At its quietest, the only notable noise source is the fan of a network switch, which could not be disabled for this experiment. The overall noise rating is just above NC20 at 27.1 dBA.

The noise source is more present in the recording than is noticed in real life. This is at least partially due to the mono nature of the MXA920 array's output.

"Low" is likely the ideal NR selection, and Medium nearly eliminates all noise without meaningful sound quality degredation.

Intelligibility is excellent at all NR settings, even without. Processing artifacts are minimal, even at High NR.

The noise source is more present in the recording than is noticed in real life. This is at least partially due to the mono nature of the MXA920 array's output.

Intelligibility is excellent at all NR settings, even without. Without NR, the background noise is an annoyance, however.

"Low" is likely the ideal NR selection, and Medium nearly eliminates all noise without meaningful sound quality degredation.

Processing artifacts are minimal up to Medium NR. Processing artifacts are increasingly noticeable at High but are easily tolerable in conferencing, capture, and streaming applications.

This noise level is beyond typically specified ideals for conferencing applications, but is none-the-less common in real life.

Intelligibility is excellent at all NR settings. Without NR, the background noise is an signifcant annoyance.

"Medium" is likely the ideal NR selection, and Medium nearly eliminates all noise without meaningful sound quality degredation.

Processing artifacts are minimal at Low NR, but background noise remains very present. Processing artifacts are increasingly noticeable at Medium NR but are easily tolerable in conferencing, capture, and streaming applications.

Background noise is all but eliminated at High NR with slightly worse processing artificats. High NR is still very useable if complete background noise removal is absolutely necessary.

This noise level is well beyond typically specified ideals for conferencing applications, but is none-the-less common in real life.

Intelligibility is acceptable at all NR settings though begins to suffer at High NR. Without NR, the background noise is an signifcant annoyance.

"Medium" is likely the ideal NR selection as residual noise level and sound quality are balanced.

Processing artifacts are minimal at Low NR, but background noise remains very present. Processing artifacts are increasingly noticeable at Medium NR but likely remain entirely tolerable in conferencing, capture, and streaming applications.

Residual background noise is similar between Medium and High NR settings, but processing artifacts are more noticeable at High.

This noise level is significantly beyond typically specified ideals for conferencing applications. The raw noise level will compete with direct sound for many in-room participants.

Speech is intelligible at all NR settings but suffers from the signifiant noise.

"High" is likely the ideal NR selection as residual noise level and sound quality are balanced.

Processing artifacts are noticeable at all NR levels but remain usable in conferencing, capture, and streaming applications.

This noise level is the absolute upper limit of usable for conferencing applications. The raw noise level will compete with direct sound for most in-room participants and is generally unacceptable at any budget or use case.

Speech is intelligible up to Medium NR settings but suffers from the signifiant noise and processing artifacts.

Surprisingly, "Low" is likely the ideal NR selection as residual noise level and sound quality are most balanced. Medium may be better in some applications, however, depending on the circumstances of far-end listeners.

Processing artifacts are noticeable at all NR levels. At High NR, intelligibility suffers unaccetably.

This noise level is beyond the usable limit for most if not all applications. The raw noise level will compete with direct sound for all in-room participants, rendering the space generally unusable.

Many syllables and vowel sounds are burried in the noise floor and are therefore suppressed by the NR algorithm. Speech is probably most intelligible without NR.

The Medium NR setting is unusable, and High NR may be the best choice if absolutely necessary for a conferencing application. Near-end participants need to be instructed to speak loudly. For archival applications, no NR may be the best choice.

This noise level is beyond the usable limit for conferencing applications. The raw noise level will overpower direct sound for all in-room participants, rendering the space generally unusable.

Few words and syllables can be discerend at any NR level as all speech is buried below the noise floor and is therefore further suppressed by the NR algorithm.