musician w electronics project -planning stage

[Phidelity]

Hello All

I am a complete newbie to this forum with very little electronics knowledge and I have a project based on an existing model that i found a paper see link below.

http://soundlab.cs.princeton.edu/resear ... rs/esitar/ (there is a pdf discussing some of the design specifics)

The project is to create an electronic sitar controller from an acoustic sitar. I will detail my design changes and reasoning below.

I have subscribed to this forum hoping to solicit advice on my adaptation of the above design ( whether the changes i need to make are feasible with my limited knowledge) and hopefully some guidance with implementation and or finding someone with the expertise to help realize the project. I am likely going to need to find assistance locally to verify that my testing and implementation is accurate. And unfortunately i live in a small community without many seasoned audio electronics specialists.

On to the specifics of the design and my needs.

First the goal is to allow me as a player to track my performance and gestures to extract pitch, amplitude and tempo information to be mapped in a number of ways. in its simplest form a straight one to one mapping to control electronic sounds' pitch, amplitude and tempo and eventually other more creative mappings. As many of you are audio industry professionals you know that pitch tracking an audio signal accurately and quickly is a very challenging process.

If you had an opportunity to view the original design - the original "esitar" controller was made using a dc voltage supplied into a resistor network that was soldered to the frets of the sitar and routed to/from a microprocessor for dealing with all the programming and logic. Creating a microprocessor capable of this is what i am hoping to avoid as that type of programming and design is well beyond my skill. However i think i have a viable work around. I own a modular synthesis program called kyma which i am confident i can utilize their prototypes/ environment to handle the programming and logic.

The first problem of the design is how to track the frequency of the sitar signal and in the original design a DC voltage was bussed via the string being played via the network back to the microprocessor. As the performer pressed the string into contact with a fret the circuit would be closed as each fret was connected to a resistor network wired in series and the amplitude of the signal was measured to obtain to a pitch. Here's where i think i can avoid the microprocessor. I was thinking why use a dc voltage - if i used an audio frequency signal then i could generate a sin wave of X frequency ( say 10 khz) from kyma and route that to the resistor network wired on the sitar and if the level of the signal were either line or mic level I would have my source signal whose amplitude would vary depending on which fret was played and that signal could be routed back into kyma for analyses. The amplitude could be easily measured in kyma and mapped to a specific pitch. Using a standard audio signal would eliminate the need for specialized electronics as i could simply use mic pres that i have at my disposal. Preferable to use a mic level signal as I read somewhere that the human body is less conductive to high frequency AC signals and the lower voltage would also pose less of a health risk ( immediate shock or long term exposure) and having a gain stage to return to kyma could be beneficial. Padding the output of a sin wave oscillators to achieve a mic level signal on the output stage of Kyma's hardware outputs is a simple matter and its even if there is added noise in the signal it is of little concern because we are not measuring the fidelity of the signal only the amplitude. This is the most critical part of the hardware design and is still a theoretical idea by me, having a very limited electronics knowledge. If everything i have mentioned above is not omitting any important theory please let me know.

The resistor network - for me this is all a bit daunting but from reading it seems there are three main design issues :

1) getting the resistor network to be scaled accurately - including an accurate measurement of the fret resistance and string resistance. I have done some DC measurements but i am not sure if these figures will change using a high frequency AC signal will alter these figures- still digging and this may not be a bit concern.
2) getting the output impedance of the signal to the resistor network circuit accurate which goes hand in hand with the next point.
3) getting the output impedance of the signal from the resistor network to the mic pre or line input to be a in the appropriate range for "healthy" signal flow.

I understand that it's best to have a low impedance output flowing into a higher impedance input but how to achieve this is another matter. I also intend to use a series resistor network as in the orignal design and i was recommended to use "military grade" resistors and i haven't any reason to deviate from this. Lastly I have a friend who is an electronics engineer ( and a synth head like myself) that is willing to help but he has got a promotion at work making his time very limited. Which is why I am here. I am hoping to do as much of the leg work as possible and get him to verify what work, measurements, etc.

I realize that this is a forum and community where it is intended as an exchange and unfortunately there isn't a lot i can offer to any other members for any assistance given other than my eternal gratitude.

So is this a much more complicated design to implement than a non electronics major could manage or should even consider managing ? FWIW i have been working my way through this site http://electronicstheory.com/COURSES/EL ... 101-42.htm

Are there any other concepts or hurdles that i have, in my ignorance, overlooked?

I will update this post with more specifics later but I thought it would be prudent to start here and see what the experts have for input.

Thanks in Advance

phidelity

[Phidelity]

oops this should have been put in the build section - mods feel free to move.

[JR.]

Too much to respond to point by point.

The esitar approach is workable. A friend of mine made a midi controller that was essentially a bass guitar with similar resistor divider fret board. (This was a commercial product manufactured by Peavey quite some time ago -19'90s).

While you could use the resistor string to divide an AC signal as easily as a DC signal, you still need to map this detected division ratio to frequency or pitch which will not be simple linear.

A cheap microprocessor seems like an ideal solution...but these kind of controllers are not simple.

