The agenda and the report of the previous meeting, held in San Francisco, 2004-10-29 were approved as written.
The chair presented a spreadsheet, originally devised by M. Opitz of AKG and expanded with additional data, showing the sensitivity and noise measurements made by each company. All companies presented sensitivity measurements, two were unable to complete noise measurements by the time of this meeting. Most of the measurements are within 10 % of the average. One microphone developed a fault and later measurements showed it becoming more noisy over time. Two microphones showed a larger than expected spread of sensitivity measurements. The discrepancies were the topic of extensive discussion. There are some difficult problems to resolve, and some simple ones. One of the latter is probably responsible for the 6 dB error band due to the use of unbalanced, instead of balanced, input conditions. The chair will draft a calibration network design that can become part of the procedure to assure that a given laboratory's sensitivity measurement results are comparable with the results of others.
Discussion continued on making the measurement setup more standardized, within the range of possibilities in IEC 60268-4. There was general agreement in the meeting that the visual data format specified in IEC 60268-1 was sufficient, and that frequency response data should be provided on-axis, at 90, 135 and 180 degrees (and at the primary null if not at one of these angles). Likewise there was a broad agreement that polar graphs, normalized to the on-axis response, were the most useful.
Smoothing of the frequency response data was somewhat more contentious. Some were in favor of 1/3-octave smoothing, others argued for more detail, at least 1/12 octave. It will probably remain the practice to use 1/3-octave smoothing for the "short form" data sheets while 1/12-octave will be sufficient for the "long form" specifications.
IEC 60268-4 specifies that frequency response be measured in free field conditions. Most members agreed that this was desirable, but it might be more relevant to include data using a defined source at a smaller distance, for instance an ITU standard artificial mouth at 25 or 250 mm distance. Distances of 0.5 or 1.0 m are commonly used today. All participants routinely used various ways of simulating free field conditions, and there is no commonly accepted way to confirm that measurements correspond to freefield conditions. There was consensus among those commenting in the meeting that both A-weighted rms noise and psophometric quasi-peak noise figures should be stated.
It was generally agreed that all specifications should include a tolerance band, being meaningless without one. For frequency response, it seemed most popular to include a marked region on the frequency response curve within which the response will fall. Output levels, impedance, etc. can be specified with a plus-minus tolerance, noise and clipping levels can be specified with "less than or equal to" or "greater than or equal to" terms.
Members were encouraged to review the data presentation formats suggested by some participants in their presentations uploaded to the document site. There needs to be agreement on the measurement setup, which should be possible within the next few meetings, and then a writing group can begin work on the data presentation format.
The next meeting will be scheduled in conjunction with the AES 119th Convention in New York, 2005-10