How Much Gain Compression is Too Much?

I made an appointment today at Costco to try the the DSL v5 formula. My fitter happened to be there and made the appointment. I am obviously her “best” customer as she does not even ask me my name. She even offered to do the fitting on Christmas eve.

Are you sure there isn’t more to this relationship than you think? :heart_eyes:

I hear what you say, but all I need is a pair of hearing aids that work as well as they can. They are pretty good right now, so I will have to eat some crow if DSL messes them up…

Obviously my attempt at humor was weak, very weak. :disappointed:

High compression helps with comfort but also hearing soft sounds. The NL2 formula generally prescribes higher compression settings as they are trying to maximize speech intelligibility.

This approach works well in quiet but breaks down in noise where the extra soft sounds amplification boosts noise arriving from a distance, and for non speech type signals like music where the compression kills important dynamic modulations.

Lower compression settings (or more linear settings) are often attributed to a more natural sound quality and may provide better performance in noise or in situations where soft sounds may be annoying. Music signals are also less constricted dynamically so less compression can help here as well.

A system that can modulate the compression is probably the ideal setup and for this reason a lot of manufacturers have adaptive compression systems.

3 Likes

I got it and chuckled. But, she is married, as am I, and she is far too young for me. But, on the other hand, I am only 70, and when you are in the hearing aid fitting business, 70 might look pretty young compared to the usual crowd.

1 Like

Well we will see what it will do in a week or so. I am a touch worried about it as the sound of my aids is doing the best they have so far. So, there is a bit of a “if it ain’t broke, don’t fix it” thought in my head too. In any case I know how the software works and it is a few clicks to put it back to where it is now. No need for REM, and I could likely get it done by just dropping the aids off for a few hours.

As far as adaptive compression goes, I believe these aids have it. But, it only seems to apply to transients, with the attack and release time constants.

Are you describing the (comfort-clairity) settings?
Thanks

I am a few hours into using DSL v5 for a prescription. My gain may still be slightly high for my left ear which does not take to lots of gain very well. Overall the impression is that sounds are quieter and more natural sounding. There is slightly less bass perhaps. For sure there is more clarity in the high frequencies. I am hearing things I did not hear with the SmartFit/NAL-NL2 prescription (they were pretty much the same). The SmartFit seemed to give a real edge to certain voices on TV, and particular ones where the original sound quality was not the greatest. That seems to be gone. So far, so good.

I was getting a touch of feedback in my left ear, so I went up from a Small to Medium closed click sleeve, and it seems to help. I noticed when the fitter was doing the testing that my left ear was more sensitive to feedback. The no-go feedback zone on the screen was larger than the right ear.

Will be interesting to see how it does in different environments. Have not tried listening to music on my stereo yet.

Interestingly the fitter played with the gains on the left ear for quite a bit of time, and in the end had a curve that looked almost exactly like the one I produced in my first post.

1 Like

No I believe you are referring to sound recover2 which is frequency compression, a different feature altogether

2 Likes

In the Phonak Target Global Tuning section there is an option to adjust the compression (gain compression) to semi-linear or linear. If someone wanted to try less gain compression that would seem to be an easy way to try it without changing the basic formula. However it changes both loud sounds and soft sounds compression. I was interested in changing the soft sounds, but not the loud.

Have now tried listening to my stereo while using this DSL v5 prescription with my music program. There is no noticeable decrease in bass. If anything the highs are clearer. I am hearing some things in the music that I did not hear before. I am going to have to go back in to get an adjustment made on my left ear as it is more prone to feedback than I like.

I get the same impression as you did with the DSL v5 Adult on my OPN as opposed to the Oticon VAC+ rationale. Things sound sharper with no noticeable decrease in the low frequency. I resort to this DSL rationale when I run into challenge speech in noise situation and want a little more boost on the highs. But normally I just use the Oticon VAC+ which sounds more natural to me, probably just because I’ve been used to it for a long while.

Compression ratios in sound studios run from 1.5 to >10 (one box would even over-compress, make big small and small big). Ratios 4:1 to 10:1 are used to restrain over-enthused singers so they “sit in the mix”. It is very probable you do NOT want 4:1. The old literature on WDRC seemed to like 1.5 maybe 2. This is also where I would go to map a wide-range performance “linearly” onto narrow DR media (cassette tape), though cheap tape generally did not sport WDRC.

How much floor-squeak and refrigerator rustle do you want to hear? While ‘normal’ ears can hear 0dB SPL, the background ambient in most homes is 35-45dB SPL; and the very young ignore or discount those very soft sounds. You have not heard them in years. Empirically you set base gain to get these sounds just-audible, and the first knee not much higher. A 2nd knee was a luxury in my old job, but would go at the louder part of the meat of the performance, to shave over-peaks. 62 seems low. I’d have to think on it but it’s late.

