Going from Oticon OPNS1 to More1 aids

Your audi is completely wrong (and demonstrated a lack of basic understanding of how it works, even though she made it sound like she really knows what she’s talking about by overriding your request) that the Detail setting on the Sound Enhancer would eliminate a lot of the noise reduction. Why? Because the Neural Noise Suppression inside the DNN is already DONE before the signal goes through the Sound Enhancer. So the relationship between the noise and the speech (in terms of the Signal to Noise Ratio) is already established before it goes through the Sound Enhancer, and this SNR remains the same and does not get changed by the Sound Enhancer. Therefore, the noise reduction doesn’t get “eliminated” like your audi alluded to.

The Sound Enhancer is simply nothing but a bandpass filter (see screenshot below) that provides more attenuation on the low and high frequency ends, with more gain in the mid frequency. The Detail setting (the top curve) gives you more gain compared to the Balanced (middle curve) and Comfort (bottom curve).

If you look at the second screenshot below, you can see the effect of the Sound Enhancer in the 3 settings, and also relative to the SNR in the OPN S. As you can see, the More provides better SNR in its Neural Noise Suppression compared to the SNR achieved by the OPN S (max noise reduction in OPN S 1 is -9 dB while max noise reduction in More 1 is -10 dB by the way). This is obvious because you can see that the speech output has more gain in the More (Balanced setting) compared to the OPN S, and the noise in the More is not as loud (in Balanced setting) compared to the OPN S.

Now all the Sound Enhancer does is boost BOTH the speech and noise up a little by the same amount of gain, as can be seen in the Detail graph, and reduce both the speech and noise by the same amount of gain in the Comfort graph compared to the Balanced graph. So NO noise reduction gets “eliminated” like your audi said. The noise reduction is already done up front before it goes to the Sound Enhancer phase, and the SNR of the noise reduction remains the same ratio regardless of the Sound Enhancer setting. All the Sound Enhancer does is to either boost up both speech and noise signals in Detail, or lower down both speech and noise signal in Comfort.

The purpose of the Sound Enhancer is to allow for individualized customization because after the Neural Noise Suppression does its job, some folks may feel like the Neural Noise Suppression is too aggressive and takes away too much noise (meaning environmental sounds that they may want to be aware of to hear more of), while some folks may feel like the Neural Noise Suppression is not aggressive enough as they’d like it too, and there’s still too much noise. So with the Sound Enhancer, folks who feel like it takes away too much noise can choose Detail to increase the output level of the noise so they can hear more of it (and speech gets boosted further, which for the most part should be OK with the user). For folks who feel like there’s still too much noise, then the Comfort setting will help suppress the gain of the noise further for better comfort, while still maintaining the speech to noise SNR to the same level as before.

Below in the third screenshot is some verbage about the Sound Enhancer in the Oticon whitepaper that basically says what I just said above.



image

To your question “How does the aid know what helps the user needs (in terms of noise reduction) without feedback from the user?”, the answer is that it may take the answers you give it in the Personalization question (see screenshot below), which your audi should have gone over with you in your initial fitting, to formulate a judgement on how much noise reduction to apply. So there IS feedback from the user in the general sense. Beside this feedback, it may also incorporate its own judgement based on Oticon’s years of experience on how to rebalance a sound scene with the appropriate amount of speech-to-noise ratio, while in parallel allowing the user’s preference to factor in there as well. Also note that most likely, this Neural Noise Suppression is probably only activated dynamically when there’s speech going on. When there’s no speech detected in the sound scene, then the Neural Noise Suppression is probably de-activated on-the-fly to enable the “open” paradigm to work its magic in conjunction with the brain hearing concept.

What it looks like is that Oticon is doing a “double talk” on the issue of what to set the max value of the Neural Noise Suppression to. On the one hand, its online Help manual (see first and second screenshots below) tells you that the HA is smart enough to determine its own speech to noise ratio and will only apply the correct amount of noise suppression, up to the max level you set. If this is correct, it infers that you should set the max level to what your premium tier level HA is, which you paid the premium for, because why would you set it to less and cripple it to a lower tier level HA that commands a lower price? You might as well buy the lower priced premium level then, if you follow their recommended default.

