Ramsey Micro-Tech 2104 and 9104: loss-in-weight feeder control, alarms and calibration

Loss-in-weight feeders are a smaller part of our work than belt weighers, but we can help with them, including the Micro-Tech 2104 and 9104 described here.

What is a Micro-Tech 2104 or 9104, and how does loss-in-weight differ from a belt weigher?

They are the same product two generations apart. The 2104 is the loss-in-weight member of the Micro-Tech 2000 family, styled by the maker as Micro-Tech 2000 Model 2104, and the documentation we work from covers the late 1990s through to about 2001. The 9104 is its 9000-series replacement, documented around 2013 and 2014, with the same measuring idea rebuilt on newer hardware.

The difference from a belt weigher is the measurement itself. A belt weigher multiplies a load signal by a speed signal to get a rate. A loss-in-weight controller has one input that matters, the hopper weight, and it derives the rate by subtracting one weighing from the next and dividing by the time between them. That single fact drives everything else on this page: the rate is a small difference between two large numbers, so anything that disturbs the hopper turns straight into rate noise, and the controller carries a stack of averaging and filtering to survive it.

A few consequences are worth knowing before you touch one.

There is no frame geometry to get wrong. A belt scale setup wants pivot distances, idler spacing, belt length and a scale code. A loss-in-weight setup wants the capacity, the number of load cells, each cell’s capacity, its sensitivity in millivolts per volt, and its measured resistance. That is a far shorter list, and it means the missing data plate problem that plagues old Ramsey weigh frames does not exist here.

The imperial units trap does. Both generations came out of a United States design house, and their manuals and factory setup work in imperial units, asking for capacity in tons per hour and load cell capacity in pounds. A metric mode exists on both, and we set them up in metric. A mixed entry, one figure left in the other family, reads as a feeder that is quietly and consistently wrong rather than as an error (more on that trap across the Micro-Tech range).

One box can run two feeders. Both generations control up to two completely independent loss-in-weight feeders, each with its own weighing channel and its own control loop, and each takes up to six 350 ohm load cells. That is why most alarms carry a scale number when two feeders are configured and drop it when only one is, so the same condition reads slightly differently on a one-feeder and a two-feeder unit.

It is not only for loss in weight. Both are documented for gain-in-weight feeders, where the hopper is filling rather than emptying, and for extraction from a silo, ratio control between several feeders, and blending. If your unit sits in a blend, the setpoint it is holding may be somebody else’s number multiplied by a ratio.

For the belt fed cousin that controls to a setpoint with a real belt and speed sensor underneath it, see Micro-Tech 9105 and 9205 weighfeeder control. For the passive belt scale integrator of the same 9000 generation, see Micro-Tech 9101 and 9201.

What do the common 2104 and 9104 alarms mean?

Before any of them, check how yours is configured. Every alarm on both generations can be set to warn, to shut the feeder down, or to be ignored entirely, and the 9104 adds a self clearing option that resets itself once the condition goes away. The same message can therefore be a note on one plant and a stopped feeder on the next, and an alarm set to be ignored is not evidence that a condition never happened.

Load cell failure. The controller no longer trusts the weight signal on a channel, and on a loss-in-weight feeder that costs the rate, the total and the control all at once. Check the junction box for moisture first, then the cabling and the terminals, then compare the millivolts at the box against the millivolts at the controller. They should agree, and they should rise as weight is added. Not a reset item (load cell and junction box faults).

Load cell imbalance. A 9104 alarm with no equivalent in the 2104 list, though the condition bites both. The cells carrying the hopper are not sharing its weight the way they should. On a hopper that nearly always means something is touching it: a rigid conduit, a hose or flexible connection that has gone stiff, a bolt run down onto a flexure, or material bridged hard against the structure. Find what is taking load before anyone touches the calibration.

Refill timeout. The refill did not finish inside the time allowed. It gets its own section below, because it is the alarm most particular to this product.

Control deviation, high and high high, positive and negative. The actual rate has sat too far from setpoint for longer than the configured delay, in one direction or the other, with a second more serious threshold above the first. This is the controller reporting the process, not itself. Look for a bridged or rat-holed hopper, a blocked outlet, a screw or valve at the end of its travel, or a setpoint the feeder cannot physically deliver. If the drive is clearly answering the output and the rate still will not come to setpoint, the measurement is the suspect.

