Hearing aid

In the UK at least the NHS digital hearing aids are not as pretty as the ones available privately but their DSP firmware is almost as good. The main difference is one of size with the private one being tiny.

The on ear and fitted to the shape of the ear fed by an acoustic tube solution is just about as good as it gets. In ear is too small and particularly fiddly for the elderly to deal with comfortably.

--
Regards, 
Martin Brown

Never the less they do when they are out of the ear and placed on a hard surface close together. Many deaf people use the howl round as a battery test (which is not necessary as it goes beep every couple of minutes for the last hour of battery life) but old habits die hard.

I agree that the units should be able to notice acoustic feedback and silence it but I know from experience that my fathers would scream all night if they were not switched off properly.

--
Regards, 
Martin Brown

g aid. But it's not something I've designed before. The musts:

e

s you can get that only run a few days on coin cells. Using AAA or AA is po ssible.

of one that would do this. The 386 has min 4-5v, Iq 4-8mA, hardly ideal, no t totally ruling it out but don't love it.

. AAA 0.3-0.5Ah at 8mA = 50 hours, and that's just I_q. No use at all.

very low voltage amp design or with sliding class A.

hnology is not likely to produce a useful design.

. ADI makes a family called Sigma perhaps? I think TI has a very similar family. They are intended for apps based on digital audio data so the arch itecture is designed to process a fixed algorithm on each sample, triggered to start when the sample arrives. It's been a while so I don't recall the details other than they are very small, require little power and are under $10.

ead. I haven't checked the other thread yet and probably won't bother.

ther performance figures are generally better but the part count is high co mpared to a 358. So on balance I think I prefer the 358. I did consider des igning a slider from scratch, but I'm not experienced with them, and the 35

8 is a quick & good solution. It's hard to get simpler, and simple means le ss time/cost to construct.

Why not pipe a cell-phone electret mike into a few BJT's, out to a crystal earphone, done?

Cheap, single-cell operation, power-thrifty, etc. Lots quieter than a '358, too.

Cheers, James Arthur

I think a hybrid solution could work well. a diff pair input stage designed for very low noise to give a portion of the gain to the mic signal, and then the second half of the gain from the op amps.

You could wire up the two op-amp sections and outputs from the diff pair collectors such that the two 358 sections are driven differentially and drive the speaker in a bridged configuration from a single supply for more power-handling capability.

Input diff pair current source transistor could possibly be used to apply AGC.

What type of discrete amp do you have in mind? The only power thrifty confi guration is class B, and on a single cell the distortion would be severe.

Crystal earpieces today are junk. They're way too unreliable to design into any product. Not to mention the appalling sound.

I'm not seeing what the upside is of adding a discrete front end.

Differentially driving the earpiece doubles current consumption, and is nor mally done when there isn't enough voltage swing available from the supply rail. It also increases v_out without changing i_out, thus requiring a high er impedance transducer. And one of the challenges here is running a transd ucer that's overly low impedance for an opamp. So I'm not seeing an advanta ge in this situation.

Yes. OTOH it's gone from 1 opamp to 2 opamps plus a discrete front end. If I do an improved version later I think one of the first things I'd go for w ould be 2 mics to enable some noise cancelling. The basic nature of these p uts them at the milder end of the hearing loss scale, where people who stru ggle to hear properly with background noise, so cancelling of the more dist ant sounds would be a real benefit.

What would be the absolute minimum way to implement agc on an opamp?

NT

LM358 is a dual opamp. I think some type of AGC is a more important feature for a bare-bones hearing aid than active noise-cancelling personally, think of all those posts people complaining about trying to watch TV late at night and the audio is too dynamic have to keep turning it up and down. Same sorta thing.

A discrete front-end can get you better noise performance (you're dealing with small input signals and the '358 is not a low noise amp!) and somewhere to inject an AGC signal.

I guess if you just wanna use one 358 section to drive the speaker the simplest AGC would be to use the other as a precision rectifier and filter and feed back a DC voltage to control the gain of a discrete diff pair, either to control the tail current or as bias on the second input that's not being fed signal.

It sounds complicated to describe in text but these aren't really complex or expensive circuits on paper

onfiguration is class B, and on a single cell the distortion would be sever e.

into any product. Not to mention the appalling sound.

ir

d

normally done when there isn't enough voltage swing available from the sup ply rail. It also increases v_out without changing i_out, thus requiring a higher impedance transducer. And one of the challenges here is running a tr ansducer that's overly low impedance for an opamp. So I'm not seeing an adv antage in this situation.

If I do an improved version later I think one of the first things I'd go f or would be 2 mics to enable some noise cancelling. The basic nature of the se puts them at the milder end of the hearing loss scale, where people who struggle to hear properly with background noise, so cancelling of the more distant sounds would be a real benefit.

