Load Cells for IV Bag Weighing: Everyone Else’s Error Is Ten Percent

1. The one bag that moves its own fluid

I have weighed a lot of containers. Tanks on legs, hoppers on modules, drums on platforms, bags on hooks. In almost every one of them the container is a passive object: something else decides how fast the contents move, and the scale simply watches. A urine bag watches. A filler nozzle controls. A tank pump delivers.

An IV bag is the exception. It hangs from a pole above a patient, it drains by gravity, and the pressure that pushes the fluid out of it is created by the height of the fluid still in it. The reading and the mechanism are the same variable. As the bag empties, the head falls, the flow slows, and the weight you are measuring is the reason. That single coupling rejects most of the standard advice about weight monitoring, and it is why three of the projects in this article turned out to be hydraulics rather than electronics.

The mechanics first, because they set everything else. A one litre bag of normal saline holds 1004.6 grams of fluid at room temperature, and an empty one-litre PVC bag with its spike weighs roughly 32 grams, so the hook carries about 1036.6 grams when the bag is full. The same bag is labelled as containing one litre. Those two numbers, 1004.6 and 1000, are the first hint that this application is a volumetric problem being solved with a mass sensor, and they are 0.46 percent apart because saline is not water.

The second hint is the direction. A urine bag counts up over three days and its total is a clinical endpoint. An IV bag counts down over one to eight hours, and its endpoint is that it stops. Nobody charts the bag’s weight; they chart the volume that went in, and the volume that went in is the volume that left the bag minus what is still sitting in the tubing. I will come back to that subtraction in section seven, because it is where a one litre bag and a fifty millilitre neonatal bag stop being the same problem.

If you are an engineer sizing load cells for IV bag weighing, or a pharmacy automation team putting a gravimetric check on a compounding bench, the useful news is that there is almost no measurement problem in this application. The signal is enormous, the required resolution is about ten bits, and the cell I would quote is fifty times more accurate than the tightest requirement anywhere in the chain. The difficult part is that everyone else in the chain is working to five or ten percent, and one of them is a nurse counting drops with a stopwatch.

2. Where my supply stops, and where yours starts

I export weighing components. I do not make infusion devices, I do not hold a medical device registration, and I cannot issue an IEC 60601 test report or a USP 797 qualification. Saying so is not a disclaimer bolted on at the end; it is the boundary that makes the rest of this article usable, because everything on the right-hand side of the table below changes the sensor requirement, and a supplier who quietly implies he covers those items will size your cell for the wrong job.

The second column is longer than a supplier usually admits, and there is a precedent for it inside our own catalogue. The published weighing solution for a hospital bed that we have carried for years stops at the junction box on purpose: the converter, the controller and the screen are drawn as customer supplied. Nobody has ever objected, because the customer was building the bed. The same is true here, with one addition. On a medical device the right-hand column is not only your work, it is your liability, and it is the part a notified body will read line by line.

So I quote a component. When I decline something, it is because I have been asked to make a decision that belongs to the device manufacturer, and pricing it would only hide where it belongs.

3. Where this sits beside the other medical weighing jobs

We have published several articles in this space already, and the fastest way to waste your time is to read the wrong one. All of them are a bag or a patient on a sensor, and none of them is the same problem. The table separates them by what the reading is for, which is the only division that changes the hardware.

Two boundaries are worth stating explicitly. The first is against the urine article. A urine bag and an IV bag look identical on a pole, and they are opposites underneath: one counts up and one counts down. The urine bag accumulates a total that a clinician reads as evidence about a kidney; the IV bag surrenders a volume that the clinician already prescribed, and the question is only whether it arrived. One is a measurement of the patient. The other is a measurement of a delivery.

The second boundary is against the syringe pump article, which is about the same room. A syringe pump pushes. A gravity bag is pushed by nothing at all — it is drained by the height of its own contents, and that is why the two applications share a data sheet but not a single design decision. If your line is pump driven, the physics in section five does not apply to you and your budget is tighter. If your line is gravity fed, almost nothing else in this article matters more than section five.

