space-rha.comShort courseChapter 5 · BOM scrub and parts review

Chapter 5 · Lesson 1 of 6 · about 12 minutes

Start the scrub with exact identity

A parts list says what was selected. The review also needs what each part does, how it is operated, the radiation it will see and what happens to the function when it drifts or fails, and it has to start while a different part, a circuit change or a targeted test is still practical.

The demonstration board

Eight constructed parts on a fictional avionics board, each with the evidence that a first search turned up. Use the board through this chapter: select a row to see what the available reports did and did not observe.

A BOM is where the assessment starts.

A parts list tells you which parts were selected and little else. To review it for radiation you also need to know what each part does in the circuit, how it is biased and loaded, what radiation reaches it and what happens to the function when it drifts or fails. Start early. If the review begins after the layout is frozen, a different part, a circuit change or a targeted test may no longer be practical.

The first product of the scrub is an inventory another engineer can check, a list of the unknowns that could matter to the mission, and a named next action for each of them.

Build a review that produces engineering work.

Freeze a BOM revision and the design it belongs to, so that everyone is reviewing the same list. For each line, tie the exact part and its procurement record to the technology, the function it serves, the exposure at its location, the radiation evidence that applies and a proposed action. Do this with the circuit designer and the parts engineer, because a radiation specialist generally cannot work out the application from a part number alone.

Keep what the evidence shows separate from the project's decision to accept a configuration. The first is a finding. The second is a disposition, and the project makes it.

The same part number does not guarantee the same device.

Record the full orderable part number, the manufacturer, the package, the revision and the procurement details. A familiar base number can hide a new die, a different foundry, a process change or a change in internal construction. A date code is useful for traceability, but by itself it does not establish that two parts came from the same wafer diffusion lot.

The question to keep asking is what identifies the silicon that was tested, and what identifies the silicon you are going to buy.

Document the match and the missing links.

Keep the part number exactly as it appears on the BOM before you simplify it for a database search, so the search can be traced back to the line. When a hit comes back, say what kind of match it is: the exact part, an alias, a package variant, a related family member or only the same technology. Each of these supports a different amount of inference. Ask the manufacturer for process-change notices and for whatever die and lot identifiers are available. Track the package on its own, because a change of package can block beam access to the die even when the silicon is unchanged.

Quality grade, automotive qualification and ILPM status describe screening and process control, which is different information from the radiation response.

Exercise on the demonstration board

Select U1 on the demonstration board and read its construction note. Then suppose the manufacturer announces a new die revision for it. Decide which of the identity and population assumptions behind its evidence reopen, and which would survive the change.

Going deeper

Make the inference chain explicit.

The sample that was tested, the population you will procure and the circuit the part goes into are three different things, and the review has to say what connects them: the process identity, how the sample was drawn, the operating conditions and the failure modes the test could not see. A finding written under configuration control states which assumptions the conclusion depends on, so that a later change can be checked against them.

The unit of review is one part, one effect, one application and one configuration (part × effect × application × configuration). One BOM row can therefore need several linked findings.

Similarity is an engineering argument.

An unchanged part number says nothing about whether the sensitive structure, the isolation scheme, the oxide process, the cell layout or the power-device geometry changed. Evidence that two parts share a design and a fabrication line can support transferring data from one to the other, within limits that depend on the effect. Lot, wafer, die and assembly identifiers mean different things, so say which link you have verified and which you are assuming.

No general rule lets you transfer every effect across a package change, a die revision or a fabrication change. Each case needs its own argument.

The common mistake

A spreadsheet with every cell filled in can still hold an incomplete radiation argument. Classifying parts by their part-number prefix, or leaning on a reliability grade, does not establish the construction of the parts that will fly.

Takeaway

  • Review each part in its circuit and its exposure, with the designer, and not from the catalog entry alone.
  • Record the exact part identity before you simplify it for a search.

Check yourself

A parts engineer notes that every active part on the BOM is automotive qualified and screened to a high quality grade, and proposes to skip the radiation scrub for those lines. What is wrong with that?

Quality grade, automotive qualification and ILPM status describe screening and process control, which is different information from the radiation response. Classifying parts by their part-number prefix or leaning on a reliability grade does not establish the construction of the parts that will fly, so each line still has to be tied to its technology, the function it serves, the exposure at its location, the radiation evidence that applies and a proposed action.

A database search on a simplified base part number returns a heavy-ion report. What has to be recorded before the hit is written into the finding?

What kind of match it is: the exact part, an alias, a package variant, a related family member or only the same technology, because each supports a different amount of inference. Keep the part number exactly as it appears on the BOM so the search can be traced back to the line, ask the manufacturer for process-change notices and for the die and lot identifiers that are available, and track the package on its own, since a change of package can block beam access to the die even when the silicon is unchanged. A familiar base number can hide a new die, a different foundry or a process change.

The flight parts and the tested parts carry date codes close together. Can the date code be used to argue that they came from the same wafer lot?

Generally not on its own. A date code is useful for traceability, but by itself it does not establish that two parts came from the same wafer diffusion lot; lot, wafer, die and assembly identifiers mean different things, and the finding should say which link has been verified and which is being assumed. The question to keep asking is what identifies the silicon that was tested and what identifies the silicon you are going to buy.

The manufacturer of U1, the shared regulator on the demonstration board, announces a new die revision. Which of the assumptions behind its evidence reopen, and which survive the change?

The single-event evidence generally follows the mask set and the process (lesson 5.4), so the nominal-condition no-SEL result and anything else tied to the die reopens, and the match between the tested sample and the parts to be procured has to be re-established. The dose and displacement evidence generally follow the lot, and parts of the new revision will not come from the tested lot, so U1's TID report, which was a lot question already, remains one. What survives is the application side of the finding: the function the part serves, the rail it feeds, the recovery that depends on it and the exposure at its location; a finding written under configuration control states which assumptions the conclusion depends on so that exactly this check can be made.

Why does the course say the scrub has to be done with the circuit designer and the parts engineer, and why does it keep a finding apart from a disposition?

A radiation specialist generally cannot work out the application from a part number alone: how the part is biased and loaded, what it feeds and what happens to the function when it drifts or fails come from the designer, and the exact procurement record comes from the parts engineer. A finding states what the evidence shows; a disposition is the project's decision to accept a configuration, and the project makes it. Keeping the two apart is what makes the first product of the scrub usable, an inventory another engineer can check, a list of the unknowns that could matter and a named next action for each, delivered while a different part, a circuit change or a targeted test is still practical.

Further reading