A pulse specification is a list of things that must not be in the bag. Stones, soil, straw, other seeds, insect-damaged seed, off-colour seed, splits, broken cotyledons, shrivelled seed. Each of those is removed by a different machine working on a different physical property, and understanding which machine removes what is the difference between a specification a supplier can meet and a specification that simply generates arguments at intake.
This guide walks the line in sequence, explains what each stage can and cannot do, and then deals with the two things buyers most often get wrong: writing tolerances without methods, and ignoring cleaning yield when comparing offers.
Why the sequence matters
Cleaning stages are ordered by how much material they remove and how expensive they are to run. Coarse, cheap separation comes first; fine, expensive separation comes last. Reversing that order does not just cost money, it damages the downstream machines. A stone reaching an optical sorter is a nuisance; a stone reaching a polisher is a maintenance event.
The corollary is that a line missing an early stage cannot compensate with a later one. If a lot arrives with heavy soil contamination and there is no effective pre-clean, the destoner and the optical sorter will both be overloaded, throughput collapses, and the reject stream carries good product with it. That shows up on the buyer’s side as a supplier who quotes a tight specification and then cannot ship on time.
Stage one: reception and pre-cleaning
Reception screening removes straw, pods, stalks, clods and anything grossly oversize or undersize. It is normally a scalper – a coarse screen with an aspiration hood – and it runs at high throughput with low precision. The point is to protect the store, not to make grade. Material that goes into a silo carrying trash and green matter raises the local moisture, and localised moisture is how an otherwise sound lot develops a mould pocket. See pulse cleaning for the terminology.
Aspiration at this stage removes dust and light chaff by air. It is the cheapest separation available and it does a disproportionate amount of the visible work, which is why a lot can look dramatically better after pre-clean while its analytical grade has barely moved.
Stage two: sieving and calibration
Nested sieves separate on size in two dimensions. Oversize goes over the top screen, undersize falls through the bottom, and the traded fraction is what remains between. This is also the stage where calibre grades are created for species where calibre is a commercial parameter – chickpeas above all, but also lentils, beans and peas. Sieve apertures should be quoted against the ISO 3310 test sieve specification, because a sieve aperture without a standard reference is not reproducible between two laboratories. Size grading explains the convention.
Sieving cannot separate two objects of the same size and different density, and it cannot separate two objects of the same size and different colour. Every buyer who has ever asked why a sieved lot still contains stones has run into that limit.
Stage three: gravity separation and destoning
A gravity table fluidises the product bed with air and inclines it, so denser particles walk uphill and lighter particles float downhill. This is the stage that removes stones of the same size as the seed, mud balls, and – importantly – insect-hollowed seed, which is lighter than sound seed of the same dimensions. Pulse destoning is usually a dedicated machine rather than a setting on the gravity table.
Gravity separation is the most under-appreciated stage on the line. It removes defects that no camera can see, because a hollowed seed looks perfect from the outside. A buyer who accepts a line without gravity separation is accepting insect-damaged seed that will pass a visual inspection and fail a cook test.
Stage four: indent cylinders and length separation
Indent cylinders separate on length rather than on width or thickness. They are the tool for removing broken cereal grains, weed seeds and split pulse fragments that are the same width as the sound product. For lentils and peas the indent is often the only machine that meaningfully reduces the split count, because splits and whole seed have similar density and similar colour.
Stage five: optical sorting
Full-colour cameras and, on modern machines, near-infrared and shape recognition reject individual seeds into an air-blast stream. This is where off-colour seed, discoloured seed, residual foreign material and any remaining stone are removed. Optical sorting and pulse colour sorting cover the technology.
Two practical points. First, sensitivity is a trade-off: tightening the reject threshold removes more defect and also removes more good product, so a very tight colour specification directly reduces yield. Second, a single pass is not the same as two passes. Most serious lines run a primary sort and then re-sort the reject stream to recover good product from it, which is what makes a tight specification economically viable.
Stage six: polishing, splitting and dehulling where applicable
Not every line runs these. Polishing improves surface appearance for retail grades. Dehulling and splitting convert whole seed into split product and are discussed in the split and whole comparison. Each additional operation costs yield, which is the central economic fact of pulse processing.
Writing a tolerance that means something
A tolerance has three parts and most contracts contain only one of them.
- The parameter, defined. “Foreign matter” is not defined until the contract says whether other food grains count, whether hulls count and whether dust counts.
- The sample mass. A count of stones per kilogram and a count per hundred grams are different specifications, and the smaller sample has far worse statistical resolution.
- The method. ISO 605 gives methods of test for pulses; naming it converts a number into a contractual figure.
A tolerance without a sample mass and a method is not a specification. It is an expression of hope, and at intake it becomes an argument that neither party can win on the evidence.
Sampling itself deserves a sentence. A composite sample drawn to a defined plan across the lot is the only defensible basis for a result. Sampling from the door of a container measures the door of the container.
Cleaning yield: the number that actually drives cost
Two offers of the same nominal grade can differ substantially in what they cost to bring to a finished specification, because the incoming material differs. The relevant ratio is cleaning yield: finished graded product as a share of the raw lot entering the line.
| Incoming condition | Typical removal load | Effect on finished cost |
|---|---|---|
| Clean, dry, single variety | Light aspiration and sieving only | Lowest; most of the lot survives to grade |
| Weedy or soil-heavy field lot | Heavy pre-clean plus gravity work | Yield loss and reduced line throughput |
| Mixed-variety or mixed-colour lot | Two optical passes | Highest sorting cost and the largest reject stream |
| Insect-active lot | Gravity plus optical, and possibly rejection | Often uneconomic regardless of price |
The reject stream is not worthless. Off-grade and off-colour fractions go to pulse flour, to feed, or to protein extraction, which is why a processor with a flour outlet can quote tighter visual tolerances than one without. The flour and isolate guide covers where those fractions go.
What to ask a supplier
- Which unit operations does the line have, in order? A list of machines is more informative than a list of tolerances.
- Is optical sorting single pass or with reject re-sort?
- Is there gravity separation, and is destoning a dedicated machine?
- What is the typical cleaning yield for the grade being offered?
- Which method is used for each declared parameter, and on what sample mass?
For the framework of defect categories see the Codex pulses standard record, and for how contract forms allocate quality risk between the parties see the GAFTA contract terms record.