Looking for a Thesis subject

[carrejans]

Hi all,
I'm studying Electronic Engineering. Next schoolyear, I have to make a Master's Thesis. My interest is off course: audio.

Most of students will do this thesis in cooperation of a company. I contacted a lot of companies, but most of them didn't had the time/place/... at the time. The few that were positive didn't have an interesting subject.
But we are also alowed to make a thesis without a company; just in our school.

Now I was thinking of doing something with DSP or something else digital. This semester I have classes about DSP and just a few labs (only some MatLab stuff). Most of the things we learn are theoretical, so I don't have any experience with DSP in practice.

So, now I'm brainstorming to find a suitable subject. And looking for some help. 
Are there any new developments the last years on the subject of DSP in audio? New ideas, new processors, ...?
Is there something that you always wanted to see made; but have no time for?
...

All help is appreciated

Thank you.




"Master's thesis: a piece of work for conclusion of the master's program. By doing so, the student shows an analytical and synthetical capacity or an independent problem-solving aptitude on an academic level. The delivered work reflects the student's general critical-reflective or research-minded attitude."

[JimRayden]

What about an attempt to divide a mono guitar signal to six different outputs? Extracting the string signals exactly would be impossible but what if it was a synth that would analyze the input signal and use up to six oscillators to achieve a similar result to six independent audio outputs or, more practical yet, as midi output.

Well it's better than nothing.

----------
Jimbo

[carrejans]

What about an attempt to divide a mono guitar signal to six different outputs? Extracting the string signals exactly would be impossible but what if it was a synth that would analyze the input signal and use up to six oscillators to achieve a similar result to six independent audio outputs or, more practical yet, as midi output.

Well it's better than nothing.

----------
Jimbo

Thank you for your tip.
I think it will be very similar to the guitar synths of roland. And I already talked with a professor about midi; and he thinks that the time for midi as an output signal is over.

Other tips, links to usefull sites, ... ?

Thanks...

[The French connection]

Don't be afraid to contact researchers, even outside of your country, you'll be surprised sometimes how they can be happy to talk about their works and futur avenue and possibilities...Sure some would not respond...but some do!

Just a suggestion, maybe it does'nt fit at all, but you'll got the idea:

http://www.ee.surrey.ac.uk/profiles?s_id=1376

Dan

[defzeppardica]

Where do you go to school? Im a EE from Georgia Tech.

[ACS]

What about an attempt to divide a mono guitar signal to six different outputs? Extracting the string signals exactly would be impossible but what if it was a synth that would analyze the input signal and use up to six oscillators to achieve a similar result to six independent audio outputs or, more practical yet, as midi output.

Well it's better than nothing.

----------
Jimbo

That's an idea that has merit - but rather than using MIDI, use six separate A/D's.  Once you've got your digital signal you can do anything you want to it.  This way you could get much better tracking for multiple notes.  Think Whammy pedal that pitch shifts an entire chord.  Or you could drop your E and A strings an octave, keeping the original signal as well, but leaving the rest untouched.  Or to only put delay on certain strings and/or notes.  Or on the fly tuning correction - so long as the strings are 'close enough' you're sweet!

Mate, the possibilities are endless!!

[JimRayden]

Thank you for your tip.
I think it will be very similar to the guitar synths of roland. And I already talked with a professor about midi; and he thinks that the time for midi as an output signal is over.

Firstly, as far as I know, Roland uses a hex pickup for their synths, not one mono signal from regular guitar pickups. Secondly, isn't MIDI the standard on all of today's synthesizer systems or have I missed an innovation to replace that technology.

---------
Jimbo

[carrejans]

What about an attempt to divide a mono guitar signal to six different outputs? Extracting the string signals exactly would be impossible but what if it was a synth that would analyze the input signal and use up to six oscillators to achieve a similar result to six independent audio outputs or, more practical yet, as midi output.

Well it's better than nothing.

----------
Jimbo

That's an idea that has merit - but rather than using MIDI, use six separate A/D's.  Once you've got your digital signal you can do anything you want to it.  This way you could get much better tracking for multiple notes.  Think Whammy pedal that pitch shifts an entire chord.  Or you could drop your E and A strings an octave, keeping the original signal as well, but leaving the rest untouched.  Or to only put delay on certain strings and/or notes.  Or on the fly tuning correction - so long as the strings are 'close enough' you're sweet!