On the other hand, the double talk is that Oticon is telling HCPs to leave it at their lower default values, because they know that there’s a price to pay for the max value -> a muddier ambient sound. I guess they don’t want to advertise this “trade-off” publicly because it’ll probably prompt the public to consider much more seriously the More 2 tier instead of the More 1 tier, if the public deems that the trade-off of a muddier ambient sound is not worth the premium cost to pay for the More 1.

But if I had already paid the extra premium for the More 1 level HAs, I would still set the max Neural Noise Suppression to the highest value possible. My reasoning is that if it gets to be THAT noisy that the HA has to resort to setting close to or up to the max amount that my More 1 can handle, then I’d still rather have clearer speech even at the expense of a muddier environment. The Genie online help manual has consistently said that the HA will apply less noise reduction when the environment is less noisy, so there’s no reason for me to believe that I’ll always have a muddied ambient sound, and the only time I’ll have to deal with that is when I get into a Very Complex environment, and only when there’s speech going on.

You can read up on this topic in much more details in this thread if you want:


The default for the Environmental Configuration (and for most other stuff) is not necessarily the same for everybody. It depends on your hearing loss and the answers you give it in your Personalization setup that your audi is supposed to go through with you in the beginning during the initial fitting

Mine has a default of Complex (see screenshot below), while yours may have a different default, depending on your hearing loss and your Personalization answers.

2 Likes

I would say that your audiologist needs better understanding of the aids, and also needs to listen to her patients requests.

@Volusiano: You know that I place great confidence in your posts, but I don’t know whether I’m buying what Oticon is saying about how much of the max 10dB gain More1s can apply at a particular instant in time - specifically the “only what’s necessary” part.

In my direct experience as a user of both More1s and More3s, and having tried both the default and the maximum settings available in More1s, I would say that whenever Neural Noise Suppression is activated, it is at the specific level of attenuation that has been selected as appropriate for the ambient noise level. There’s no gradual ramp-up of attenuation, and you’re stuck at that level until the noise level subsides and NNS goes to the next detent level down.

In other words, what I hear is that there are various thresholds or elbow points that activate a particular level of noise suppression triggered by the noise level, and you’re stuck there until you hit the next knee point. There’s no smooth application of power, as there is with an automobile accelerator. Uh uh! When I had my NNS value set to 10dB, I seem to remember reporting at the time that it seemed to muddy things up. Further to that, I’d also say that it was a “herky-jerky” kind of a ride, and not a smooth application of power, as Oticon is trying to imply.

This is my report based solely on what my ears have heard. It’s not based on any scientific measurements or learning on my part. It’s purely subjective.

But I find Oticon’s explanation of what’s happening to be at variance with my senses.

[Note: I’ll keep returning to this post and improve upon my explanation as better turns of phrase come to mind. I find the sensation difficult to define.]

Thanks for sharing this, Jim. I think what you’re observing is not inconsistent with what Oticon claims. It’s only because of some confusion that the expectation is not seen. Let me explain.

First and foremost, there are 2 types of noise reduction done by Oticon, both by the More and also similarly by the OPN and OPN S. I probably have mentioned this in previous posts before, but maybe didn’t make a clear enough distinction so it’s not well understood and causes confusion.

The very first stage of noise reduction is done via beam forming, using a technique called MVDR (Minimum Variance Distortionless Response) that would attenuate well placed noise sources. Below is a screenshot showing how it works, where the sounds from the 2 cars get attenuated while the voices are preserved. In the More, this noise reduction is placed in the Spatial Balancer (see second screenshot to see where it is in the flow). In the OPN/S, it is in the Balance module (see third screenshot).

THIS type of noise reduction is most likely what you hear and notice as you move from one environment to the next, where well placed noise sources are noticed by you to be attenuated. Although the Balance module does get updated 125 times per second (almost continuously smooth as claimed by Oticon), the well placed noise sources are pretty much static, and thereby the noise reduction level is “perceived” to be the same for you, because although it does get updated 125 times per second, the well placed noise sources most likely never change during that one second anyway, so it seems static.

