Why your charcoal smells sour — and which fault it actually is

This page is a diagnosis, not a pitch. If you work through it and conclude that your current supplier is fine and the fault is in your own warehouse, the page has done its job. Both of the tests below run on charcoal you already own, without us, and the one question in §5 is worth more to you than anything we could sell you.

§1 · What you are actually smelling

Someone opened a carton and something was wrong. The words people reach for vary more than the faults do: sour, fermented, like bread dough, like a struck match, like a bonfire in the room. Those are not five ways of saying the same thing. They are different faults, produced at different points between the carbonising kiln and your warehouse, and they implicate different people — one of whom may be you.

This matters commercially before it matters technically. A brand owner who escalates the wrong fault either absorbs the cost of a genuine supplier defect or fires a competent supplier over something that happened in their own store room. Both are expensive and both are avoidable, because the character of the smell and the moment you notice it are between them enough to tell the cases apart.

So the word you instinctively used is the most useful diagnostic information you have. Start there rather than with the word “quality”, which tells a supplier nothing and invites the answer that this is normal for coconut charcoal.

§2 · The differential

Four odour types, four causes, four different conversations to have. Find the row that matches what you smelled and when you smelled it.

What it smells like	Probable cause	Mechanism	When you notice it	Buyer test
Sour, fermented, bread-like, “rotten wheat”	Ungelatinised starch binder	Starch granules were never fully cooked, survive into the finished cube, and break down when the cube is heated.	On ignition, in the first minutes	Heat one cube alone, with no bakhoor on it. A sour note means the binder.
Sour, and already present in the sealed box	Storage or transit — our assessment, not established fact	Residual moisture and residual starch, plus weeks in a hot container. Organic acids develop inside the sealed carton before anything is lit.	Before ignition, on opening	Open one inner carton from the container centre and one from the door end. A difference between them points to transit.
Acrid, smoky, “burning wood”	Under-carbonisation	Volatile matter was never driven off in the kiln, so it pyrolyses at the burner instead of in the factory.	On ignition, and it persists	Compare against a cube you know to be low in volatile matter, and ask your supplier for the VM figure with its method.
Chemical, sulfurous, “match head”	Ignition accelerant	Nitrate or sulfur compounds off-gassing at light-up, by design rather than by fault.	Immediately at ignition, then fades	Check what you actually bought. This is quick-light contamination or a mislabelled product — a product-type fault, not a batch fault.

The pair that decides who pays is rows one and two. They smell the same to a human nose — both read as sour — and they have opposite responsible parties. Row one is a production fault your supplier owns. Row two happened after the cube left the factory, which in most contracts means it is yours. Everything in §6 exists to separate them.

One caution on row two, because a buyer may confront a supplier on the strength of it. That a pre-ignition sour smell points to storage or transit rather than production is our assessment, not an established finding. The underlying chemistry is standard for starch-containing goods in warm humid storage, but we are not aware of it having been demonstrated for charcoal briquettes specifically, and we have not demonstrated it ourselves. Treat it as the hypothesis worth testing first, not as a verdict to put in an email. The container-position test is how you check it.

§3 · Why ungelatinised starch smells

Tapioca starch does not melt, it gelatinises, and it does so across a window rather than at a point. That window has been measured directly by differential scanning calorimetry: onset at 60.8 °C, peak at 66.7 °C, and conclusion at 77.2 °C. Onset is where the first granules begin to swell. Peak is where most of the population goes. Conclusion is the point past which effectively none of it is left raw.

The consequence for a briquette is direct. If the binder paste never reaches the conclusion temperature throughout — not on average, not at the probe, but throughout — then a fraction of the granules is still raw starch when it goes into the mix. Raw granules bind poorly, which is a strength problem, and they survive drying intact, which is your problem. They are still in the cube when it reaches you. Heat them on a burner and they break down, and what you smell is that breakdown: sour, fermented, faintly like bread.

Two limits on those figures, stated because they change how you should use them. They were measured on isolated starch in excess water, and a binder paste in a charcoal plant is neither isolated nor swimming in water; concentration and everything else in the mix shift the window. And they describe the starch, not the briquette. So they are the reason a temperature target has to exist. They are not a setpoint to hand to a factory, and any supplier who recites them back at you as their specification has misunderstood them in the same direction.

Why we do not publish a binder ratio

The question every buyer asks at this point is what percentage of binder is correct, and we are not going to answer it — not because it is proprietary, but because the answer does not exist in the form the question assumes. Published briquette work uses a wide spread of figures depending on feedstock and press, and two of our own source documents disagree with each other. Naming one number would hand you a figure that was never a standard and invite you to challenge a supplier on it.

The reframe is the whole point of this section: the ratio is not the diagnostic, the preparation temperature is. A correctly gelatinised binder outperforms a raw one at any sensible ratio, and a raw binder is a defect at every ratio. If you take one thing from this page into a supplier conversation, take that rather than a percentage.