JR

[JR.]

I have no idea of the complete history,, only that I knew one guy who made a commercial product "bass to midi" controller that worked, and also was playable as a normal bass guitar.
=====
Back in the '70s I had a roommate who made a guitar to synth interface, and he wrested with many of the expected problems trying to convert wiggling strings to accurate pitch information. His worked too, AFAIK.
=====
Knowing what I know about vibrating masses, just reinforces that this isn't trivial to do. My midi-bass friend who was still messing with such things made an updated guitar to synth controller, but he took it off the market a few years ago...Peavey long ago obsoleted theirs, so apparently the market for these is limited.

JR

[Phidelity]

Too much to respond to point by point.

The esitar approach is workable. A friend of mine made a midi controller that was essentially a bass guitar with similar resistor divider fret board. (This was a commercial product manufactured by Peavey quite some time ago -19'90s).

very cool

While you could use the resistor string to divide an AC signal as easily as a DC signal, you still need to map this detected division ratio to frequency or pitch which will not be simple linear.

A cheap microprocessor seems like an ideal solution...but these kind of controllers are not simple.

JR

yes however i will have a finite amplitude to represent the frequency and once i have a specific data set for the range of the voltages that cover the range of frets then that is a manageable problem within kyma ( something i own and have knowledge of - thus saving time and money)

[Phidelity]

You might find that pitch analysis is a bit more challenging than simple switches will allow.
A lot of systems that need accuracy and speed are using RADAR, laser or IR transducers to measure string vibrations and convert to absolute pitch.

Im not familiar with this technology. i will google and get back to you

With Sitar or Sarod, you do not want quantization, those instruments are commonly played with very slight pitch shifts that quantization (ie wired frets) can not follow. You will probably need to have live data to analyze the absolute frquency of each string individually, process in a microcontroller and send to a DAC to control a synth or whatever tone generator you will use.

yes, - but playing ornaments ( meends - slow pitch bends and gammaks - rapid ones ) and having that aspect of performance tracked accurately will be an evolution of the controller and within the kyma environment there are pitch tracking modules . If i can help the pitch tracker by giving it a lower finite pitch and a range that it looks for the performance of the module improves dramatically. The first step is to give the program the range and this is the purpose of the resistor network. At this point in my consideration of this project the resistor network/audio preamp ADC solution seems like the fastest and cheapest sollution obtainable with my skill set.

but your point is one that i am hopeful is solveable but i wont know until i can start supplying a signal to the synth

and your point regarding live data - yes i will have a mic'd signal too ( it will be needed to extract amp envelop information )

[Phidelity]

Phidelity,
If the controller can extrapolate nuance, that's half the battle. I worked on a live recording of a Sarod trio led by Bruce Hamm (bassist Stuart Hamm's brother)... Although I know nothing of classical terminology of the genre, what I do know is that the nuance of these stringed instruments are absolutely fluidic. Would be a real mind twister to translate into a serial data stream I would think.

yes absolutely and that is the dream controller!!

On that note i found a paper that electrical conductivity does change with tension. As you probably know when pulling pitchbends on a sitar the string is pulled down and when pulling a min3 to P5 there is an significant increase in the tension - different from the sarod). but i haven't found any specific data sets on how much tension increase is required for the effect to be significant ( i was hoping/ imagining) if the range of this effect could fall within the range of two adjacent resistors in my network then that may be my solution as i wouldn't require pitch trackers.

I would agree that you should start with the simple solution first, especially if the CV input on the synth is an exponential response... Makes CV generation a whole lot easier, and I think nearly all synths are exponential CV input nowadays. Then you can look at the necessity for response curves to impose on the CV output.

My old PAIA modular stuff I built in the 70s was linear CV input, so the keyboard had the curve, and that was a bear!! Then I built a keyboard scanner that had an 8 bit linear DAC on the output after getting into the newer oscillators that had exponential inputs, and that was much more stable than the old R string approach, after I got the data latches to calm down that is...

Got any links to the modules you are using?
rf

very interesting accomplishments you have outlined !!

Kyma is a modular, hardware accelerated, sound synthesis engine - imagine max with a hardware component to offload the sound synthesis from your computer's cpu. Its a sound synthesis environment in which you can go as deep as your knowledge, skill and dedication can take you. you can write code and develop your own prototypes, or use OOP to write scripts to acheive your goals but that is not required to utilize its huge library of protypes and sound structures . it is a really deep audio synthesis tool. R string approach and data latches are terms i am not familiar with. but converting a signal from log to linear is a simple as dragging a prototype with that capablity into the signal chain ( but getting the log scale correct is the trick). FWIW -the lovely thing is i can use any audio signal as a cv and at sample rate!!

http://www.symbolicsound.com/cgi-bin/bi ... ny/WebHome

I was thinking about your point, the voltage from the controller, and imagining whether it would be easier to use the same size resister on each fret then deal with any scaling internally or would it be better to try to scale the cv amplitude changes (via the resistor network first) beforehand making the interpretation of the data easier. As i said i haven't any formal electronics training so this is all a bit nebulous. thanks for sharing your experience with these considerations as it is a problem i am due to tackle.

