Good IC for making stable oscillators?

[Top Top]

I am trying to design an effect that requires oscillators tuned to standard pitches (chromatic notes). They need to hold an accurate pitch and stay stable once they are tuned. So far I have only designed things with oscillators that needed to be roughly tunable, nothing conventionally "musical."

I would prefer to be able to keep fixed resistors on them once the values are figured out, but if I needed trimmers, that is ok as long as they will stay stable once the trimmers are set - ie: if you turn the power off and turn it back on, the pitches will be the same, also doesn't drift over time.

More oscillators out of one chip = better. Also, less external parts per oscillator = better.

I was thinking 40106 because I have used it, it has 6 oscs per chip, only needs 1 resistor and one cap per osc, and designed basic things with it before... but I don't know how stable it is.

Any other suggestions?

Any waveshape is fine.

[Rectangular]

I'm a firm believer in the ICL8038, you get three waveshapes out of it simultaneously. I don't know if your application requires multiple synced oscillators, or if you want to be able to individually tune each one.

but its a great chip, and if you look up the thomas henry circuit for it, you can get a full 20-20k range out of it. really pure sine.

[slacker]

There's some interesting stuff about using CD40106s for this sort of thing at this forum http://deathlehem.com/php/viewforum.php?f=23, Rich over there did some pretty decent tests and decided they were stable enough to be usable.

[R.G.]

The conventional wisdom in electronics is that the usability of oscillators for musical notes is determined by the quality of the frequency setting components, not the oscillator itself.

1. Resistors and capacitors by themselves don't make it without special trouble taken and more or less continuous retuning.
2. L-C (inductor-capacitor) timing networks are just barely good enough, and need periodic retuning.
3. Long term "good enough" requires a ceramic oscillator (marginal at best) but preferably a crystal.

The reasons involve both the temperature/time stability of the timing parts and the sharpness of their selection for the single frequency needed.

[Top Top]

There's some interesting stuff about using CD40106s for this sort of thing at this forum http://deathlehem.com/php/viewforum.php?f=23, Rich over there did some pretty decent tests and decided they were stable enough to be usable.

That is fantastic info. I will comb through it closely.

I should be clear...

I would like this to stay in tune for a week or two at a time  - I guess like a guitar or something. If it needs to be tuned about that often, that is fine. This is for my own use, not a commercial product, so some tinkering is ok... I just can't have it going out of tune as I use it.

[cloudscapes]

I like the XR2206 personally
lots of waveshapes. few external components. voltage and cap controlled. I'm using a 12-bit DAC to get it to keep notes.

[PRR]

Musical-tune is a VERY strict spec.

From one fret to the next is about 6%. Already we know you can't drop 10% resistors in and get near any specific fret.

"In-tune" is relative. Sub-fret pitch is often specified in "cents", 1/100th of a semi-tone. We all know musicians who don't know they are 50 cents off, but many musicians hear smaller errors. You will probably object to a 5 cent error, if not alone then with other notes either perfect or slightly-off.

5 musical cents error is 0.3% frequency error. So we know that you can't drop 1% resistors in and get "small" pitch errors.

And 1% caps are VERY expensive, yet not good enough.

Just to "get to pitch" you need a trimmer. That's why gitars have knobs, why xylophone makers have power grinders.

> I would like this to stay in tune for a week

Everything drifts with time. I don't know many guitar players who would perform on a guitar which has not been tuned in a week. Piano tuning is expensive, so piano makers strive for tuning stability, yet I've seen a piano need a touch-up at halftime.... Dave just goes out on stage while the audience is on smoke-break and dink-dink-dings it back to self-harmony.

Read up on the very first MOOGs. Despite being very precise electronics, it could take an hour to tune-up for a 1-minute monophonic passage, and then tuned-up again to lay another monophonic layer. ARP got even crazier with precision, but still more tuning than playing.

> I was thinking 40106

Hex Schmitt? The threshold is the ratio of self-voltages of two opposite FETs. The P-type are not even made the same as the N-type. There is some cancellation, but still the threshold drifts a lot more than sub-percent with temperature and supply.

