensor architecture behind attribute 22 differs between families and shapes how we read the raw field. HGST and WD Ultrastar helium platforms calculate concentration indirectly by measuring the thermal conductivity of the internal gas through a heated thermistor. When the resistance needed to hold the thermistor at its target temperature rises beyond the factory baseline, the firmware infers that denser atmospheric air has entered the chamber. Seagate Exos drives use a discrete MEMS sensor that reports temperature, pressure, and relative humidity directly, and on Exos platforms the F3 terminal exposes the underlying partial-pressure value (for example, 684 mBar) alongside the normalized SMART attribute. Toshiba MG08 and MG09 expose this telemetry through SMART Attribute 23 (Helium Condition Lower) rather than Attribute 22, and reserve the raw field; the underlying measurement is read through the vendor diagnostic interface rather than through SATA.

A drive that shows attribute 22 holding at 100 with attribute 188 maximized and attribute 1 pinned tells us the chassis can stay sealed: the failure is electrical and the workflow routes to PCB triage and Service Area firmware repair rather than to the ULPA clean bench. A drive showing attribute 22 trending downward while attributes 5 and 197 climb together tells us the seal is failing and the platters are accumulating real damage; that drive is queued for immediate multi-pass imaging followed by reseal and refill if the heads survive the first capture window.

Multi-Pass Imaging Strategy for High-Density Helium Media

Helium drives run at tighter track pitch and higher areal density than air-filled drives in the same form factor. The read channel uses Partial Response Maximum Likelihood (PRML) or Extended PRML detection. When original heads degrade or a donor head is installed, the internal channel calibration no longer matches the new head signature, and the on-drive Viterbi detector reports elevated bit error rates during imaging.

On PC-3000 and DeepSpar Disk Imager, we work the levers the tools actually expose: head-selective imaging so a weak head never blocks a strong one, seek-from-far to settle the servo before each read, per-sector read timeouts, soft-resets between passes, and off-track read attempts at micro-jog offsets above and below the nominal track center. A bad sector that returns CRC errors on the first pass is skipped and re-queued for a later pass under different conditions rather than hammered in place.

The same multi-pass strategy is applied to every drive imaged through our lab. On standard hard drive data recovery cases the BER margins are wider, so fewer passes are required. On helium platforms with high areal density, staged per-head imaging with off-track retries is what separates a successful image from a stalled one.

Post-Head-Swap Helium Refill and PC-3000 Initialization Procedure

The sections above cover donor matching, helium purity, and SMART-based loss diagnosis. The procedure below documents what happens at our Austin lab between the moment the donor head stack assembly is seated and the moment the drive is handed off to the DeepSpar Disk Imager for sector capture. Skipping any step of this chain on a high-capacity Seagate Exos, WD Ultrastar HC5xx, or Toshiba MG08/MG09 patient produces a stable-looking spinup followed by cascading read failures inside the first hour of imaging.

Step 1: Head Stack Transplant on the 0.02 Micron ULPA Clean Bench

The hermetic seal on a Seagate Exos, WD Ultrastar HC5xx, or Toshiba MG08 chassis is a continuous laser weld around the perimeter of the top cover. We score the cover roughly 2 millimeters inboard of the factory weld using a precision straight blade, then lift it free. CNC milling of the seam is rejected as a procedure even though it is faster; milling generates aluminum and stainless-steel particulates that bypass the recirculation filter inside the head-disk assembly and settle onto the platter surfaces. Once those particles are inside the chamber, every subsequent read is at risk of head-particle contact, and the case stops being a standard hard drive data recovery job and becomes a platter cleaning job.

The donor head stack is seated using the matching criteria documented in the donor-matching tables above (model, site code, firmware family, preamp revision, microcode generation). During this window the internal chamber is fully exposed to atmospheric air. The slider air-bearing surface on the donor head was etched at the wafer-fab step for the aerodynamic lift profile of helium at roughly 4 g/mol average molecular weight. Air sits at roughly 28.96 g/mol; if the spindle is energized in this state, lift force at the slider increases, the head pitches outside its calibrated fly-height window, and contact with the platter surface occurs within seconds. Every step that follows exists to prevent that spinup condition.

