These procedures were designed and tested for internal use by OSHA personnel. Mention of any company name or commercial product does not constitute endorsement by OSHA.

The recent change in the OSHA Time Weighted Average (TWA) Permissible Exposure Limit (PEL) for CO from 50 to 35 ppm (5.1), the inclusion of a Ceiling PEL of 200 ppm (5-min sampling period) (5.2), and the addition of a maximum Instantaneous limit of 1,500 ppm (5.3) stimulated a review of the methods used for the analysis of CO in workplace atmospheres. This review included both direct-reading monitoring procedures and classical collection procedures.

The previous OSHA sampling and analytical method for CO required the use of direct-reading procedures for monitoring (5.4). One direct-reading procedure involved the use of CO short-term detector tubes (5.5). An evaluation at the OSHA Salt Lake Technical Center (OSHA-SLTC) was conducted for several brands of these tubes (5.6). Short-term detector tubes offer only spot checks of the environment, and sampling procedures capable of determining long-term CO concentrations are preferred. A long-term detector tube for CO was evaluated by OSHA and was determined to have significant lot-to-lot variability, and results were affected by sampling temperature (5.7). For TWA compliance measurements, a long-term direct-reading method was performed by OSHA compliance officers using an electrochemical detector (Ecolyzer, Energetics Science, Inc., Elmsford, NY). However, this instrument required constant calibration, readings were subject to drift, and TWA determinations were difficult to calculate. Due to the size of the instrument, only area samples were normally taken. Personal monitoring could be conducted by taking air samples in gas sampling bags and then attaching them to the Ecolyzer at the end of the sampling period. Air samples collected in gas bags or canisters have also been analyzed for CO using infrared absorption spectrophotometry (5.8), electrochemical means (5.9), gas chromatography (GC) with a flame ionization detector (FID) (5.10), or GC with a helium glow discharge ionization detector (DID) (5.11). A recently developed OSHA Method ID-210 (5.12) involves the use of gas bags followed by analysis by GC-DID. This is a reliable and sensitive method; however, it has the disadvantage of requiring sample shipment for laboratory analysis, and storage stability of the collected samples has to be addressed.

Recent developments in direct-reading instrumentation provide increased ease in personal monitoring of workplace atmospheres. These instruments are advertised as accurate and precise, and capable of storing data points for specified sampling periods (each period may be as small as 1 min). One such monitor is the Draeger Model 190 CO Datalogger (National Draeger, Inc., Pittsburgh, PA), which was the instrument evaluated for the current method. Other instruments may be substituted for monitoring CO provided they meet validation requirements.

Samples for monitoring exposures in regards to the TWA and 5-min Ceiling for CO are taken using a portable direct-reading instrument. A datalogger system can be used to reduce data to a TWA or 5-min Ceiling value. Other labor intensive methods such as hand-calculated averaging of readings, strip-chart-recordings, etc. can also be employed to reduce data if necessary.

Note: During the evaluation, six instruments were used for each experiment. Imprecision was likely greater than if only one instrument were used and the experiment repeated six times; however, the statistical values obtained will be more representative of OSHA field office use. Many OSHA field offices are supplied with five or more monitors and an industrial hygienist may randomly take one or more instruments into the field for monitoring. As previously discussed in Section 1.3.4, both truncated and hand-calculated averages were used to determine instrument performance. The hand-calculated average only appears beneficial if detection limit is a significant factor as shown below.

Note: The following paragraphs list specific interferences associated with the Draeger Datalogger. If other types of instruments are used, always consult with the manufacturer or manufacturer-provided instructional manuals concerning potential interferences before general use. Additional information regarding interferences concerning the Draeger Datalogger can be found in reference 5.13

The following interferences occur with the CO-specific filter attached to the monitor. Acetylene and hydrogen have small positive interferences (97 ppm acetylene will indicate 10 ppm CO, 1,000 ppm hydrogen will indicate 40 ppm CO ) (5.13). During the evaluation of the Draeger Datalogger, it was noted high concentrations of the lower alcohols produced large positive biases when the monitor was exposed to headspace concentrations. Dataloggers were also placed near the open tops of glass bottles containing the chemicals listed below and monitor response was noted:

