After a looong telephone call to my audi, we enabled the silencer-checkbox and controlled the automatic-options-page.
Silencer is on.
Virtual ear is aware.
Directionality is neural.
Tinnitus-Mask is off.
… and the (low) noise is unchanged, although i’m geting so used to it, that - except immediately after putting my HAs in in the morning - i only get aware of it, when i’m listening.
So if everything else fails, i go to the extreme: “i try to think”:
:
oticon claims, that they split the input of each microphone (and the auxiliary input) to seperate 64 frequency channels. Since this is done by a signal processing chip, I think they use some sort of polyphase quadrature filter (this is the same procedure as the first step of mp3-coding).
I don’t want to torment you all with the mathematics of digital filters, but in principle they work by delaying the signal step by step and summing up scaled portions of each delayed step.
So if the addends are scaled, there will be an error in the lower bits, small but ubiquitous. A polyphase quadrature filter stacks and combines the output of several filters, what again adds some of this “quantization noise”.
So if in the end - after processing all those clever comfort features - if they sum up the channels again, they will sum up not ony the analog noise of the microphones (until now we only considered this) but also the mathematical noise of 3x64=192 channels. And since noise is a nasty fellow ('think i mentioned it
, especially if you quantizise phase) it sums up merciless. This may quite possibly add several db of noise.
Of course this can be reduced by upscaling, but since there have to be a plethora of adders and multipliers, this increases the technical effort enormously (there are signal processors that work with float numbers, but they are expensive in terms of transistors per chip and energy demand from the batteries. And they would trade in quantizision noise for non-representability errors).
This is interesting, since real uses an even more advanced chip - but since it is almost a miracle to create a 64-filter bank that is delay- and phaseneutral enough to not sound like a drainpipe after re-summing, i guess, they did not touch it.
I’m afraid, as long as they do not employ Harry Potter as an engineer
, we will have to work with existing physics.
So in the end:
Use the “mental signal processing” to just not percive the noise. Our head is still more powerful than any computational device!