Technical Note · Random and Mixed-Mode Control
Limit Channels, Notching and Alarm and Abort Lines
How a limit channel holds a response point to its allowable level by notching the drive, how an abort channel watches a point with alarm and abort lines in both directions, and why, in a mixed-mode test, notching never switches an abort off.
Key takeaways
- A control channel defines the test. A limit channel caps the response at one point by reducing the drive there. An abort channel watches a point and stops the test if its response leaves the window you set.
- Limiting works line by line: wherever a limit channel would exceed its limit, the drive on those lines comes down, and it recovers as soon as the response falls back. The limit is absolute by default, so the notch deepens as the test level rises.
- Abort channels in Random, Sine-on-Random (SoR) and Random-on-Random (RoR) can carry a full set of lines: the profile becomes a reference with ±alarm and ±abort dB at every corner, judged in both directions and scaled with the test level. A profile without tolerances keeps the classic ceiling.
- The high abort is judged all the time, through the start-up ramp and every level change. The low side and the alarms are judged at a steady level after a short settle, so a ramp cannot trip them.
- In SoR, RoR and SoRoR the control's own alarm and abort lines stay active on notched lines. A notch that would pull the control below its low abort line ends the test: limiting never loosens an abort.
- Sine tones are taken out of limit and abort channels' spectra and judged against their own peak-amplitude lines. A channel without a tone profile still judges its total response against its high abort line, so a tone cannot hide from an abort.
Three jobs for a response channel
Every input channel in a Panther random-family test has a type. The type decides what the controller does with that channel's spectrum. Three types act on the test; a fourth only records.
| Type | Its profile is | What it does to the drive | How it can end the test |
|---|---|---|---|
| Control | The test specification: a reference with ±alarm and ±abort dB | The drive is equalized so the control spectrum matches the reference. Several control channels are combined: average, maximum or minimum | More lines than allowed past the control's abort lines |
| Limit | The most the point may see: an allowable response level | Where the response would exceed the limit, the drive is notched on those lines | Through the control: in SoR, RoR and SoRoR a notch that pulls the control past its own abort lines ends the test. A tone over its tone limit by the channel's abort margin also ends it |
| Abort | A window: a reference with ±alarm and ±abort dB, or a classic ceiling | None. It only watches | More lines than allowed outside its abort lines |
| Measure | None | None | Never. It is displayed and saved only |
Limiting: how notching works
A limit profile is a PSD entered for the limit channel, usually the allowable response at a critical point: a component's qualification level, an interface acceleration limit, or a measured field response. On every control loop Panther compares each limit channel's averaged spectrum with its limit, line by line. Where the channel is over its limit, that line's error becomes the limit channel's error instead of the control's, and the drive is corrected with the control's own law and loop gain:
- Only the lines that need it. The notch is as wide as the region where the limit channel is over its limit, and as deep as it takes to bring it back. The rest of the band runs at full level.
- It follows the response. When the response moves, for example a swept narrow band crossing a mode, the notch moves with it and releases behind it.
- Absolute by default. The limit is not scaled with the test level. A limit that bites at full level may not bite at all at −6 dB, and the notch grows as the schedule steps up.
- Several limit channels: on each line, the most restrictive one governs.
In sine-on-random, random-on-random and SoRoR
Mixed-mode tests drive the broadband random through a compensation filter, measured during system identification and updated during the test. Panther applies the notch as a per-line gain on that filter, so it acts on the broadband and on any narrow bands riding on it. When automatic compensation updates are on, each update aims at the notched target, so it cannot undo the notch. RoR's Dynamic method, which has no compensation filter, notches the drive directly, as Random does.
Note the deepest control line in Figure 1: −6.1 dB, on the control's −6 dB abort line. How Panther treats that line is the subject of a later section, Notching and the control's own lines.
Swept tones on a limit channel
A PSD limit has no meaning for a sine tone: a tone's power lands in one or two lines and reads as a sharp peak far above any random allowable. So in SoR tests Panther takes the fitted tones out of the limit channel's spectrum before comparing it with the limit. Without that, the random under every tone would be notched.
Each tone is then limited in its own terms, against a tone limit for the channel: peak g against frequency. While a tone is in a region where the limit channel would exceed its tone limit, the tone's target becomes the lower of its reference and the level the limit allows. That level is predicted from the transfer between the control and the limit channel, which the broadband measures at every line before the tone arrives, and trimmed by the limit channel's measured tone. The event log records when a limit channel takes control of a tone and when it lets go.
Because every tone is generated through the same compensation filter as the broadband, a tone sweeping through a notched region would be cut by the broadband's notch too. Panther raises that tone's drive by exactly the notch's cut, so the notch acts on the random alone and the tone stays on its own target.
A limit is part of the test specification. Response limiting lowers the input exactly where the article responds most, so it changes the test. Base each limit on an engineering rationale (component allowables, interface limits or measured field response) and agree it with whoever owns the specification. Our Resonance note shows why limits are usually needed at the article's modes.