4. The error stack, and why the sensor is not your problem

Every infusion project I have been asked to quote arrives with a sensor question attached, and the sensor question is almost never the question that matters. So before we discuss any cell, here is the error budget of an infusion, written by the people who own each line of it rather than by me.

Percentages hide the ordering, so convert every row to grams on a one litre bag and sort them. The bag holds 1004.6 grams of saline, so one percent of it is about ten grams.

That last sentence is the commercial case for the whole product, so it is worth saying plainly. A volumetric pump knows what it pushed, which is not the same as what arrived. A nurse knows what she counted, which is not the same as what fell. The chart records what somebody wrote down. At no point in that sequence does anything measure the fluid that left the container. Your pharmacy already solved this problem on its own bench: USP <797> requires a documented gravimetric accuracy assessment of every automated compounding device on each day of use, and the published result of that practice in hospital IV workflow systems is a 90 to 99 percent reduction in compounding errors. The weighing method is not exotic in this building. It is the last place without one that is the ward.

The measurement side, to close it out

A common adult maintenance infusion is one litre over eight hours, which is 125 millilitres an hour. Convert it and the numbers get large very quickly: 2.09 grams a minute, or 34.9 milligrams a second. A fifteen minute window — the sort of window a rate display would use — contains 31.4 grams of signal.

Now ask what resolution the job needs. Resolving one millilitre on a full one litre bag means resolving one part in 1037, which is 10.02 bits. A twenty-four bit converter delivers 16185 times that resolution. If you fit a sixteen bit converter because it is cheap, you are still 63 times finer than the question. I have quoted load cells for IV bag weighing into hardware where the signal conditioning was the most expensive line on the bill of materials, and in every one of those projects the money belonged somewhere else.

Resolution is also not the constraint you think it is at the slow end. The quietest line anybody monitors is a keep-vein-open drip, and the smallest rate a fifteen minute window can resolve is set by the cell’s own error floor, not by the converter. Here is what each range can actually see.

Even the ten kilogram option resolves a paediatric line. The consequence is that there is nothing technical to argue about, and the choice of cell is decided by mechanics: what the bag weighs, where it can hang, and what happens on the day somebody lets it drop onto the hook.

5. The hanging height sets the flow, and the bag lowers it

A gravity infusion is driven by the column of fluid between the surface of the liquid in the bag and the point where the fluid enters the patient. Every millimetre of water is worth 0.0736 millimetres of mercury, and the arithmetic is stable enough to quote. The UK regulator’s own guidance on infusion systems states that a bag may be hung up to 1.5 metres above the infusion site, and that this height provides about 110 millimetres of mercury of driving pressure; it also notes, in exactly these words, that simple gravity drips slow down when the patient sits up, because the head has fallen. One and a half metres at 0.0736 gives 110.3. The published figure and the arithmetic agree.

The international infusion set standard uses a one metre reference head for its own flow test, which is 73.6 millimetres of mercury. So one metre is not my invention; it is the industry’s reference condition.

Now the part that surprises people. A full one litre bag hangs about twenty-five centimetres tall, and that twenty-five centimetres is inside the head. As the bag drains, the fluid surface falls by the full height of the column, so the driving pressure at the end of the bag is lower than at the start by twenty-five centimetres divided by whatever the head was. Hang the bag high and the bag’s own column is a small fraction of the total. Hang it low and the same column is a large fraction.

Read the last column carefully, because it contains the sentence I have to say most often. Hanging the bag lower does not merely slow the infusion; it makes the infusion slow down more. A bag on a 1.5 metre pole loses 16.7 percent of its rate over its own life. The same bag on a 0.77 metre pole loses 32.5 percent. The nurse who lowers a bag to make a patient more comfortable has also changed the shape of the whole hour.