§4 · What correct binder preparation looks like

You are unlikely to audit a binder line in person, so what you need is a frame for judging a supplier's description of their own process. Roughly, a controlled process has four movements: the starch is cooked in water into a paste before it meets the charcoal, rather than dry-blended into the mix; the paste is held hot enough, long enough, for the whole mass to pass the conclusion point rather than just the surface; the wetted material is formed; and the formed briquettes are dried slowly and warm for a long time, rather than fast and hot.

For a sense of scale on that last movement, one documented briquette method dries at 105 °C for 15 hours. Read that as an order of magnitude and nothing more. It is a single documented approach rather than an industry standard, it was recorded on hardwood char rather than coconut shell, and we are not presenting it as our specification. We carry the drying schedule from that source and deliberately leave its binder ratio behind, for the reason in §3.

What this gives you is a way to hear an answer. A supplier who describes a sequence with temperatures and holds in it is describing a process. A supplier who answers with adjectives — premium binder, food-grade starch, advanced formula — is describing a brochure, and has told you that nobody there is measuring the variable that matters.

§5 · The question to ask your supplier

Do you use native or pregelatinised tapioca starch, and what temperature is your binder paste?

That is one message. It can go out on WhatsApp this afternoon and it does not require you to be a chemist, because the value is in how the answer partitions suppliers rather than in the answer itself.

Native starch arrives raw and has to be cooked to bind at all, which is where the whole failure mode in §3 lives. Pregelatinised starch has already been cooked and dried before it is sold; it disperses in cold water and needs no cooking step in the factory. So a plant using pregelatinised starch cannot produce the raw-granule fault, because there are no raw granules to survive. The defect is only available to plants using native starch. One question, and you know whether the fault you are chasing is even possible at that factory.

How to read what comes back

“Pregelatinised.” Row one of the differential is off the table for that supplier. If your cube still smells sour, you are in row two or the smell is not the coal at all — go to §6.
“Native, and our paste runs at this temperature for this long.” A real answer. They are measuring the variable that matters. Ask for the hold time if they only gave you a temperature, because a temperature without a duration does not establish that the whole mass got there.
A ratio instead of a temperature. They answered a different question, most likely because the ratio is the number they have. Ask again.
No answer, or “that is our proprietary formula.” Binder type is not a formula, it is a purchase order line. A supplier who will not say which of two commodity inputs they buy is usually a trader who does not know.

Our own answer

Ours is native tapioca starch, with the paste held at 80 °C for 30 minutes.

That is the question above, answered in the form it asks for: a type, a temperature and a hold time, in one sentence. All three are published with the rest of what we measure on the specification page, which is where they change if they ever change.

Read against §3, the temperature is the part that matters. It is above the conclusion point of the gelatinisation window, and the hold is long enough for the mass to reach it rather than for the surface to, so the mechanism this page describes is not one we are running. Native starch makes the fault available in principle — that is what the first bullet above means — and the paste temperature is what takes it off the table.

The limit of that claim, since this page is about reading other people's claims: the gelatinisation figures in §3 were measured on starch in excess water rather than in a charcoal slurry. What a committed paste temperature establishes is that the process sits on the right side of the transition. It is not a count of granules in a batch, and the test in §6 is still the one that settles what a particular cube is doing.

§6 · Two tests you can run today

Both of these run on charcoal you already have. The first tells you whether the coal is the culprit at all. The second tells you which of rows one and two you are in.

The coal-alone test

Light one cube in your usual burner with nothing on it — no bakhoor, no chip, no oil. Let it come up to a steady burn and smell it directly.

If the coal alone is clean, the coal is not your problem, and you should stop investigating it. Plenty of blended bakhoor contains synthetic aromatics that read as chemical or sour when they are heated, and the coal is the visible object in the burner so it collects the blame. We would rather say that plainly than sell you a replacement for a part that was working. If the coal alone is not clean, you have confirmed the coal is involved and the character of the smell sends you to a row in §2.

The sealed-jar test — our proposed method

This one is ours. It is not an industry standard and it has not been validated against an instrument; we are publishing it because it is cheap, it is falsifiable, and it separates two things a nose cannot separate on an open pallet.

Take three cubes straight from a carton. Do not handle them more than you must and do not leave them out first.
Seal them in a clean, dry glass jar at room temperature. Glass because plastic contributes a smell of its own; dry because you are testing the cube, not the jar.
Leave it 24 hours, out of direct sun.
Open it and smell the headspace immediately, before the air in the jar mixes with the room.

The reasoning is that volatiles which are far too dilute to detect coming off a pallet will concentrate in a sealed headspace overnight. That is our inference and it is the weak point of the method — see the ledger in §9.

Sour headspace, before any heat. Something is present in the product as it sits. That is row two: our assessment is storage or transit, and the container-position test is your next step. Run the jar test on cartons from the container centre and from the door end and compare — a gradient supports transit, uniformity across the container points back at production.

Clean headspace, but sour on ignition. Nothing volatile is sitting in the cube; the smell is being produced by heating. That is row one, the binder, and §5 is the message to send.

§7 · What cannot be fixed

If the fault is the binder, the container cannot be rescued. Airing it out will not work, nor will drying it, resting it, re-packing it into fresh cartons, or storing it somewhere better. The raw starch is inside the briquette and it is structural. It was baked into the product before it was boxed, and the only thing that happens while you wait is that you lose a season.