Step 2: Hermetic Reseal Pathway on Welded-Lid Drives

The original laser weld cannot be reproduced inside a recovery lab. Reproducing it would require focusing weld energy onto the chassis after the platter stack is reinstalled, which introduces thermal stress on the head stack assembly and generates metallic particulates that fall directly onto the platters. No recovery lab attempts this on a patient drive. The pathway we use is mechanical bonding of the cover with a high-viscosity lab-grade adhesive applied continuously along the cover-to-chassis interface. The adhesive is selected for low outgassing because outgassing compounds condense onto the platter surfaces during the multi-day imaging window and degrade subsequent reads.

The perimeter bond cures before the helium injection step in Step 3 begins, leaving the chamber as a closed volume with the factory breather port as the single inlet. The reseal does not need to match the lifetime specification of the original laser weld; it needs to hold the helium atmosphere stable through the imaging window, which on a 9-platter helium drive with eighteen heads can run several days to several weeks of staged per-head imaging. The head-swap pricing tier of $3,000–$4,500 reflects the consumable cost of the adhesive, helium, donor stack, and the bench time required to execute the purge and reseal under ULPA-filtered conditions. Helium cost: $400-$800 additional for head swap and surface damage tiers. This covers the helium refill required after opening the sealed chamber.

Step 3: Atmospheric Purge and Helium Refill

Passive backfilling of helium into an open chamber does not produce a clean helium atmosphere. Helium molecules are small and low-density; nitrogen, oxygen, and ambient water vapor inside the chamber are heavier and settle below the helium layer rather than being displaced by it. Powering a drive after a passive backfill leaves a mixed-gas atmosphere with a density gradient across the platter stack, and the air-bearing equilibrium under the slider oscillates with rotational position. The heads drift off-track during seek operations and the read channel reports continuous loss-of-sync.

The procedure used at our Austin lab is a controlled positive-pressure helium purge through the factory breather port, not a passive backfill. With the perimeter bond from Step 2 cured, the chamber forms a closed volume with the breather port as its single inlet. The breather port is the original factory fill port located on the chassis wall and sealed at manufacture with a secondary weld or polymer plug. We open this port, attach the helium supply line, and run a continuous flow that displaces the heavier residual gases out through the same opening before pressure is allowed to equilibrate. Grade 5.0 helium (99.999 percent purity) is used throughout; trace nitrogen or moisture from a lower-grade fill raises the composite gas density above the 4 g/mol design target and produces the contamination signature documented later in Step 4. Pressure inside the chamber is equilibrated against the OEM target for the family before the inlet is closed with an impermeable adhesive plug.

Step 4: PC-3000 Initialization Before DeepSpar Imaging Begins

A resealed and refilled drive is not connected to a host operating system directly. The donor heads are not factory-calibrated for the patient platters, so direct ATA enumeration either fails outright or, worse, triggers internal defect remapping against the wrong calibration set. The drive is brought up on PC-3000 Portable III with the external SAS adapter for SAS-interface enterprise drives, or PC-3000 Express for SATA-interface variants, and the four-stage initialization below is executed before control is handed to DeepSpar Disk Imager.

