Have a question about multimedia or audio? Ask our Expert, Ken Boyce

Hi Nate,

I don't know of a single device of the type you describe.  If one existed, it would have to have separate ability to program the EEPROM.

TI has some digital input Class D amps, but high power ~ 20W.  The digital input is in the I2S format, commonly used in consumer electronics audio equipment.  This type of part  does not have any internal EEPROM that could be programmed.

Maybe the best bet to prove the concept would be to use a D-A after the external A-D and EEPROM, then feed the new HOH analog input a part like the National LM48310  or other analog input Class D amp at   http://www.national.com/cat/index.cgi?i=i//277  or some similar parts at TI  (TPA2001) at  http://focus.ti.com/paramsearch/docs/parametricsearch.tsp?family=analog&familyId=923&uiTemplateId=NODE_STRY_PGE_T.  You could hear the result on an 8 Ohm speaker,  or thru a headphone if you chose a headphone amplifier from either of these 2 suppliers.

Ken

Hello Ken:   From your old friend Nate, the 90+ y/o guy.   When I talk to experts in the hearing aid field I end up proceding down the wrong pathway.  They have the solid belief that improved hearing can be accomplished in the present manner, namely though the control and adjustment of the frequency response characteristics..

     Since I come out of the field of radar, my thoughts are  that hearing is in some important details quite analogous. In both there is a detection phase and a lock-on follow-up.  In the hearing process there is similarly a detection phase where sound is first heard, following which as the sound gets nearer it can be judged to be speech and might now be understood.  The entire concept hinges on the fact that there are two components of human speech, two separate languages, interleaved to form into a code of sorts of noise sounds and tones, consonants and vowels.  The big prolem for the deafend is that the noise-like sounds of the consonants are 10,20, 30 dbs lower than vowels.  And that is where the problem lies.  The consonants are the first to be 'lost' as we become deafend.  You have to hear both sides of the speech code in order to understand.  Like in the Morse Code, one must hear both the dits and the dahs in order to understand the message.  Frequency is hardly involved in the process of understanding.   I have spelled it out in the Wikipedia web page under the title  "user:roselamb"  if you have the interest.

Finally my question.  Help me to understand how the ADC functions.  Does it take as the input an analog signal, as a voice for example,  and convert it into discreet levels, numbers,  corrensponding to the amplitude of the  input signal?  That would account for the amplitude consideration.  But what happens to the corresponding frequencies of the input signal.  They have to be brought into the equation somehow.  That is where I get lost. Please help me to understand how both amplitude as well as frequency result from the ADC conversion process.    Thanks again.   Nate Almond

Message was edited by: nathan alMOND

Hello Nat,

Take a look at en.wikipedia.org/wiki/Digital_audio

There are loads of references on Wiki to sampling and sampling theory but most are rather confusing.

If you lookat the diagrams on the page I refer to you can imagine that the signal is represented by numbers like 0,0.5,0.866,1,0.866,0.5,0,-0.5,-0.866,-1,-0.866,-0.5,0 (that's one cycle of a sine wave at 12 samples per cycle).

You can see that if you plot it out it makes  a sine wave.

But the bit you are missing is that from just one sample you can tell (almost) NOTHING about the signal. You need to look at several samples to estimate what the signal would look like in the frequency domain.

This shouldn't be a surprise - if you look at your sine wave plot though a narrow vertical slot you can see only one point, and you can no longer tell what the shape of the rest of the signal is.

This is why ALL digital filters need more than one sample (usually a lot more) to function and all analogue filters use some kind of energy storage to keep a "memory" of the preceding signal.

Hope this helps.

Michael Kellett

Thank you very much Michael for the fine understandable explanation.  It seems so logical once you explain it.  In other words, coming out of the ADC are numbers corresponding ( example 0 to 255 for an eight bit ADC) to the input amplitude levels, the twelve levels you used to explain it to me.  Like you said, those levels correspond to just one cyle.   The frequency then is the number of these cycles per second, the hertz, of the input to the ADC.

Now that I have an understanding, what I wish to accomplish is to reverse this output.  By this I mean, that where the output might be 255, I wish to convert it to a 0.  And do this at each bit level which should end up with the loudest input sounds becoming the quietest and vice versa.  That is what I am trying to accomplish.

A long time ago Ken Boyce gave me some practical suggestions on how to go about implementing  such a goal.   Don't know it I can still find the email he sent. Can you clue me in again on how to reverse the output numbers of the ADC?  It probably would involve creating a simple arithmetic algorithm which might amount to a software programming exercise.    Give me more of your ideas, if you would please.

Thanks once again for you highly understandable explanation of the ADC.

Hello Nate,

Long time no hear from :>)

I hope all is well with you.

A very good tutorial on ADC's is at the following URL.

http://www.hardwaresecrets.com/article/How-Analog-to-Digital-Converter-ADC-Works/317/2.

There are 10 pages, and the link above takes you to page 2 of the tutorial where the meat of the explanation takes place, but you should start on page 1 ( http://www.hardwaresecrets.com/article/How-Analog-to-Digital-Converter-ADC-Works/317/1 ).

