Looking for ADC for cochlear implant signal

Thread Starter

aaron_m

Joined Apr 27, 2014
13
Hi there. I'm an "amateur hacker" with a few years of informal experience, and have spent a bit of time looking into this, but I figure it's now more helpful to ask the experts for assistance. I started reading through Art of Electronics for information on ADCs and that's a deep rabbit hole, to put it mildly.

I'm trying to decode an AC signal to the coil of a cochlear implant (CI): I wear one, but have found that the newer models are awful in terms of sound quality, so I'd like to create something of my own. The base frequency of the CI coil is 2.5 MHz. This is actually an older model - Cochlear Spectra 22 - and the newer models run at 5 MHz. The signals are sent in packets *1 and to test the output of my homegrown "processor," I need to be able to decode the packets coming from the sound processor to the coil that I wear on my head. So far, the problem I'm having is that the decoder "microprocessors" that I've used - NXP's LPCXpresso55S69, and MIMXRT1060-EVK - don't seem to be able to capture the ADC signal at the 5 MHz speed I need to satisfy Nyquist.

Given a signal frequency of 2.5 MHz, but with the likely need to increase that to 5 MHz, what should I look for to handle these speeds? Should I look at a specific ADC chip, or is there a particular microcontroller that will adequately do the job? What do I even need to know to properly evaluate a chip / µC? Any particular tools that might help? Or, possibly, does a decoder like this even exist already?
 

Attachments

MrChips

Joined Oct 2, 2009
30,720
The trace on the oscilloscope which you have presented represents a digital waveform. You don't need an ADC to capture this information. What you need is a fast digital processor somewhat similar to I2C. You would need information on both the low level hardware protocol, i.e. how the data is encoded in the digital signal, and the protocol at the information layer, i.e. how to extract the information.
 

Thread Starter

aaron_m

Joined Apr 27, 2014
13
The trace on the oscilloscope which you have presented represents a digital waveform.
From this link, on page 122 - which is actually only 7 pages down because it's from an article in a journal - does the description of the "Radio Frequency (RF) Link" square with what you wrote? Without a doubt, the level of detail in the trace as presented isn't enough to show the whole wave.

https://cpb-us-e2.wpmucdn.com/facul...2008-Zeng-Rebcher-Sun-IEEE-RBME-CI-review.pdf
 

crutschow

Joined Mar 14, 2008
34,285
That RF signal is digitally modulated so an ADC of that signal would serve no purpose and give no useful information.
You need a circuit that can decode the ASK modulation of the RF.
 

Deleted member 115935

Joined Dec 31, 1969
0
@arron_m ,
well done you for experimenting,
Can I make a suggestion, as your dis satisfied with the implants, and want to experiment,
you a perfect person whom the manufacturers would love to work with,
Have you tried talking to the manufacturers, may be a email to the CEO ?
 

Thread Starter

aaron_m

Joined Apr 27, 2014
13
@andrewmm, at the risk of getting off-topic, no, I haven't. From what I've seen from Cochlear's products, literature, and support groups, combined with my decades of experience with product development - I used to work on technical products for Fortune 100 companies; just not on the electronics side - my sense is that Cochlear hit a home run and owned the market with their initial model, and now they're coasting along by adding "features," at the expense of fundamental quality. Is it a fair assumption? Maybe. Maybe not. Cochlear would not be the first company to take that approach. However, it's very unlikely that one unsatisfied customer, by contacting the CEO, is going to cause them to re-evaluate their approach to their established, profitable product line. If I can develop a better solution on my own, that might become the edge I need to prompt a re-evaluation. If nothing else, my sound quality would be directly under my control.
 

Deleted member 115935

Joined Dec 31, 1969
0
@aaron_m,
thank you
I do admire your work, and I wish you the very best , it sounds like its about time the market was shook up.

I was not so much thinking of a letter of complaint, more one of offering to be a help to them, may be an experimental employee.
I see from a quick google that there are at least 5 manufacturers of implants,
I don't know anything about them apart from what I have read. I'm guessing they are all incompatible, so you have to stick to the company you have the implant from.

good luck to you and let us knwo how things go,
 

Phil-S

Joined Dec 4, 2015
238
Erring on the safe side, I don't think I would experiment on implants, however passive the approach.
I have a life-saving implant and wouldn't go near it with a bargepole.
I would voice your concerns about sound quality to the implant team who would be in a better position to contact the manufacturers.
An incredible amount of work goes into designing and fitting these devices, but as @andrewmm says, your input as a user might be more valuable
 

du00000001

Joined Nov 10, 2020
117
From this link, on page 122 - which is actually only 7 pages down because it's from an article in a journal - does the description of the "Radio Frequency (RF) Link" square with what you wrote? Without a doubt, the level of detail in the trace as presented isn't enough to show the whole wave.

https://cpb-us-e2.wpmucdn.com/facul...2008-Zeng-Rebcher-Sun-IEEE-RBME-CI-review.pdf
Hi aaron_m,

your reference clearly states "ASK" for 3 manufacturers listed (Table II, p. 122), so basic decoding is the same as for AM broadcast: a rectifier (aka e.g. diode) followed by an appropriate low-pass. This is independent from the modulation scheme applied (whether OOK, PWM or Manchester coding).
The frame scheme you supplied seems to indicate that within the frame data is inserted in some time-division scheme. Could be you have more information about the timing within the frame.