The OTHER type of noise reduction (the Neural Noise Suppression in the More, and the Noise Removal module in the OPN/S as seen in the 2nd and 3rd screenshots below) really cannot be noticed by you, because it’s only activated when there’s speech going on, and ONLY during speech. And it’s not well placed noise sources (which are taken care of already by the Balance module), but is more like “diffused” noise, meaning the noise doesn’t have a precise well placed location, like the echos, the ambient sounds that permeate all around you and you can’t really tell where it comes from.

So it’s VERY dynamic. As soon as a continuous sentence is over (maybe in a fraction of a second), THIS type of noise reduction stops. The only thing you can perceive is that the speech may seem clearer to help you understand it better. You won’t be able to perceive any suppressed noise during the speech because all you’re focusing on is the (clearer) speech part, you don’t have time to notice if there’s noise in between the words in that sentence anyway

That is why what you observed as noise reduction and it seems more static and not as dynamic as Oticon claims, is not inconsistent with how it works. You usually and most likely are only able to observe and discern attenuation of well placed and usually mostly static noise sources, as attenuated by the Spatial Balancer. And what we were talking about is in a different zone, the second noise suppressor in the DNN zone.

So in theory, if Neural Noise Suppression only happens during speech and stops right after speech, how would a very high (max) amount of Neural Noise Suppression, even if/when applied, cause the ambient sound to become muddied? Bad enough to force Oticon to shy away from using a higher default max value? Well, we won’t really know, and all we can surmise is that maybe there’s a side effect to apply too much Neural Noise Suppression. Maybe the DNN is not fast enough to keep up with the speech when/if it has to be too aggressive, and the Neural Noise Suppression lingers on and doesn’t get deactivated quickly enough, causing users to notice a muddier ambience even after speech stops.

My personal experience with my OPN 1 is actually consistent with what was said above. For example yesterday I went to the supermarket and when I stood in front of the fish tanks (very noisy pumps going on), I hear all the pump sounds. As soon as I turned my back to them, I notice that the pump sounds got attenuated. So I know that this is done via the first stage of the noise suppression in the Balance module.

Now I’ve also tried to notice (very hard) between a speech sentence to see if I could discern suppressed diffused noise in between the words of a sentence or not, but for the life of me, I can’t tell. But I don’t expect to be able to tell. All I know is that I hear a more clear sentence, and at the end of the sentence, I still hear almost all of the surrounding sounds (with the well placed noise sources still attenuated, of course). I do have my max value set to -9 dB for my max Noise Reduction in Complex environment, but I don’t really notice the ambient sounds to be muddied in very complex noisy places. But that’s the OPN 1 and me, it may not be the case with the More and other folks.



@Volusiano: As usual, that’s a great explanation, and - now that I understand it better (I guess I only barely understood your previous post, which I remember) - I agree with your explanation of what I’m hearing, and I understand that NNS is happening concurrently with speech, and not in between syllables (is that correct?). If that’s the case, perhaps the NNS is taking enough of the edge of consonants and sibilants to create the perception of muddiness?

If that’s, in fact, the case, it would explain why aim preferring the next level down of NNS, and kicking myself for buying More1s when More2s would have served me just as well.

Thanks again for taking the time to craft an easily-understandable explanation.

1 Like

It may very well be the case, Jim (the bolded part above). We just don’t know and we can only guess.

We should remember, though, that the DNN is only breaking down and re-creating the sound components of the sound scene (at least that’s my educated guess, we’ll never know unless Oticon reveal their secret). How faithfully it can recreate the sound components is also Oticon’s secret. The training part of the More DNN, in my personal opinion, probably focuses more on the rebalancing of the sound components so that the sound scene is rebuilt to be as accurate as possible. But I never heard of any mention of the training to focus on the accuracy of the re-created sound components, but maybe they do maybe they don’t, we just never know.

But below is @flashb1024 's sharing of what Oticon told him. They never said that the person you hear would not be clear if the max NNS is applied. They only said that the ambient sound may seem distorted. So go figure…

Your explanation of ON/OFF latency could account for the distortion of intersyllabic noises, couldn’t it, V?

The only thing is that the ON/OFF latency in theory shouldn’t be a variable dependent on the amount of noise suppression applied. But then the DNN is just a black box to us and maybe there’s enough complication in there to make the ON/OFF latency be a variable dependent to the NNR value.