I will answer your question about capturing the frequency information in the following way:

If you take a look at the article, and at Michael Kellett's reply to you, you already know that the input analog signal gets converted to a digital number by a process known as sampling, which simply means that you take the conversion a some specific time and what ever the signal level is at that time is the level that will be converted to a digital number.

Intuitively you know that if I only sampled that signal one time, it would not be correct representation of the true input signal. Right?  Kinda like describing an elephant by  looking at a fly on its back.   So, to get a truer representation of the input signal I am going to have to sample it quite often.

Also intuitively you know that if I sample it more than once, it would be better to sample it at very close intervals rather than at far apart intervals.  Closer interval sampling means that I can capture the smaller variations in the signal better.

Now imagine sampling a 1KHz sine wave.  Should I sample that once every second? No...since I would only pick up one value on one cycle of the sine wave and miss all the rest.  Ideally, I would want to sample at a rate which is closer to the expected highest frequency of the input signal so I don't miss much during the conversion.

Fortunately, this problem has been extensively studied, and way back when, a gentlemen by the name of Nyquist developed his famous theorem (paraphrased as follows) which says that if you want fully recover in the analog domain a digitally sampled signal, then you must have orginally sampled it at 2X the highest frequency of interest.

So for Hi Fi audio which has an upper bandwidth of 22KHz, the ADC samples that signal typically at 44.1 thousand times a second, e.g. a 44.1KHz sample rate.  The beauty of the Nyquist Theorem is that you could feed those 44.1K samples directly to an analog low pass filter that had a upper cutoff of 22KHz and Voila!, out comes your original analog input signal.

But sample rate is only part of the problem.  You would want to know how closely the sampler actually got the right value, and you would also want to have sufficient resolution in the sampler in order to distinquish correctly one value from another.

So the ADC has to have a resolution aspect to the number, or what is the precision of the number.  So all ADC's have a resolution specification which says how the digital number (almost always binary) is represented.  For typical audio, the typical resolution per sampled number is 16 bits.  However, professional audio engineers often go to 24, 32, and even higher, bits to get an even closer representation of the input signal.

And I will add at this point that although I mentioned 44.1KHz sample rate, professional audio is often sampled as high as 384K samples a second which also gives a closer representation of the input signal.

Modern audio ADC's usually have a selection of bit resolutions and sample rates that enable them to be used with almost any audio input signal.   The higher the bit resolution and sample rate used, the better the input signal in it's entirety is captured.  That means you not only capture the signal amplitudes, you also capture the signal frequency components.....so long as the sample rate is a minimum of 2X the highest audio frequency you want to capture.

DAC's do the reverse....taking in whatever the digital representation is and converting it back to analog, as well as incorporating digital filters which do the job of the analog filter I was referring to earlier, and Voila! clean analog signal out.

CODEC's combine both ADC and DAC in one package.

I hope this helps you.  Enjoy the article.  You will learn perhaps more than you intended.

Kind regards,

Ken

Hi Nate,

The suggestion a long time ago was to feed the ADC sample output into a content addressable memory which basically used the ADC output values as the memory address, and for that address, the memory value would be programmed ahead of time to be the opposite of the address.

So if the sample value was 1111 1111 1111 1111, in the memory at that location would be the number 0000 0000 0000 0000, and so on.

I believe Nicholas Grey also provided an alternative way.  You should find both by looking at the element14 expert entries under the persons name.

One thing to keep in mind in that most ADC's have serial data output, not parallel data output.   The problem might be as simple to solve as inverting the bit stream data, then feeding it to a DAC.

Ken

Ken and Michael:  Can't thank you two enough for your detailed information with the wonderful tutorial explanations of how the ADC works. It was more understandable than any text might do.   I will follow up on all the web sites you and Michael  provided me in order to help in my greater understanding.

What is so sad is that all I ever wanted to do was to go out and buy a hearing aid that emphasized the consonants.  The ones I tried don't provide the understandability that is needed. Didn't seem like it was asking too much.  I could not find one word of advertisement  that said the magic words "this hearing aid emphasizes the consonants."   Just that simple.  I had to delve into this further to get the answer.   For persons  with perfect hearing which may describe your own situation, to see this defect in action, try using the system for speaking into a microphone using the Nuance company device called Dragon Speaking (I'm guessing at the exact title).  The written output is typed out on the computer.  The results are pure humor..   Just like the deafened world, it does not detect/hear consonants with the resulting written words and sentences that are hilarious. Yes it is most laughable too when deaf individuals, like the Nuance system,  try to guess at the consonants.

Since consonant sounds are constructed in a completly different fashion than the vowels, these differences can be used to amplify the consonants. Consonants are always the weaker of the two languages of speech. So what I wish to do in the easiet possible manner (for me at least) is to simply reverse the order of the output of the ADC so now the loudest portion of the speech input, the vowels,  become the quietest. Aah, just what I was looking for.  What is accomplished by this reversal of sound level is to make the consonants loud enough so that the two parts of speech can always be heard and thus understanding by the deafened is accomplished.   Never once was an ear deficiency, such as diminished frequency response addressed. What I have  accomplished by this exercise is to provide an infinitely  better input signal to the ear, or to the hearing aid if one is being used..

