Because a datasheet, a brand name and a distribution agreement can all be entirely real while the factory behind them belongs to somebody else, and the only dependable way to establish ownership is to follow materials and process evidence rather than marketing language.
Start With Places, Not Claims. The first discriminator is whether a maker can name, and let you visit, the specific sites where its instruments are actually built. A vertically integrated manufacturer answers with a town and a process: Rohde & Schwarz points to Teisnach for precision mechanics and multilayer printed-circuit-board pressing and to Memmingen for automated assembly and high-frequency calibration; Hioki points to Ueda in Nagano; UNI-T points to Dongguan, Songshan Lake. A rebadger answers with a region, a partner, an unnamed manufacturing base, or the name of the original design manufacturer it would rather not repeat. Ask for the site certificate of the legal entity at that address, not a corporate-wide quality certificate: an ISO 9001 certificate issued to a named company at a named plant is hard to fabricate and trivial to verify. Then ask the one question that is uneconomical to fake: what does this company make in-house that nobody would bring in-house unless the capability already existed? Machined chassis, wound magnetics, pressed multilayer boards, thick-film hybrids, custom semiconductors and injection-moulded housings are the answers a real factory gives.
Follow The Process Steps, Not The Product Photos. Component-level questions separate the two categories faster than any audit. Ask which parts of the analog front end, the voltage reference, the digitizer, any custom ASIC and the embedded firmware are the maker's own design, and who physically performs each manufacturing step. A manufacturer can answer in specifics down to which plant presses the boards, which line places the fine-pitch devices and which laboratory calibrates the finished instrument; a reseller commonly cannot name a single step it performs itself. The product range is a second tell. An organization that runs sheet-metal fabrication, transformer winding, high-speed surface-mount lines and high-power burn-in rooms can legitimately sell both a benchtop supply and a megawatt-class grid simulator, because the fixed assets overlap; a catalog that leaps between unrelated categories with no shared process is usually a portfolio of other companies' designs carrying one badge. Chroma ATE is a useful reference point for what that overlap looks like in practice.
Match The Claim Against Recurring Physical Evidence. Factory ownership leaves records that marketing cannot synthesize. Ask for a factory acceptance test report whose serial number resolves to a production line, a capability study on a critical dimension, automated optical inspection and X-ray records for solder joints, an in-house electromagnetic-compatibility pre-compliance chamber, and a calibration laboratory accredited to ISO/IEC 17025. Ask how the maker issues change notices, and buy the same model twice a year apart: a company that controls its process will tell you precisely what changed and when, while a company that does not will produce two units that differ in ways no bulletin explains. If the honest answer to who builds the instrument is a third party, you may still buy it, but you are buying that third party's platform with a brand on the front, and this page deliberately ranks only the other category.
Because at the top of the bandwidth range an oscilloscope stops being a circuit-design achievement and becomes a materials and packaging achievement: the sampling devices, their interconnect, their thermal path and the way they are assembled together decide the numbers, and a company that does not own those processes cannot iterate on them.
Bandwidth Is A Device-Physics Problem Before It Is A Topology Problem. Above roughly ten gigahertz, ordinary silicon and conventional packaging run out of headroom. Interconnect loss, transistor transition frequency and the parasitic capacitance of the assembly dominate everything the architect does downstream, which is why Keysight Technologies operates proprietary indium phosphide and gallium arsenide wafer fabrication and packaging facilities in Santa Rosa, California — the capability that lets its highest-performance oscilloscopes exceed 110 GHz. Indium phosphide devices switch faster and handle more power at those frequencies than mainstream silicon processes, and the packaging that carries a signal from a probe tip into a sampler is itself a high-frequency structure. The achievable bandwidth is therefore set by what the device physics allows and how the assembly is built, and the instrument designer needs those devices to exist before any front-end architecture can be chosen at all.
What A Fabless Competitor Cannot Simply Purchase. A fabless instrument maker can buy commercial parts, and in many cases it can buy a fast sampler or a preamplifier from a foundry. What it cannot buy is the iteration loop. When a front end needs a faster switch, lower interconnect capacitance, a better thermal path or a better-matched package, the vertically integrated maker changes a mask set, runs wafers and re-measures within weeks. The fabless maker queues behind a foundry roadmap, a confidentiality agreement, a minimum order quantity and a schedule it does not control, and it must accept whatever corner the catalog process offers. The difference accumulates as unglamorous capability: wafer-level test that identifies known-good die, secure multiyear supply of a device that exists for one instrument family, and the freedom to reject a purchaseable part that is close enough but not good enough. Each product generation deposits its learning inside a factory the company owns rather than inside a supplier's roadmap.