Mate, the possibilities are endless!!

Very, very interesting idea. But how to divide the audio signal into 6 different outputs?

- If you just use a normal guitar cable; you will have to filter the one audio signal into an output of each string. But this is impossible. You cannot 'see' in your audio signal which frequency is coming from wich string. Because you can play one same note on different strings; but the frequency will be the same.

- If you make a pick-up and an audio cable with six different channels; you will face another problem. The six different microphones in your pick-up will not only pick-up the signal of the string above, but also of the other strings.

- You could make a digital pick-up; but that's been done before. (Roland synths)

So, it is an interesting idea; but I think impossible to make.

Please, keep the solutions and ideas coming.

Thank you all

[carrejans]

Don't be afraid to contact researchers, even outside of your country, you'll be surprised sometimes how they can be happy to talk about their works and futur avenue and possibilities...Sure some would not respond...but some do!

Just a suggestion, maybe it does'nt fit at all, but you'll got the idea:

http://www.ee.surrey.ac.uk/profiles?s_id=1376

Dan

I emailed about 80 companies over the world, that I was interested in. Only 12 replied. 
This weekend I'm going to visit an AES convention in Amsterdam. I hope I can find a company willing to cooperate; and maybe some inspiration for a subject.

[carrejans]

Thank you for your tip.
I think it will be very similar to the guitar synths of roland. And I already talked with a professor about midi; and he thinks that the time for midi as an output signal is over.

Firstly, as far as I know, Roland uses a hex pickup for their synths, not one mono signal from regular guitar pickups. Secondly, isn't MIDI the standard on all of today's synthesizer systems or have I missed an innovation to replace that technology.

---------
Jimbo

You are totally right.
But like said in a few posts above this one; it is not possible to extract the signals from the different strings out of one mono signal. (I think)

But thank you for your thoughts; they are very usefull. I appreciate it.

[carrejans]

Where do you go to school? Im a EE from Georgia Tech.

In Leuven, Belgium.

[gaussmarkov]

What about an attempt to divide a mono guitar signal to six different outputs? Extracting the string signals exactly would be impossible but what if it was a synth that would analyze the input signal and use up to six oscillators to achieve a similar result to six independent audio outputs or, more practical yet, as midi output.

Well it's better than nothing.

----------
Jimbo

That's an idea that has merit - but rather than using MIDI, use six separate A/D's.  Once you've got your digital signal you can do anything you want to it.  This way you could get much better tracking for multiple notes.  Think Whammy pedal that pitch shifts an entire chord.  Or you could drop your E and A strings an octave, keeping the original signal as well, but leaving the rest untouched.  Or to only put delay on certain strings and/or notes.  Or on the fly tuning correction - so long as the strings are 'close enough' you're sweet!

Mate, the possibilities are endless!!

Very, very interesting idea. But how to divide the audio signal into 6 different outputs?

- If you just use a normal guitar cable; you will have to filter the one audio signal into an output of each string. But this is impossible. You cannot 'see' in your audio signal which frequency is coming from wich string. Because you can play one same note on different strings; but the frequency will be the same.

the fundamental frequency will be the same.  but you don't need an oscilloscope to know that each string has its own sonic personality.  if i were trying to do this, i would see whether you could "train" the software to recognize each string in analogy with training for voice recognition.  i think it's doable.

[ACS]

What about an attempt to divide a mono guitar signal to six different outputs? Extracting the string signals exactly would be impossible but what if it was a synth that would analyze the input signal and use up to six oscillators to achieve a similar result to six independent audio outputs or, more practical yet, as midi output.

Well it's better than nothing.

----------
Jimbo

That's an idea that has merit - but rather than using MIDI, use six separate A/D's.  Once you've got your digital signal you can do anything you want to it.  This way you could get much better tracking for multiple notes.  Think Whammy pedal that pitch shifts an entire chord.  Or you could drop your E and A strings an octave, keeping the original signal as well, but leaving the rest untouched.  Or to only put delay on certain strings and/or notes.  Or on the fly tuning correction - so long as the strings are 'close enough' you're sweet!