The responses indicated lower alcohols gave positive responses and others structurally similar to CO (i.e. acetone) could give a response if the CO-specific filter is removed or the filter has diminished capacity. Placement of the monitors into atmospheres saturated with each substance giving a positive response gave very large false positive readings for CO and also tended to fatigue the sensor. In addition, fine mists or droplets of the positive-interfering substances will produce similar false positive results. These interferences were produced using near-saturated or saturated atmospheres and should not be significant at concentrations near the PELs for these compounds. Concentration estimates for those substances giving positive responses at saturated vapor conditions are shown in Appendix A (Note: These are only rough approximations). The industrial hygienist should note whenever there is a significant amount of these compounds present (i.e. some confined space monitoring may have near-saturated or saturated atmospheres), or operations where droplets or fine mists of these substances could contact the surface of the filter/sensor of the Datalogger. If large amounts or aerosols of these substances are expected, monitoring for the interfering substances should take precedence, and an alternate method such as OSHA Method No. ID-210 can be used for CO monitoring if necessary. It is highly recommended that the CO-specific filters be changed periodically.

Early reports from other Datalogger users indicated a positive bias was observed when the monitor was placed near liquid aerosol streams. To assess if small liquid droplets could trigger a positive reading, aerosols of water or ethyl acetate were sprayed directly above the monitor sensor and produced no response.

With the single exception of CO2, the total yearly emissions of CO exceed all other atmospheric pollutants combined (5.16). Potential sources for CO emission and exposure are listed (5.16, 5.17):

Note: Information contained within this section is a synopsis of the present knowledge of the physiological effects of CO and is not necessarily intended to be used as the basis for OSHA policy.

Carbon monoxide has over a 200-fold greater affinity for hemoglobin than has oxygen (5.18, 5.19). Thus, it can make hemoglobin incapable of carrying oxygen to the tissues. The presence of CO-hemoglobin (COHb) interferes with the dissociation of the remaining oxyhemoglobin, further depriving the tissues of oxygen (5.15,5.16).

The signs and symptoms of CO poisoning include headache, nausea, weakness, dizziness, mental confusion, hallucinations, cyanosis, and depression of the S-T segment of an electrocardiogram. Although most injuries in survivors of CO poisoning occur to the central nervous system, it is likely that myocardial ischemia is the cause for many CO-induced deaths (5.18).

The uptake rate of CO by blood when air containing CO is breathed increases from 3 to 6 times between rest and heavy work. The uptake rate is also influenced by oxygen partial pressure and altitude (5.20).

Carbon monoxide can be removed through the lungs when CO-free air is breathed, with generally half of the CO being removed in 1 hour. Breathing of 100% oxygen removes CO quickly.

Acute poisoning from brief exposure to high concentrations rarely leads to permanent disability if recovery occurs. Chronic effects from repeated exposure to lower concentrations have been reported. These include visual and auditory disturbances and heart irregularities. Where poisoning has been long and severe, long-lasting mental and/or nerve damage has resulted (5.15).

The following table gives the levels of COHb in the blood which tend to form at equilibrium with various concentrations of CO in the air and the clinical effects observed (5.21):

Adults (non-smokers) normally have about 1% COHb in the body. Cigarette smokers generally have blood levels of 2 to 10% COHb (5.20).

In examining the CO levels in an occupational environment, CO generated from tobacco smoking may need to be considered. These amounts may ordinarily be small, but when added to the amounts generated by occupational activities, may aggravate conditions from an already existing high concentration of CO (5.22,5.23).

Carbon monoxide is flammable and can be a dangerous fire and explosion risk. The flammable limits in air range from 12 to 75% by volume (5.19).

Procedure: Three sets of samples (at 0.5, 1, and 2 × TWA PEL) were collected and evaluated. Samples taken at 5 and 10 ppm CO (50% RH) were also prepared and analyzed. Six Dataloggers were evaluated at each of these sampling conditions. Samples were collected according to the procedure listed below.