A remark: as hearing with a Cochlea implant has to be learned (not very different from "normal" hearing), you might just not be accustomed to the "audio experience" of newer encoders/algorithms (that might be really better when trained). Is there no opportunity to train a newer device ?
(I have no implant, but I'm somewhat informed about the operation of Cochlea implants as well as "vision aids" (basically an implant for the eye - not that different from the Cochlea implant, although more electrodes at a lower update rate).)
 

MrChips

Joined Oct 2, 2009
30,720
Honestly, I know very little about cochlear implants and I can only imagine that it is a very complex system.
I would not be messing around with the electronics of a cochlear implant.
 

metermannd

Joined Oct 25, 2020
343
Keep in mind that there are two parts to a cochlear implant, and the portion OP is looking to work with is the external processor, not the internal implant.

It's not just the newer implants that don't work in the same way as the older ones did. I use a hearing aid in my one 'good' ear, and the type of loss I have simply needs a raw dB boost so I can hear properly. The newer digital hearing aids are more processing of sound and not raw oomph - my current device doesn't even have a volume control!
 
Last edited:

Thread Starter

aaron_m

Joined Apr 27, 2014
13
Why not peek inside your unit to look at what they use and start upgrading from there?
I have - a few times. It's an old unit, out of warranty, with no replacement parts, and if anything goes wrong, I fix it as best I can. The main processor is a Motorola DSP chip, from what I can tell. The markings are:
AWM
AU85410
5410
9452

But, I figure since that was done back in late '80s, early '90s, the processors we have now should be plenty powerful enough to do the job without having to delve into a DSP-specific chip.
 

Thread Starter

aaron_m

Joined Apr 27, 2014
13
Hi aaron_m,

your reference clearly states "ASK" for 3 manufacturers listed (Table II, p. 122), so basic decoding is the same as for AM broadcast: a rectifier (aka e.g. diode) followed by an appropriate low-pass. This is independent from the modulation scheme applied (whether OOK, PWM or Manchester coding).
The frame scheme you supplied seems to indicate that within the frame data is inserted in some time-division scheme. Could be you have more information about the timing within the frame.
Based on this and the earlier comments, that's the direction I'm heading.

A remark: as hearing with a Cochlea implant has to be learned (not very different from "normal" hearing), you might just not be accustomed to the "audio experience" of newer encoders/algorithms (that might be really better when trained). Is there no opportunity to train a newer device ?
I remember way back in '97 when I first got the implant with the first mappings that it sounded pretty true to what I remembered from when I lost my hearing in '84. That was with the Spectra 22. But with the newer Freedom, the sound was awful right off the bat. I told the audiologist between map tests that something was wrong. She said, "You'll get used to it." Disappointed but willing to try, I wore it for two weeks and it was as bad as not having a cochlear implant at all. I could hear that a noise had occurred but not what it was. Conversation required that I read lips, again. My wife was so grateful when I resumed wearing the Spectra 22.

Anecdotal and definitely a low sample, but I spoke with another cochlear implant wearer I'd met and in our discussion he said he also thought the newer units were a vast step backward in terms of sound quality. Take that for what you will.
 

MrChips

Joined Oct 2, 2009
30,720
I have gone through three generations of hearing aid designs from analog to digital.

I have moderate hearing loss in my left ear and severe hearing loss in my right ear, -80dB owing to a tumor impacting on the auditory nerve.

My current pair of hearing aids are vastly superior to the previous ones and as a classical guitar player I can now enjoy the quality of sound that I had previously enjoyed.

While there is no volume control on the devices themselves I can still control the level from a remote Bluetooth controller. One of the nice things with the technology is I can stream audio directly from my iPad. The devices also have different preset programs for different audio environments.
 

du00000001

Joined Nov 10, 2020
117
I have gone through three generations of hearing aid designs from analog to digital.

I have moderate hearing loss in my left ear and severe hearing loss in my right ear, -80dB owing to a tumor impacting on the auditory nerve.

My current pair of hearing aids are vastly superior to the previous ones and as a classical guitar player I can now enjoy the quality of sound that I had previously enjoyed.