1. SA module read test. The drive is energized through the PC-3000 intelligent power supply and held in safe mode. PC-3000 reads the Service Area modules located on the reserved negative-cylinder tracks. A successful read confirms that the donor heads are electrically compatible with the patient's preamp and controller circuit. A failed read indicates a preamp-channel or head-map mismatch that must be corrected before any further step is attempted.
2. Per-head stability evaluation. PC-3000 commands the drive to read sample sectors across the LBA range for each physical head position. Error rate is logged per head. Any head that produces unrecoverable reads at a rate inconsistent with the rest of the stack is flagged in a RAM head map and disabled for the initial imaging pass; data behind that head is captured later under different read conditions rather than blocked from imaging entirely.
3. Adaptive parameter transfer per family. The donor's factory adaptives are wrong for the patient platters; the patient's original adaptives must be re-injected before read-channel lock is achieved. WD Ultrastar HC5xx drives are built on the HGST Command Code Based (CCB) architecture inherited from the HGST acquisition, not the Marvell-platform module taxonomy used on WD consumer drives. The CCB adaptives (per-head microjog offsets, read-channel gain profiles, head map, and servo calibration values) are accessed through the PC-3000 HGST CCB utility and the dedicated CCB hardware adapter; the patient's adaptives are read out before the donor stack is seated and written back to the Service Area once the donor heads are mechanically installed. On Seagate Exos drives, the equivalent values live in SysFile 4 (SAP servo adaptives), SysFile 6 (RAP read-channel amplifier, CTAF shaping, and FIR coefficients), and SysFile 7 (CAP); these are re-read or re-written through the F3 serial UART terminal after a Tech Mode unlock patch is loaded into RAM. On Toshiba MG07, MG08, and MG09 drives, the Configuration Pages carrying the adaptive parameter block are re-injected through the PC-3000 Toshiba module; if the internal translator is unstable, a virtual translator is constructed in PC-3000 host memory so the controller is not forced to walk its own G-list during imaging.
4. Read channel calibration and contamination check. Once adaptives are in place, a short read pass is run with the on-drive channel telemetry exposed. SNR at the Continuous Time Analog Filter and raw bit error rate are validated against the LDPC or Reed-Solomon correction headroom for the family. A contaminated helium fill or a partially purged chamber shows up here as cascading LBA read-error interrupts, climbing pending-sector counts, and eventually a busy-state firmware panic; if that pattern appears, the imaging window is aborted before donor heads are wasted and the drive returns to Step 3. On a clean fill with marginal donor adaptives, the PC-3000 system is used to trim per-head VGA boost, FIR equalizer coefficients, and MR bias current in RAM to push raw BER back below the ECC threshold. Only when the channel locks across the active head map does control hand off to the DeepSpar Disk Imager, which then runs the staged per-head, off-track-retry strategy documented in the multi-pass imaging section above for the actual hard drive data recovery capture.

PCB-Level Triage on Exos and Ultrastar Helium Drives

A helium Exos or Ultrastar that arrives dead with zero current draw is rarely a head failure. The dominant failure on enterprise helium PCBs is an electronic fuse latched open on the 5V or 12V rail, usually traceable to a backplane power event or a modular ATX cable mismatch upstream. The triage below covers component-level diagnosis on Seagate Exos and WD Ultrastar boards before any chassis opening. The hermetic seal stays intact throughout this PCB-only workflow.

E-Fuse Failure Signature: NIS5232 and MP5018 on Exos PCBs

Seagate Exos PCB revisions in current circulation (including 100833707 REV B and 100852967 REV C) route incoming power through dedicated electronic fuses rather than relying on transient voltage suppression diodes alone. The 12V rail is gated by an ON Semiconductor NIS5232, a 12V 4.2A electronic fuse in a DFN10 package, marking 232. The 5V rail is gated by a Monolithic Power Systems MP5018, a 5V current limit switch rated for 1A to 5A with reverse current blocking and an output overvoltage clamp, in a QFN12 package, marking AMAy. Both parts are designed to sacrifice themselves when an upstream event pushes voltage or current outside the specified envelope, protecting the downstream motor controller and preamplifier interface.