Procedure: In order to evaluate the Dataloggers at CO levels commonly encountered during IAQ investigations, test atmospheres at about 5 and 10 ppm CO (50% RH) were prepared and Datalogger samples were taken, using the same equipment and the sensor conditions described in Section 4.2. The effects of the gas flow rate over the sensor, as well as the CO background concentration in the generator diluent air, were accounted for in these studies. Six Dataloggers were evaluated at each concentration.

Results: Recoveries for the samples generated at approximately 5 and 10 ppm CO are given in Table 2a (Datalogger truncated TWA output values) and Table 2b(hand-calculated averages of the 1-min ppm values collected by each Datalogger).

Results at 5 ppm gave a large OE value; however, this concentration level is very low in relation to the TWA PEL for CO. The hand-calculated results in Table 2bare an improvement over TWA output values given in Table 2a. This is to be expected because, at these low levels, recovery values based on hand calculating are taken to the nearest 0.1 ppm (Table 2b). The TWA output values (Table 2a) are truncated to the next lowest whole ppm.

Procedure: In order to evaluate the Dataloggers at CO levels in the vicinity of the Ceiling PEL (for a 5-min exposure period), generated samples near 200 ppm CO (50% RH) were also prepared and analyzed. For this experiment, six dataloggers were calibrated using the 40-ppm CO gas previously mentioned, and then mounted in a Du Pont gas flow chamber [E. I. Du Pont de Nemours and Co., Inc., Wilmington, DE. An excellent description of a similar chamber and the mixing chamber (Figure 1a) can be found in reference 5.28] designed for passive monitor evaluation. The chamber was connected to the generation system using Teflon tubing. The cross-sectional area in the chamber was 83.87 cm2 (5.29). The gas flow rate in the chamber during the experiment was estimated by calculation as 1.1 m/min (3.6 ft/min). The experimental procedure consisted of placing the Dataloggers along a rigid Teflon tube in the chamber before assembling the system. The Dataloggers were mounted with the sensors facing upward, toward the direction of air flow through the chamber. Datalogging was initiated immediately before the monitors were mounted in the chamber. After placing the monitors in the chamber, air flow was started, then the 0.496% CO was turned on.

After the CO concentration in the chamber had stabilized (determined by observing the monitor LCDs through the glass chamber wall), two gas bag samples of the CO/air mixture were taken for GC analysis from a sampling manifold connected to the Du Pont chamber. At about 50 min after the CO concentration had stabilized, the gas flows were stopped and the Dataloggers were removed and datalogging was halted.

Results: Recoveries for the samples generated at approximately 200 ppm CO are given in Table 3. The hand-calculated averages of the 1-min ppm values collected by each Datalogger for the time period of interest were used. Results are given for:

Procedure: Samples were also generated at 30% and 80% RH using the same equipment and conditions described in Section 4.2. An experiment was not conducted for 2 × TWA PEL at 30% RH.

Results: The results for the generated samples collected at the three RH levels are listed in the following Tables:

Two "B" experiment tests (1 × PEL at 30% RH and 0.5 × PEL at 80% RH) showed a positive bias in the mean recoveries obtained. These were initial "B" experiments, and were conducted with calibration flow rates lower than generation system flow rates. The increased flow across the sensor face as compared to the calibration flow was suspected of causing the higher bias.

As shown in Table 6, an analysis of variance (F test) (5.30) was performed on the data to determine any significant difference among or within the various RH groups. Variance at each concentration (0.5, 1, and 2 × TWA PEL) was compared across the three RH levels (30%, 50%, and 80% RH). The variance among and within the different concentration groups gave high calculated F values at 0.5 and 1 × PEL. The large calculated F values appear to be mainly due to variability in mean recoveries, and are most likely related to variance from sensor face velocity differences within the experiments.

Recalculation of F values after making estimated corrections for face velocity differences between calibrations and generation system measurements reduced the F values considerably.