Alarm and abort lines on abort channels
An abort channel watches a location that is not controlled, such as a sensitive component, a second fixture point or the far end of a long article, and stops the test if its response leaves the window you set. Panther's abort channels take the same kind of profile as a control channel: a reference curve with four tolerances at every corner.
| Line | Typical | When more lines than allowed are past it |
|---|---|---|
| High abort (+abort) | +6 dB | The test aborts |
| High alarm (+alarm) | +3 dB | An alarm; the test continues |
| Low alarm (−alarm) | −3 dB | An alarm; the test continues |
| Low abort (−abort) | −6 dB | The test aborts |
The lines are built the way the control's are: each corner's level scaled by its dB, interpolated between corners as the profile is. They follow the test level, so at −6 dB the whole window sits 6 dB lower. Lines outside the channel's checked frequency range are not judged. Lines past the abort lines, high and low together, are counted per channel and compared with that channel's own number of lines allowed out; the test aborts when the count exceeds it, the rule the control uses. The alarms use the same count.
When each line is judged
A random test does not start at full level. It ramps up, equalizes and often steps through a schedule. While the level is moving, a response that is correctly low for the moment would read as a low-side failure, so the four lines are not all judged all the time.
- The high abort is judged on every loop: the start-up ramp, equalization, level changes and full level. It reads the abort-channel average set for the test, so it responds as quickly as the classic ceiling does. During a level change it is judged against the higher of the old and new levels, so a response moving between levels never reads falsely high, while a real overshoot still trips it.
- The high alarm runs whenever the control's own line checks run.
- The low abort and low alarm are judged only at a steady level, after a short settle once averaging restarts. A ramp from a low initial level would otherwise trip every low line on the way up.
- The alarms and the low side read a smoothed average, as the control's own checks do, so a single noisy frame cannot set them off.
Why the low side matters
A ceiling can only see too much. Many of the failures an abort channel is there to catch read as too little: an accelerometer that has worked loose, a fixture bolt that has backed out and decoupled a mass, or a mode that has moved away from where the specification put it. With a low abort line, each of those shows up as a channel falling below its window. In Panther's validation tests on the simulated test system, abort channels built to read low stopped Random and SoR tests a few seconds after full level was reached, with hundreds of lines below the low abort and none above the high one: a failure a ceiling cannot see.
Profiles without tolerances: the classic ceiling
An abort profile whose tolerances are all 0 dB behaves as Panther abort channels always have: the curve itself is the abort line, high side only, at the level entered and not scaled with the test level. Test files written before two-sided lines existed keep exactly that meaning. A side whose tolerances are all 0 dB is not judged, so one channel can be set up as a full window, as a high side with alarms, or as a classic ceiling.
Notching and the control's own lines
When a limit channel notches the drive, the control on those lines is deliberately below its reference. A controller can treat those lines in one of two ways. Panther Random has always taken notched lines out of the control's alarm and abort count, since the control is meant to be low there. In SoR, RoR and SoRoR, Panther keeps them in the count, as Panther Sine always has: the control's tolerances stay in effect while a limit channel is active.
The consequence is simple. The control's low lines cap the notch. A notch deeper than the control's low alarm line raises an alarm; one deeper than its low abort line, on more lines than allowed, ends the test. That is the intent: the control's abort lines are part of the test specification, and a notch that deep is no longer a notch but an under-test of the very point the specification was written for.
Limiting never loosens an abort. Every hard abort in the setup, on the broadband, a band or a tone, on a control, limit or abort channel, is honored whatever the limiting is doing. If a limit and an abort line conflict, the abort wins and the test stops. The conflict is resolved in the setup, not by the controller.
| Random | Sine-on-Random | Random-on-Random | SoRoR | |
|---|---|---|---|---|
| Broadband notched by limit channels | Yes, per line | Yes, through the compensation | Yes, through the compensation (Dynamic method: per line) | Yes, through the compensation |
| Narrow bands notched | – | – | Yes, with the broadband | Yes, with the broadband |
| Tones limited | – | Yes, against a tone limit | – | Yes, against a tone limit |
| Control's lines on notched lines | Taken out of the count | Kept | Kept (Dynamic method: taken out) | Kept |
| Two-sided abort channels | Yes | Yes | Yes | Yes |
| Tones on abort channels | – | Their own tone profile, or the total response against the high abort | – | As sine-on-random |
Tones and bands on limit and abort channels
Sine-on-random and SoRoR tests put deterministic tones on top of the random, and every limit and abort channel sees them. A PSD line cannot judge a tone fairly, so Panther treats the two parts of the channel separately.
- The random part. The fitted tones are removed from limit and abort channels' time data before their spectra are formed, exactly as on control channels. The PSD limit and the alarm and abort lines judge the random alone. The lines right under a tone, where the removal also takes out a little of the random, are left out of the low side so they cannot read falsely low; the high side still judges them.