The diagnostic signature is deterministic. With the PCB on the bench and 5V applied to the input pad, we read voltage at the Vin pin of the MP5018 and at its Vout pin. If Vin sits at 5V and Vout reads 0V, with no short to ground anywhere on the downstream 5V network, the MP5018 has latched open and the rail is dead at the fuse itself. The equivalent check on the 12V rail uses the NIS5232: 12V at Vin, 0V at Vout, no downstream short. A multimeter continuity check, a FLIR thermal scan during a brief power-up to confirm there is no hot spot indicating a downstream short, and a resistance check from Vout to ground complete the bench triage. When the signature matches, the fuse is the single point of failure and the rest of the PCB and the head-disk assembly inside the sealed chamber are intact.

E-Fuse Bypass vs ROM Adaptive Transplant

The terms e-fuse bypass and ROM adaptive transplant describe two different repairs for two different states of PCB degradation. Competitor write-ups routinely conflate them. We treat them as separate procedures with separate decision criteria.

E-fuse bypass

Board-level microsoldering on the original patient PCB. When the bench triage confirms Vin present and Vout dead at the e-fuse, with the downstream network clean, the damaged fuse IC is removed using hot-air rework (Atten 862 or equivalent) and the Vin and Vout pads are bridged with a short solder jumper or fine micro-wire under a Hakko FM-2032 on an FM-203 base station. The original PCB, ROM, and motor controller all stay on the drive. No donor board is needed. The fuse is removed from the protection path; the operator now assumes responsibility for clean power input upstream during imaging.

ROM adaptive transplant

A separate procedure for a different failure state. When the overvoltage event walked past the e-fuse and destroyed the main controller or motor driver downstream, the patient PCB cannot be restored at the component level. The repair pathway requires an identical donor PCB; the 8-pin SPI flash ROM is desoldered from the patient board and soldered onto the donor board so the donor controller boots against the patient drive's unique adaptive parameters. A direct donor PCB swap without the ROM transplant fails for the reasons described in the next subsection.

Choosing between the two procedures comes down to where the energy went. If the e-fuse caught the event and latched open with the downstream network intact, the bypass is the correct repair and the donor PCB stays on the shelf. If the e-fuse was overwhelmed or shorted through and the downstream silicon was damaged, the ROM transplant is the only path, because a generic donor PCB cannot drive the patient's head stack at calibrated fly height.

Why a Direct PCB Swap Fails on Enterprise Helium Drives

Modern enterprise helium drives operate at an active magnetic spacing of 1 to 2 nanometers above the platter surface during reads and writes, with Thermal Fly-height Control trimming the slider clearance from a baseline aerodynamic fly height of a few nanometers. To hold that gap across temperature, altitude, and per-head variation, every drive is calibrated at the factory and the calibration data is written to the SPI flash ROM on its PCB. The ROM is keyed to the individual head stack assembly inside the sealed chamber. A donor PCB ships from the donor drive with the donor drive's calibration, not the patient's. The calibration that matters for a clean spin-up and a successful image includes:

• Micro-jog offsets. The physical separation between the read element and the write element on a single head varies head to head because of photolithography tolerances at the wafer-fab step. Micro-jog offsets are the per-head corrections the servo applies so the read element sits over the track centerline when the servo locks to a write-aligned reference. On WD and HGST drives, the offsets live in the ROM-resident shadow of Module 47 alongside the Module 0A head map; the master copy of Module 47 sits in the Service Area on the platters. On Seagate Exos, the equivalent corrections live in the ROM-resident RAP, SAP, and CAP adaptive blocks (Read, Servo, and Controller Adaptive Parameters). Toshiba MG08 and MG09 store their per-head corrections in Control Program overlays whose ROM-resident portion the controller loads at spin-up.
• Thermal Fly-height Control voltage curves.Each slider carries an embedded heater. The voltage applied to that heater thermally protrudes the read and write elements toward the platter and trims the fly height in flight. The voltage-to-protrusion curve is unique to each head and is stored on the patient's ROM. A donor PCB running its own TFC curves against the patient's heads either flies the elements too high (no read) or pushes them into the platter (head crash and surface damage).
• Preamplifier gain and channel tuning.VGA boost, MR bias current, FIR equalizer coefficients, and Continuous Time Analog Filter shaping are calibrated per head at the factory. The donor PCB's tuning was written for the donor heads. Mismatched tuning leaves the read channel unable to lock even when the heads physically fly correctly.

