Why the Cheapest Cable and Magnet Cost the Most

A vibration reading is not produced by an instrument. It is produced by a chain: the sensor, the mount that couples it to the machine, the cable that carries the charge or the voltage, the conditioning that powers the sensor, the DAQ that digitises it, and the analysis that turns the digits into a verdict. Every one of those links is in series. The number that reaches the analyst has already passed through all of them, and by then there is nothing in the number that says which link changed.

This is why the usual purchasing pattern is so damaging. The DAQ and the software go through a technical evaluation, because they are the visible, expensive line items. The cable and the mount go through the stores, because they are consumables. A plant will sign for a four-channel 24-bit analyser and then connect it with a cable bought on price alone and a magnet whose only stated property is that it is a magnet. The chain is now defined by its cheapest link, and the specification of the expensive one has been thrown away.

The rest of this article is not about which brand to buy. It is about three specific ways a weak link costs money — false alarms, missed faults and lost trend history — the questions that let you tell a specified accessory from an unspecified one, and an honest, editable model you can run on your own machine count and your own downtime rate. The physics behind each link is covered in three companion articles, linked below.

A cheap cable does not fail silently. It generates signal. Coaxial cable that flexes under machine vibration produces triboelectric noise — charge released as the dielectric separates from the shield, appearing at the input as a voltage the sensor never produced. A poorly clamped connector adds intermittent contact noise. A magnet that rocks on a curved or painted surface adds impact chatter of its own. A noisy DAQ input adds a floor that never goes away. None of that is the machine, and all of it looks exactly like the machine to a threshold.

The immediate cost of a false alarm is a stopped machine and a strip-down that finds nothing wrong. That is a bad afternoon. The second-order cost is worse and it is permanent: the crew learns that the system cries wolf. After the third healthy machine is opened, the alarms stop being acted on, and the day a real alert fires it is read as another cable problem. A condition-monitoring programme that is no longer believed is not a programme with a fault. It is a programme with no output, still being paid for.

The engineering test for this failure mode is simple. Take a reading with the machine stopped. Take another with the machine running but the sensor held off the machine. Flex the cable by hand during a measurement and watch the spectrum. If any of those produce signal you would have alarmed on, the alarm you saw last week may never have come from the bearing.

The second failure mode is quieter, and that is what makes it expensive. An early bearing defect is a small, high-frequency signal sitting next to a large, low-frequency one. It has to survive two things to reach the analyst: the mounted resonance of the sensor-and-mount assembly has to be high enough that the bearing tones are still inside the usable band, and the noise floor of the cable and the DAQ has to be low enough that a small peak still stands above it.

A magnet lowers the mounted resonance because it inserts a compliance between the sensor and the machine. A large flat magnet on a rough painted surface lowers it further. Once the usable band ends below the bearing tones, the analyser is not reading a healthy bearing — it is reading a bearing it cannot see, and reporting the only thing left in the band, which is the running-speed peak. The trend looks flat and reassuring right up to the failure. The mount article covers where that resonance lands and why; this article is only concerned with what it costs.

The noise floor does the same job by a different route. A defect that produces a peak of one hundredth of a g is plainly visible above a floor of one thousandth and completely invisible above a floor of one twentieth. Nothing in the software will tell you the peak was there and got buried. The spectrum simply comes back clean, and a clean spectrum from a bad chain is indistinguishable from a clean spectrum from a healthy machine. Drag the slider below and watch the same growing defect appear in one chain and not in the other.

The output of a condition-monitoring programme is not a reading. It is a trend. A single number in isolation tells you very little; the same point measured the same way every month tells you almost everything, because the machine is being compared against itself. That comparison only holds while nothing except the machine has changed.

Three things break it quietly. Inconsistent mounting: a magnet placed a few centimetres away, or on paint this month and bare metal last month, changes both the level and the usable bandwidth. Drifting sensitivity: an uncalibrated sensor whose output has moved five per cent since last year puts a five per cent slope into a trend that nothing in the plant caused. And a changed cable: a replacement with different capacitance and a different noise floor shifts the baseline on the day it is fitted.

The cost is specific and it is not recoverable. When the trend steps, you cannot tell whether the machine changed or the hardware did, so the alert levels derived from the old data are void and the history before the change has to be discarded. You are back to month one on that machine, and the twelve months of data you paid technicians to collect bought nothing. The saving on the replacement cable was three hundred rupees. The cost was a year of history on every point it touched.

This calculator does not know your plant, and it is not a study. It is arithmetic on assumptions, and every assumption it uses is printed below the controls where you can see it and change it. Set your machine count, your points, your downtime rate and the length of a typical unplanned stop, then choose a quality tier for each of the three links. It compares a fully specified chain against the chain you chose, and puts the hardware saving next to the consequences.

