GN Hear & Now

TECH IN A SEC

'DFS ULTRA II WITH WHISTLE CONTROL' FOR DUMMIES​

Q: What is Whistle Control?
A: Whistle Control is the ReSound name for Adaptive Feedback Gain Correction (AFGC).
Q: What is AFGC?
A: AFGC stands for Adaptive Feedback Gain Correction.

Q: How does it work?
A: AFGC monitors the feedback path and estimates the amount of excessive amplification due to residual feedback. Gains are adjusted in all 17 bands of the Warp filter bank to restore normal loudness.
Q: Why is there residual feedback?
A: The feedback canceller has limited accuracy to approximate the true feedback path.
 
Q: Isn’t the gain reduction causing loss of audibility?
A: Under normal operation AFGC reduces the gain just enough to correct the excessive loudness that is introduced by the residual feedback. Only when the system is somehow tricked into believing there is more residual feedback than there actually is there may be some loss in audibility.
 
Q: What about time constants?
A: There are several time constants in AFGC. We have time constants for smoothing the residual feedback estimate and for smoothing gain corrections. The residual feedback is tracked nearly as fast as the feedback canceller can adapt. For all practical purposes, gain reductions are applied as soon as the feedback canceller observes a significant change in the feedback path. Increasing the gain is done much slower; after the residual feedback drops the gain returns to normal at a rate of about 15dB/s.
 
Q: What are the attack and release times of the feedback canceller?
A: The feedback canceller does not have explicit attack and release times. Changes in AFGC depend on the rate at which the external feedback path changes and the maximal adaptation rate at which the internal feedback model can keep up. The maximal adaptation rate is configurable and depends on a trade-off between static and dynamic performance.
 
Q: Do the attack and release times depend on the AFGC setting in the fitting software?
A: No.
 
Q: What is the difference between the mild, moderate, and strong settings?
A: AFGC relies on an estimate of the residual feedback. If the feedback canceller performs well there is low residual feedback and the mild setting suffices. For difficult devices, ears, or poor calibrations the feedback canceller may perform worse than expected and the mild setting would underestimate the amount of residual feedback. To correct for this the moderate or strong setting can be selected to indicate that there is more residual feedback.
 
Q: How much attenuation is provided for mild, moderate, and strong settings?
A: The attenuation depends on various factors such as the gain, the feedback model and the AFGC setting. Unless the system is close to instability all settings should provide very little or no attenuation at all.
 
Q: What is the width of the notch for the attenuation of the feedback?
A: We do not use notch filters. AFGC operates in the 17 warp bands and the feedback canceller does not use a notch filter either.
 
Q: Is the method of gain reduction employed by the whistle control AGC or Peak Clipping?
A: No, currently AFGC only looks at the residual feedback loop gain. Signal levels, amplitudes, and power estimates are all discarded to minimize algorithm complexity.
 
Q: Can the sensitivity difference of AFGC be represented by dB?
A: Yes, but it is difficult to understand by laymen because it represents an internal estimate of the amount of residual feedback relative to the expected performance of the feedback canceller. There is no one-to-one correspondence with the observed output level because AFGC performs gain *correction* (in contrast to gain reduction).
 
Q: What is the sensitivity difference between Mild/Moderate/Strong?
A: Currently the offset in the estimated residual feedback between mild and moderate is usually 4dB, and between moderate and strong usually 5dB.
 
Q: What is the impact on speech intelligibility?

A: Trials so far have not shown significant negative impact on speech intelligibility. For normal settings (not the strong demo mode) the possible negative effect of too much gain reduction should be less than the negative impact of feedback.