Powering a helium Exos or Ultrastar with an unmodified donor PCB drives the patient head stack with the donor's adaptives. The servo cannot lock, the drive clicks or spins down, and on a worst-case TFC mismatch the heads contact the platter surface and score tracks the customer was paying us to recover. A direct PCB swap is not a faster shortcut; it is the procedure that converts a firmware-tier repair into a head-swap and platter- cleaning case. The ROM transplant exists because the alternative writes off the data.

SMART Telemetry Differential Diagnosis Across Helium Drive Families

A single SMART attribute ID does not mean the same thing on a Seagate Exos as it does on a Western Digital Ultrastar HC5xx or a Toshiba MG08 and MG09. Each vendor implements the underlying sensor, threshold, and reporting cadence differently, and helium-specific attributes layer on top of that variance. Reading HDD recovery telemetry without normalising for vendor and family is the fastest way to misdiagnose a slow helium leak as a preamplifier failure, or the reverse. The table below maps the attributes we read on every helium intake against per-family behaviour, so the technician on our bench picks the correct PC-3000 module and the correct read path the first time.

Helium Loss Event Signature

A slow leak and an acute preamp failure both produce read failures, but their telemetry fingerprints diverge. Slow-leak progression looks like this across the three families:

1. Seagate Exos: attribute 22 drops in single-digit steps over weeks; attributes 1 and 188 remain quiet; attribute 197 begins to climb only after density has crossed the fly-height tolerance band. The MEMS sensor gives the earliest warning and is the first attribute to move. See the Seagate Exos 10TB failure page for the specific signature we see on intake.
2. WD Ultrastar HC5xx: attribute 22 stays flat until the thermistor-derived density estimate crosses an internal threshold, then drops in a single large step. The first symptom on telemetry is usually a rising attribute 1 caused by changes in fly height before 22 reports the leak.
3. Toshiba MG08 / MG09: attribute 22 reports in discrete steps with long quiet periods in between; the first warning is a small non-zero 187 combined with a normalised 1 that drifts down. The Toshiba MG08 failure page documents the leak-versus-head decision tree for this family.

Preamplifier Failure Signature

Acute preamp or HSA failure produces a different shape:

1. Seagate Exos: attribute 22 stays at factory fill, attribute 188 rises in a single bench session, and attribute 197 jumps from zero into the tens as one head loses the ability to lock. Identifying the failing head index requires querying internal error logs through the F3 SATA terminal.
2. WD Ultrastar HC5xx: 187 and 197 increment together in a tight correlation; 188 follows as the preamp's head-select line fails to respond. Attribute 22 is unchanged. Parsing the service area defect logs via the PC-3000 HGST CCB utility attributes the failure to a specific head.
3. Toshiba MG08 / MG09: 188 climbs in sustained increments while 22 holds; 197 follows once internal retries are exhausted. Because Toshiba does not expose per-head attribution through standard SMART, head isolation requires the PC-3000 Toshiba module rather than telemetry alone.

Where to Read These Attributes Per Family

Standard SMART tooling does not surface the vendor-specific raw fields these decisions depend on. On Seagate Exos we read telemetry through the F3 SATA terminal in diagnostic mode, which exposes the raw MEMS environmental sensor reading behind attribute 22 and allows direct querying of the drive's internal error logs to isolate the failing head. On WD Ultrastar HC5xx we read through the PC-3000 HGST CCB utility, parsing the service area defect logs to obtain per-head G-list attribution for attribute 5 and internal flying-hours data referenced by attribute 240. On Toshiba MG08 and MG09 we read through the PC-3000 Toshiba module, which provides direct access to the service area and translator tables necessary for diagnosing advanced degradation. Telemetry interpretation is the first read after intake; it determines which subsequent procedure runs on the bench.