This same technique is usable in all speech applications, whether is be in a broadcast studio or elswhere.  No more over-modulation of speech monitoring is required in the broadcast world..

I offered my ideas to Nuance Communication, to  my University of Southern California,   to Staley Hearing Aid Co., to the State of Israel, got not an ounce of interest.  Now I am looking for Venture Capital to proceed further.

So I will continue my slow progress in learning how to make this happen, knowing full well that there are millions of technical people like you and Michael infinitely more suited to doing this project.

Thanks once again for your great patience and wonderful helpfulness.    Nate

Ken and Michael:  Can't thank you two enough for your detailed information with the wonderful tutorial explanations of how the ADC works. It was more understandable than any text might do.   I will follow up on all the web sites you and Michael  provided me in order to help in my greater understanding.

What is so sad is that all I ever wanted to do was to go out and buy a hearing aid that emphasized the consonants.  The ones I tried don't provide the understandability that is needed. Didn't seem like it was asking too much.  I could not find one word of advertisement  that said the magic words "this hearing aid emphasizes the consonants."   Just that simple.  I had to delve into this further to get the answer.   For persons  with perfect hearing which may describe your own situation, to see this defect in action, try using the system for speaking into a microphone using the Nuance company device called Dragon Speaking (I'm guessing at the exact title).  The written output is typed out on the computer.  The results are pure humor..   Just like the deafened world, it does not detect/hear consonants with the resulting written words and sentences that are hilarious. Yes it is most laughable too when deaf individuals, like the Nuance system,  try to guess at the consonants.

Since consonant sounds are constructed in a completly different fashion than the vowels, these differences can be used to amplify the consonants. Consonants are always the weaker of the two languages of speech. So what I wish to do in the easiet possible manner (for me at least) is to simply reverse the order of the output of the ADC so now the loudest portion of the speech input, the vowels,  become the quietest. Aah, just what I was looking for.  What is accomplished by this reversal of sound level is to make the consonants loud enough so that the two parts of speech can always be heard and thus understanding by the deafened is accomplished.   Never once was an ear deficiency, such as diminished frequency response addressed. What I have  accomplished by this exercise is to provide an infinitely  better input signal to the ear, or to the hearing aid if one is being used..

This same technique is usable in all speech applications, whether is be in a broadcast studio or elswhere.  No more over-modulation of speech monitoring is required in the broadcast world..

I offered my ideas to Nuance Communication, to  my University of Southern California,   to Staley Hearing Aid Co., to the State of Israel, got not an ounce of interest.  Now I am looking for Venture Capital to proceed further.

So I will continue my slow progress in learning how to make this happen, knowing full well that there are millions of technical people like you and Michael infinitely more suited to doing this project.

Thanks once again for your great patience and wonderful helpfulness.    Nate

Ken and Michael:    Once again this is Nate Almond, the Ancient  One.  By now I have reached the ripe age of 91y/o and have been sidetracked in my efforts at producing an improved hearing aid, what I choose to call an understaning aid.   The reason for the side tracking was the fact that I began studying the structure of human speech and found it had evolutionary impact such that it caused homo sapien to go from just another tribe of something like ape-man to what he now labels himself as "human."  It was the cause of the Cognizant Revolution, call it if you wish, the "missing link,  in the field of paleontology. Ancestors of the hominids used the same language as all other surviving creatures, the common feature being that it was produced sound by means of the vocal chords.  Air from the lungs blowing past the vocal chords, (vibrating membranes)  can produce huge sound levels, right from the instant of birth. It was part of the Darwinian theory of survival of the fittest.  In this case fittist meant developing vocal chords as a protective means.   This continued for more than 2.5 million years of hominid existance in the same manner for all animals excepting homo sapien. About 100 thousand years ago this hominid developed a second  language using  means other than the vocal chords.  The distintively different sounds came from a new source, from the mouth, tongue and lip combination.  And now  the genius idea was born.   The new language was not simply an add-on feature to the old language but was interspersed into the first language to produce the concept of time element into the two languagesystem.  It could now be articulated  such that 'before', 'now' and 'in the future' was part of the new language.  Also, sounding for ideas as greater than, same as, more than, could be articuated.  The new sounds, what is by then  called speech,  became the basis for modern history, science and mathematics of present human existance. Cognizant revolution, indeed! This was a first audible 'publication'. Many thousands of years later, the developemnet of hierographics attempted to put those sounds into a permanent form.

Nate,

A BIG CONGRATULATION to you on your 91st birthday.  How many don't make it that far?  Furthermore, you still have your brains, and a quest for knowledge and a desire to somehow make the world better for yourself, and in the process, better for others.  Now that is a fullfilled life.  Aside from that I bet you are a real character!

I really am sorry no company has taken up your ideas for possible further development, but overcoming the status quo is always a challenge.  Sometimes you win - other times you don't.

Language is a great thing but the twisting of it we now see in many areas of human endeavor is heart breaking.