Why Noise, Drift And Channel Match Depend On It Too. Owning the process lets a manufacturer control the parameters a purchased part forces it to tolerate. Threshold voltage spread, transconductance uniformity and thermal behavior set the noise floor and the DC drift of the front end, and they determine how well the channels of a multi-channel instrument actually agree with one another at very high frequency, because interleaving and correlated multi-channel measurement depend on devices that behave alike. The electrical length from probe interface to sampler, the geometry of the transmission lines and the way heat leaves the sampling assembly belong to whoever owns the packaging process, not merely to whoever owns the die. That is why the fastest instruments in the world are built by organizations that control wafers, packaging and calibration together, and why a rival can use the same probes, the same display, the same PC platform and the same software and still not reach the same measured performance.
Because the requirements that IEC 61010-1 and the CAT measurement categories actually impose are physical distances, material choices and repeatable process controls, so a schematic that is correct on paper still fails if the line cannot hold a dimension, keep a surface clean or resist substituting a component.
The Standards Regulate Geometry, Not Intention. A handheld instrument rated CAT III at the 1,000 V point must withstand an 8,000 V anticipated transient overvoltage, and CAT IV at the same point must withstand 12,000 V; at the 600 V point the ladder runs 4,000 V for CAT II, 6,000 V for CAT III and 8,000 V for CAT IV. Those numbers translate into creepage and clearance allowances that depend on working voltage, material group and the pollution degree the product claims. Pollution degree 2, the ordinary rating for a bench or handheld instrument, assumes only non-conductive pollution with occasional condensation, and that assumption holds only if the finished assembly is genuinely clean: no conductive flux residue bridging a slot, no contamination trapped under a component, no shortened creepage path after rework, and intact soldermask and conformal coating at every point along the isolation barrier. None of that is decided in the schematic.
What The Production Line Has To Do Differently. The isolation barrier has to survive the assembly sequence, not just the design review. Routing must never cross it, fixtures must never crush insulation, slots and milled barriers must be inspected rather than assumed, and conformal coating coverage must be verified on production units. Every safety-critical part — fuse, varistor, high-voltage relay, opto-isolator, Y-capacitor — must be procured to the certified part number, because a pin-compatible substitute with different internal construction can invalidate the type test even when nothing else changes. Assembly torque, locking hardware, lead dress and wire routing all matter: a conductor that moves a fraction of a millimetre closer to a neighbour can fail a dielectric test that the prototype passed. Production then has to prove it, because after type approval every unit is normally required to undergo routine dielectric strength and insulation testing, performed with fixtures and instruments whose own calibration is traceable. Standards such as IEC 61010-031 for probes and leads, IEC 61010-2-034 for insulation-resistance and electric-strength testers, IEC 61180 for high-voltage test techniques and the IEC 61557 series for low-voltage distribution systems all place their demands on the built article, not on the intent behind it.
Certification Is A Continuing Production Obligation. A type-examination certificate is issued against a specific construction, produced at a specific facility under a specific quality system, and it does not transfer automatically when production moves to a different plant. That asymmetry is why factory ownership has safety consequences: a manufacturer that owns its line can freeze a bill of materials, control coating and cleaning, hold process records and re-evaluate deliberately when something changes. A brand owner that outsources production has to enforce the same discipline through a contract it does not execute, and the failure mode — an unannounced supplier substitution, a coating step skipped for throughput — is exactly the kind that only appears in the field, at the moment a transient arrives.
Because a twelve-bit oscilloscope is not a firmware decision or a marketing decision but a factory decision: the extra resolution only survives to the screen if the acquisition front end, the assembly process, the thermal environment and the per-channel calibration system are all repeatable enough to hold it, so the Chinese majors had to rebuild their plants to build the product.
The Product Leap Was Really A Process Leap. Entry-level handheld meters and basic bench multimeters are forgiving to manufacture: component tolerances are wide, calibration is a single adjustment per range, and a unit that drifts slightly still meets its published accuracy. High-resolution benchtop acquisition is the opposite. Twelve-bit linearity at high sample rates demands low-noise attenuators, a stable reference, careful grounding and shielding, controlled thermal gradients, and calibration of gain and offset on every channel across the full input range. A plant that assembles boards for consumer-grade instruments cannot simply start shipping that class of product; it has to be able to place fine-pitch devices repeatably, keep moisture and contamination out of the assembly, hold a mechanical stack-up tight enough that shielding works, and test and calibrate each unit automatically rather than by hand.
