Post by thread:guitar tuner, PIC & MIDI (was:Freq.counter)

>... David's guitar tuner project ..... how about amplifying,low-pass filtering ( 3rd order), and clipping the guitar signal with a cheapo 324 quad-opa. Generate the desired reference freq. with the RTCC. Make a software-PLL style frequency- comparator ( edge-triggered RS-flip-flop )which is set by ref.-timing and reset by input signal . Connect a 'sharp' LED to the Fin>Fref port and a 'flat' LED to the Fref>Fin port. As the input freq. comes up/ down towards the ref.freq. the flat/sharp LED will dim. Just a quick idea; don't know if it makes sense. Another idea, not mine , saw it on Frankfurt Musik fair about 10yrs ago: A guy fed a bright LED with a microcontroller generated reference frequency and used the stroboscobe effect to visually tune the strings without making noise - nice ! >It would be interesting to also pursue the generation of MIDI from a >guitar signal .... and from other real world events. In my free-time ( not much) I currently work on such a project : a microphone is placed at each end of a 1m plastic tube. If you hit the tube with e.g. a drum-stick the mikes generate pulses with a runtime delay. A 16C54 along with a 324 opamp digitizes these pulses ( MIDI key velocity ) and calculates a MIDI-note-# from the trigger delays. I 'll put schematics, layouts + code related to MIDI, musicelectronics, PICs and other controllers on my ftp site as soon as possible ( late summer ?) Will be announced here and on MIDI.boards P.S. I too have some code that sends MIDI. MIDI IN is tough. a 20MHz PIC with HW-interrupt could do it. Somebody tried it ? Anybody else doing MIDI with PICs ? Uwe Sch"uler __\ /__ ein genie Physiologie --\,,,/-- das sich nicht geniert Gmelinstr.5 /" "\ ist wie ein rentier D-72076 T"ubingen \_V_/ das sich nicht rentiert Tel: (49) 70 71 29 - 30 72 Fax: - 30 73

Your suggestion for dealing with the harmonics is great. But my tuner attempts to determine by itself which string is being tuned. I suppose with expensive programmable filters you could detect the pitch, then dial in the filter real quick to protect against the harmonics, which intensify with time. But if you put in a filter sufficient to chop the harmonics of the low E string (86 Hz or so) you'd never pick up the high E two octaves up. My hackbuddy and I have decided to return to this project and see if we can get it to work. The idea will be: count on the fundamental being strong for the first several cycles (seems to be safe), then start ignoring any transitions that aren't close to the expected timings. For example, if you're tuning the 110 Hz A string, you'd initially get a 110 cycle square wave. Then as the vibrations decay, you'd start seening double, triple, or some oddball combination. If I remember right, the second harmonic is the most prominent and can be isolated. But I'll have to look again. Anyway, it might be as simple as ignoring every odd pulse, once the expected pitch is established. Owell I'll try it and post the results! ps: about the code I've offered to put up... I was going to make it available via ftp this morning but I forgot to move the files. I'll do it tomorrow when I'm dialed in from home. David -- Their address sums up their attitude: One Microsoft Way http://www.rt66.com/dthomas/

