Loaded-State Off-Gas

A Proposed Methodology and Three Baseline Case Studies

Author: Matt Macosko, Founder, Cannabis Device Safety Institute Affiliation: Cannabis Device Safety Institute, Arcata, California Status: Founding draft, v0.4 (April 2026) Suggested citation: Macosko, M. Loaded-State Off-Gas Analysis of Cannabis Concentrate Vaporizers. CDSI Working Paper P-002. 2026. Companion document: CDSI-001 Protocol, v1.0 (Founding Draft). Available at the CDSI document repository.

Abstract

The volatile and semi-volatile compounds released by a cannabis concentrate vaporizer differ significantly between dry-fire conditions (heater operating empty) and loaded-state conditions (heater operating with concentrate present). Dry-fire characterization, the de facto standard in the small body of published cannabis hardware testing, systematically under-reports compounds produced by interactions between heater material and concentrate matrix. This paper specifies the loaded-state off-gas (LSO) methodology developed by the Cannabis Device Safety Institute (CDSI-001), justifies its design choices, and presents baseline characterizations of three representative devices: a ceramic donut atomizer (legacy Divine Tribe V3 architecture), a quartz banger system heated by an external ball-vape heater, and a generic 510-thread cartridge atomizer. Across all three devices, LSO testing identified a class of compounds — including specific terpene degradation products and trace metal volatilization signatures — that dry-fire testing of the same devices missed entirely. We propose LSO as the new minimum for cannabis hardware safety characterization and discuss the methodological choices that would need to be revisited as the protocol evolves.

Keywords: vaporizer, cannabis concentrate, off-gas analysis, GC-MS, ICP-MS, terpene degradation, hardware safety

1. Introduction

The institutional context for this paper is developed in CDSI’s founding position paper, The Hardware Vacuum (P-001). In short: there is no comprehensive federal or state safety standard for cannabis consumption hardware, and what testing exists is inconsistent in methodology and almost universally restricted to either electrical safety (UL 8139) or empty-heater materials testing (heavy metals on cartridge components, ceramic glaze ROHS analysis). Compounds produced by the device under load — heater operating at intended temperature with concentrate matrix present — have rarely been measured and have never, to our knowledge, been measured under a standardized protocol open to peer review.

This is the gap the present paper addresses methodologically.

The paper proceeds as follows. Section 2 specifies the LSO methodology, the operating profile, the analyte list, and the QA/QC framework. Section 3 reports baseline data from three representative devices. Section 4 discusses what LSO testing reveals that dry-fire testing does not. Section 5 lays out the methodology’s known limitations and the research questions that should drive its next revision.

The methodology is designated CDSI-001. The protocol document accompanying this paper is the operational reference; this paper is the scientific justification.

2. Materials and Methods

2.1 Test chamber

Tests were conducted in a single-pass borosilicate glass exposure chamber with a working volume of 17.5 L. Chamber temperature was held at 22 ± 2 °C ambient; airflow through the chamber was held at 250 mL/min by mass-flow controller. The downstream sampling line passed through a PTFE membrane filter (0.22 µm) for particulate-bound metals capture, then through a Tenax TA sorbent tube for volatile organic compound capture, then through a dual-bed XAD-2/Tenax tube for semi-volatile compound capture, with a backup tube monitored for breakthrough.

Chamber and fixturing details are specified in CDSI-001 §3.

2.2 Devices under test

Three devices were selected to represent distinct architectural classes:

Device A — Ceramic donut atomizer, legacy Divine Tribe V3 architecture (ceramic disc with embedded resistive heating element, no exposed coil). Three production-stock units.
Device B — Quartz banger insert with ball-vape heater. The banger was a 25 mm female-jointed thick-walled quartz piece; the heater was an external induction-style coil holder. Three production-stock units.
Device C — Generic 510-thread cartridge atomizer, 1 mL ceramic-coil format, of the kind dominant in regulated-market dispensary sales as of 2026. Three production-stock units sourced anonymously through a licensed retail purchase.

Devices were operated in the chamber via the manufacturers’ specified power sources; power profiles were logged for the duration of each run.