Mate, the possibilities are endless!!

Very, very interesting idea. But how to divide the audio signal into 6 different outputs?

- If you just use a normal guitar cable; you will have to filter the one audio signal into an output of each string. But this is impossible. You cannot 'see' in your audio signal which frequency is coming from wich string. Because you can play one same note on different strings; but the frequency will be the same.

- If you make a pick-up and an audio cable with six different channels; you will face another problem. The six different microphones in your pick-up will not only pick-up the signal of the string above, but also of the other strings.

- You could make a digital pick-up; but that's been done before. (Roland synths)

So, it is an interesting idea; but I think impossible to make.

Please, keep the solutions and ideas coming.

Thank you all

What about using a piezo system?  This would be far easier to isolate each string.  Yes, it would require a custom piezo solution (afaik they con't have six outputs as standard - but I may be wrong!!), but then you can easily run six outputs from there to your A/D's.  In fact, if you set the guitar up with a USB connection to the processor, you could put you A/D's on board at the guitar, drawing their power from the USB bus...

[JimRayden]

You are totally right.
But like said in a few posts above this one; it is not possible to extract the signals from the different strings out of one mono signal. (I think)

Well I believe it's pretty doable, in one form or another. Would need some massive amounts of frequency spectrum analyzing and a whole bunch of lines of code.

One way could be identifying peaks in the frequency spectrum - one would just have to start from low frequencies, and when detecting a major peak, start an oscillator of that frequency. Thereafter it would subtract the assumable harmonics of that note from the rest of the frequency spectrum and continue moving upwards with the same process.  And as gaussmarkov suggested, one could set up the machine by playing different notes on the guitar as the program prompts, so it could adjust its harmonics database to your guitar. Pretty utopic but, why the heck not?

One wouldn't even have to do it at a great rate since the oscillators take care of the sound fidelity, a few hundred times a second would be sufficient I think.

Besides, using hex pickups would be just plain boring.

---------
Jimbo

[carrejans]

the fundamental frequency will be the same.  but you don't need an oscilloscope to know that each string has its own sonic personality.  if i were trying to do this, i would see whether you could "train" the software to recognize each string in analogy with training for voice recognition.  i think it's doable.

What is exactly this "sonic personality"?
But wouldn't this "sonic personality" be different on each guitar and brand of strings?
If I understand you correct, you have to use the audio signal for the "training" and not the frequency?

[carrejans]

What about using a piezo system?  This would be far easier to isolate each string.  Yes, it would require a custom piezo solution (afaik they con't have six outputs as standard - but I may be wrong!!), but then you can easily run six outputs from there to your A/D's.  In fact, if you set the guitar up with a USB connection to the processor, you could put you A/D's on board at the guitar, drawing their power from the USB bus...

Hmm, very good thinking. I will investigate the piezo system this weekend. Thank you.

[carrejans]

Well I believe it's pretty doable, in one form or another. Would need some massive amounts of frequency spectrum analyzing and a whole bunch of lines of code.

One way could be identifying peaks in the frequency spectrum - one would just have to start from low frequencies, and when detecting a major peak, start an oscillator of that frequency. Thereafter it would subtract the assumable harmonics of that note from the rest of the frequency spectrum and continue moving upwards with the same process.  And as gaussmarkov suggested, one could set up the machine by playing different notes on the guitar as the program prompts, so it could adjust its harmonics database to your guitar. Pretty utopic but, why the heck not?

One wouldn't even have to do it at a great rate since the oscillators take care of the sound fidelity, a few hundred times a second would be sufficient I think.

Besides, using hex pickups would be just plain boring.

---------
Jimbo

Few, I had to read this twice to understand what you mean. Theoretical it sounds that it could work; but if it is doable in practice. I have to think about this for a while. 
Do you think that the output will still sound nice; and not artificial?


Thank you all, for contributing to this topic. Very usefull.

[JimRayden]

Few, I had to read this twice to understand what you mean. Theoretical it sounds that it could work; but if it is doable in practice. I have to think about this for a while. 
Do you think that the output will still sound nice; and not artificial?