David, here are a few suggestions for a solution to your problem with harmonics: 1) Use a dual speed PLL to initially lock onto the strong fundamental in 'fast' tracking mode and then automatically switch to 'slow' mode to stay locked onto the decaying fundamental (decrease capture range). The 'fast' speed would be optimized to lock onto the fundamental quickly and the 'slow' speed would be fast enough to track tuning variations but too slow to 'jump' over to a harmonic. The fundamental could then be measured by coupling the vco to the pic. 2) Use zero crossing detection. It has the unique property of allways digitizing the lowest frequency, even if its not the strongest. A well designed front end, and possibly threshold tracking to maintain sensitivity could be used to extend its range. 3) Use a switched capacitor filter with external clock input configured as a 'comb' filter. Lock on to the fundamental and feed it into the filter's clock, thereby attenuating its harmonics. Something creative would need to be done here to aquire the fundamental in the first place. 4) Use the pic to implement an FIR or IIR filter. 5) Use the pic to integrate its measurement of frequency and then use that to detect the rate-of-change. If the rate of change (slope) exceeds a pre-set value then we know the detection circuitry has switched from the fundamental to a harmonic. At that point in time divide the new frequency measurement by the fundamental to calculate the harmonic factor and use that factor in subsequent calculations. This idea would work best if the signal were processed first by a PLL to lock onto a single spectral component. In fact, you could use the lock detection circuitry of the PLL to detect when the switchover to a harminic occurs. 6) Can you tell if a note is sharp or flat from its harmonic as well as from its fundamental? If so, then use a PLL front end to purify the spectrum (one frequency at a time) and do your function on the VCO signal. 7) Use an analog input and implement a PLL in pic firmware, using a combination of the above ideas. I really know absolutely nothing about music, but I hope these ideas are usefull, or stimulate the formulation of new ones. Regards, Dana Frank Raymond - Foxtrot Systems Ltd. Internet: .....dana.raymondKILLspam@spam@canrem.com. Compuserve: 73362,3052

On Sat, 8 Jul 1995, Dana Raymond wrote: > 1) Use a dual speed PLL to initially lock onto the strong fundamental Would it really stay locked to the weak fundamental? > 2) Use zero crossing detection. It has the unique property of allways > digitizing the lowest frequency, even if its not the strongest. It can't be that easy, can it? I'm already amplifying and heavily clipping the signal. Isn't that the same thing? Maybe I need to read up on zero crossing detection. > 3) Use a switched capacitor filter [...] Impractical but clever. > 4) Use the pic to implement an FIR or IIR filter. What are these? > 5) Use the pic to integrate its measurement of frequency and then use > that to detect the rate-of-change. What I was planning was simpler. Once the harmonic kicks in I'll be able to recognize it because transitions will suddenly arrive way too early. I can simply discard transitions that aren't within a few percent of where I expect to find them. > 6) Can you tell if a note is sharp or flat from its harmonic as well as > from its fundamental? If so, then use a PLL front end to purify the > spectrum (one frequency at a time) and do your function on the VCO > signal. How does one "purify a spectrum" with a PLL? What does that mean? But yes of course it is possible to track changes in a harmonic as readily as the fundamental. > 7) Use an analog input and implement a PLL in pic firmware, using a > combination of the above ideas. I've done this for another project and it worked *very* well. However I was able to isolate the fundamentals with analog filters ahead of time. Thanks for all your cool suggestions. David -- Their address sums up their attitude: One Microsoft Way http://www.rt66.com/dthomas/

