辐照灭菌高压电源在医疗器械与生物样本处理中的剂量
Sterilisation by radiation is a dose, not a voltage. The dose is the product of the beam's energy deposition and the dwell time, and the supply that accelerates the beam sits at the origin of both terms.
A dose of twenty-five kilogray is the common target for medical devices, and the regulatory window around it is narrow, typically from fifteen to thirty kilogray depending on the product and the validation. A supply that drifts by five percent moves the delivered dose by five percent, and the validation no longer covers the process.
The beam is produced by electrons accelerated through a few hundred kilovolts to a few megavolts, and the supply's contribution is the accelerating voltage's stability. Electron energy determines penetration, and penetration determines whether the dose reaches the interior of a packaged device.
Energy below the design value produces a surface dose that is high and a centre dose that is low, and the product fails the dose audit even though the surface would pass. The failure is invisible without dosimeters at depth, which is why the validation uses them.
The supply's second contribution is beam current, which sets the dose rate. Dose is the integral of rate over time, so a current that is constant to one percent yields a dose that is constant to one percent if the conveyor is constant to the same degree.
Conveyor speed is the time term, and the conveyor's accuracy is often the limiting factor rather than the supply. A conveyor that varies by three percent produces a dose that varies by three percent regardless of the supply, and it is worth measuring before blaming the high-voltage side.
Transverse uniformity is the third term. The beam is scanned across the product, and the scan's uniformity combined with the supply's stability sets the dose uniformity in the direction perpendicular to the motion.
Scan magnet current is usually supplied by a separate unit, and the two supplies interact through the common earth. A scan supply that shares an earth with the accelerator supply will modulate the beam's position when it switches, and the modulation appears as a dose ripple along the product.
X-ray conversion adds a case. A high-energy electron beam can be converted to X-rays by a target, and the X-ray dose depends on the electron energy to a higher power than the electron dose does, so a one percent energy change becomes a two or three percent dose change.
For X-ray sterilisation the energy stability requirement is therefore tighter than for direct electron irradiation, and the supply's specification must reflect the conversion rather than the direct case.
Shielding does not affect the dose directly but affects the machine's design. The thicker the shield, the more the machine is constrained in layout, and a compact layout puts the supply close to the accelerator, where the supply's electronics see a radiation field that degrades semiconductors over years.
Radiation-tolerant design in the supply is a matter of placement and of margin. Placing the control electronics outside the shield, with only the high-voltage output penetrating it, removes the problem; where that is impossible, shielding the electronics locally and derating the semiconductors extends life.
A device sterilisation case shows the accumulation. A plant delivering twenty-five kilogray to packaged sutures found that a subset of packages failed the dose audit at the centre. The centre dose depended on energy, and the energy depended on a divider that had drifted by four percent over two years of operation.
The divider was recalibrated, the energy restored, and the centre dose returned to the validated value. The drift had been invisible because the top-of-chain meter was also reading through the same divider, so the supply believed it was correct.
A second case involved the scan supply. Products showed a periodic dose ripple with a wavelength of ten centimetres along the conveyor, and the period matched the scan supply's switching frequency divided by the conveyor speed. Synchronising the scan supply to the accelerator's trigger removed the ripple.
A third case involved a beam that lost current gradually over a shift. The cause was the cathode's emission falling as it aged, and the supply was in voltage mode so the current fell with it. A current-regulated mode, or a dose-rate feedback loop, held the dose until the cathode could be replaced.
Dose validation requires records that a supply can support. Voltage, current, conveyor speed and dosimeter readings should be logged together, so that a future audit can reconstruct the process from the data rather than from memory.
Maintenance intervals follow the validation. The divider is checked twice a year, the current sensor quarterly, and the complete dose audit annually with dosimeters at three depths.
Spares are modest: one divider, one current sensor, one trigger interface. The accelerator's tube is the item that dominates the spares budget, and the supply's role is to avoid the tube failures that a drifting voltage would cause.
Cost has to be set against the validation. A supply with a few parts in ten thousand stability costs perhaps a third more than a basic unit, and a failed dose audit costs a week of production plus a revalidation, which is far more than the difference.
Limits should be explicit. No supply can compensate for a conveyor whose speed varies, and no stability specification can promise a dose at depth if the product's density varies within a batch. The supply holds energy and current; the rest is mechanics and product.
The design intent is short. Hold the energy so that the penetration is the one that was validated, hold the current so that the rate is the one that was validated, and let the conveyor and the dosimeters carry the rest of the proof.
Ozone is a by-product that the supply indirectly controls. The amount produced depends on the energy deposited in air, and a plant that runs at a higher voltage than necessary produces more ozone and needs more ventilation. Running at the validated energy is therefore also an air-quality measure.
Biological sample treatment differs from device sterilisation in its tolerance. Samples in sealed containers tolerate a wider dose window but a narrower temperature window, and the beam heats the sample. A supply that delivers the same dose in a shorter time raises the temperature, so the dose rate becomes a process constraint rather than only a throughput parameter.
Dose rate control means the supply must be able to run at reduced current without losing stability. A supply whose regulation degrades at ten percent of rated current will produce an uneven dose at the low rate that a heat-sensitive sample requires.
Low-current stability is therefore a specification item, not an afterthought. It is verified by running at ten percent of rating for an hour and measuring both the current and the energy, and a supply that fails that test will fail in the sample line rather than in the device line.
Beam scanning at low current is also different. The scan field's uniformity depends on the scan waveform, and at low current the space charge that spreads the beam is absent, so the beam is narrower and the scan must be adjusted. The supply's energy must remain the same while the scan changes.
Accelerator vacuum is a precondition rather than a variable. The supply should interlock on the vacuum level, because a beam that is accelerated into a poor vacuum will scatter and deposit energy in the accelerating structure, and the structure is not designed for that heat.
The interlock chain deserves a diagram. Vacuum, cooling water, shield door, conveyor, dose monitor, and only then the high voltage. Each of the five conditions should be a separate contact, and the chain should fail to a safe state if any contact opens.
A single combined interlock signal is convenient and dangerous. When the signal opens, the operator cannot tell which condition failed, and the tendency is to bypass the chain rather than to diagnose it.
Cooling water is the condition that most often trips at start-up. The supply's power stage and the accelerator's window both need water, and a flow switch that is set too close to the normal flow will trip on a warm afternoon. A switch with a ten percent margin avoids nuisance trips without removing the protection.
The window is a consumable, and its failure mode is a leak. A window that has lost a fraction of its thickness will pass the beam but will also pass air, and the vacuum interlock will trip. Recording the window's age and replacing it on schedule is cheaper than the downtime of a sudden failure.
Operator interface should show the dose, not only the voltage. An operator who sees the accumulated dose can stop the line before a product is over-irradiated, and an operator who sees only a voltage cannot.
Records that support an audit should be immutable. A log that can be edited is not evidence, and a log that is written to a removable medium weekly is evidence that survives a controller failure.