What the fall costs in minutes

If the flow is proportional to the head, and the head falls linearly as the bag empties, the volume against time curve is not a straight line. It is an exponential decay, and it has one consequence that matters commercially: a rate computed from the start of the bag is not the rate the bag will run at. Working the model through for a one metre head, a one litre bag and 250 millimetres of fluid column:

Thirteen point one percent is 2.6 times the tolerance IEC 60601-2-24 demands of a volumetric pump, and it is produced by the geometry of a pole inside a room where nothing has gone wrong. Published clinical engineering material puts the field figure higher still — a gravity infusion can lose as much as 40 percent of its rate over two hours while the bag empties — and 40 percent back-calculates to a pole so low that the bag sits only about 625 millimetres above the site. The difference between my 25 percent and their 40 percent is not a contradiction. It is the set: PVC tubing creeping under a roller clamp, and a fluid that got colder than it was in the pharmacy.

The reason I lead with this is not to attack gravity infusion, which is cheap, safe and appropriate for most maintenance fluids. It is that a customer who has just bought a weighing module will report the falling curve to me as a sensor fault, and I will have to ask for their log and their pole height before I answer. The test takes an afternoon and it is worth doing before you write a purchase order:

6. Where to hang the cell, and where not to

If the sensor is fifty times better than it needs to be, then the selection is mechanical, and the mechanical question is one sentence long: what must the cell see, and what must it not see? Five positions get used in practice, and four of them have a failure mode that is worth knowing before you draw the bracket.

The third row is the one worth arguing for. Putting the cell in line below the bag means the reading includes the fifteen millilitres or so that live in the administration set, and that fluid has already left the container. The difference between what the bag lost and what the patient received is exactly that hold-up, and it is a systematic bias rather than a random error, which means it can be designed out. On a one litre bag it is 1.5 percent and it disappears into the budget. Section seven shows what it does to a fifty millilitre neonatal bag, where it is thirty percent, and that single comparison is the most useful thing in this article for anyone building a small-volume product.

Two mechanical details are worth dispatching with arithmetic rather than opinion. The first is the cell’s own deflection. If the mount moves half a millimetre under the full bag, that is half a millimetre out of a one metre head, which is 0.05 percent of the driving pressure. The compliance of a load cell is not a hydraulic problem. The second is the effect of the cell’s position on the head: moving the sensor five centimetres down the hanger changes the head by five centimetres, which is 5 percent of a one metre column, and that is why a monitor should be calibrated with the sensor fitted, not on the bench.

The two seconds nobody designs for

A bag is not placed on a hook; it is spiked, lifted and let down onto one. That gesture produces a transient that is far larger than any of the error terms we have been discussing, and it is the only event in this application that can physically destroy the sensor. A one litre bag weighs 1.0366 kilograms, and if it is released five millimetres above the hook and stopped in ten milliseconds, it arrives at the hook as 4.35 kilograms.

7. Empty is not a weight, it is a weight plus a residue

Every monitor in this category needs to answer the same question: when is the bag empty? The intuition is that empty means the cell reads the dry tare of the bag. The intuition is wrong, and it is wrong in a way that scales badly in exactly the direction where the patients are smallest.

Three masses have to be separated. The first is the dry tare: the bag and its spike, around thirty-two grams for a common one litre PVC bag, with a spread of a few grams between cartons. The second is the residue: the fluid the bag physically cannot deliver, because its outlet is not at the lowest point of a collapsed pillow. The third is the set prime, the fifteen millilitres or so that fill the administration tubing and stay in it after the bag has finished. The cell reads the sum of all three when the bag is done, and only the first of them is a number you can look up.

The consequence is a commissioning instruction rather than a specification. An alarm set at the dry tare will never fire, because the cell never reaches it. An alarm set at a round number will fire early, and the difference between those two behaviours — the hold-up — is a property of the disposable, not of my sensor. So the number has to be measured per set model, and it has to be re-measured when the customer changes set supplier, which is a sentence I have had to write into three quotations.