> From @uga.cc.uga.edu:owner-piclistKILLspamMITVMA.MIT.EDU Fri Jul 7 14:28:45 1995 > From: Uwe Schueler <.....uwe.schuelerKILLspam.....UNI-TUEBINGEN.DE> > Subject: Re: guitar tuner, PIC & MIDI (was:Freq.counter) > > >... David's guitar tuner project ..... > > how about amplifying,low-pass filtering ( 3rd order), and clipping > the guitar signal with a cheapo 324 quad-opa. Generate the desired > reference freq. with the RTCC. Make a software-PLL style frequency- > comparator ( edge-triggered RS-flip-flop )which is set by ref.-timing > and reset by input signal . Connect a 'sharp' LED to the Fin>Fref port and > a 'flat' LED to the Fref>Fin port. As the input freq. comes up/ > down towards the ref.freq. the flat/sharp LED will dim. > Just a quick idea; don't know if it makes sense. This has some possibilities. But you need to guess the reference freq. from early edge crossings to get the correct string. Help me out now, what would the algorithm look like in pseudo-code? wait for an edge crossing. start a timer. next edge crossing is 1/2 period. estimate string, get reference period, set up reference period in RTCC. loop wait for an edge or RTCC timeout. if edge set sharp LED. ;; you'd want to re-sync the RTCC? if RTCC set flat LED. ;; you'd have to re-sync the RTCC to the next edge? if (too many RTCC's with no edge) break, restart So when you're close to 'in-tune', both LED's would be on? > >It would be interesting to also pursue the generation of MIDI from a > >guitar signal .... > > and from other real world events. In my free-time ( not much) I currently > work on such a project : a microphone is placed at each end of a 1m plastic > tube. If you hit the tube with e.g. a drum-stick the mikes generate pulses > with a runtime delay. A 16C54 along with a 324 opamp digitizes these pulses > ( MIDI key velocity ) and calculates a MIDI-note-# from the trigger delays. This sounds like a really fun midi-trigger. I'd like to get a hold of some of your results. Are you digitizing Mike output with RC discharge time? > I 'll put schematics, layouts + code related to MIDI, musicelectronics, PICs > and other controllers on my ftp site as soon as possible ( late summer ?) > Will be announced here and on MIDI.boards I'll look forward to this. > P.S. > I too have some code that sends MIDI. MIDI IN is tough. > a 20MHz PIC with HW-interrupt could do it. Somebody tried it ? 32 usecs/bit doesn't leave much time. Setting up an 'hc11 is quite a bit easier. > Anybody else doing MIDI with PICs ? Not yet, but when my (non-existent) free time allows .. > Uwe Sch"uler __\ /__ ein genie Tim Braun | Continental Healthcare Systems Canada | Voice: 204-942-2992 ext 228 1900-155 Carlton St | FAX: 204-942-3001 Winnipeg, Manitoba, Canada R3C 3H8 | Email: EraseMEtimspam_OUTTakeThisOuTchs.mb.ca

On Mon, 10 Jul 1995, Tim Braun wrote: > This has some possibilities. But you need to guess the reference freq. > from early edge crossings to get the correct string. Help me out now, > what would the algorithm look like in pseudo-code? > > wait for an edge crossing. > start a timer. > next edge crossing is 1/2 period. Are you sure you can assume 50% duty cycle? Depending on the DC level when you clip, the 1/2 cycle transitions will hopefully be near the middle but you can't count on the exact timing. I time from one low-to-high transition to the next to unask this question. The rest of your algorithm looks good. What I do is wait for the edge, then peek at the RTCC and look up the count in a table. See http://www.rt66.com/dthomas/pic/pic.html if you want to see my tuner so far. David -- Their address sums up their attitude: One Microsoft Way http://www.rt66.com/dthomas/