High and low weight. The hopper contents crossed a threshold set as a percentage of capacity. Low weight usually means the refill is not keeping up, or is not starting at all. High weight often means the refill is not stopping, or the hopper was hand filled past where the controller expects. Read both as level alarms rather than weighing faults.

High and low rate. The delivered rate crossed a threshold set as a percentage of rate capacity, again with its own delay. These are site configured process alarms, so establish whether the plant condition is real before suspecting the weighing. A low rate alarm with material clearly flowing points back at the measurement.

Warm start, cold start, and power lost during a calibration. A warm start says power went away and came back. A cold start says the setup data did not survive it, which on these controllers means the memory backup battery or the board, and it is always a call out because it is a recommission of the weighing and the control loop together. The documentation is blunt that removing the battery loses everything, which is exactly why those settings need to exist somewhere other than inside the box. A calibration interrupted by a power loss cannot be trusted either: run it again on a settled, empty hopper and compare against the last record.

Calibration time elapsed. A timer the site sets itself, reminding somebody that the zero and span have not been checked in a long while. Nothing is broken, and it is worth leaving switched on.

Math error. A divide by zero or an overflow inside the calculation, meaning a setup value is not physically possible. On these units the usual cause is a units mix up or a capacity figure that does not match the hardware. Treat it as a call out, because the rate and the total are wrong while it stands.

Communication error, printer error, and a fieldbus watchdog. A timeout or handshake failure on a serial link, a printer disconnected or out of paper, or a Profibus interface that has seen no valid exchange inside its watchdog interval. On a feeder this matters more than it does on a belt weigher, because the setpoint itself may be arriving over that link.

Overflow totaliser. The pulse output driving a remote mechanical counter has run past what it can represent, or the master total has rolled over. Check the actual rate against what the pulse divider was set up for.

Why do the 2104 and 9104 alarm numbers not match?

Because the numbering was rebuilt between the two generations, and it never matched the belt scale integrators either. This is the honest answer to the search that brings most people here, so it is worth being plain about it: a Micro-Tech alarm number is only meaningful against the exact model and software version that produced it.

Three separate things go wrong if you assume otherwise.

The two generations disagree with each other. Comparing the 2104’s alarm list against the 9104’s, only the first couple of entries kept their positions. Most of the rest shifted down the list by a place or two, and at least one moved further than that, so a number read off an old unit lands on the wrong condition when you look it up in the newer manual. The shift is not a constant offset you can correct for.

The loss-in-weight controller does not share numbering with the belt scale integrator of its own generation. The 2000-series belt scale integrator numbers its alarms in its own sequence, and that sequence includes conditions a hopper feeder has no use for, such as speed sensor and belt speed alarms, while leaving out the refill and deviation alarms a feeder needs. The result is that the same number means different things on two boxes sitting in the same substation, from the same maker, in the same generation.

Neither manual’s numbered list is the complete alarm set. The 9104’s alarm configuration table skips a run of positions that the same manual’s communications register map clearly shows are in use, including external alarm inputs, a hardware configuration changed flag for each expansion slot, and a batch deviation alarm. The 2104’s own description of what its diagnostics detect names conditions its numbered list does not carry either. So a number missing from the table is not proof the condition does not exist on your unit.

What to do instead is simple. Trust the wording on your own screen, write it down exactly as it appears including any scale number after it, and note the software version the unit reports and the revision printed on your manual. Quote those three things when you ring, and the number becomes unnecessary.

What does a Micro-Tech loss-in-weight calibration involve?

Two measurements, in the same order as any scale: a zero on an empty hopper, then a span against a known load. What is different is what is absent. There is no belt to run empty, no revolution counting, and no calibration chain, because there is nothing to lay one along.

Safety comes first, and it is a real consideration on these units rather than a formality. The controller’s output drives a feeder, so it can start machinery, and the documentation warns in as many words that equipment may start when certain setup and test functions are used. Isolate and lock out the feeder drive and the refill device to the site procedure before anyone begins. The field mount enclosures on both generations carry mains voltages inside, so opening one is work for trained, authorised people only.