They mean different things to different people. However “Dual” used to be a very specific meaning. A semi-syllabic, 0.1 second, time-constant so peaks did not distort (tape, transmitter, phono). And a longer, 1 sec, TC so a sustained loud sound would be ducked and not pop-up at every break. The two together can work Amazingly Well. I am so glad I invented it in the 1970s. (Based on a $$$ product from the 1960s, which turns out got it from a product in 1938…)

YOU want to play with those compression ratios. However I agree completely that this +plus+ the strong EQ and a rather opaque “set of knobs” means you will make a muck-up before you get good at it. I have spent an hour+ setting-up compression on a 10-minute track. With less controls than modern HAs have hidden away in their bowels.And some fairly awkward artifacts.

The outcome of going to DSL v5 is that overall sounds seem more natural, and less harsh, except for the occasional female “s” sound. There might be a slight loss in overall volume as I have noticed that I have tweaked the TV volume up some from where I was with NAL-NL2 (SmartFit). We were out to a restaurant a couple of nights ago that was fairly loud and I did quite well hearing conversation at our table of 4. Didn’t have to resort to my Noise/Party program, or narrow down the microphone focus, or turn the volume down.

The unique feature of the DSL v5 at least with the type of loss I have is that compression is very low and approaching 1.2 for the soft sounds, but there is still some significant compression in the loud sounds, with up to 1.7 for compression. But still way less than what I was getting with NL2.

I can’t detect any impact of any changes in the time constants. I believe they are supposed to be adaptive in any case, so who knows what they really are…

2 Likes

Note that in some situations this is “what you want”. Say I recorded a conference, talkers soft and loud near and far. For listening comfort we want everybody about the same loudness. (We have a family reunion tape with this problem, and I was able to smooth it a bit.)

But you are interfered. Somehow we need a machine to know “wanted” from “unwanted”. People usually want the “near” sound. If it is also louder (because nearer) then simply don’t compress softer sounds up all the same. Lower compression ratio.

However if the far sounds are just loud, a simple compressor won’t help. Multi-microphone arrays can sense distance, but when as small as an HA they are physically limited what they can sense. If the interfering sound is non-speech, there are good (or promising) algorithms to sense typical speech pitch pace and dynamics, and separate from non-speech sounds.

Judging time-constants on running sound, either the artifacts are obvious, or it takes deep experience with familiar sound to guess the time constants.

An example is an old film on a lazy TV broadcast. The loud parts are consistently loud, banging on the limiter. If the sound goes “silent”, it is quiet for a second and then the film-hiss starts to rise, sometimes near as loud as speech. Then when a sound happens the hiss ducks out, then rises in the next silence. If it is a very simple compressor you can judge the recovery time.

But even lazy TV stations use dual constants. One sub-second, one several-second. Used well, the fast constant is only working for a part second before the long constant brings the average gain down.

The non real-time way to measure time constants is to apply a soft, a loud, and a soft sound, and see/hear/measure what happens. I generated a steady tone and trimmed it to jump -26dB -6dB -26dB. It happened to be 10 seconds (VERY long for speech) so I used long single time-constants.

Oh, and recordings call “zero” the MAXimum level, while audiometry calls normal threshold “zero”, so the numbers are all negative.

At 2.6 seconds the level jumps up 20dB, to 4dB higher than the -10dB “knee” that I set. Ratio is 4:1, so the 4dB over is compressed to 1dB over. 3dB reduction. But not instantly! It drops from 4 to 1 in 250mS. When the over-loud ends at 6.4 sec, the compression recovers along a 2,500mS time constant; -29dB to -26dB at about 2.5sec after loud part ends.


Aside from avoiding annoyance, a prime concern for speech (especially into a poor listener) is to not let the loud vowels pump-down subsequent sibilants or fricatives. Say “off”. It really sounds like “OOf”. If the vowel is gained-down, the “f” is lost. OTOH a very fast time constant will let the “f” back up; but then all sounds are the same loudness and important clues are lost. In this case the vowels and fricatives dominate different frequency bands, and a smart machine can process them differently.

1 Like

Discussion of why fast or slow compression makes sense depending on the situation (largely the user need and preference).

To MY experience, a many-second TC on loud sounds reminds me of unattended TV programming, where the compressor is constantly gaining up/down to keep the long-term average “good”. Especially when you have an uncompressed path also (better ear, or large vent), and no great video to watch, the balance shift is obvious and may be unsettling. OTOH it preserves short-term dynamics which are important in speech, and on-average keeps them in the user’s dynamic range.

At the other end: I just got my aid and I can hear the "Noise Reduction"s time-constants. They are not wrong; the release could be slower at the cost of audible noise-rise after a sound.