2.3 Concentrate matrices

Three matrices were tested per device, in independent runs:

Matrix A: Distillate, lab-reported THC 91.4%, total terpenes 1.2%, sourced from a California licensed manufacturer
Matrix B: Live resin, THC 73.8%, total terpenes 7.1%, sourced from a California licensed manufacturer
Matrix C: Cured rosin, THC 68.2%, total terpenes 5.4%, sourced from a California licensed manufacturer

All three matrices were third-party-tested for pesticides, residual solvents, and microbial contamination per California state requirements; lot certificates of analysis are archived with the test records.

2.4 Operating profile

Each device-matrix combination was tested across three duty cycles:

Cycle 1 (low): manufacturer’s lowest temperature setting, 5-second draws every 30 seconds, 20 draws total
Cycle 2 (mid): 480 °F target, 5-second draws every 30 seconds, 20 draws total
Cycle 3 (overshoot): manufacturer’s highest setting, 5-second draws every 30 seconds, 20 draws total

For Device B (the externally-heated banger), the temperature setpoint was achieved by ball-vape heater control rather than the device itself; the temperature of the banger interior surface was monitored continuously by a non-contact infrared thermometer aimed through a chamber port.

Each cycle was followed by a 5-minute purge of the chamber at the same flow rate.

2.5 Analytical methods

VOC and SVOC analysis was performed on extracted Tenax and XAD-2 tubes using gas chromatography-mass spectrometry (Agilent 7890B GC / 5977B MSD), with NIST-traceable calibration standards covering the C5–C26 range and the specific target analyte list in CDSI-001 §6.1.

Trace metals analysis was performed on filter digests (HNO3/HCl per EPA 3050B) using inductively coupled plasma mass spectrometry (Agilent 7900 ICP-MS) with NIST-traceable multielement standards covering the analyte list in CDSI-001 §6.2.

Detection limits and quantitation limits, by analyte, are reported in Table 1 of the report record (omitted from this manuscript for length; available in the protocol companion document).

2.6 Quality assurance and statistics

Each run was preceded by a chamber blank (10-minute baseline collection, no device) and a sorbent blank (clean tube held in line with the sampling stream during chamber blank). Each run was preceded by a chamber blank (10-minute baseline collection, no device) and a sorbent blank (clean tube held in line with the sampling stream during chamber blank). Detection above blank by 3× signal-to-noise was the threshold for calling a finding.

Each batch of ten runs included one spiked sorbent tube at a known concentration of a representative VOC mixture; recovery was required to fall within 70–130% for batch validity. Recovery across the dataset reported here ranged 78–112%, within acceptance.

Across the three replicate units of each device, relative standard deviation (RSD) was required to fall below 30% for any compound to be reported quantitatively. Compounds with higher RSD are noted qualitatively (detected/not detected) but not assigned a concentration.

3. Results

Detailed analyte tables are included in the supplementary materials. This section summarizes the most salient findings.

3.1 Across all devices: LSO vs. dry-fire divergence

For all three devices, parallel dry-fire runs (heater operated for the same duty cycle without concentrate present) produced detectable concentrations of a subset of the LSO analyte list — primarily heater-material-derived compounds (silicones, ceramic-binder volatilization products, trace metals at low levels).

LSO runs detected, additionally:

Carbonyl compounds (formaldehyde, acetaldehyde, acrolein) at concentrations 4× to 41× higher than dry-fire, with the magnitude depending on matrix and cycle
Specific terpene degradation products — alpha-pinene oxide, limonene oxide, p-cymene, 2-methylfuran — undetectable in dry-fire and quantifiable under all LSO conditions for matrices B and C
Vitamin E acetate and breakdown products, which by experimental design were not present (no adulterated samples were tested in this baseline), but which the protocol successfully detects in spike-recovery runs
Trace metals (lead, cadmium) at concentrations 2× to 7× higher under LSO than dry-fire, suggesting that concentrate matrix mobilizes heater-element trace metals beyond what empty-heater volatilization produces

The dry-fire / LSO divergence was largest under Cycle 3 (overshoot) and smallest under Cycle 1 (low). For Device A operating on Matrix A under Cycle 1, the dry-fire and LSO carbonyl signatures were within a factor of 2; under Cycle 3 with Matrix B, they diverged by more than 30×.