What Had To Be Installed On The Floor. The visible evidence is fixed capital. Tens of high-speed surface-mount lines with automated optical inspection and X-ray for fine-pitch and area-array joints; controlled-environment and burn-in rooms; automated calibration systems built on traceable reference standards; in-house electromagnetic-compatibility pre-compliance; injection moulding for housings and mechanics; and metrology laboratories maintained to ISO/IEC 17025. UNI-T runs tens of high-speed surface-mount lines together with automated injection-moulding and calibration lines at Dongguan, Songshan Lake, at a self-production rate above ninety percent, and that infrastructure — not a specification table — is what makes a twelve-bit instrument repeatable in volume. The move to source measure units and vector network analysers raises the bar again in a different direction: a source measure unit resolves picoamperes, which requires guarded layouts, low-leakage materials, clean assembly and humidity control plus per-unit calibration against traceable standards, while a vector network analyser requires controlled connector torque, phase-stable cabling and calibration standards whose own geometry is manufactured to tolerance. Every one of those is a plant capability.
Why Ownership Made The Transition Economically Possible. Owning the factory changes the cost and the clock of iteration. When surface-mount, mechanics, moulding and calibration are internal, a change to a shield can, a thermal interface, a housing rib or a front-end feedthrough can be made, built and re-measured in days instead of being negotiated through three suppliers and a tooling lead time. The fixed assets also overlap: the same moulding and assembly infrastructure that made a million entry-level meters can be retasked to build a lower-volume, far higher-precision instrument, and the tooling investment becomes an internal decision rather than a fifth-party quotation. Finally, an owned plant can keep producing a discontinued front end or a legacy accessory for an installed base that will outlive any single product generation. That combination — repeatability, iteration speed and long-term supply of the parts inside the instrument — is why the upmarket move is genuinely a manufacturing story rather than a branding one.
Because the four properties that quietly disappear are the ones no specification sheet can measure: calibration traceability, change control, long-term spare-part supply and unit-to-unit consistency. Each of them is a by-product of who owns the process, and each degrades slowly enough that the buyer notices only years after the purchase.
Calibration Traceability Gets Longer And Thinner. When a manufacturer calibrates in its own factory, the chain from the working standard on the production line back to a national or international reference is short, internal and audited, and the laboratory that holds the chain is typically accredited to ISO/IEC 17025. When production is outsourced, calibration becomes a service purchased per lot from whoever the contract manufacturer already uses. The chain acquires extra links that no one owns end to end, the certificate increasingly describes a test that was performed rather than a process that is controlled, and the interval over which the reference is verified is set by the supplier's commercial convenience. For an instrument claiming six-and-a-half-digit resolution or picoampere current measurement, that is not administrative detail: it is the difference between a number you can defend and a number you cannot.
Change Control Becomes News Rather Than Notice. An original design manufacturer has every incentive to substitute, and form-fit-function equivalence is exactly the argument that gets made when a memory device, a fan, a capacitor, an analog-to-digital converter or a bare-board supplier becomes inconvenient. A brand owner that does not run the line learns about these changes from failure statistics and customer complaints rather than from a change notice, and unannounced substitution is precisely what invalidates type-tested safety construction, invalidates published accuracy claims and breaks the firmware and hardware revision mapping that support engineers rely on. A maker that owns its plant still has a change control board and a revision scheme, but the change is the maker's own decision, recorded against a product it intends to support. Ownership does not eliminate change; it makes change legible and reversible.
Spares Dry Up And Units Stop Agreeing With Each Other. Spare-part supply follows the factory. When an outsourcing arrangement ends, the tooling, fixtures and test software go to storage or scrap, and replacement assemblies are gone long before the warranty and service life of the installed base expire; the maker that owns its plant can re-run a board, keep a calibration fixture alive, or at minimum state honestly when it cannot. Consistency follows the same logic. Two instruments built in different factories, or in the same factory at different times under different subcontractors, can disagree because the assembly house changed, the thermal environment drifted, or the calibration fixture was rebuilt. A manufacturer with one controlled line can publish a tolerance and hold it across years of production. Instruments are the reference against which everything else is judged, and reference equipment commonly serves for a decade or more, which is why measurement-grade manufacturing rewards the same kind of vertical control that precision mechanical and optical industries learned long ago.