The several function generators are a bit better. Thresholds are often ratio of three quasi-identical resistors, drift is much lower than two different FETs.

Capacitor drift is another big killer. You can get low tempco caps in pFd sizes, but you will be wanting stuff made with paper/plastic or salted-ceramic dielectrics. Tempco of a few percent full temperature range will be many musical cents error with normal room temperature variation.

But... are you re-inventing a wheel? The standard late-1970s electronic organ had this problem nailed. A MHz oscillator was divided by 12 odd divisors to give a "Top Octave" which was sufficiently perfect for most musicians. Flipflop chains divided the octaves down from this. It was ALWAYS in tune with itself.

For that matter: very fine electronic organs turn up in rummage sales etc. You don't care about the cabinet, the speakers, the key and stops switchery, the filters. Either it is a master/divider chain or it is a dozen discrete oscillators and octave dividers: this stuff rarely fails. The master oscillator type is easier to re-tune to match a piano or xylophone. The 12-oscillator type is fun to pull just a bit "off" for "flavor". (It can also be set for one of the many perfect tunings instead of the equi-tempered "scale".)

[PRR]

Man, there was a time I combed my hair with top-8ve chips, but those days are gone.

http://www.organservice.com/crm/topdividers.htm


http://www.users.cloud9.net/~pastark/sotnot7.htm

http://www.synthdiy.com/show/?id=172
http://www.synthdiy.com/files/2001/mk50240.pdf

[chi_boy]

I'm a firm believer in the ICL8038, you get three waveshapes out of it simultaneously. I don't know if your application requires multiple synced oscillators, or if you want to be able to individually tune each one.

but its a great chip, and if you look up the thomas henry circuit for it, you can get a full 20-20k range out of it. really pure sine.

That chip looks interesting.  What have you done with it?  And is it strictly bipolar?

[Top Top]

That looks fantastic. Perfect even for what I am going for...

BUT

Where in the world do you still get something like that? Does ANYONE still manufacture them? I MIGHT find an organ to rip it out of, but that seems like such a waste, if I do even find an organ at a price that is worth it.

[jimbeaux]

TOG's (top octave generator) were made by Mostek (50240) & went out of production some time ago (25 years ago?)

PAIA had several kits that used them (Organtua, Strings N' Things, etc,..) & you could use a crystal oscillator & get a stable frequency output.

I've seen a work around replacement for them & if you do a search for "50240" on this forum you'll find a post that references using a PIC - programmed to replicate one (only 5 notes I think) - but with a faster clocking speed (and the right PIC) you could probably get all twelve notes.

[jimbeaux]

This forum discussion at electro-music.com has the PIC links (& assembly code) that I talked about above

http://electro-music.com/forum/post-188525.html

[Top Top]

I think I am going to just start with trying some 40106 and see if it gets me close enough.

The PIC idea looks awesome, but I am not ready to dive into that at this point - don't have the neccessary stuff or knowledge to burn them, and it is too much for me to get into for a simple experiment of an idea.

I also can't afford to fry a $40 NOS top octave divider.

[PRR]

Summa those PICs are really cheap and small. You could buy a dozen for $40.

Hey.... REALLY cheap! PIC10F202 is $0.64 each, ten for $5.

Say you want to divide by 239. KISS, count to 118, set output high, count to 118, set output low, repeat forever. You need to fudge the "118" to account for test and jump overhead.

Do this with 11 more PICs using other numbers. Feed all from one master ~~1MHz clock. (Maybe 8MHz to account for 4 clocks/instruction and 2 instructions/loop.)

While the PICs are cheap, the programmer may be $200. There ARE cheaper ways; the fancy programmer is much more than you need. For code this small, you could almost program with a push-button (back in the day, we did). But I'm not the digi-head to know what the lowest total cost modern platform is.

[R.G.]

While the PICs are cheap, the programmer may be $200.

And the programmer may be as cheap as a few resistors and caps; many programmers will be under $50.

What's expensive is either (a) the programming language or (b) the time to learn to do it.