Look at the last column as a column and it becomes obvious. The same fifteen millilitres of tubing is a rounding error on an adult maintenance bag and a third of the dose on a neonatal one. It is twenty times the relative burden, from the same disposable, changed only by the size of the container on the hook. That is why the line-prime measurement is not an optional refinement in small-volume work; it is the dominant term, and no choice of load cell changes it.

The alarm rule follows from the same reasoning. If the residue is set by the disposable and the tare spread is a few grams, then the only thing you can alarm on reliably is the slope. When the rate goes to zero the bag is finished, and that condition is independent of which bag you hung and what its dry weight was. A monitor that alarms on a weight threshold will need a new threshold every time procurement changes set supplier. A monitor that alarms on the rate going flat will not.

8. Volume against weight, and why it is easier here than it looks

An IV bag is labelled in millilitres and a load cell reports grams, so at some point somebody has to multiply. This is the exact point where a urine bag project becomes hard, and it is the exact point where an IV bag project becomes easy, and the difference is worth spelling out because people carry the same anxiety from one to the other.

Normal saline is 9 grams of sodium chloride per litre, and one millilitre of it weighs 1.0046 grams at 22 degrees. Water at the same temperature is 1.0000. Whole blood sits around 1.055. So the entire spread of fluids you will ever see on this hook is about five percent, and the correction for the common case is half of one percent.

Five percent is exactly the tolerance IEC 60601-2-24 gives a volumetric pump, which means a firmware that hard-codes saline and is then handed a unit of blood has spent the entire accuracy budget of the device on the conversion, before the load cell or the hydraulics get a turn. Fifty grams on a litre is not a small number; it is a tenth of the bag.

The reason this is nonetheless the easy chapter is that the fluid is on the label. Nobody has to infer it, and nobody has to average over a population. The bag says 0.9 percent sodium chloride or 5 percent dextrose or red cells, and each of those has a published specific gravity. When I look at the IV bag application next to the urine bag application I worked on last month, the contrast is stark: on a urine bag the specific gravity ranges from 1.003 to 1.030 and only the lab knows today’s value, so a whole percentage point of uncertainty is baked into the sensor’s job; here the same correction is a lookup with four rows in it.

9. The room and the pharmacy both move the number

Two entirely different things are called temperature error in this application, and buyers routinely conflate them. One is inside my data sheet. The other is inside a refrigerator. The useful move is to put both in the same unit — grams per degree, and then as a share of the fifteen minute rate signal, which is 31.4 grams.

Two conclusions fall out of that table, and the second one took me a project to learn. The first is that the largest per-degree term in the whole assembly is not the bridge inside the load cell; it is the point where the cable meets it. A pair of dissimilar-metal junctions at slightly different temperatures generates an offset directly in series with the bridge output, and at two and a half microvolts per gram, a few microvolts per degree becomes two grams per degree. If the two ends of your connector differ by 1.6 degrees, you have reproduced the cold-bag effect with a piece of wire.

The second conclusion is the useful one. Everything in that table except the fluid is a zero offset, and a zero offset is deleted by the next re-zero. The fluid’s viscosity change is a change in the actual flow, and no re-zero will touch it. So the design priority is not to buy a better temperature coefficient; it is to put a little thermal mass and insulation around the connector so its pins stay together, and then to accept the fluid as a clinical fact. On a ward running three or four bags a day, the schedule does the zeroing for you.

10. Case one: the cell that drifted, and the test I gave instead

A monitor manufacturer in northern Europe wrote to me in February with a complaint rather than an enquiry. They had built a bedside fluid module around one of our 2 kilogram cells, and their bench results were unacceptable: the weight trace fell away smoothly over the first hour of a bag and finished around twenty-five percent below where a straight line from the start would have put it. Their words were that the sensor was drifting under load, and they wanted a quotation for a higher grade cell, or a different technology.

I did not quote anything in that reply. What I asked for instead was their raw logger file, in full, plus a photograph of the pole with the bag on it and a tape measure held against the wall. I have learned to ask for those two files before I ask a single technical question, because the log tells me what is happening and the photograph tells me where the sensor is.

