So, I've been wrestling with this project for a while. I am making a tuner for musical instuments. At first, I was planning on using a simple 555 timer. But I learned from an electrical engineer I talked to today, that it wasn't a sine wave. So here is the plan using the 555 (Which I'm not using now):
I have a 555 chip. I spend a long time solving systems of equations and solve for the frequency I want. I had this schematic to select between the 3 values, (R1, R2, and C). Then you select the note to tune and a comparator tells whether sharp or flat.
The new idea though, is to use a pll. (Phase Locked Loop) This changes things a lot becuase it must use a binary counter, which is stimulated by 88 push buttons instead of the selector. (Really busy on solving that 88 button problem!) The counter's whole goal is to divide the tones down into their respective ratio. Here is a list of the frequencies I need to nail down: (Sorry it's kinda sloppy:()
Octave0 Octave1 Octave2 Octave3 Octave4 Octave5

0 A 55.000-110.000-220.000-440.000-880.000-1,760.000

1 A#/Bb 58.270-116.541-233.082-466.164-932.328-1,864.655

2 B 61.735-123.471-246.942-493.883-987.767-1,975.533

3 C 65.406-130.813-261.626-523.251-1,046.502-2,093.005

4 C#/Db 69.296-138.591-277.183-554.365-1,108.731-2,217.461

5 D 73.416-146.832-293.665-587.330-1,174.659-2,349.318

6 D#/Eb77.782-155.563-311.127-622.254-1,244.508-2,489.016

7 E 82.407-164.814-329.628-659.255-1,318.510-2,637.020

8 F 87.307-174.614-349.228-698.456-1,396.913-2,793.826

9 F#/Gb92.499-184.997-369.994-739.989-1,479.978-2,959.955

10 G 97.999-195.998-391.995-783.991-1,567.982-3,135.963

11 G#/Ab103.826-207.652-415.305-830.61-1,661.21-3,322.438

12 A 110.000-220.000-440.000-880.000-1,760.000-3,520.000





So, the counter will divide it down using complex ratios. I need to have Double Bit Precision on this. I need as many decimal points as possible, So I can measure cents(a pitch unit).
On a simpler issue, I want to make this 7 color LED light that changes with the mood of the note. The whole color thing. It would be a feature that no other tuner has. That would be cool... I need some advice on 7 color leds though. I have never used one. PLLs are also new to me, so if there is any advice on that, I'd love it. I have plenty of time to learn new things. BTW, NO PROGRAMMABLE CHIPS! I loathe them, and can't blow the budget on some fancy programmer again....


Edited by Tanner

What's wrong with using a square wave to tune? But yeah, a PLL is a more dependable tool to use than a 555. So do you basically want a tone generator at each of those frequencies?

Also: please use the button.

Square waves sound terrible Kerm! And yes, I want a tone generator.

Really? You can hear the difference between a sine and square wave? Then toss in a little rounding capacitor to smooth off the edges.

A square wave sounds harsh to the ears, not pure. A smoother wave form is far more pleasing to hear. Trust me Kerm. A capacitor would slow it down, and make it go flat, plus it wouldn't be a proper sine wave, now would it.
And how could I solve the whole 88 button problem?

May I ask why you don't want to use a microcontroller? That would turn a "super" hard project into a relatively straightforwards one. A dsPIC would seem like a logical choice as they are fast, cheap and have audio hardware on board.

False. A sufficiently small capacitor would merely do a tiny bit of phase shifting, as well as smooth out the vertical transitions. Obviously you'd need to pick a possibly-different value for each frequency. There are plenty of simple and naive ways to turn a square wave into a pseudo-sine wave. And I like Ben's idea; why make your life more complicated than it needs to be?

Oh, you most definitely can. The four main wave forms are sine, square, triangle, and sawtooth, and they all have a different sound (timbre). Think of a flute playing a C and and oboe playing a C. A flute has a very "pure" sound (it's very close to a sine wave, with some fundamentals) whereas an oboe is less smooth. If you listen to a lot of chiptunes, you can hear when some notes are smoother than others, and those are generally sine waves.
You can hear them here:
sine wave
square wave
sawtooth wave
triangle wave
It doesn't have them all at the same pitch, but sine and triangle are, and you can hear the difference there.

Indeed, and I do listen to a lot of chiptunes. Fine, then perhaps my question should have been, how important is the tone sounding audibly-pleasing to the overall purpose of using the tone to tune an instrument?

I don't think it's to do with how nice it sounds as opposed to how pure it sounds. A sine wave produces a pure tone, whereas a square wave is made up of a range of harmonics.

A fair point (gogogo Fourier series!), but one that doesn't nullify your original point about using a DSPic, or pretty much any uC connected to a DAC.

The dsPIC has an audio DAC built into it. You can top up its FIFO buffer with an interrupt or use a DMA module to top it up automatically. It shouldn't be too tricky to generate a sine wave in software when you've got 40 MIPS to play with!

The problem is the programmer. I don't have a PIC programmer to use. I have a SIMPLE Atmel one I made for the last robot I made. But anyway, I simply don't have funds to buy a programmer, I would rather spend more time in design and make something more complex, but cheaper. I just need some help with the ideas I have, not new ideas I've probably considered already. No offense to anybody's opinion.

Definitely understood; I think we just feel that your expenditure in time and components will work out to be more in the long run if you go the non-uC route than if you simply get a dsPic and a programmer.

I got an idea from another professional design engineer. Use a frequency to voltage converter (lm2907) to make the wave a voltage. Then use a specific and calibrated voltage source (Possibly from a variable voltage regulator, with varying R values) to get the reference. Then just a comparator to finish of the circuit. Then display on a bar graph LED thing. I need help implementing the schematic, which after its redesign, I will post a picture here.

Sure, that would work, although you'll need a significant amount of filtering and smoothing circuitry.

That's not what he said! He said:


BTW, totally different forum site. EEVBlog. I don't have a reliable way to program these fancy chips. Mine glitched out and failed at least 10 times before it transfered! So I'll ask around. There's no excuse not to use a chip except for the programmer. Anyway, the simple thing to do is to follow the LM2917 idea. So, he gives some answers to the problem. Anyway, any ideas on what the schematic is going to look like. I'll go and scan what I have already tonight after I eat supper in like 10 minutes. Then Kerm, you could tell me whats wrong and needs to change. It'll be up for sure by the time I get out of after school rehearsal tomorrow, if everything decides to go wrong with the scan machine, which I need to hook up... (timer beeps in the background, and I go get my casserole out of the oven...)

AVRs are very easy to program if you have a parallel port on your computer, it's just a case of connecting a few pins directly from the parallel port to the AVR. See avrdude.conf for more information (under "# Parallel port programmers.")

I wouldn't waste your time building a serial port programmer (I used one for a while and they are significantly slower than parallel port ones); if you're going to make the effort, build a USB programmer (that particular one requires a 32-bit OS as it uses an unsigned driver). Of course, you'll still need a way to program the ATtiny2313 in that design...

I did it with a serial port. It didn't turn out half as well as I'd like... It was unreliable in Baud rate, and screwed up many transmissions .

For Kerm:
Here is the schematic. SO SORRY FOR THE SIZE!!!
http://img268.imageshack.us/img268/7894/schematicz.jpg
By adept2 at 2010-09-19
Make any corrections and reply

That's the easiest way to think about it, but remember that the input is nowhere near a perfect wave; it's a giant mish-mash of frequencies superimposed. I feel like you'll need some smoothing and perhaps latency/derivation in there.