If you use the 2nd one, current consumption goes up. As far as I can see, B TL isn't needed here. In fact if more P_out were wanted I'd parallel the op amps.

sure. I wasn't expecting it to be a significant issue though, will have to see.

I'm more inclined to prefer a tr/FET, 2 resistors, 1 cap, 1 diode as agc. I f there's a simpler way I can't think of it. There is if P_out were higher, but it isn't.

NT

bitrex wrote

A simple JFET can be used as AGC, a diode works too at low levels.

Fun, any idea how they did the mixing? Diode ring mixer?

George H.

Diode ring is just one way to construct a doubly-balanced mixer.

The signal needs first to be shifted pretty far up in frequency to get rid of undesired mixing products in the passband. It can then be filtered and mixed back to near baseband.

An I/Q image-suppression mixing scheme helps a bit, but it does not provide a way to completely avoid the undesired products in analog mixing.

In digital domain, it may be possible to get the frequency shift directly by creating an analytic signal from the input and mixing it with an analytic frequency-shift signal.

--

-TV

I'm a fan of bare-bones engineering but not ultra-bare-bones, I guess. At the low end the performance gain you can get by spending another couple pennies are massive compared to the performance gain you can get by spending another couple dollars on something that costs $50 in parts already.

I don't buy everything off-the-shelf but design time is worth something, and it's often difficult and time-consuming for mere mortals to leverage trash into treasure.

e.g. a while back someone posted a design for a low-cost micropower boost converter based off a 4093 in an unusual configuration, i've breadboarded it several times at this point with slightly different part values and inductors of slightly different specs and 4093s from different mfgrsm it's cute but only one of three of my attempts worked. It's very difficult to simulate too, why it should work in one of these configurations (which wasn't even the configuration specified in the schematic) but not the slightly different other ones is mysterious.

It's cute and cool when it works but I think way too fiddly to actually be good for anything produced in quantity. It took time to figure that out which could've been better spent clicking a button and ordering a few hundred off-the-shelf boost converter ICs

Appalling sound might still be helpful to someone who is struggling to hear and make sense of speech. You only need 300Hz to 3kHz and a bit of distortion might not necessarily affect intelligibility. The brain is rather good at hearing the fundamental based on a series of harmonics.

Better noise performance AOTBE.

I recall seeing an old simple design for AGC using a JFET and 358. Here:

No idea if it is any good. Might limit choice of supply voltage though.

The original article seems to be missing the circuit diagram so I have linked to a discussion of it instead.

--
Regards, 
Martin Brown

Used crystal earpieces are hen's teeth. New ones are garbage, with around 3

0% not working even when new. They also have no hope whatever of doing 300H z-3kHz.

I used piezo discs in a speaker design once, highly efficient. Whether I ca n get them to behave in an earpiece I'm less confident.

air

nd

But will I need it?

-agc-circuit

One extra transistor :/. Will see if I can do without the bipolar first. I' m not building it yet though. A BJT in lieu of the fet would be a plus, I w as vaguely wondering about putting one across the input. I probably don't n eed wide gain swing.

NT

On a sunny day (Tue, 18 Dec 2018 11:18:08 -0800 (PST)) it happened George Herold wrote in :

There are 2 different things, pitch change, and frequency shift.

Changing the pitch is like playing a tape faster and slower but it stays 'musical', for example

1000 Hz becomes 1010 Hz and 2000 Hz become 2020 Hz. Here you need to drop or repeat samples I'd think. The harmonic relations are preserved.

Shifting the frequency makes 1000 Hz 1010 Hz and 2000 Hz 2010Hz and destroys the harmonic relationship, like in a SSB receiver. There the ring modulator may comes in.

Digital: In Linux these days almost every distribution has 'sox' and in sox you can adjust speed, pitch and tempo all from the command line. Digital has latency (a lot often).

I have played with sox, and with SSB.

There seem to be LSI chips, the HT8950:

Here about hearing aids:

That is all I think I know :-)

Google may help you more, has a lot on this subjct including circuit diagrams.

Grin... thanks. GH

The hand-wavey explanation used in the software world is "profile and see." It's almost as easy to "profile" simple analog circuits as it is to re-factor and re-compile profile software configurations for testing.

The in-ear type is more popular among young people with hearing loss who don't want to look prematurely elderly ;)

Sure, but for the most common form of high frequency hearing loss, piezo works just fine. It's the lower frequencies, which require larger piezo elements, that are the problem. Todays hearing aids go up to about 8KHz so users can properly hear music. Some researchers want even more which is another reason to use piezo: "Considerations in Fitting Hearing Aids with Extended Bandwidths"

"A Piezoelectric Transducer for a Hearing Aid Using PZT Thin Film" The article is from 1998. I couldn't find anything on the results of the study.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558