I may have mailed this incorrectly. If not, sorry for the duplicates. . . {Quote hidden}> David, here are a few suggestions for a solution to your problem with > harmonics: > > 1) Use a dual speed PLL to initially lock onto the strong fundamental > in 'fast' tracking mode and then automatically switch to 'slow' mode > to stay locked onto the decaying fundamental (decrease capture > range). The 'fast' speed would be optimized to lock onto the > fundamental quickly and the 'slow' speed would be fast enough to > track tuning variations but too slow to 'jump' over to a harmonic. > The fundamental could then be measured by coupling the vco to the > pic. > > 2) Use zero crossing detection. It has the unique property of allways > digitizing the lowest frequency, even if its not the strongest. > A well designed front end, and possibly threshold tracking to > maintain sensitivity could be used to extend its range. > > 3) Use a switched capacitor filter with external clock input configured > as a 'comb' filter. Lock on to the fundamental and feed it into the > filter's clock, thereby attenuating its harmonics. Something > creative would need to be done here to aquire the fundamental in the > first place. > > 4) Use the pic to implement an FIR or IIR filter. > > 5) Use the pic to integrate its measurement of frequency and then use > that to detect the rate-of-change. If the rate of change (slope) > exceeds a pre-set value then we know the detection circuitry has > switched from the fundamental to a harmonic. At that point in time > divide the new frequency measurement by the fundamental to calculate > the harmonic factor and use that factor in subsequent calculations. > This idea would work best if the signal were processed first by a > PLL to lock onto a single spectral component. In fact, you could use > the lock detection circuitry of the PLL to detect when the > switchover to a harminic occurs. > > 6) Can you tell if a note is sharp or flat from its harmonic as well as > from its fundamental? If so, then use a PLL front end to purify the > spectrum (one frequency at a time) and do your function on the VCO > signal. > > 7) Use an analog input and implement a PLL in pic firmware, using a > combination of the above ideas. > > I really know absolutely nothing about music, but I hope these ideas are > usefull, or stimulate the formulation of new ones. > > Regards, Dana Frank Raymond - Foxtrot Systems Ltd. > Internet: dana.raymondspam_OUTcanrem.com. Compuserve: 73362,3052 Please help my confusion - it seems to me that we are trying to over engineer what could be an easy solution. Wouldn't the normal tuning process be: 1. "Pluck" the string. 2. Listen for it being sharp or flat. 3. Tune the string. 4. Re-pluck the string. If this is the case, why worry about decaying harmonics? Just count the number of zero crossings in 100 msec & compair to a table of expected values. If it is high or low out-of-range discard it & try again. Even if you tune on the fly ( i.e. pluck-listen-tune as a single process); I would think you could still get a good frequency sample after 2 sec. I could be wrong here - I've never tried it. I don't know much about guitars. But I think the difference between the 1st string (low E) and the 6th (high E) is 2 octaves. With a string way out- of-tune, I think it is possible to tune to the frequency of another string. Taking a string this far out-of-tune would be undetectable by your circuit. You may want to consider using 8 LEDs: 1 per string, plus a high and low indication. This would let you know which string the circuit thought you were tuning. Jon Poland Sun Microsystems Inc. St. Louis, MO (314) 569-4716 @spam@jon.polandKILLspamcentral.sun.com

On Mon, 10 Jul 1995, Jon Poland SE Sun St Louis wrote: > Please help my confusion - it seems to me that we are trying to over engineer > what could be an easy solution. > > Wouldn't the normal tuning process be: > 1. "Pluck" the string. > 2. Listen for it being sharp or flat. > 3. Tune the string. > 4. Re-pluck the string. No. The way guitar players generally tune is to hit the string once, good and hard, then adjust the peg as the string is decaying. That way you can adjust a continuous pitch until it's right. If you try to tune with only "that was too high" type information, it takes way too many tries. Being off by even a few cents is enough to make the instrument sound sour. > Even if you tune on the fly ( i.e. pluck-listen-tune as a single process); I > would think you could still get a good frequency sample after 2 sec. I could > be wrong here - I've never tried it. 2 sec. is enough time to know if you're right or not, but it's not enough time to bring the string in. > I don't know much about guitars. But I think the difference between the > 1st string (low E) and the 6th (high E) is 2 octaves. <nod> > With a string way out- > of-tune, I think it is possible to tune to the frequency of another string. Guitarists are smarter than this. Very few guitar players have any trouble at all getting it close, even very close, by ear. The tuner is just used for fine tuning. Also, guitars tend to drift some but not a lot. A guitar that was packed up a month ago is unlikely to be so far out of tune that an autodetect mechanism would identify the wrong string. And most musicians wouldn't fall for it even then. Still, your idea of providing a display that shows which string was detected is a good one, provided you can afford to add it in your particular application. David -- Their address sums up their attitude: One Microsoft Way http://www.rt66.com/dthomas/