Procedure: As mentioned previously in Section 4.2.1, tubing devices similar to the Draeger calibration adapter were used to connect the Dataloggers to the generation system. Under these conditions, generation system flow rates over the Datalogger sensor faces (linear face velocities) tended to be high, and therefore, higher CO sample concentration readings were obtained.

An investigation of face velocity dependence of gas flowing through the calibration adapter or similar tubing was conducted with six Dataloggers which were calibrated with the 104-ppm CO certified standard mentioned in Section 4.2.4. Four different sets of tests were conducted:

Results: For test (1), the Datalogger readings as a function of gas flow across the face of the sensor (face velocity) are given in Figure 2. It is evident that the readings tend to rise as the linear flow increases, and a "steady state" is achieved where the readings stop rising at a certain point.

Gas flow rates over the sensor in metric units are given below.

For test (2) the rise in observed readings was less pronounced when tubing similar to the calibration tubing was replaced with the loose-fitting glass tube. It should be stressed in either test, the flow of the CO/air mixture is being applied directly at the sensors.

The Datalogger dependence on face velocity appears important during calibrations and appears partially due to increased pressure on the sensor face. Although the curvature shown in Figure 2 is important during calibration, it follows a similar pattern noted with other passive monitors. Face velocity dependence has been noted previously (5.29), and it is not unusual for a monitor to have:

This effect of air flow rate on the results for passive monitors has been discussed further in a report from the National Institute for Occupational Safety and Health (5.31). Typical face velocities in general industry, according to the NIOSH report (5.31) are 10 to 150 cm/s (approximately 6 to 90 m/min or 20 to 300 ft/min).

In test (3) where the nitrogen flow after zeroing was altered, there was no significant change from zero as the flow rate of nitrogen increased. This indicates that pressure alone is not responsible for the elevated readings noted in tests (1) and (2).

In test (4) where the monitors were exposed to the rapidly moving turbulent air, the monitor readings leveled off in the range of 103 to 114 ppm CO. The readings were similar even when the fan was turned off. It appears the monitors are only slightly affected by turbulent air flows of 15.2 to 91.4 m/min (50 to 300 ft/min) where a positive bias of approximately 5 to 10% was noted. The monitor appears more affected by linear flows. It is expected in many industrial settings the air flow will move in a turbulent fashion.

The manufacturer has reported the following calibration flow deviations when CO-containing gas flow rates different from the recommended 0.20 L/min are used:

These manufacturer-determined values appear similar to those shown in Figure 2. The flows are clustered close to the recommended calibration flow rate of 0.2 L/min. According to the manufacturer, experiments in which the air velocity was varied from 0 to 6 m/s (1,200 ft/min), the sensitivity was <±5% of the measured value (5.32). It is not known whether the test was conducted using linear or turbulent flows.

Both series of internal and manufacturer-performed experiments discussed above demonstrate a significant relationship of Datalogger accuracy and calibration flow rate. Significantly different results may be obtained during sampling if the Datalogger is calibrated at flow rates different than recommended (0.2 L/min).

Procedure: Both qualitative and quantitative limits for the analysis of CO by the Dataloggers were calculated using the International Union of Pure and Applied Chemistry (IUPAC) method for detection limit determinations (5.33). The data collected at 50% RH and 5.6, 11.1, and 16.1 ppm (Tables 1a, 1b, 2a, and 2b) were used. Also included were data from a blank generation in which no CO was added. The blank generation was performed when the background CO level in the air was at a very low level (0 to 1 ppm).

Results: The detection limit results are listed in Table 8a (for the truncated TWA output values) and Table 8b (for the hand-calculated averages of all of the 1-min ppm values). A derivation of the IUPAC method was used for the data in Table 8a because the blank readings were zero. The standard deviation of the lowest concentration standard was used in place of the blank for data in Table 8a. The results are summarized as follows:

The lower detection limits for the "Hand-Calc. Ave." of all the 1-min ppm values is to be expected, since fractional values for the ppm determination are calculated, averaged, and rounded. This type of data (rounded rather than truncated) is also more attuned to the gaussian statistical model used. Also, hand-calculated averages are indicative of the ultimate CO measuring capabilities of the instrument at low concentrations.