The log is worth describing, because the shape is diagnostic. The trace was flat at the start of every bag and decayed smoothly towards the end. It reset on every bag change. The level was correct: when they took the bag off, the cell returned to the empty hanger weight within half a gram. And the decay was monotonic — it never went back up.

That combination rules out almost every sensor failure mode I know. A cell with a creep problem drifts under a constant load and stops drifting when the load comes off; it does not start again from a different baseline every six hours. A cell with a thermal problem follows the room, not the bag. A cell with a bad connection produces steps, not curves. What follows the bag exactly, resets when the bag is replaced, and leaves the level intact is the load path itself changing as the bag empties, and there is only one mechanism that does that.

The photograph confirmed it. Their pole was a low one; the fluid surface sat about a metre above the insertion site, which gives 73.6 millimetres of mercury at the start. A full one litre bag hangs about 250 millimetres of fluid column, so by the end of the bag the driving pressure had fallen by a quarter. Section five sets out the whole calculation, and the arithmetic predicts that a one metre head should show a 25 percent decay and an average rate 13 percent below the starting rate. Their trace showed 24.8 percent.

I tell this story first because it is the most common shape of the problem in this application. A gravity set produces a curve that looks exactly like a slow sensor failure. If you are buying load cells for IV bag weighing and your acceptance test is a straight line, you will reject the correct part.

11. Case two: the piggyback hook I would not quote

A hospital equipment distributor in the Gulf asked for a pole scale. The requirement, as they wrote it, was elegant: one load cell under the pole hook, one number on a screen, tracking two bags at once. They wanted a hundred units and they wanted them in a month, and it was the sort of email that is easy to answer quickly and wrongly.

I asked for the administration layout drawing first. Not the mechanical drawing of the pole; the drawing of the line. What came back showed a one litre maintenance bag on the primary line and a hundred millilitre antibiotic piggyback on a secondary line above it, joined at a Y-site. That one drawing turned a simple quotation into a refusal.

Here is why the hook cannot work. While the secondary is running, the hook carries both bags: about 1036 grams of maintenance fluid plus about 132 grams of antibiotic, so 1168 grams together. When the secondary finishes and the primary resumes alone, the hook carries about 1036 grams. A single sensor cannot tell those two states apart from a primary bag that simply got 132 grams lighter through normal delivery, and the rate your firmware computes during the transition belongs to neither drug. Worse, the two bags are deliberately hung at different heights, because a secondary has to sit above the primary for it to drain first, which means the two fluids are being delivered at different driving pressures as well as different concentrations. A hook is not a fluid. The sum of two doses is not a dose.

There is a second problem in the same drawing, and it is the one that changed the project. Their alarm for the end of the bag was going to be a weight threshold, and to set it they needed the hold-up of the administration set. They had three set SKUs from two suppliers across their hospitals, and one of the suppliers had changed the tubing specification eighteen months earlier without changing the part number. The hold-up, which is roughly fifteen millilitres for a standard adult set, is not my number to supply; it belongs to their disposable, and it changes when their procurement changes.

12. Case three: the resolution that was two hundred times too fine

The third project came from a neonatal accessories company building a monitor for total parenteral nutrition in a fifty millilitre bag. Their specification asked for a display resolution of 0.1 millilitres, an empty-bag alarm, and a sensor that would fit inside a hanger they had already tooled. It was the last requirement they cared about and the first one I had to argue with.

A fifty millilitre bag of saline holds 50.2 grams, and with a small bag and spike the whole thing weighs about 82 grams. One tenth of a millilitre is 0.1 grams, which is 0.12 percent of the total load. That is not an unreasonable thing to ask of a 300 gram cell. It is simply not what they needed, because three other errors in their own system are far larger than the resolution they had specified.