-> > 1) Use a dual speed PLL to initially lock onto the strong -> Would it really stay locked to the weak fundamental? Yes, I believe so. A PLL CAN track a signal outside its capture range if it has been locked onto it in the first place and its loop characteristic are such that it resists change that is too fast (a change from the fundamental to a harmonic can be considered a fast change). When the harmonic comes into the capture range (close enough in frequency and strong enough compared to the fundamental) the PLL won't move to it because its loop characteristics inhibit fast change. -> It can't be that easy, can it? I'm already amplifying and heavily No, you're right, I goofed. I'm still trying to find the reference in "The Art of Electronics" that got me confused on that one (An excellent reference BTW. I highly recommend it). -> > 4) Use the pic to implement an FIR or IIR filter. -> -> What are these? Digital filters. You need a very fast CPU with an analog front end. -> How does one "purify a spectrum" with a PLL? What does that mean? -> But yes of course it is possible to track changes in a harmonic as -> readily as the fundamental. What I mean by that is that A PLL will lock onto one spectral component (frequency) if it is coherent, stays within the tracking range, and doesn't change too fast (slew rate). If you take a standard audio range PLL and pluck a guitar string the PLL's oscillator will be one frequency; The base note (terminology correct?), and the phase should be shifted at around 90 degrees from the fundamental. The PLL will either stay locked onto the fundamental or switch abruptly to a harmonic if that harmonic is strong enough. By measureing the PLL's oscillator frequency, you arn't getting a lot of noise due to harmonics. Its only one frequency, changing over time. Good luck with your project. Regards, Dana Frank Raymond - Foxtrot Systems Ltd. Internet: KILLspamdana.raymondKILLspamcanrem.com. Compuserve: 73362,3052

On Tue, 11 Jul 1995, Dana Raymond wrote: > Yes, I believe so. A PLL CAN track a signal outside its capture range if > it has been locked onto it in the first place and its loop > characteristic are such that it resists change that is too fast (a > change from the fundamental to a harmonic can be considered a fast > change). Okay yeah this makes sense. I just need to make the loop fast enough to lock onto the fundamental before the harmonic gets too strong (500 ms response is plenty good enough) but slow enough that the almost-instant switch to the harmonic doesn't -- scuse the pun -- phase it. I'd really like to try this approach. What would be some audio PLLs to use? Perhaps a 565? 567? Or are there more modern, better ones? I've worked extensively with RF type PLLs (gotta whole drawer full of motorola chips) but never worked with audio PLLs. Incidentally, for anybody wanting to use a PIC to tune an RF circuit, I highly recommend the MC145170 chip. It's cheap (about $8), readily available (Hamilton Hallmark, Newark), and miraculous. Programmed with 3 wire interface. Reference modulus can be any number from 4(?) to 32767. Programmable divider from 40 to 65535. Can be programmed to be single-ended or dual-ended. Ref. output available. Sign of loop can be inverted by programming. Can directly synthesize up to 160 mc. XTAL ref. osc. on-board. 16 pin package. I've used these with PICs and they work great. David -- Their address sums up their attitude: One Microsoft Way http://www.rt66.com/dthomas/

> Date: Tue, 11 Jul 1995 21:46:25 -0500 > From: Doug Sellner <spamBeGonenewsspamBeGoneEMBAY.COM> > Subject: Re: guitar tuner, PIC & MIDI (was:Freq.counter) > All one needs is an audible tone refference to tune the guitar. As the two > tones get closer they cause a 'RING' which allows a semi skilled musician > to tune-a-fish. Using 'beating' to tune a guitar is fine and good when you can hear the guitar and reference ... There are many commercial tuning devices on the market, any which I've looked inside are based on microcontroller technology. This is a great application for a PIC. > Doug Sellner > Internet: TakeThisOuTdsellnerEraseMEspam_OUTembay.com Tim Braun | Continental Healthcare Systems Canada | Voice: 204-942-2992 ext 228 1900-155 Carlton St | FAX: 204-942-3001 Winnipeg, Manitoba, Canada R3C 3H8 | Email: RemoveMEtimTakeThisOuTchs.mb.ca

On Tue, 11 Jul 1995, Doug Sellner wrote: > All one needs is an audible tone refference to tune the guitar. As the two > tones get closer they cause a 'RING' which allows a semi skilled musician > to tune-a-fish. Yeah that "ring" is called beating. It's the difference between the two frequencies. And you are right that lots of musicians use this technique all the time and it works great. However, a lot of musicians prefer to use electronic tuners, at least in some situations. Particularly, stage musicians like to have them. You're playing and a string breaks -- you finish the song, then hit a foot pedal and suddenly your tuner (with visual indicator) is in circuit and your amp is not. You tune up without bothering the audience, then you're back in business. Or, let's say another band is playing, and your band is up next. You probably couldn't hear yourself to tune, perhaps even with a headset. I personally am very accustomed to relying on my ears, a lot more than electronic measuring devices. I've never tried to play an instrument on stage that I didn't first check with my ears. But a lot of pro musicians do this sort of thing all the time. David -- Their address sums up their attitude: One Microsoft Way http://www.rt66.com/dthomas/