- The tones. Each limit or abort channel can carry a tone profile: peak g against frequency, with ±alarm and ±abort dB per corner, judged at each tone's current frequency on the averaged tone amplitude. On an abort channel the high tone abort is judged all the time and the low side and alarms at level, as for the PSD lines. On a limit channel the profile is the tone limit described above.
- The safety net. An abort channel without a tone profile still judges its total response, random plus tones, against its high abort line. A tone that over-tests at an abort channel is caught either way.
- What you see. The Tone Extract setting (control channels, or all channels) decides whether limit and abort channels are displayed and saved with their tones or without them. The checks always work on the separated parts.
Narrow bands in random-on-random and SoRoR
A narrow band is random, so it is judged in the channel's PSD like the broadband. On a limit channel that needs nothing special: the notch follows the band across a mode and releases behind it. On an abort channel the lines are judged exactly as you set them, so set them with the band in mind. If a band sweeps through the channel's checked range, a window drawn at the broadband level reads high wherever the band is, and a window drawn around the band's level reads low wherever it is not. Either keep the checked range clear of the band's sweep, or use the channel as a ceiling there: low tolerances at 0 dB, and the profile raised over the band's whole sweep range.
Setting it up
- In the channel table, set the channel's type to Limit or Abort.
- Each limit and abort channel gets its own profile tab. Enter the limit (limit channel) or the reference (abort channel) as a PSD.
- For an abort channel, enter the four tolerance columns at each corner. Leave a side at 0 dB to switch it off; leave all four at 0 dB for a classic ceiling.
- Set the channel's checked frequency range and its number of lines allowed out.
- In a sine-on-random test, add a tone profile for any limit or abort channel whose tones should be judged in their own terms.
- Start the schedule at a low initial level. Because the limit is absolute, you see where it will bite before full level, and the abort channels' low side is only judged once each level is reached.
Common misconceptions
| Misconception | Reality |
|---|---|
| "A limit channel protects the article." | It caps one point, in its range, from above. An abort channel with a low line also catches a point that reads too little, which is how a loose sensor or a failed fixture often shows up. |
| "While a limit is notching, the control's abort lines don't apply." | In SoR, RoR and SoRoR they do. A notch that pulls the control past its low abort line, on more lines than allowed, ends the test. |
| "Low aborts will trip while the test ramps up." | The low side is judged only at a steady level, after a short settle. |
| "Abort lines are fixed in g²/Hz." | With tolerances entered they follow the test level, as the control's lines do. Only the classic ceiling stays fixed. |
| "A sine tone on a monitor channel will trip its PSD abort." | Tones are removed from the PSD and judged against their own peak-amplitude lines, or, without a tone profile, as part of the total response against the high abort line. |
| "A deeper notch is a safer test." | Below the control's low abort line it is an under-test, and Panther treats it as one. |
Quick reference
| Item | Rule |
|---|---|
| Notch on a limited line | g ← g · [min(L/R, 1/g)]k, never above 1 |
| Limit scaling | Absolute (not scaled with the test level), by default |
| Several limit channels | The most restrictive governs each line |
| Abort-channel lines | Each corner's level × 10dB/10, interpolated as the profile; scaled with the test level |
| Abort count | Lines past the abort lines, high and low together, per channel; aborts when more than the lines allowed |
| High abort | Every loop, through ramps and level changes |
| High alarm | Whenever the control's line checks run |
| Low abort and low alarm | At a steady level, after a short settle |
| No tolerances entered | Classic ceiling: high side only, at the level entered |
| Notched lines, SoR / RoR / SoRoR | Stay in the control's alarm and abort count |
| Tones on limit and abort channels | Removed from the PSD; judged by a tone profile, or by the total response against the high abort |
How Panther helps
- One profile model for every check: a reference with ±alarm and ±abort dB per corner, for control channels, abort channels, narrow bands and tones.
- Two-sided abort channels in Random, Sine-on-Random and Random-on-Random, with every existing test file keeping its meaning.
- Line-by-line notching in every random-family test, applied through the compensation filter in mixed-mode tests so it holds with automatic updates on.
- Tones and bands in their own terms on limit and abort channels, in sine-on-random, random-on-random and SoRoR tests.
- Every hard abort honored: limiting never loosens one.
- A log that says what happened: which channel alarmed or aborted, on which side, on how many lines, and when a limit channel took and released control of a tone.
References and further reading
- MIL-STD-810H Change 1, Environmental Engineering Considerations and Laboratory Tests, Method 514.8, Vibration (control strategies, multi-point control and response limiting).
- NASA-HDBK-7004C, Force Limited Vibration Testing.
- NASA-STD-7001B, Payload Vibroacoustic Test Criteria.
- C. M. Harris and A. G. Piersol (eds.), Harris' Shock and Vibration Handbook, 5th ed., McGraw-Hill.
- Spectral Dynamics Technical Library: Resonance; Power Spectral Density (PSD) and Grms; Statistical Degrees of Freedom; Tracking Filters in Sine Vibration Control.