Deep-Stack Platter Physics on 10+ Platter Helium Drives After Lid Breach

Current-generation helium enterprise drives in the WD Ultrastar HC570 and HC580 families, the Seagate Exos X22 and X24 families, and the Toshiba MG10 family carry 10 platters in the same 3.5″ chassis. When a customer drive arrives with a damaged hermetic seal or a lid that has already been pried by another lab, the chamber is no longer at the original helium partial pressure. The drive is now operating in a mixed atmosphere, and the physics of fly height, head azimuth, and read-channel signal-to-noise differ for the top platter versus the deepest platter in the stack. Imaging order on the bench matters because the deeper layers degrade fastest once the helium atmosphere is lost.

Why Deeper Platters Degrade Fastest in Atmospheric Air

Each platter and head pair sits inside a thin air-bearing layer trapped between the slider and the magnetic substrate. In a sealed helium chamber at 99.999% purity, that air-bearing layer behaves uniformly from the top of the stack to the bottom because the gas density (approximately 0.166 kg per cubic metre at room temperature) is the same at every platter. Once atmospheric air replaces the helium, two effects stack with platter depth. Air density is roughly seven times higher than helium density, which raises the aerodynamic load on every slider and shifts the operating fly height downward. The deeper platters additionally sit closer to the spindle motor and farther from the top vent, so any residual turbulence from spindle acceleration, head-stack motion, and chassis vibration couples through the entire stack and is largest near the bottom. The result is that the bottom platter in the stack runs with a tighter air-bearing margin and a lower read channel signal-to-noise ratio than the top platter, before any media damage has occurred.

HSA Azimuth Tolerance Versus Platter Depth

The head stack assembly is a rigid comb of suspension arms machined for a specific chamber gas density. Each suspension carries a slight pre-load that places the slider at its design fly height under the air-bearing pressure the chamber was sealed to produce. When the chamber is in air, the fly height drops, and the slider sits at a steeper effective azimuth relative to the track because the suspension pre-load now pushes through a denser gas. The change is small at the top arm and compounds toward the bottom arm because the bottom arm carries the most cumulative load through the actuator pivot. On a 10-platter HSA out of an HC570 or Exos X22, the head on the bottom platter loses its track-following margin first, and microjog offsets calibrated against a helium chamber no longer center the read element over the track. The PC-3000 utility is used to read the per-head microjog offsets from the Service Area and confirm which head is closest to the track-following edge before the bench is staged.

Per-Platter Read-Channel SNR Drift

The read channel inside an enterprise helium drive uses an LDPC inner code concatenated with a separate outer code for iterative decoding; the channel can tolerate a few decibels of SNR loss before uncorrectables begin to climb. In a breached chamber, the SNR margin on the bottom platter is consumed first because the deeper air-bearing layer is the most turbulent and the head sits closest to the platter surface. The middle of the stack drifts next, and the top platter holds margin longest. The practical consequence on the bench is that any imaging attempt in air pulls clean sectors off the top platter while uncorrectables climb on the bottom platter; reading order has to account for that asymmetry rather than treating all heads as equivalent.

Bench Staging Order for Imaging in a Breached Chamber

When a customer drive arrives breached and we have decided to attempt a brief diagnostic read before transferring to a refilled donor chassis, the imaging order is bottom-up. PC-3000 head selection is used to disable every head except the one over the deepest platter, an imaging pass is run against that head only, and progress is monitored against the per-head error counters. If the bottom platter still returns clean blocks at the deepest point, the next head up is added; if the bottom platter starts throwing uncorrectables, the head is disabled and the pass continues against the next platter up. This ordering captures the most fragile surface first while it still has SNR margin and prevents the situation where every head except the bottom completes a full pass, the drive is then stalled on a single failing head, and the bottom platter has already lost margin by the time it is its turn. DeepSpar Disk Imager runs the per-head session against the head map exposed by PC-3000 after Service Area access is established.