Procedure: In order to confirm the accuracy of the CO concentration of the 0.50% and 0.496% CO sources used in the generation system, as well as the accuracy of generation, one or two gas bag samples were taken from the generated gas stream during each Datalogger study. Bags employed were 5-L five-layer aluminized gas bags (Calibrated Instruments, Inc., Ardsley, NY). Analysis of the CO content of the gas bags was performed with a Tracor Model No. 540 GC equipped with a Model No. 706 DID according to OSHA Method ID-210 (5.12). Standards were prepared from the cylinder of 104-ppm CO in nitrogen previously discussed inSection 4.2.4.

The CO concentration of this gas cylinder had been confirmed during previous work by reaction with iodine pentoxide and titration of the subsequent iodine produced (5.12).

Results: The GC results were compared with theoretical values (e.g., "GC Results" = GC Recovery/Theoretical Value). Theoretical values were calculated from generation system flow settings. As shown in Figures 3 - 5, good agreement is noted between theoretical and GC results. The Datalogger results ("Datalogger" = TWA Output Values, Calc Average = Hand-Calculated Average Values) are also illustrated for comparison purposes. It should be noted in Figure 4 the TWA output value (Datalogger result) for the Ceiling determination is not present. In this case the results were hand-calculated as previously mentioned in Section 4.4.

The experimental results indicate this direct-reading monitor is an excellent means for sampling and analyzing CO in the workplace if properly calibrated and if air flow is not directly at the face of the monitor sensor (linear flow). Obtaining accurate analytical results for other-than-normal conditions existing in the working environment may be contingent on matching the flow of the calibration gas. Although the precision and accuracy results were still acceptable when the calibration/generation system flow ratios were different by a factor of two (Tables 4a and 5a), the monitors should normally be calibrated using the manufacturer-stated flow settings. If the face velocity in the sampling environment simulates a linear flow (i.e. measurements inside of air ducts), calibration flow can be matched to sampling flows to provide for more accurate results. It should be noted that turbulent flow at the face of the sensors did not produce the magnitude of positive bias noted to occur during the linear face velocity tests. The positive bias is likely partially due to increased pressure on the CO sample directly above the sensor face. The Dataloggers should be calibrated at the same altitude as the site where sampling will be conducted.

Because the instrument automatically performs integer math for calculations (truncating to the next lower integer ppm value), rounding-off error for the resulting TWA output values is possible. This does not appear problematic for CO concentrations normally encountered in industry, and is aptly shown by comparable precision and accuracy data obtained when testing in the vicinity of the PEL (Tables 1a and 1b). The error does become significant at single digit levels (i.e., <10 ppm); however, these levels are well under the present TWA PEL. It should also be pointed out that in workplace environments the CO level will not be static and may quite often rise to high concentrations and fall to single digit values in one monitoring period.

Determinations near the TWA PEL were well within NIOSH and OSHA accuracy and precision guidelines (5.26, 5.27). For the truncated TWA output values at 50% RH, the pooled coefficient of variation (CV2) was 0.040 and the overall recovery was 98.4%. An effect from humidity is not apparent. In two cases where the OE values appear somewhat high (1 × PEL at 30% RH and 0.5 × PEL at 80% RH), face velocity and increased pressure at the sensor face appear as contributory factors. The manufacturer of the Datalogger may need to further investigate interferences caused by low molecular weight alcohols and aldehydes which appear structurally similar to the CO molecule. These interferences are not expected to be significant in normal workplace monitoring but may cause problems during confined space evaluations.

The OE values at approximately 5 and 10 ppm CO were somewhat high, but within NIOSH and OSHA guidelines if non-truncated values are used.

The Dataloggers are very convenient to use. There is minimum interference with work procedures by employees. The exposure data are stored and easily recovered to produce a minute-by-minute record of the working environment.

This sampling and analytical method is capable of accurate and precise measurements to determine compliance with the 35-ppm TWA or 200 ppm Ceiling PEL for CO exposures.