Zero. The hopper is emptied and left to settle, and the controller averages the weight for about a minute. It can be cut short from the keypad, which is a temptation worth resisting on a hopper that is still swinging. During the routine the unit reads weight at a finer resolution than it does in normal running, which is why a zero result can look noisier than the running display ever does. An empty hopper is genuinely empty: material stuck to the walls or bridged above the outlet is being learnt as part of the tare.

Span. Two methods are offered, and only two. Test weights hang a known load on the hopper and put real force through the cells, the mountings and the structure. The electronic reference, R-Cal, unbalances the load cell bridge with a precision resistor so the controller sees a simulated load with nothing on the hopper at all, which proves the cells, the wiring, the converter and the constants, and proves nothing whatever about the mechanics. The documentation’s own preference is the weights method, with a fallback of loading the hopper with a known weight of material where the capacity makes weights impractical (which method suits which scale, test weights, chains and billets).

Tie the electronic reference to the real one. Once a weights span has established a proven number, the electronic reference is factored to agree with it. Skip that and later electronic checks are being judged against a figure nobody ever proved, which is the most common reason a hopper passes its checks for years while reading wrong.

The numbers, and the discipline. On a first setup, an electronic span error beyond roughly three quarters of a per cent means something needs finding before you go on. Routinely the bar is tighter, of the order of half a per cent. Accept a new zero or span on the initial setup only. After that, run the check, record the error and decline the change, because the repeat tests are measuring repeatability, and a hopper that cannot repeat has a problem no adjustment will fix. Run several zeroes and several spans, not one of each (zero, span and repeatability explained).

Correction across the range. Both generations can trim the weight reading at several points rather than relying on one straight line, which matters on a hopper that is honest when full and out when nearly empty, or the reverse.

What a calibration does not do. It sets the weighing right. It does not tune the control loop, and the two are separate jobs with separate evidence. A loss-in-weight feeder can be perfectly calibrated as a static scale and still feed badly, because the number the plant cares about comes from the rate of change of that weight, not from the weight itself. On our visits we bring test weights for the hopper when a span needs them, set the weighing side right, and hand loop tuning back to the site’s control people.

On frequency, the manufacturer’s own guidance for a scale whose weights carry commercial value is a zero every day and a span check every week, with a weekly look at the mechanical installation for material building up around the load cells. Most sites run less often than that and lean on trend records instead (what happens during a calibration visit).

What happens during a refill, and what does a refill timeout mean?

During a refill the controller stops measuring and starts guessing, and understanding that is most of what separates a loss-in-weight feeder from anything else on site.

Two weight thresholds run the cycle. When the hopper falls below the low one, a refill output closes and starts the filling device. From that moment the weight is going up, so the rate can no longer be derived from it, and the controller switches to volumetric operation: it holds the control output where it was and lets the feeder run open loop, trusting that a screw turning at the same speed delivers the same rate it delivered a moment ago. Where the cascade mode is enabled the output still tracks setpoint changes, but nothing is being measured. When the weight passes the high threshold the refill output opens, a settling delay runs, and normal control and rate calculation resume.

Two things follow from that, and both surprise people.

The tonnes booked during a refill are estimated, not weighed. Outside a refill the total is the difference between successive weighings, which is as good as a static weighing. Inside one it is the setpoint integrated over time. That is why the documentation is firm that a system should be sized so the refill takes only a tiny fraction of the running time, of the order of one per cent. A feeder that spends a quarter of its life refilling has a total nobody should lean on.

Material density changes during the fill. A hopper filling up compresses what is underneath, so the same screw speed delivers a different rate at the end of a refill than at the start. Both generations offer compensation for it, from a simple correction applied to the output at the end of the fill through to learning the relationship between weight and output across the fill and replaying it. A feeder that consistently runs heavy or light for a minute after every refill and then settles is showing you this, not a calibration fault.

Refill timeout means the refill did not complete inside the time allowed. The order to check it in:

• Is the filling device actually running? A failed conveyor, a stuck rotary valve, an unopened slide gate.
• Is there material above it to deliver? A bridged or rat-holed silo is the commonest cause by a distance.
• Is the refill input wired and being made? Where that input is used, the controller waits for it, and a failed contact or a broken wire back from the filling machine leaves the controller waiting for a confirmation that never arrives.
• Is the timeout realistic? A time set for a fast fill will alarm every cycle once the supply degrades, and it is the sort of setting that gets shortened during commissioning and never revisited.
• Did the high threshold ever get reached? A drifting zero can leave the controller unable to see the hopper as full.