Artificial? Oh heck yeah, definitely not as pretty as using hex PU. It would certainly have at least some glitches and "jumps", especially when pulling off double stops and such. My idea behind this is not to make it control digital patches of different guitar models or any sorts of realistic instruments, but rather use it as an input for an (perhaps virtual-)analogue synth where excellent tracking and guitar-y feel aren't that important. It would certainly need a bit different approach to playing also, just like EHX's monophonic synth pedals and octavers, where you'll just mess up the tracking if you don't take it gentle and back off the tone knob.

The problem with frequency analyzing is that it needs a pretty long sample of the signal, so that sets the "refresh rate" of this device quite low (not influencing oscillator fidelity, as I mentioned before). The sixth string is 83Hz, so one would need to aim lower than that. Then again, I might be way off here, I'm not familiar with the exact algorithms.

---------
Jimbo

[gaussmarkov]

the fundamental frequency will be the same.  but you don't need an oscilloscope to know that each string has its own sonic personality.  if i were trying to do this, i would see whether you could "train" the software to recognize each string in analogy with training for voice recognition.  i think it's doable.

What is exactly this "sonic personality"?
But wouldn't this "sonic personality" be different on each guitar and brand of strings?
If I understand you correct, you have to use the audio signal for the "training" and not the frequency?

i think JimRayden and i are thinking along the same lines.  as an example, each string has different amounts of each of the overtones that accompany a fundamental.  this is one of the things that makes the strings sound different even though you may play the same note on each.   so when you pick out a fundamental from the signal, you also pick out its overtones and look at the relative amplitudes.  that may be enough to identify which string is making the note.

and yes, each set up would probably require some training with the software so that it can memorize the relative amplitudes.  because it seems a lot simpler than word recognition, i am guessing that a relatively small amount of training would be needed.  who knows?  maybe you play 4 notes on each string and you are calibrated.

[carrejans]

Thank you guys for all your input.

I've made an overview, to show to my thesis-partner tomorrow to discuss the possibilities.

Code Select Expand


PROCESSING ELECTRIC GUITAR OUTPUT PER SEPERATE STRING


The ultimate goal with this product:

- pitch shifting each string seperately (no need for tuning your guitar for alternate tunings)
- put delay on only those strings, chosen by the user
- put any effect on only those strings, chosen by the user
- ...

Two possible ways of dividing the guitar signal into 6 seperate outputs

Possibility 1: - keeping original pickups
                    - connection between guitar and the processing box with normal instrument cable
        - dividing the audio signal into the 6 different signals inside the box
                    - processing the seperate signals inside the processing box (pitch shifting, delay, ... with the help of DSP)

Possibility 2: - using/making pickups with outputs for each string
        - using/making a cable for the 6 seperate signals
                    - or putting ADC's inside the guitar/pickups and sending the signals via ethernet or USB cable
                    - processing the seperate signals inside the processing box (pitch shifting, delay, ... with the help of DSP)

Solutions for the problems of splitting the signal

Possibility 1: - problem: same note on different strings have the same frequency
                    - possible solution: each string has its own sonic personality. The fundamental frequency will be the same, but each string                                                 
                   has different amounts of each of the overtones that accompany a fundamental. So we pick out a fundamental from the signal 
                   and pick out its overtones; then we have to look at the relative amplitudes.
                   - another problem: this sonic personality will be different on each guitar and probably each string brand
                   - possible solution: before using the box, the user has to "train" the software to calibrate the system (cfr. voice recognition 
                   software)

Possibility 2: - problem: when using a normal magnetic pickup, each polepiece will not only pick up the vibrations of its corresponding
                    string
                    - possible solution 1: isolating each string and its accompanying polepiece from the others
                    - possible solution 2: using piezoelectric pickups
                    - possible solution 3: using optical pickups (fairly recent development (1969) that works by sensing the interruption of a light 
                    beam by the string

Advantages & disadvantages

Possibility 1: + no extra pickup has to be placed, originals can be used
                    + no special cable is necessary, normal instrument cable can be used
                    - theoretical possible; but is it possible in practice? Difficulty?
                    - user needs to train the software
                    - will the sound loses its character? Will it sound a bit artificial?

Possibility 2: - extra pickup needs to be placed
                    - requires a special cable
                    + easier to implement in software
                    + no need to train the software
                    + no excessive filtering; sound will be less artificial