I declined to quote a 0.1 millilitre resolution, and I said why in one sentence: you are asking for a resolution two hundred times finer than the largest error already in your own system, and if I sell it to you it will hide your real problem for a year. A specification like that is worse than a loose one, because it is verifiable, it will pass, and it will be believed.

What I proposed instead had three parts, and the second is the one that mattered. First, claim resolution to 0.5 millilitres — one percent of the bag, which is five times finer than the pump standard and twenty times finer than the hold-up you cannot remove anyway. Second, move the cell so that it sees the bag and the fluid in the line down to the patient, which makes what left the container and what entered the patient the same measurement instead of two that differ by a third of the dose. And third, alarm on the slope rather than the level, so the threshold survives a change of set supplier.

13. Which law applies, and which one does not

This is the section I would ask a customer to read before they write a purchase specification, because the reflex in this industry is to reach for trade metrology and the reflex is wrong.

Two things in that table are worth more than the rest. The first is that the reference method inside IEC 60601-2-24 for measuring flow accuracy is gravimetric: you collect the delivery on a balance and convert mass to volume. China’s JJF 1259-2018 does the same thing with a balance. The same is true of the pharmacopoeia’s requirement for compounded sterile preparations. So a weighing device in this application is not introducing an unfamiliar measurement into a regulated space. It is the method the regulators themselves already accept as the referee, moved from the calibration bench to the bedside or the compounding bench.

The second is new to me and it is a design constraint rather than a legal one. If your product is labelled a pump or a controller, then the weighing subsystem’s error becomes part of the pump’s ±5 percent. In every other application I write about, the load cell sits outside the regulation and only has to be good enough for the customer’s own purposes. Here, quoting a resolution instead of an error is not a marketing liberty; it is a defect. When we quote this family I will give you a maximum error in grams at your capacity, with the temperature range it holds over, because that is the number that survives an audit.

One practical consequence follows for any customer already in the field with an approved device, and it is the same rule that applies to every medical component I quote. A change of critical component supplier can itself be a substantial change, which means that once your device is approved, the material, the heat treatment, the capacity, the rated output, the cable and connector, the potting and the ingress construction, the matching tolerance, and the marking and part number are all frozen. If you need a different capacity for a second product, it gets its own part number and its own approval; it never ships under the approved one. And what I ask in return is three things in writing: a change-notification window, a last-time-buy arrangement, and a named engineer on your side who owns the interface.

14. Sealing, chemistry, and the connector nobody budgets for

Four of the five cells I quote for load cells for IV bag weighing declare IP65. The fifth, the WST103, declares no ingress figure at all, which is not the same as being unsealed — it means the number has to come from me in writing rather than from the page. On an infusion pole that matters less than it does on a urine bag, because the fluid is inside a closed system and the cell is above the wet zone. What matters more is the chemistry, and no ingress standard covers it. Every test behind an IP65 rating uses fresh water. Your ward uses alcohol, chlorhexidine and quaternary ammonium wipes several times a shift, and a swollen silicone seal is a leaking seal. Ask me which elastomer is in the cell before you commit to a disinfectant that your cleaning protocol insists on.

A damp connector is a shunt resistor, and here is how many grams

A leakage path across one arm of the bridge behaves exactly like a shunt calibration resistor, and the only unusual thing about it in this application is that the signal is continuous. On a 1 000 ohm bridge with a one millivolt-per-volt cell, the arithmetic is short and worth memorising.

Twenty-five megohms is not a short circuit. It is a slightly damp connector, or a cable gland that has been wiped four hundred times, or a drop of cleaning fluid that has crept down a strain relief. At that point the cell reports twenty grams too high on a two kilogram range, which is sixty-four percent of a fifteen minute window, and the failure has no signature: it looks like the infusion is running fast. The WST201 is the only cell on this page with any margin here — it declares 5 000 MΩ at 50 V instead of 2 000.