Well, I've had this response on my desk-top for a few days, better get it away, though some of the comments I make here have been made already. {Quote hidden}> Date: Mon, 10 Jul 1995 12:23:20 -0500 > From: Jon Poland SE Sun St Louis <EraseMEjon.polandSUNARCH.CENTRAL.SUN.COM> > Subject: Re: guitar tuner, PIC & MIDI (was:Freq.counter) > > David, here are a few suggestions for a solution to your problem with > > harmonics: > > > > 1) Use a dual speed PLL to initially lock onto the strong fundamental . . . > > spectrum (one frequency at a time) and do your function on the VCO > > signal. > > > > 7) Use an analog input and implement a PLL in pic firmware, using a > > combination of the above ideas. > > > > I really know absolutely nothing about music, but I hope these ideas are > > usefull, or stimulate the formulation of new ones. > > > > Regards, Dana Frank Raymond - Foxtrot Systems Ltd. > > Internet: RemoveMEdana.raymondEraseMEEraseMEcanrem.com. Compuserve: 73362,3052 > > Please help my confusion - it seems to me that we are trying to over engineer > what could be an easy solution. Being a guitarist, and having used a tuner, and having attempted to design a tuner, once, I'll share a bit of my experience. I don't think we're over- engineering yet. > Wouldn't the normal tuning process be: > 1. "Pluck" the string. > 2. Listen for it being sharp or flat. > 3. Tune the string. > 4. Re-pluck the string. Guitarists pluck the string and adjust the tuning while it sounds, and I for one would not consider it useful to have to stop the string and re-pluck to check the adjusted frequency. > If this is the case, why worry about decaying harmonics? Just count > the number of zero crossings in 100 msec & compair to a table of expected > values. If it is high or low out-of-range discard it & try again. The upper harmonics do decay quite quickly. The first and second harmonic can be a problem. > Even if you tune on the fly ( i.e. pluck-listen-tune as a single process); I > would think you could still get a good frequency sample after 2 sec. I could > be wrong here - I've never tried it. I normally don't get a useful tuning measurement out of my commercial tuner until about 1/2 second after the string is plucked. If you can't get a good frequency sample for 2-5 seconds/pluck, it is really laborious to tune. {Quote hidden}> I don't know much about guitars. But I think the difference between the > 1st string (low E) and the 6th (high E) is 2 octaves. With a string way out- > of-tune, I think it is possible to tune to the frequency of another string. > Taking a string this far out-of-tune would be undetectable by your circuit. > You may want to consider using 8 LEDs: 1 per string, plus a high and low > indication. This would let you know which string the circuit thought you > were tuning. > > > Jon Poland > Sun Microsystems Inc. > St. Louis, MO > (314) 569-4716 > RemoveMEjon.polandspam_OUTKILLspamcentral.sun.com The two E strings are indeed 2 octaves apart. You can indeed tune a string down to the next one, if you try. The harmonics on a given pluck on a string won't let you simply count zero crossings, as you can pluck to generate harmonics or not. The zero crossings should be harmonically related, though. The low E is about 82.41 Hz, the A is about 110.0 Hz, the high B is 247.2 Hz, high E is 329.6 Hz. Referenced to A-440Hz. It's very useful to be able to adjust the reference of a tuner. Acoustic pianos are out most of the time in most places. Tim Braun | Continental Healthcare Systems Canada | Voice: 204-942-2992 ext 228 1900-155 Carlton St | FAX: 204-942-3001 Winnipeg, Manitoba, Canada R3C 3H8 | Email: RemoveMEtimTakeThisOuTspamchs.mb.ca