Can anyone recommend a simple linear VCO?

[george2525]

Hello

ive been trying lots of voltage controlled oscillators but they are either not accurately linear or are just so complicated troubleshooting is hard

All I want is a sine or triangle wave that depends on input control voltage. a square wave is better than nothing

ive looked at voltage to frequency converters but these seem to always be pulse output which is annoying. also theres noone using these for VCOs that I have found

I was thinking maybe a 555 timer but im having trouble finding good circuit example that are clearly explained. I found one that responds to voltage but frequency decreased when control voltage is increased. This is no use!

I made one that was suggested but all it does is change pulse width. i dont want this either

so...

if anyone knows of a good simple acurate linear VCO and has a link to a book or a page that would be great.

 

[Alec_t]

How much non-linearity can you tolerate?
What voltage, frequency range, current capability?

 

[george2525]

its for sound so ideally it wants to be less than 0.059% out

frequency range is about 37Hz to 185Hz

control voltage can be from small values of a few mV up to a max of around 4V. the control range can be anywhere within these limits

output would ideally be small at around 15mVp-p but im ok to attenuate it later if not possible

current capability - well i want it to use a little as possible. 1-2mA or so

the supply is +/-6V

but all ideas welcome

 

[BobK]

its for sound so ideally it wants to be less than 0.059% out

I don't understand the use of the word "so" in this sentence. Why does that fact that "it is for audio" imply that the linearity should be better than 0,059%?

And, if indeed you need that kind of accuracy, you are talking about a lab precision instrument here.

What is it you are trying to do?

Bob

 

[george2525]

its for a music keyboard

it would be nice if the oscillator could be less than one cent out because its around that amount that humans can tell if the note is out.

each note on the music scale differs by a factor of 2^(1/12) or 5.946....% in pitch and a cent it 1/100 of that = approx 0.059%

obviously the control voltage must also be within this error too. But the oscillator should be as accurate as possible so that combined errors are minimised

Im not expecting it to be perfect but that is the ideal target. an overall error of <5.9% would be ok at this stage

im currently working with this one from the art of electronics book.

just wondered if anyone has a good knowledge of these and has a better suggestion

 

[(*steve*)]

Is your plan to have each key on the keyboard produce a different voltage which of feed into a VCO to produce the appropriate note?

If so, a better solution is to simply trim the voltage produced by each key to produce the correct frequency.

However, if you're going to that effort I'd wonder why you don't just use the keys to switch various resistors to an RC oscillator.

But, both of these methods mean you can only play a single note.

Have you researched how it's done on synthesisers?

 

[george2525]

I quite like that second idea. very simple. might use that one day but to go there now would have to tear down the whole project ;-)

but its a strange design that doesnt involve a conventional keyboard. i dont want to explain the whole thing as it will take a while!

Basically I have my control voltages sorted out, increasing at the correct ratio for the note scale. now im looking to find an oscillator that responds accordingly and with max accuracy.

 

[Alec_t]

I think you'll be hard pressed to manage even 1% linearity by analogue methods. Have you considered a digital approach?

 

[(*steve*)]

The standard way of doing this is 1V per octave. Is that what you're doing? (you did mention linear -- I hope you really didn't mean that)

 

[george2525]

I think you'll be hard pressed to manage even 1% linearity by analogue methods. Have you considered a digital approach?

I dont want to use digital. Ive never liked it. i should probably try again one day

 

[george2525]

The standard way of doing this is 1V per octave. Is that what you're doing? (you did mention linear -- I hope you really didn't mean that)

yes its supposed to be linear. ive already set the control voltages so they represent the notes, so now the oscillator should be linear

 

[BobK]

Just out of curiosity, how did you get the control voltage to be accurate to within 0.06%?

I am quite expeienced and I cannot imagine how to do it in a way that it would not drift more than that.

Bob

 

[george2525]

Just out of curiosity, how did you get the control voltage to be accurate to within 0.06%?

I am quite expeienced and I cannot imagine how to do it in a way that it would not drift more than that.

Bob

I havnt. im only using multisim which is not accurate because it doesnt simulate thermal voltage in BJT's

The idea is to get it accurate in multisim then when I make the real circuit I will use temperature compensating resistors and see what happens. the control voltage is set via 2 matched BJTs so that the saturation current is not a big issue, which leaves the VT problem.

other methods include heating the transistors to a constant temperature eg 40 degrees and tuning the circuit so that it operates at that level. so each time you turn it on it has to heat up for a minute or so.

my target is just a target. i wont be surprised if it fails

 

[Alec_t]

Basically I have my control voltages sorted out, increasing at the correct ratio for the note scale.

