I’ve studied signal processing extensively in grad school, I can’t speak specifically to the biological side of this, but I think I understand what’s probably going on and try to explain it in lay terms.
I’m sure you’ve seen waveforms for music or recorded voices or whatever before so you have some intuitive understanding of what they mean. Our brains are generally very good at identifying frequencies and separating this out, so if we have a nice clean signal, it’s not terribly difficult to isolate certain frequencies. Our brain’s ability, or lack thereof, isn’t what’s affected by hearing loss, but it’s our ear’s actual ability to faithfully detect the signal and translate it into an electrical signal that our brain can interpret. As a result, our brains get less useful information and we result in having hearing difficulties.
But this affects our ability to hear things by virtue of how it affects the degradation of the signal. Different frequencies will be more susceptible to different types of interference. I’d suspect that, in a lot of cases, basically when we have hearing damage, those nice sharp wave forms aren’t as clean anymore. When this frequency is considerably louder than the surrounding background, it’s still easy to account for that, but this is different when we start getting background noise. When the frequencies are nice and clean, a steady background noise doesn’t make much difference in detecting it, because it’s modelled by adding across all frequencies, so the prevelant ones will still stand out. But once your signal isn’t clean, it’s not just a single frequency, it’s not got small associated frequencies, and the greater the background noise, the more of those smaller frequencies will get lost in the noise.
I’ll try to do an analogy using vision, let’s use black and white for simplicity. We can liken hearing damage to like adding a solid grey value across the board to the whole image, so whites get darker and blacks get lighter, like turning the contrast WAY down, and background noise is the same as the snow on an analogue TV. So using this analogy, if we’re looking at a solid white circle on a black background, it would take a whole lot of noise and/or a whole lot of damage to make it difficult to discern. But that’s like detecting someone using a tuning whistle. Speach would be like trying to do this with an object that has some depth and complexity to it. Imagine now, rather than a solid white circle, a ball with a light source. Without the background noise, we may be able to tell it’s a ball even with a lot of grey washing out a lot of the detail. But now when you add the snow, all of the wash out is completely overwhelmed by the noise. If the snow is intense enough, you’ll be lucky to make out generic shapes and only because they’re separated by significant differences in intensity.
As a fairly simple experiment, try having someone talk to you in a way that’s a bit muffled, like just covering their mouth and nose with their hands, or talking into a cup or something. It’ll be a bit harder than usual in a quieter environment, but you’ll notice if you then try to talk to them with the TV, a fan, the dishwasher, and whatever else going on at your house, it’ll be noticeably more difficult because some of those subtle things you could pull out just muffled or unmuffled but with noise, get completely lost.
As for what’s going on, this is probably a lot easier with some basic wave forms and their corresponding FFTs, which is how we translate signals into frequency space. If we have good hearing we see our signal should be fairly high in the expected space and low over the rest of the space basically like a normal curve that has a high, narrow peak with the average at the signal. As our hearing gets worse, we’ll receive that signal, with the same area under the curve, but with greater variance. Thus, the peak gets lower and broader. Broadband noise is basically just adding random values across all frequencies, or modeled by basically adding the same value to all frequencies. So if you add a flat amount across a narrow, tall normal curve, that point still stands out reasonably well, but the broader and shorter that peak is, the more quickly it will get lost in even moderate noise.
But basically, your ears just get crappy at gathering the signal. Still, if there’s only one, even if it’s bad, we can still generally figure out the signal without too much effort, but once you add broadband noise, the clean signal will degrade a lot less quickly with background noise than an unclean one.