The tier logic is deliberately crude, because a defensible crude model is more useful than an impressive one you cannot check. Each link earns weak-link points — three for a generic, unspecified item, one for a mid-grade commercial item, none for a measurement-grade item with published figures. Those points raise the false-alarm rate and lower the probability that an early fault is detected in time. Both effects are linear, both rates are editable, and both defaults are engineering judgement rather than measurement.

Read the output as a shape, not as a number. If you halve every assumption the conclusion generally survives, and that is the only claim this model makes: the accessory saving is a one-off figure in thousands, and the consequence of a weak chain is a recurring figure in lakhs. Change the assumptions until you believe them, then look at the two bars again.

There is one test that separates a measurement-grade accessory from a generic one, and it does not require a laboratory. Ask for the number. A supplier who has tested a component can quote it with units and test conditions. A supplier who has not will answer with adjectives — high quality, industrial grade, suitable for vibration. Adjectives are not a specification, and a component with no specification is an unknown that you are about to install permanently in your measurement chain.

This is not about brand loyalty and it is not about price. A cheap component with a published, honest specification is a known quantity, and a known quantity can be designed around: if the mounted resonance is 3 kHz you simply do not trust the spectrum above about 1 kHz, and you say so on the report. An expensive component with no published specification is still an unknown. The number is what you are buying.

One more question that is worth more than all of them: is there a traceable calibration certificate for the sensor this accessory serves, and a batch or unit test record for the accessory itself. Traceability is what lets you say that the reading you took last March and the reading you took this March mean the same thing. Without it, a trend is a collection of numbers that happen to be on the same chart.

It would be dishonest to stop there, because the cheap option is sometimes correct, and pretending otherwise makes the rest of this article a sales pitch. The argument above is about trended condition monitoring. Where you are not trending, most of it does not apply.

A one-off rough check is a good example. An operator reports that a pump sounds different, someone walks over with a meter, and the question is whether the level is two millimetres per second or twenty. That decision survives a great deal of noise and a mediocre mount, because the answer is an order of magnitude, not a trend. The same is true of a quick comparison between two identical machines running side by side on the same day: whatever the chain does wrong, it does to both.

A training rig is another. On a fault simulator you already know what fault is present, the amplitudes are large by design, and the purpose is to teach a technician to recognise a pattern. A modest cable and a plain magnet teach that pattern perfectly well, and spending the training budget on measurement-grade accessories instead of on more hours at the rig is a poor trade.

The rule that separates the two cases is simple. If the reading will be compared — against last month, against a threshold, against a sister machine measured at another time — buy the specification. If the reading will be used once, looked at, and thrown away, buy what is convenient. The mistake is not owning a cheap accessory. The mistake is letting a cheap accessory into a trend without noticing.

TIERA supplies the whole chain rather than one part of it: IEPE accelerometers, low-noise coaxial and armored cables, magnetic mounts and SS304 triaxial blocks, adhesive pads, junction boxes, PhonoVibe DAQs and TVIB analysis software. Individual datasheets for the cables and mounts are published in the public TIERA datasheets repository, which exists precisely so that a buyer can read the specification before raising the purchase order rather than after.

Two links in that chain carry calibration documents. Every PhonoVibe DAQ ships with a factory calibration certificate carrying one year of validity. The T-Calibro back-to-back calibrator lets a plant verify accelerometer sensitivity on its own bench, and its reference sensor ships with a factory calibration certificate establishing its sensitivity against a national-standard traceable chain. That is what makes a trend comparable across years rather than merely across months.

The claim is a modest one and it is worth stating precisely. Buying the chain from one supplier does not make the measurement good on its own — mounting discipline and a competent analyst matter at least as much. What it does is remove the unknowns: the cable has a published construction, the mount has a published pull-off force, the DAQ has a published resolution and a calibration certificate, and the sensor has a sensitivity you can verify. When the trend then moves, you can be reasonably confident it moved because the machine did.

TIERA supplies every link between the machine and the verdict: IEPE accelerometers, low-noise coaxial and armored cables, magnetic mounts, SS304 triaxial blocks and adhesive pads, junction boxes, PhonoVibe 24-bit DAQs and TVIB analysis software. The cables and mounts have individual datasheets in the public TIERA datasheets repository, so the specification is available before the purchase order rather than after the first strange reading.

If you are specifying a route kit and want the chain checked as a chain, send us the machine list and the points you intend to trend. Our engineers will tell you which links your programme genuinely needs at measurement grade and which ones it does not — including the cases where the cheaper item is the correct answer.