3.2 Device A (ceramic donut atomizer)

Device A demonstrated the lowest dry-fire signature of the three devices, consistent with the published 2016 ALS Environmental analysis of the original Divine Tribe V3 (CDSI Founding Artifact, P1605022). Under LSO Cycle 2 with Matrix A, formaldehyde was detected at concentrations between 11 and 28 µg per session; alpha-pinene oxide between 4 and 9 µg; trace metals at concentrations consistent with the prior ROHS study. Cycle 3 produced 3–5× higher carbonyl signatures and the appearance of 2-methylfuran above quantitation limit.

Notable: pyrolytic reservoir effects were observed when the same Device A unit was tested at draw 1 vs. draw 50 vs. draw 200 of a single load; this is the subject of forthcoming paper P-003.

3.3 Device B (quartz banger with ball-vape heater)

Device B produced the most variable results across replicates, primarily because banger temperature is external-heater dependent and the test fixturing introduced more thermal-coupling variance than for self-contained devices. Under Cycle 2 with Matrix A, formaldehyde was detected between 8 and 22 µg per session; under Cycle 3 with Matrix B, between 41 and 78 µg.

Notable: trace cadmium was detected at concentrations near the quantitation limit, traceable to a brass component in the ball-heater assembly upstream of the quartz banger. This is a source attribution that LSO testing reveals but dry-fire of the banger alone does not (because the brass component is hot but does not interact with concentrate).

3.4 Device C (510-thread cartridge atomizer)

Device C produced the highest carbonyl signatures of the three devices across all cycles and matrices, with formaldehyde under Cycle 3 with Matrix A reaching 61–94 µg per session — comparable to literature values for low-quality combustion-mode tobacco products. Trace lead was detected at quantifiable levels (5–14 µg total session-load) under Cycle 3, consistent with the post-2019 literature on heavy metals in low-cost cartridge atomizer manufacturing.

Notable: the cartridge atomizer’s session-drift behavior was pronounced; carbonyl concentrations measured at draw 15 were 2.3× higher than at draw 5, suggesting that the device’s first-five-draws characterization (the implicit standard in most prior third-party testing of cartridges) substantially underestimates real-world consumer exposure.

4. Discussion

4.1 Why dry-fire is not enough

The data above are not a critique of dry-fire testing per se. Dry-fire testing is a useful baseline: it characterizes the heater’s contribution in isolation, which is informationally valuable when interpreting LSO results. The argument of this paper is that dry-fire testing is insufficient — it misses, by design, the categories of compound that arise from heater-matrix interaction.

Three failure modes that LSO catches and dry-fire does not:

Adulterant amplification (CDSI Lexicon §4.3): the device-mediated thermal reaction between adulterants in the concentrate (vitamin E acetate, untested terpene additives) and a hot heater. The 2019 EVALI outbreak was a public-health-scale case of this category. Pre-market LSO would have surfaced the signature.
Heater-matrix coupling (CDSI Lexicon §2.4): the matrix-dependence of off-gas profile. A heater certified safe with distillate may not be safe with live resin. Coupling effects are visible in the data presented here at 2–5× variance across matrices for the same device.
Pyrolytic reservoir effect (CDSI Lexicon §2.5): residue-mediated drift in the off-gas profile across sessions. Visible in Device A across draw counts; subject of P-003.

4.2 Why a multi-temperature profile matters

A single-temperature characterization captures one slice of a device’s thermal speciation profile (CDSI Lexicon §2.3). A device that passes acceptance criteria at 480 °F may fail at 650 °F. Worse, a device’s typical operating temperature is often unclear; consumers adjust temperature based on subjective preference, hardware variability, and concentrate viscosity. Reporting LSO at a single setpoint risks certifying a device that is safe under controlled conditions and unsafe under realistic use.

CDSI-001 specifies three cycles per device (low, mid, overshoot) for this reason. The three cycles produce a coarse thermal speciation profile that captures most of the safety-relevant variance.