An occasional timeout on a struggling supply is a plant problem. One every cycle is a call out, because the feeder is spending its life in volumetric mode, and in volumetric mode nothing is being weighed.

Why does the feed rate wander on a loss-in-weight feeder?

Because the rate is the difference between two weighings taken a second or two apart, and on a hopper holding a tonne of material that difference is a tiny number sitting on top of a large one. Anything that moves the hopper, or takes some of its weight, corrupts the rate long before it is big enough to show up as a weight error.

The physical causes, roughly in the order we would look at them:

Something is touching the hopper. A conduit run rigidly onto it, a hose or bellows that has hardened, a cable tie-wrapped to the structure, a walkway or handrail bolted across, a lagging strap. Every one of them takes a share of the weight that changes with temperature, pressure and vibration. A person leaning on the hopper does the same thing for as long as they lean.

Vibration. A screen, a compressor, a nearby crusher, or the hopper’s own vibrator or air cannon fitted to stop bridging. The controller sees each shake as weight arriving and leaving.

Bridging and collapse. Material that hangs up and then drops does not lose weight smoothly, so the rate reads low, then briefly reads high, and the loop chases both. A feeder with a chronic deviation alarm and a vibrator working hard is usually telling you about the material, not the electronics.

Draught and pressure. An extraction system, a dust filter, or a sealed hopper under even slight pressure will pull on the structure. It is one of the few faults that appears and disappears with a fan starting.

Mounting and sizing. A structure that is not rigid enough under the hopper, or a load cell set so large that a full hopper only uses a fraction of its range, both throw away the resolution the rate calculation depends on. So does an extremely long discharge time: the documentation itself flags anything over about three hours as a case where small vibrations start to matter.

The controller has three defences against all this, and each buys steadiness at the cost of speed. It averages a configurable number of weighings before using one, at a configurable interval, and the two multiplied together set how long the rate takes to catch up with a genuine change. It can apply a further damping filter to the calculated rate. And it offers a jump cut-off that watches for a stable loop, then freezes the rate at setpoint for a set delay whenever a disturbance kicks it outside a band, releasing it afterwards; the documentation’s own worked example is a vibrator that periodically shakes the hopper. Every one of those settings hides a symptom. Wound too far, they make the feeder blind to a real change for as long as they run.

The useful diagnostic separates measurement from control, and it takes five minutes. Put the loop in manual and hold the output at a fixed value. The switch between manual and automatic is bumpless on both generations, so the plant will not lurch. Then watch. If the rate is jumping with the output frozen, the problem is on the weighing side: mounting, vibration, something touching the hopper, filtering set too fast for the installation. If the weight trace is smooth and the loop still hunts once it is back in automatic, the problem is the tuning or the process it is fighting (common faults across the range).

What load cell and wiring points matter on a 2104 or 9104?

The same three that matter on any Micro-Tech, plus one trap peculiar to the change of generations.

Excitation is not the same voltage on both. The 2104 runs its load cells at around ten volts and the 9104 at around five. Measure a 9104 expecting the older figure and you will condemn a healthy board. Check which unit you are standing in front of before deciding the excitation is wrong.

Cable length decides the wiring, and the jumpers must agree with it. Both generations take up to six 350 ohm cells in parallel per channel. Four-wire wiring is good for a short run, of the order of sixty metres; six-wire with sense leads carries several hundred. The sense jumpers inside the controller have to match what is actually wired in the field, and a mismatch reads as a slow calibration error that moves with temperature rather than as a fault. The detail is the same as on the belt scale integrators of the same family, and it is covered in Micro-Tech 9101 and 9201 and the belt weigher load cells guide.

The setup asks for measured resistance, so measure it. Both generations want each cell’s signal resistance entered during setup. Taking that reading properly is a free check on the cells and the junction box, and a cell reading well away from its nameplate figure has usually already failed.

Millivolts should rise with weight, and should agree end to end. The reading at the junction box and the reading at the controller terminals should match. If they do not, the cable or a terminal is the problem, not the cell and not the box. If they match but the weight is wrong, the problem is in the setup data or the mechanics.

