State of Minnesota, Respondent, vs. Dwight Thomas Traxler, Appellant.

may not be cited except as provided by

Minn. Stat. § 480 A. 08, subd. 3 (1996).

 STATE OF MINNESOTA

 IN COURT OF APPEALS

 C8-96-2433

State of Minnesota,

Respondent,

vs.

Dwight Thomas Traxler,

Appellant.

 Filed September 16, 1997

 Reversed

 Peterson, Judge

Scott County District Court

File No. 9516622

Hubert H. Humphrey III, Attorney General, 1400 NCL Tower, 445 Minnesota Street, St. Paul, MN 55101 (for respondent)

Thomas J. Harbinson, Scott County Attorney, Susan K. McNellis, Assistant County Attorney, 428 South Holmes Street, Shakopee, MN 55379 (for respondent)

John M. Stuart, State Public Defender, Stephen K. Warch, Assistant Public Defender, 2000 Metropolitan Centre, 333 South Seventh Street, Minneapolis, MN 55402 (for appellant)

Considered and decided by Amundson, Presiding Judge, Norton, Judge, and Peterson, Judge.

 U N P U B L I S H E D O P I N I O N

 PETERSON, Judge

Dwight Thomas Traxler challenges the sufficiency of the evidence to support his conviction for first-degree sale of a controlled substance. We reverse.

  FACTS

While conducting surveillance early one morning, Southwest Metro Task Force agent Roger Roatch observed appellant Dwight Thomas Traxler leave the property under surveillance in a pickup truck. Roatch called for another police unit to make a traffic stop on the truck. During a search of the truck following the traffic stop, Roatch found two white bottles containing ephedrine tablets, drug paraphernalia, a document listing items that Roatch knew were used in the production of methamphetamine, a propane bottle and torch head, and a bag containing 43 coffee filters.

Roatch obtained a warrant to search a stall in a five-stall garage located on the property where he had observed Traxler. In the garage stall, Roatch and Drug Enforcement Agency (DEA) agent Richard Ripley found many items associated with the production and distribution of methamphetamine. Those items included a large bottle of red phosphorus, lyes, rubber tubing, stoppers, glass tubes, a large electric frying pan, a square frying pan with bottles in it, a jar containing a liquid byproduct of the methamphetamine production process, funnels, unused coffee filters, scales, nine empty ephedrine bottles, and coffee filters with a red residue on them. Ripley testified that the only methamphetamine found during the search that could be characterized as finished product was trace amounts on the coffee filters.

Five samples from the liquids found in the garage stall, the coffee filters containing the red residue, and the coffee filters found in the truck were sent to the DEA lab in Chicago for analysis. Gerald Skowronski, a forensic chemist for the DEA laboratory in Chicago, testified that producing methamphetamine generally is a four-step process. First, ephedrine is extracted from tablets containing ephedrine. Second, ephedrine is combined with iodine and red phosphorous, and the mixture is cooked. Third, unreacted ephedrine, red phosphorous, and iodine are removed. The red phosphorous is filtered out, and a liquid remains. A strong base, such as lye, and an organic solvent are added to the liquid. The base and the solvent are emissible liquids, which separate into two layers when poured together. The unreacted materials dissolve in the base, and the methamphetamine dissolves in the solvent. Fourth, liquid methamphetamine is converted into solid form by adding HCL gas or a gas made from rock salt and sulfuric acid.

To extract ephedrine from tablets in the first step of the process, the tablets are ground into a fine powder and mixed with a solvent, such as water or isopropyl alcohol. The ephedrine dissolves in the solvent, but starch that acts as a binder in the tablets does not dissolve. When the mixture is poured through a filter, the dissolved ephedrine passes through the filter in the liquid solvent and the starch is trapped in the filter. At that point, the solvent is evaporated, leaving solid ephedrine. The solid ephedrine can be cleaned up by mixing it with a solvent, such as acetone, and pouring the mixture through a filter. Skowronski believed that the 43 coffee filters found in Traxler's pickup were used in this process of cleaning up the ephedrine.

Skowronski testified that he visually inspected 42 of the coffee filters found in the truck. He estimated that about two-thirds of the filters had ephedrine residue on them. Skowronski tested one of the filters and found ephedrine hydrochloride and a small amount of methamphetamine on it. Skowronski then scraped the residue from each remaining filter paper that had residue on it, ground the collected residue together, and ran tests that showed the mixture contained ephedrine hydrochloride with a purity of 80% or greater. Tests also showed methamphetamine in the mixture.

Based on his knowledge regarding methamphetamine laboratory operations, Skowronski assumed that the filters had been filled between one-third and two-thirds full when used to filter the ephedrine mixture. Skowronski filled one of the coffee filters one-third full of ephedrine hydrochloride and calculated the amount of ephedrine in the filter to be five grams. He also filled one filter two-thirds full of ephedrine hydrochloride and calculated the amount of ephedrine in the filter to be 10 grams.