If you are able to tweak those voltages then you could compensate for any non-linearity.

 

[hevans1944]

if anyone knows of a good simple acurate linear VCO and has a link to a book or a page that would be great.

Well, there used to be a nice integrated circuit available from Intersil, the ICL8038 whose datasheet you can download for ideas. There may be a few of these still floating around on eBay and such.

I used this IC in the 1970s to build a musical note generator by programmable division of a high-frequency oscillator, the divided frequency then being used to phase-lock the VCO in the ICL8038. The divider ratios were set up to produce an equally tempered chromatic scale, so the "programming" was actually hard-wired into the dividers. You only adjusted the high-frequency oscillator to set the pitch of one note and all the other notes "automagically" fell into place, each note being a tap selected on the divider chain. Some math and ingenuity required to design the divider and select an appropriate oscillator frequency to divide. I didn't need absolute pitch accuracy, so I designed a simple LC oscillator that I could tune over a narrow frequency range. A crystal oscillator would have been more accurate and stable, but the LC oscillator was "gud enuf".

If you do decide to try the PLL approach, it is important that the last stage of the divider chain be a simple divide-by-two flip-flop to ensure a symmetrical square wave to compare with the VCO square wave in the phase comparator.

The thing that attracted me to the ICL8038 was its sine-wave shaper applied internally to the triangle wave output. That allowed for reasonably good-sounding musical notes... one at a time of course. The key to the design was getting the phase-lock loop to track quickly to changes in note selection, which could be problematical for very low frequency audio tones. Phase lock loop circuits are plentiful on the Internet, and so are triangle-to-sine-wave converter circuits. But designing a linear VCO for musical notes requires more precision than most folks are capable of mustering. And, temperature compensation not withstanding, it won't stay in tune. If you use a crystal oscillator and divide its frequency properly, then phase-lock the VCO to the divided frequency, it will stay in tune, last a long time, cost a little, do a lot. Too bad you eschew a hybrid digital/analog approach.

Lots of design idea links at this Google results page.

 

[(*steve*)]

The only thing I can think of is a constant current source controlled by the "linear" voltage. This could be used to charge a capacitor. If the discharge is sufficiently rapid, the frequency would be fairly linearly related to charge current.

You would get a sawtooth or square wave output

If you could both charge and discharge with the constant current you could get a triangular waveform.

 

[george2525]

the ICL8038 is actually available and cheap. looks interesting but i think ill have to start from scratch. I like the idea though and will bear it in mind.

Can someone tell me why the circuit I posted will not be any good?

in simulation it works pretty well but i havnt built it yet

 

[(*steve*)]

yeah, sure. Give it a go. It essentially matches my suggestion of a constant current charge/discharge of a capacitor.

 

[hevans1944]

The only thing I can think of is a constant current source controlled by the "linear" voltage. This could be used to charge a capacitor. If the discharge is sufficiently rapid, the frequency would be fairly linearly related to charge current.

You would get a sawtooth or square wave output

If you could both charge and discharge with the constant current you could get a triangular waveform.

That is exactly what the posted circuit tries to do. There are many sources of "error" that will affect the linearity, but in principle it works perfectly. The op-amps selected are good for this application: very high input impedance, low bias currents, decent slew rates, yada, yada, yada. If I were the OP, I would go ahead and build this circuit and compare its performance against the Multisim simulation. This is an excellent demonstration of the usefulness (or not) of simulation versus the real deal, especially since the only criterion for success is the actual frequency output versus voltage input, and that is easily measured (and perhaps trimmed on a note-by-note basis). Everyone should recall the origin of SPICE simulation. An entire generation of analog engineers used it to successfully design analog integrated circuits.

 

[george2525]

The op-amps selected are good for this application: very high input impedance, low bias currents, decent slew rates, yada, yada, yada. successfully design analog integrated circuits.

I was going to use TL072's or even LM6142's my supply must be +/-6V. id be interested in your opinion on these. the TL072 seems the best bet to me

Also the circuit above is all positive output. not realy sure why they made it like that. Im guessing the speaker wont like that very much so I was going to use a dual supply and stick a capacitor on the triangle output. This leaves the triangle amplitude much less than the square but it does work in the software. maybe just amplify the triangle a bit later. All this does it make the 150 into 75 in the formula. Then im lowering the cap value to increase the frequency for my range. does that all sound sensible enough? Not trying to be a pain in the ass, just save some time where I can!