The battery is decided by the bridge, not by the cell count

Most IV monitors I am asked about are pole-mounted with a mains lead, and the ones that are not are the interesting ones. Excitation is the whole battery argument, and the four miniature cells have a quiet advantage: a 1 000 ohm bridge at five volts draws 5.00 milliamps, which is 25.0 milliwatts of self-heating. The WST201’s 406 ohm bridge draws 12.32 milliamps at the same five volts, 61.58 milliwatts, or 2.46 times the heat. If somebody has specified twelve volts for headroom, that same bridge becomes 354.7 milliwatts — fourteen times the miniature cell’s dissipation, sitting inside the housing you are trying to keep isothermal. On a 3 000 milliamp-hour pack, five milliamps buys 600 hours of continuous excitation and the 406 ohm bridge buys 243.6. Duty-cycle the excitation to two seconds in sixty and you are at eighteen thousand hours, which is roughly seven hundred and fifty days, and the battery stops being a design constraint at all.

What I cannot do from my side is the enclosure. I seal the cell; the housing that survives a ward is yours, and the ingress rating of your product is the weakest of every joint in it — the lid gasket, the gland, the display window, the battery door and the cell’s own seal. A monitor rated IP65 on the strength of its load cell alone is not rated at all. Two things I can supply for the parts that are not the sensor: the WST1513 stainless multi-channel junction box if you are summing a cradle and a hook, and the WST1508 or WST1510 transmitter to carry the millivolts up the pole to the display. There is a junction box wiring guide on the site for teams new to four-wire cells. None of that is a medical device. All of it is what makes the medical device’s weighing channel work.

15. Five load cells for IV bag weighing, five different jobs

These are the five cells I would put in front of an infusion monitor builder, and each one carries the limitation that matters as well as the specification. I have kept the comments as blunt as the numbers, because a cell chosen for the wrong reason costs more than a cell chosen badly. Note what is missing from this list: nothing here is a medical device, and nothing here carries a medical certification. What they carry is a millivolt output, a zero coefficient and an overload figure, and those are the three numbers your risk file will actually argue about.

The five, side by side, with the limitation attached to each

There is one column that deserves its own table, because it is the number nobody reads and the number that decides whether your firmware’s zero is stable. The data sheets state it in millivolts per volt, which is not a unit anybody can argue about, so here it is in grams.

Zero balance is not an error, and I do not want it read as one. It is a permitted offset, and every cell here is specified as a ratio to its own output, which means the offset is proportional to the capacity you bought. That is the trap: choosing a 5 kg cell to be safe costs you a 1 579 gram permitted offset, which is one and a half litre bags’ worth of tare. In this application that is tolerable for exactly one reason, and it is the structural advantage I described in section seven — the bag is replaced, and the empty bag is the natural zero. Re-zero on every empty bag and the offset column stops mattering. Skip it once and the offset column is your whole reading.

If you are also building the readout side, the WST1502 indicator and the WST1505 process indicator both give you a stable gross and a rate, which is what your monitor needs from me. If your existing numbers have already gone strange, the damage diagnosis checklist is the fastest path, and the selection guide covers the mechanical part in more depth than I can here. Everything in those three pages applies to a weighing channel on a pole exactly as it applies to a hopper.

16. The ten questions, and the habit that matters more

These are the ten things I ask for on load cells for IV bag weighing, in the order I ask them. They take a customer fifteen minutes, and they have changed my recommended capacity in four of the last six projects — twice upward because of an accident case, twice downward because a tape measure settled the hanging head before I had quoted anything.

The routine that keeps the number honest afterwards is short and cheap. A certified weight hung in the loaded geometry at the start and end of every shift, not in the empty one, because the loaded geometry is where the load path actually changes. The two-condition head test run monthly — the same bag, hung high and then hung low, compared against the ratio in section five — because that is the only cheap early warning that a bypass path has appeared. The dry bag weighed at every delivery, because the tare is a purchased item and it drifts with the supplier. And the one I would insist on: keep the raw, unfiltered log for every alarm, with the nursing notes alongside it.

17. Questions I get asked about load cells for IV bag weighing