In calculating the amount of methamphetamine Traxler's laboratory was capable of producing, because more than one-half of the coffee filters contained residue, Skowronski assumed that 22 of the 42 coffee filters had been used as filters. Skowronski multiplied 22 x 5 to determine that if each of the filters were filled one-third full of ephedrine hydrochloride, the total amount of ephedrine poured into all of the filters was 110 grams. Skowronski multiplied 22 x 10 to determine that if each of the filters were filled two-thirds full, the total amount of ephedrine was 220 grams. Skowronski explained that one gram of ephedrine produces .92 grams of methamphetamine. Thus, 110 grams of ephedrine would produce 101 grams of methamphetamine, and 220 grams of ephedrine would produce 202 grams of methamphetamine. Skowronski testified that in his opinion, to a reasonable degree of scientific certainty, Traxler's lab was capable of producing between 100 and 200 grams of methamphetamine.

  D E C I S I O N

When the sufficiency of the evidence is challenged, this court must review

the record to determine whether the evidence, when viewed in a light most favorable to the conviction, was sufficient to permit the jurors to reach the verdict which they did.

 State v. Webb, 440 N.W.2d 426, 430 (Minn. 1989). This court must assume "the jury believed the state's witnesses and disbelieved any evidence to the contrary." State v. Moore, 438 N.W.2d 101, 108 (Minn. 1989).

Where a conviction is based on circumstantial evidence, the verdict will be sustained on appeal when the reasonable inferences from such evidence are consistent only with defendant's guilt and inconsistent with any rational hypothesis other than guilt.

 State v. Alton, 432 N.W.2d 754, 756 (Minn. 1988).

A person is guilty of a first-degree controlled substance crime if

on one or more occasions within a 90-day period, the person unlawfully sells one or more mixtures of a total weight of 50 grams or more containing methamphetamine.

Minn. Stat. § 152.021, subd. 1(3) (1994). The definition of sell includes manufacturing. Minn. Stat. § 152.01, subd. 15a (1994).

Traxler argues that the state failed to prove that he manufactured 50 or more grams of methamphetamine. Weight is an element of the offense of which Traxler was convicted. When weight is an element of an offense, as with other elements, the state has the burden of proving weight beyond a reasonable doubt. State v. Robinson, 517 N.W.2d 336, 339 (Minn.1994). In Robinson, the supreme court held that taking samples from only 7 of 13 packets suspected of containing cocaine was insufficient to show that the total amount of the substance in all 13 packets weighed at least 10 grams when the total weight of cocaine in the tested packets was estimated to be less than 10 grams. Id. The court stated that in the case of substances not homogeneously packaged, extrapolation was insufficient to establish weight and that the state was required to test a sufficient quantity of the mixture to establish weight beyond a reasonable doubt. Id. There was evidence that drug traffickers sometimes substitute placebos for cocaine, so the court was concerned that the untested packets could have contained placebos. Id. at 338-39. The court also noted:

Where a penalty as serious as 10 years' imprisonment may hinge on a gram, it seems not too much to require scientific testing to establish the requisite weight.

 Id. at 340.

In Robinson, the state's proof of weight depended upon the validity of its assumption that the substance contained in the untested packets was the same as the substance contained in the tested packets. Here, the state's proof of weight depended upon the validity of Skowronski's assumptions about the method used by Traxler to manufacture methamphetamine. Skowronski's calculation of the total amount of ephedrine hydrochloride processed by Traxler assumed that the residue on the 22 filters was produced by 22 different quantities of the ephedrine and solvent mixture. But Skowronski did not testify that Traxler used only one filter at a time when cleaning the mixture, or that Traxler filtered the mixture only once. Skowronski also did not testify that pouring the mixture through more than one filter at a time would leave residue on only one of the filters. If a single quantity of the mixture was poured through more than one filter at a time and left residue on more than one filter, or if a single quantity of the mixture was filtered more than once by pouring it through different filters and left residue on each of the filters, fewer than 22 different quantities of the ephedrine and solvent mixture could have left residue on 22 filters. If fewer than 22 different quantities of the mixture left residue on 22 filters, Skowronski's calculation overstated the total amount of ephedrine hydrochloride processed through the filters, and the amount of methamphetamine that Traxler could have manufactured.

Under Robinson, when weight is an element of a controlled substance offense, circumstantial evidence that a defendant could have manufactured the required amount is insufficient to prove that the defendant manufactured the required amount. Here, the state proved only that Traxler could have manufactured 50 or more grams of methamphetamine, and the evidence presented by the state is consistent with the rational hypothesis that Traxler manufactured less than 50 grams of methamphetamine. The state therefore failed to prove beyond a reasonable doubt that Traxler manufactured the required amount, 50 or more grams, of methamphetamine.

  Reversed.