While there is no volume control on the devices themselves I can still control the level from a remote Bluetooth controller. One of the nice things with the technology is I can stream audio directly from my iPad. The devices also have different preset programs for different audio environments.
A Cochlea implant is way from a hearing aid in terms of "hearing":
While your hearing aid can be considered an amplifier plus an equalizer plus some speaker (complex enough to get this right), the Cochlea implant equals a spectrum analyzer with some speech processing. And while your aid's actuator is a continuous spectrum speaker, the implant's actuator is no more than 22 electrodes locally triggering a bundle of nerves each (each electrode assigned to some part of the audio spectrum).The auditory experience certainly is by no way comparable, although it still makes a big difference whether you "hear" absolutely nothing or 22 "strings" played now and then.
 

du00000001

Joined Nov 10, 2020
117
I have - a few times. It's an old unit, out of warranty, with no replacement parts, and if anything goes wrong, I fix it as best I can. The main processor is a Motorola DSP chip, from what I can tell. The markings are:
AWM
AU85410
5410
9452

But, I figure since that was done back in late '80s, early '90s, the processors we have now should be plenty powerful enough to do the job without having to delve into a DSP-specific chip.
Motorola and 5410 doesn't look like a Motorola DSP (these would usually start with 56 - known as the 56k family).
Although I couldn't find this exact number, it might be an MCF5410 - which would be a Coldfire V4 architecture. Basically a RISC successor of the 68k family. Might incorporate a bit of a DSP extension (the MAC).

Today, you can find DSP extensions on a lot of MCUs/MPUs as the MAC is really versatile for the algorithms of filters, FFT and some more signal processing. So I'd expect to see a core with DSP extensions on a recent Cochlear processing unit as well. ("Pure-bread" DSPs seem to be a dying race other than in highest-end applications, but DSP extensions are spreading farther and farther.)
 

Thread Starter

aaron_m

Joined Apr 27, 2014
13
The auditory experience certainly is by no way comparable, although it still makes a big difference whether you "hear" absolutely nothing or 22 "strings" played now and then.
According to the literature, the processor takes the top 6 signals per sample and encodes the channels and volumes for those. It doesn't stimulate all 22 channels per sample.
 

du00000001

Joined Nov 10, 2020
117
<...snipped...>
I remember way back in '97 when I first got the implant with the first mappings that it sounded pretty true to what I remembered from when I lost my hearing in '84. That was with the Spectra 22. But with the newer Freedom, the sound was awful right off the bat. I told the audiologist between map tests that something was wrong. She said, "You'll get used to it." Disappointed but willing to try, I wore it for two weeks and it was as bad as not having a cochlear implant at all. I could hear that a noise had occurred but not what it was. Conversation required that I read lips, again. My wife was so grateful when I resumed wearing the Spectra 22.

Anecdotal and definitely a low sample, but I spoke with another cochlear implant wearer I'd met and in our discussion he said he also thought the newer units were a vast step backward in terms of sound quality. Take that for what you will.
Just checked about the learning/adaptation curve etc. @ wikipedia:
Basically, the learning curve slows down with age and might take years. I assume, having experience with "some" Cochlea implant might lower acceptance for newer development . trying to avoid another learning process. (No, this is not meant to insult you - just an assessment.)

Weird that there is no backward compatibility mode in recent processors: should not be too difficult. OTOH, when it comes to a device meeting its fate, you just cannot avoid another learning phase :confused:
And while the basic algorithms (at least the filtering ones) are not too difficult to implement, there's quite a bit to do to achieve the performance of even your old device. The band-pass/filter approach is quite well documented - not that sure about the rest of the processing task...

Good luck with that!
 

du00000001

Joined Nov 10, 2020
117
According to the literature, the processor takes the top 6 signals per sample and encodes the channels and volumes for those. It doesn't stimulate all 22 channels per sample.
6? Already in the Spectra 22? Not that bad.
Channel encoding won't be that difficult, "volume" control might be: I'm not sure how volume would be controlled as this is not above just increasing some voltage. The "nerve interface" is quite different from that. Anyway - if you have access to some documentation and are able to measure a bit (you should even be able to "tap" the transmission), I assume the air interface is just about transmitting some "volume level" while the real task (to create some nerve stimulus from the volume level) is the job if the device that resides under your scalp :)
 

Thread Starter

aaron_m

Joined Apr 27, 2014
13
Just checked about the learning/adaptation curve etc. @ wikipedia:
Basically, the learning curve slows down with age and might take years.
:oops: Years?! I already spent 13 years not being able to understand people in a conversational setting before I got my implant. I'd really like to avoid something like that, again.

I assume, having experience with "some" Cochlea implant might lower acceptance for newer development.
That was the problem: If it were just po-tay-to / po-tah-to - minor differences - I'd try to make the new ones work. But it was like going back to no implant at all, really. With the Spectra 22, I can carry on conversations with my wife while driving in a car; I can focus on driving and just listen to her. With the newer models, I'd have to look at her to read her lips, taking my eyes off the road, which is obviously not good.

OTOH, when it comes to a device meeting its fate, you just cannot avoid another learning phase
I can *try* to avoid it: a) by taking very good care of my Spectra 22; and b) by trying to figure out how to process sounds on my own. Go down trying, right?

And while the basic algorithms (at least the filtering ones) are not too difficult to implement, there's quite a bit to do to achieve the performance of even your old device. The band-pass/filter approach is quite well documented - not that sure about the rest of the processing task...

Good luck with that!
Thanks!
 
Top