The only clear thing is that Oticon sells the More 1 at a premium with an advertised -10 dB NNR for complex environment, while it only wants people to use -6 dB NNR top (not even at the -8 dB NNR that it shows in the menu). See the screenshot below where the default is showing only 6 dB NNR max for the More 1 while 8 and 10 is available as well.

The More 2 has a max of -6 dB NNR for complex. So if you prefer to leave your More 1 at the Oticon’s recommended default of -6 dB, it’s probably cheaper to buy the More 2 up front instead.

Yes, that’s the model I should have bought, but - and I hate to give the Oticon Haters traction - Oticon needs to apply a Marketing-Speak Suppression Module to itself so that laypeople like me can actually understand what the Tiers are about without having to get weeks and months of private instruction by an MSEE in order to comprehend the differences.

I’ve never used brands other than Unitron and Oticon, so I can’t speak for the others, but I remember that the Unitron hierarchy was no easier to fathom than what we’re talking about here.

Well, there are other things that are better in the More 1, though, Jim. So you’re not completely put out by going for the More 1. The Spatial Balancer (what we talked about earlier as the first stage of MVDR beam forming attenuator for well placed noise source) is at 100% full functionality for the More 1, and only 60% full functionality for the More 2 and 3. This is an important noise reduction module as well, so it does help make a difference as far as NR is concerned.

There are other differences as well between the More 1 and 2, although not as significant enough to mention here. You can refer to the snapshot below to spot the other differences.

I’ve read that chart over quite a few times, but I can’t say that I understand what it all means in practical terms.

Great info. Thank you.

First, I was not asked the personalization questions, so this is news to me. I have an appointment Monday morning and will ask her about them. She has been my audiologist for a long time and is someone I trust, and our conversations are just that. If I request a setting I will get it, but we typically talk about it first. Since I am just coming to the More 1’s, I don’t know enough to request anything except, I do know what I hear; I edit videos professionally among other things and am very conscious of sound quality. The aids were initially set up with Balance in the Virtual Outer Ear setting and I was not getting enough ambient sound, so I asked that they be changed to Aware, as I said in an earlier post and it has been an improvement, but the tradeoff is that noise is more dominant and makes me work harder to understand speech. So next, I am going to ask that she bump up the neural noise suppression a notch, to see what difference it makes. I believe in going slowly through settings, so I have a chance to see how they work.

When I first got the aids, the voices sounded crunchy and I went back and she adjusted them to give me more access to the top end, she said, and that did get rid of the crunchy, but it did remove some intelligibility.

It does sound like it might be beneficial to get the programming gear, so that I can try modifying these automatic settings. I tried it once before when I had Alta Pros and before that a couple of Costco aids and before that with Oticon Deltas, but stopped because I had no idea what I was doing and I do have a life, which these are supposed to help me enjoy and not become my new work! I was so frustrated with the fitters at Costco (aids were not the Kirklands) and I tried a couple of different Costcos and fitters … that I tried to do it myself. But in this case, trying out various automatic settings may be the way to go. I would need to get her buy-in so I am not left in the cold should I still need her help. Something to think about anyway. … I am getting distracted here.

In any case, thanks for the good info. It is very helpful.

I see that the Virtual Outer Ear setting in the two lower Oticon levels have no other option than what the default it is. Limited.

Sorry, Jeremy, but what do you mean by "automatic settings "? “Crunchy” voices also has me scratching my head.

I see anything outside of the frequency adjustments as being automatics.

Crunchy to me means that the top end of voices is not smooth, but sounds distorted and possibly over modulated. I can understand them just fine, but the voice quality is not pleasant to listen to.

Okay. Thank you, Jeremy.


The Environmental configuration is simply the choice between Very Simple, Simple, Moderate, Complex and Very Complex. The More 1 and 2 have 5, the More 3 has 3 (lacks the Very Simple and Very Complex -> no big deal). These let you define which of them are part of the Easy Environment and which are part of the Difficult environment so that the parameters in the MoreSound Intelligence would fall under which and which.