The memory backup battery holds everything. It is a lithium cell on the motherboard, and both manuals say plainly that removing it loses the configuration, the calibration and the control settings together. It is the single component most likely to turn a working feeder into a recommission, and it fails silently until the next power outage.

How does a 2104 or 9104 talk to the plant?

Through more paths than a belt weigher needs, because a feeder controller receives instructions as well as reporting results.

The analogue side carries the control output to the drive plus rate, net weight and gross weight, each with its own damping and delay, and takes analogue inputs for a remote setpoint or a moisture correction. The 2104 offers two current outputs as standard and up to four with an extra board; the 9104 arranges the same idea across its own expansion slots.

The digital side is where the feeder-specific signals live, and they are the ones worth tracing when a feeder misbehaves without an alarm:

• Refill output and refill input. The output starts the filling device; the input, where used, is the filling device confirming it is running. A refill input assigned but never made leaves the controller waiting.
• Feeder running and regulation interlock. The 9104 documentation states both contacts are required. Between them they tell the controller whether the feeder is turning and whether it is allowed to control. Get one wrong and the controller either keeps controlling into a stopped feeder or refuses to take control of a running one.
• Ready. On both generations this output is true only when the unit is calibrated, has no failure alarm, no shutdown condition and no calibration in progress. If the plant is interlocked on ready, calibrating the feeder will drop it, which needs planning for rather than discovering.
• Local and remote, automatic and manual, alarm and shutdown summaries, a hardware fault output, increase and decrease contacts where the control element is stepped rather than analogue, and a totaliser pulse for a remote counter.

On the data side the two generations are a decade apart in exactly the way you would expect. The 2104 offers RS-232, RS-485 multidrop and a passive current loop, with Allen-Bradley Remote I/O and Profibus DP as boards, plus a serial printer that can print totals, the setup, alarms and an audit trail on command, at set times of day or at intervals. The 9104 keeps two serial ports and Profibus DP, drops the older Allen-Bradley remote link in favour of DF1 and Modbus RTU, and adds Ethernet on the motherboard carrying Modbus TCP and EtherNet/IP with an embedded web server, and a USB port that data can be downloaded to.

Both are slave devices on every one of those links. They answer a master and never start a conversation, so a link that has gone quiet is a question for the far end first (getting weighing data into your PLC or SCADA).

Looking for the Micro-Tech 2104 or 9104 manual?

Both books are Thermo Fisher’s copyright, published first under the Ramsey name, and the copy you want sits with the maker or with whoever supplied the feeder; we hold neither here to hand out. Before you download anything, be clear about what you are matching against. One digit separates the loss-in-weight books from the 9101 belt scale and 9105 weighfeeder books of the same family, the 2104 and the 9104 number their alarms differently from each other, and the wording shifted again between revisions, so the document reference and revision letter on the cover and the software version on your screen are what decide whether a book describes your feeder at all. We do not publish access codes for these or any other controller either, and nor should anyone else (locked out of your integrator?).

Rather than chase it, send us the model, the nameplate details and the software version the screen reports, and we will pin down which book covers your generation and revision so you can ask the right people for the right one. Our technicians work from the full documentation for the equipment we service, and we keep a configuration record for the units we look after, which counts for a lot on a controller whose backup battery can take the whole setup with it. Use contact us, ring 1300 101 666, or email admin@accurateindustries.com.au.

When should you call us rather than keep resetting it?

Call when the alarm comes straight back after a reset, because that is the controller telling you the condition never went away. Call sooner on a feeder than on a belt weigher: a belt weigher that is wrong misreports the tonnes, while a feeder that is wrong actually feeds the plant at the wrong rate, and downstream that costs product rather than paperwork.

The clear cases are a load cell fault that survives a dry junction box, a load cell imbalance, a refill timeout on every cycle, a deviation alarm with the drive plainly answering the output, a zero that will not repeat after the hopper has been properly emptied and left to settle, a cold start, and anything at all that means opening the enclosure. Add one that is particular to this product: a feeder whose booked total no longer agrees with what the downstream process actually consumed, even though every alarm is clear.

We service Thermo Ramsey equipment and Micro-Tech controllers across Australia, on every brand a site happens to run, and we can look at the weighing, the wiring and the controller on the same visit. If you have a loss-in-weight feeder giving you trouble, talk to us and describe what the screen says (booking a service).

Frequently asked questions