PC-3000 HGST Module Versus PC-3000 WD Module on HC5xx and HC6xx

The Ultrastar HC5xx and HC6xx families inherit firmware lineage from the HGST helium platform. After the merger the families continued under the WD brand, but the on-disc firmware architecture, the Service Area structure, and the per-head adaptive parameter format remained on the HGST side rather than the WD consumer Marvell-controller side. On the bench, PC-3000 exposes two distinct modules for these drives, and selecting the correct one is the first decision after intake. The PC-3000 HGST module uses the CCB (Command Code Based) adapter to read per-head defect logs, per-head microjog tables, and the internal flying-hours register. The PC-3000 WD module is built around the WD-specific T2 translator and Marvell controller commands used on consumer WD Blue, Red, and Black drives, and it does not surface the CCB structures that the HC5xx and HC6xx firmware depend on. Reading an HC560 or HC570 through the WD module exposes only the standard ATA SMART page; reading the same drive through the HGST CCB module surfaces the per-head G-list, the per-head defect attribution, and the partial-pressure raw value behind attribute 22. The HGST module is the correct choice for every drive in this lineage, including HC560, HC570, and HC580 examples that arrived after the brand transition and carry WD labels on the chassis.

Partial-Pressure Telemetry That Standard SMART Hides

Standard host-side smartctl output reports attribute 22 only as a normalized percentage of factory fill. The PC-3000 HGST CCB module reads the underlying partial-pressure register that the firmware uses to compute that normalization. On an HC560 or HC570 the register reports the residual helium partial pressure in mBar and the thermistor-derived density estimate in counts; comparing the two against the factory calibration values stored in the Service Area is how a slow leak is distinguished from a thermistor that has drifted out of tolerance. A drive that reports attribute 22 at 100 in standard SMART can show a partial-pressure register that has drifted 40 mBar below the factory point, which is enough margin loss to explain rising attribute 1 even though the host-visible SMART looks healthy. The same register is read after a head swap and helium refill to confirm that the chamber held the refilled gas at the correct partial pressure before the imaging pass begins.

False Positive: Low Helium Reading With Healthy Chamber

The inverse situation appears on intakes that have spent extended time on a shelf at elevated temperature. The thermistor behind attribute 22 measures thermal conductivity of the chamber gas; if the thermistor itself has drifted, the firmware computes a lower helium density than the chamber actually holds. The host-side SMART then reports attribute 22 dropping into the warning band even though the chamber seal is intact and the heads are flying at their calibrated height. The bench check is to read the partial-pressure register through the HGST CCB module; if the partial pressure is at the factory value while the normalized SMART attribute is low, the failure is in the telemetry path rather than in the chamber, and the recovery plan does not require opening the lid. This is the case where opening the chamber would convert a no-action diagnostic into a full head-swap, refill, and platter cleaning job for no reason.

Refill Verification After a Head Swap on a Deep Stack

After a head swap on a 10-platter helium drive, the helium refill is checked against the same partial-pressure register that diagnoses leaks on intake. The refill target is the factory calibration value stored in the Service Area, not a generic helium density figure. The chamber is sealed, the drive is allowed to thermally equilibrate on the bench, and the partial-pressure register is read through the HGST CCB module before the imaging pass begins. If the register reads below the factory value, the seal is re-inspected and the chamber is refilled before the heads are loaded onto the platters for the first time. The same procedure runs on Seagate Exos through the F3 serial terminal and on Toshiba MG drives through the PC-3000 Toshiba module. The imaging session opens only after the partial-pressure register confirms the chamber is at the correct operating point for the head stack that was just installed.

Helium Drive Recovery FAQ