The Virtual Outer Ear has Aware, Balanced and Focused for the More 1, and only 1 configuration (Balanced) for the More 2 and 3. This basically models the Outer Ear for the case of the Simple environment configuration. It’s really no big deal to miss the Aware and Focused. You hear more around you with Aware, you hear less (and more in front) with Focused. The Virtual Outer Ear is how to filter sounds before sending the sound scene signal into the DNN for processing, in the case the environment is classified as “Easy”. In the case the environment is classified as “Difficult” by the user, the Virtual Outer Ear is not used. Instead, the Spatial Balancer (noise reduction of well placed noise sources) is used for Difficult environments.

The Spatial Balancer was already covered in the previous posts. It’s an important MVDR beam forming functionality that would help attenuated well placed noise sources. The More 1 has 100% functionality, while the More 2 and 3 have only 60% functionality. It’s not clear what the percentage means. I’m assuming that it probably means the amount of attenuation -> the More 1 attenuates 100% of what it should attenuate, while the More 2 and 3 attenuates less than they should/can.

The Neural Noise Suppression for Difficult/Easy for the More 1 are 10/4, More 2 are 6/2, More 3 are 6/0. Note that these are max levels. It doesn’t mean that the More 1 always get more aggressive NNS than the More 2 and 3. Actually for the most part, they all get the same NNS until more than 6 dB is needed, then the More 1 MAY get more BUT ONLY IF you select more than the 6 dB default that Oticon recommends for the More 1. But if set at the default 6 dB, then the More 1 doesn’t do any more NNS than the More 2 or 3.

The Sound Enhancer has 3 configurations (Detail, Balanced and Comfort) for the More 1. I covered what they mean in great details in a post earlier in this thread. The More 2 only has Balanced and Comfort. The More 3 only has Comfort.

The Spatial Sound gives the More 1 4 Estimators while the More 2 and 3 have only 2 Estimators. Spatial Sound LX is the feature to give the patient a sense of direction on where the sounds are coming from. Without going into the details of what an spatial estimator is, the 4 estimators allow a separation of the spatial cues into 4 equally distributed frequency channels so that you can get more details on the spatial cues if they happen to be in different frequencies. With only 2 estimators in the More 2 and 3, there’s less details on the spatial cues in only 2 frequency bands. I would say this may or may not be important to you depending on whether it’s important to you to know where the sounds are coming from. You do have that with the More 2 and 3, but the More 1 will give you the best spatial cues available.

The fitting bandwidths of 10 KHz for the More 1 and 8 KHz for the More 2 and 3 is not really a big deal, because most of us can’t really hear worth anything beyond 8 KHz anyway. It may only help for those who still have very good high frequency hearing to be able to discern this kind of difference.

The 64 processing channels on the More 1 vs 48 channels on the More 2 and 3 is probably not really significant either. It used to be that 16, then 24 processing channels was plenty good, so I’m not sure if your ears can really discern the details that 64 channels can bring over 48 channels anyway.

Transient Noise Management has OFF/Low/Medium/High for the More 1 and only OFF/Low/Medium for the More 2 and 3. Transient noise management is used to remove discomfort due to sudden loud sounds without sacrificing speech audibility. I don’t think it’s a big deal if the More 2 and 3 can give a little bit less comfort on the More 1 here.

The More 1 has 24 fitting bands, the More 2 has 20, and the More 3 has 18 bands. It used to be that the OPN 1 has only 16 fitting bands. Fitting bands is the granularity of how you can adjust the gain across the frequency range for your Fine Tuning section. I think 16 is more than enough and 24 is overkill anyway.

2 Likes

@Volusiano: I sense that what I’m about to ask can be a bit technical, but could you please say more about spatial estimators?

I am going to say something that may sound strange, I really don’t care how my aids work as long as they work so I can hear and understand speech. That may sound crazy coming from someone that spend his working life in highly technical positions. But it is the truth, when I retired I was freed from the need to know the technical stuff. I am now an artist, and also someone that is trying to accomplish the things I never had time for before.

1 Like

Yup, that’s great, Chuck, and there’s nothing wrong with anything about that. But - ever since I was a child, taking toys apart and destroying them in the process - I’ve wanted to understand how things work the way they do.

There’s nothing wrong with that, either.