We state that flatly because the most common failure we see is not a misdiagnosis, it is weeks spent airing out a container that was never going to improve, followed by an escalation to the supplier that arrives too late to be commercially useful. Once the coal-alone test and the jar test put you in row one, the decision in front of you is commercial, not remedial: what you can recover from the supplier, and what you ship in the meantime.

§8 · Questions buyers actually ask

My supplier says a sour smell is normal for coconut charcoal. Is it?

No. Normal is a coal that contributes nothing detectable of its own when it is lit with nothing on it. Run the coal-alone test in §6 and you will have a specific claim instead of an impression. What the supplier probably means is that it is normal for their process, which is a different statement and a more useful one — it says the fault is systematic rather than a bad batch.

Will airing out the container help?

Not if it is the binder. See §7 — the starch is inside the briquette and it does not leave. Airing can help a little where the problem is surface moisture picked up in storage, which is one reason to run the jar test before you commit weeks to it.

It smells fine in the carton but sour the moment I light it. Which row is that?

Row one. Nothing volatile is sitting in the cube at room temperature, so the smell is being produced by heating, which is what raw starch breaking down looks like. Send the question in §5.

Only some of the cartons smell. What does that mean?

That pattern is the reason the container-position test exists. A production fault is usually uniform across a run, because every cube went through the same mixer on the same day. A gradient across the container — worse in the centre, better at the doors, or the reverse — points instead at what happened in transit. In our assessment that favours row two, though it is an assessment rather than a finding, and cartons from different production dates in one container can mimic it.

Is this caused by too much binder?

Probably not, and the question quietly assumes the ratio is the control variable. It is not: a correctly cooked binder is fine at a generous ratio and a raw one is a defect at a stingy one. Ask about the paste temperature instead, which is the question that distinguishes suppliers. §3 explains why we decline to publish a ratio figure at all.

Reviewed by Mohamed Noor, Director — 23 September 2026

§9 · What supports each claim on this page

Two of the claims doing real work here are not measurements, and one of them is the one a buyer is most likely to act on. So rather than level everything up to sound equally certain, here is the basis for each and the form in which we are willing to state it.

Claim	What supports it	How we state it
Native tapioca gelatinises between 60.8 °C and 77.2 °C.	Direct instrumental measurement by differential scanning calorimetry.	Stated as fact, with the figures and their transfer limits.
Below the conclusion temperature, a fraction of granules stays raw.	The same measurement — it is what a conclusion temperature means.	Stated as fact.
A plant using pregelatinised starch cannot produce this fault.	Established process distinction: the starch is cooked before it is sold.	Stated as fact.
Raw starch surviving into a cube produces a sour note on heating.	Our production research, consistent with what switching buyers describe.	Stated as fact, with the mechanism shown in §3.
A sour smell already present in the sealed box points to storage or transit rather than production.	Standard chemistry for starch-containing goods in warm storage, but not demonstrated for charcoal briquettes, including by us.	Our hypothesis. Labelled as ours in §2 and §6, and the container-position test is how you check it rather than take our word.
Container position changes a carton's thermal history in transit.	Documented behaviour of loaded containers in transit.	Stated as fact.
Under-carbonisation gives an acrid note rather than a sour one.	Our production research, plus the volatile-matter mechanism in row three.	Stated as fact.
Our own paste is held past the conclusion point, so this fault is not one we are running.	Our own process, committed as a standard rather than a range we run. The gelatinisation figures it is read against were measured on starch in excess water rather than in a charcoal slurry.	Stated as fact about our process, with that transfer limit stated beside it in §5 rather than left for a reader to find.
A sealed jar concentrates volatiles enough to make them detectable.	Our reasoning about headspace. Not validated against an instrument.	Our proposed method, labelled as such in §6.

What we do not publish: a binder ratio, for the reason in §3. Our own type, paste temperature and hold time are published in full, though not here — they are on the specification page with everything else we measure, and §5 renders them. Where figures appear on this site at all, the governing values for your shipment are the ones on the certificate of analysis issued for your lot rather than the indicative values on a web page.

§10 · If you want a cube to compare against

If the differential points at your own storage, you do not need anything from us. Fix the storage, re-run the jar test on the next container, and keep the supplier you have. If it points at your supplier's binder and they cannot answer the question in §5, then a reference cube is useful, because “this one does not do it” is a much stronger position in a supplier conversation than “ours smells wrong”.

We send 1 kg of our bakhoor-grade cube for that purpose, on the terms set out on the sample request page. Run the coal-alone test and the jar test on ours and on yours, in the same room on the same day, and you will have a comparison rather than an impression. We manufacture under your brand at container scale, with a minimum order of 1 × 20 ft FCL, which is explained alongside how an order works end to end. If your programme is smaller than that we are the wrong factory — the diagnosis above is still yours to use.

How we test for this in production, if you want to see the method rather than the claim.
Other checks you can run on a carton you already hold.
What changes if you have it made under your own brand, for buyers who have diagnosed this and want out.
Request the sample, and see what arrives with it.