I will literally have to sit down with pen and paper. After doing this i should create a resister on my sitar and plug it and get some real numbers and if necessary reconfigure the network. i mean soldering 15 resisitors isnt the hard part.


thanks again for the thoughts

phi

[mediatechnology]

I'm not qualified to post in this thread other than to say it's extremely interesting. Thanks for joining us Phidelity and this contribution.

I did do some work in pitch extraction for speech and if you can get a close estimate from the fret board I think the job of estimating the actual pitch would be made far easier.

[Phidelity]

Thanks for the words of support and encouragement from the members here and Nice rack Raf!

So I spent a good amount of time over the weekend preparing a spreadsheet with a combination of empirical data ( dBm and dBu scales as reference voltages, along with some measurements of the resistance of the frets and strings on the instrument). In conjunction i plotted a few formulas to allow me to change the value of the resistors int he network in order to be able to quickly make adjustments to the design. This has led to a few questions.

1 during all my synthesis / digital audio studies and practices i realized i have taken for granted what is actually being sampled during the ADC process and so i dug into a few texts and dbl checked online to confirm a change in amplitude in an audio wavefrom corresponds to a change in Voltage in the electrical system representing the waveform. Which leads me to the to reason for my confusion next - shouldnt the current be varying in an analogous fashion? From my understanding when you add resistors (load) to a circuit your are impeding the flow of electrons not the force driving those electrons. But even as i write the previous sentence i realize with all the reading on the subject that a resistance to flow of electrons must also create an impediment to the force driving them too.

it has me confused because when i attempted to calculate the voltage drop in the circuit when adding a resister (resisters) i came to an impass. how do you calculate the voltage drop or difference between a circuit's voltage between two resisters in series compared to the voltage after to two resisters? i originally thought it was a simple matter of using Ohms law but there is a flaw in my logic/ understanding because according to my calculations the voltage doesn't change only the current does (as more total resistance is incorporated into the circuit).

So either my conceptualization is wrong or i have made an error in my methodology in calculations.

thanks

phi

[Phidelity]

So I realized I was being totally daft in my conceptualization and calculating in that i forgot I total = I 1 = i 2 = i3 in a series network which affected all my subsequent voltage figures. So now my chart is fully functional and my comprehension/ knowledge is improving

now i have two problems to tackle. The first is at low resister values the current is very high, ie at 50 Ohm and nominal (+4 dBu /1.23 v) voltage supplied to the circuit the current is really high 24.55 mA which is way above the typical line level inputs (according to my calculations using dBm scale as a reference a line input having a large headroom of 20 db -a big assumption as that is a lot of headroom- would mean the current would be in the range of 12 mA). Conversely, if i increase the resister values by the time i have added enough resistors to cover all the frets the resolution has decreased to less than a half a dB per fret in the final half octave of frets. Aside from the curve of my resistor network not being scaled in a way that matches the way analog signals are scaled it has the added problem that i am expecting a change in the voltage when plucking the string. I have been told that the distortion in the shape of the conductor/ string is going to create a distortion in the source signal as it create a physical change in the conductive quality of the metal, and I am hoping that change is going to be less than the value of my resister. If i am using a 50 Ohm resister ( the string is conductive with a nominal measured resistance ) does the voltage change , with a loud pluck of the string more than the voltage drop that the 50 ohm resistor will create between adjacent frets ( gates)? I wont be able to know this without wiring my network and testing it but I cant just yet because of the next problem.



The other problem is that a low resister levels it creates a situation where the impedance matching from the source signal to the resistor network is close to ideal (as the output from the synth is approx 100 Ohms and the ) but the output resistance of the controller at the opposite end of the network is near 2200 Ohms which is above the rating for both mic preamps that i own. From my readings it seems best to create a situation where my controller is impedance bridged from the source ( a higher input impedance than 100 Ohms but an output impedance considerably less than 1200 + ohms of the typical mic pres) and impedance bridged to the output device - the mic pre or line amp for ideal voltage transfers between devices. To that end of been reading up on Lpad, Pie pad etc but i am hoping someone can point me in the right direction here in terms of a method that would be best to study.


Is there a rule of thumb for a ratio that is considerred good impedance bridging ie 1:5 source to load impedance ratio?

in my case would it be more prudent to impedance bridge the input and try to match the impedance on the output to the mic pre? I am wondering if I will have a reflection of my source signal back creating cancellation due to the laws of super position?

Out of curiousity - a tranformer acts as an impedance changing device and it changes level as well but is it possible to find ones that could function in rather one capacity or the other - ie are they made to change impedance but not level?

i am also reading up on buffer amplifiers as an option but i was hoping to keep my resister network passive ?

I am thinking out loud here and if anyone has any feedback over and above my thoughts i am open to it. no doubt those of you who have been following are bemused at my process! i knew this was a challenging project given my knowledge but the amount of reading i have done seems to be increasing as every turn keeps leading me deeper into the electronics labyrinth.

TIA

phi