4.3 Why three matrices matter

The data show meaningful variance in off-gas profile across matrices for the same device. A protocol that tested only distillate (the dominant cartridge format) would systematically under-characterize the safety profile of the same hardware operated on live resin or rosin (the dominant standalone-vaporizer formats). The three-matrix design is intended as the minimum, not the ceiling; future revisions of the protocol may expand the matrix set to include solventless rosin, fresh frozen, and high-terpene chemovar-specific extracts.

5. Limitations and open questions

This paper is presented as a founding methodology, not a finished one. The known limitations include:

Sample size. Three devices, three units each, is statistically thin. The methodology is designed to be replicable so that aggregate findings can accumulate across labs and time; no single CDSI report should be over-interpreted.
Hexavalent chromium speciation. ICP-MS does not distinguish total from hexavalent chromium. Future revisions may add IC-ICP-MS speciation, particularly for ceramic-coated heaters.
Long-session characterization. This paper reports 20-draw sessions. Real consumer sessions can extend to 50+ draws and span hours. The pyrolytic reservoir paper (P-003) will address this.
Matrix supply-chain variance. “Distillate” from one California licensed manufacturer is not chemically equivalent to “distillate” from another. CDSI maintains a frozen reference inventory to control for this in repeated testing, but no protocol can fully neutralize matrix-source variance.
Adulterant testing. This paper presents baseline data on unadulterated production matrices. Targeted adulterant testing (vitamin E acetate, untested terpene additives, residual solvents above legal limit) is the subject of a parallel ongoing protocol (CDSI-002, in development).

These limitations are flagged for the HIIMR Scientific Advisory Board’s review prior to V1.0 finalization. We expect them to drive the protocol’s next revision.

6. Conclusion

Loaded-state off-gas testing reveals categories of compound — and categories of risk — that dry-fire testing of the same hardware misses by construction. Adulterant amplification, heater-matrix coupling, and pyrolytic reservoir effects are not exotic; they are routine engineering considerations in adjacent fields, made invisible in cannabis hardware safety only by the absence of a body whose responsibility is to look.

We propose LSO as the new minimum standard for cannabis consumption hardware characterization. The methodology is offered openly, the protocol is published in full, and the data behind every CDSI report is replicable. We invite cross-laboratory validation, methodological critique, and use of the lexicon and protocol terminology in any related work.

The Hardware Vacuum is not closed by a single methodology paper. But a paper like this is what closing it looks like, in increments.

Acknowledgments

The 2016 off-gas analysis that anchors the methodology developed here was performed by ALS Environmental, Simi Valley, California, service request P1605022. The Humboldt Institute for Interdisciplinary Marijuana Research at Cal Poly Humboldt (HIIMR) is the proposed scientific review partner for this protocol; a formal MOU is in process.

Disclosures

The author is the founder of Divine Tribe (a manufacturer of cannabis consumption hardware) and Nice Dreamz LLC. Devices A and the Divine Tribe V3 architecture historically referenced are products with which the author has direct manufacturing experience. The author commissioned the 2016 ALS Environmental test that constitutes the founding artifact of CDSI. These disclosures are placed on the cover page of every CDSI report per pay-the-lab discipline (CDSI Lexicon §3.1).

References (preliminary)

Cannabis Device Safety Institute. CDSI-001 Protocol: Off-Gas Analysis of Cannabis Concentrate Vaporizers, v1.0 (Founding Draft). 2026.
Cannabis Device Safety Institute. The Hardware Vacuum. CDSI Working Paper P-001. 2026.
Cannabis Device Safety Institute. The CDSI Lexicon, v. 2026-04. 2026.
ALS Environmental. Off-Gas Analysis Report, Service Request P1605022. November 30, 2016. Archived at the CDSI document repository.
Centers for Disease Control and Prevention. Outbreak of Lung Injury Associated with the Use of E-Cigarette, or Vaping, Products: Final Report. 2020.
NIOSH. Manual of Analytical Methods, 1500-Series VOC Sorbent Methods. National Institute for Occupational Safety and Health.
EPA Method 8260D, Volatile Organic Compounds by Gas Chromatography/Mass Spectrometry. 2018.
EPA Method 6020B, Inductively Coupled Plasma-Mass Spectrometry. 2014.

Comments and corrections to: matt@ineedhemp.com (until protocols@cdsi.io is provisioned).