Looking for an IC with the Ability to Output Many Sine Wave Oscillations

Hi everyone,

I'm searching for a way to cheaply build an array of possibly hundreds of low amplitude sine wave oscillators. Since their outputs will be summed up to produce a line level signal, I currently believe that a 20mV output would be sufficient per oscillator, provided that it is not too close to the noise floor. For example, if there were 150 oscillators outputting 20mV each, that would leave you with a 3V output to work with summed up - plenty sufficient for a line signal (it could be attenuated if it is too high for the input stage of the amplifier, for example).

So, because of these requirements, it seems to me that an IC chip could fit the bill because of the sheer number of oscillators I need, and the fact that they don't need to produce a very strong signal.

The grand purpose of this is an esoteric piece of test equipment. I am open to alternatives to an array of sine wave oscillators, but I am for the most part convinced that this is the only way to do what I need to do.

What do you guys think, does an IC like this exist? Advice?
 
Yes, filtering could turn a digital signal into an "analog" sort of thing. However, it does not completely fulfill my design requirements.

The bandwidth would be simply the audio band, with each oscillator tuned to some increment, logarithmic, within the 20 - 20k range. For example something like this

1 - 20 Hz
2 - 40 Hz
3 - 100 Hz
4 - 200 Hz
5 - 400 Hz
6 - 1 kHz
...
x - 20 khz

But the increments would be smaller for more precision... I haven't worked out the actual numbers. But no less than 20 and no more than 20k.
 

CopperTop

Member
2009-02-28 9:40 am
UK
If it really could only be achieved with analogue means, could the oscillators free-run with frequencies derived from RC networks, for example, or would the frequency & phase need to be accurately derived? What frequencies are required?

It could be expensive, but 'the gold standard' of generating an individual sine wave at an accurate but adjustable frequency is a sine wave VCO locked to an incoming clock signal using a frequency divider and PLL i.e. a frequency synthesiser with a sine wave output. You could look at ICL8038 (probably fairly obsolete these days) and a 4046 PLL for example. It could cost you a few hundred GBP to build what you want, however, and consume a fair bit of power...

An intermediate solution could be analogue low pass filtering of accurate square waves with op amps or you might be able to shape triangle waveforms using diodes (what THD is required?)

Free running oscillators include Wien bridge, for example,
http://en.wikipedia.org/wiki/Wien_bridge_oscillator
http://www.4qdtec.com/singen.html

or Colpitts
http://arxiv.org/pdf/physics/0507133.pdf

You can use Operational Transconductance Amplifiers:
http://technologyinterface.nmsu.edu/fall98/electronics/grise/griseota.html

Maybe there's a way of using a couple of CMOS gates in a 4000 series IC to generate a sine wave - that would be cheap.
http://www.edn.com/design/power-man...-inverter-generates-low-distortion-sine-waves

It's all a bit nasty though...
 
Last edited:

CopperTop

Member
2009-02-28 9:40 am
UK
Any mileage in ripping apart a vintage electronic organ..? I should think there are various types, but some with drawbars may effectively give you a large number of harmonically-locked sine wave generators. Modern ones will synthesise the output from a DAC of course (very sensibly! :) )
 
a long time ago a friend of mine built an electronic keyboard from scratch. He built 12 oscilators for the top octave and then divided each ouput by two to get each lower octave and so on. He then passed the square waves through a LC filter to get his sine waves. You could do the same, start at the top 10khz, 12khz, 15khz, 20khz, 40khz divide each output by 10 and then 10 and then 10 again to get your frequencies. LC filter your outputs and voila. Another aproach is to use quad op amps, 1 op amp per frequency. Depending on the number of frequencies you need will determine the number of opamps. Getting all the frequemcies to be in phase is another matter. Remember a square wave is nothing but odd sine waves added up. A 10hz square wave will give you 10, 30, 50, 70 90 etc hz after proper LC filtering. A tringle wave is even harmonics added up. The obvious as mentioned before is to use some software to generate the sine waves and output via sound card.
 

CopperTop

Member
2009-02-28 9:40 am
UK
A "programmable function generator" or "arbitrary waveform generator" is what you need. As I remember, Wavetek used to produce one, perhaps others too (HP/Agilent?).
That will also use a DAC, and so doesn't meet the requirement.

If DAC outputs were allowed, a PC headphone socket would do fine for audio frequencies. The waveform could be synthesised on the fly, or saved as a WAV file. Or a CD could be burned and the waveforms played from a CD player.
 
Interesting links, CopperTop. Seems like I will want to purchase a bunch of op amps and then design a Wein-bridge oscillator of specific frequencies around each one. Is there any op amp IC greater than the octal op amp? If it is possible to buy more than 8 packaged in one chip, it would be nice.

But it seems like this is going to be the way to do it.
 
Octal op-amps in one package? Never seen such a thing. Stick with inexpensive industry standard quads!
To go back to the OPs original quote: "if there were 150 oscillators outputting 20mV each, that would leave you with a 3V output" that to me is addition. Summing generally (to my understanding) is that, for every doubling of sources with the same impedance, you half the signal output and you need make-up gain at the end. E
 
Hm. Yeah I think the addition like I did only applies to active mixing (which I believe involves difference amplifiers?), whereas I will most likely be mixing the signals passively to reduce the number of components. In passive mixing, the different channels form a voltage divider with one another, so the output will be considerably less than 3V, still assuming 20mV output. I haven't looked into how the math would actually work out, but I think that I wouldn't lose too much signal from passive mixing.

Also, I have found 10 and 12 op amp models, which are interesting but prohibitively expensive. Quads are probably the way to go.
 
I understand you don't want to use a digital arbitrary signal generator; can you tell me why not?
You are going to a lot of expense and trouble to have all these generators, so what is the reason?

Edit: even easier: make all the oscillators in a simulator like LTspice, with all the levels and phases set as you want, then add them all, and save the output it as a .wav file.
You can then play that into your equipment.
Cost: zero, nada, nothing.

jan
 
Make 150 Wien Bridge oscillators set to wherever you want to, and actively mix their outputs.
To ease requirements you may sub-mix blocks of 12 oscillators into a single signal and then finally mix those 12 or 13 blocks into a final output signal.
It's the practical solution:
1) only 41 x TL074
2) only some 250 mA power consumption, and best of all:
3) you don't need a to design a nightmare PCB, just a 1 TL074 quad oscillator which you will repeat 13 times and a 13:1 x 4 mixer, which you will repeat 12 times.
Piece of cake.
You will need a trimmer per oscillator for frequency fine adjust and maybe another for distortion .
Oscillator phase will not be locked, but apparently that's not a requirement.

100% analog, of course :p
 
Jan.Didden: I understand that there are significantly more practical ways to do approximately what I am trying to do. I love analog equipment. I am doing this for the joy and the experimentation of it, however it could also be used as an instrument or a signal generator for testing something down the road, who knows.

JMFahey: Thanks for the advice. I did not think of submixing - it is the perfect way to balance the losses of passive mixing and the expense of active mixing. Also, I believe that the added component expense of phase locking each oscillator would not pay off, but I actually don't know all the reasons why PLL's are implemented. Would having 150 un-locked oscillators be a problem?
 
Last edited:

benb

Member
2010-04-24 1:52 am
Hi everyone,

I'm searching for a way to cheaply build an array of possibly hundreds of low amplitude sine wave oscillators. Since their outputs will be summed up to produce a line level signal, I currently believe that a 20mV output would be sufficient per oscillator, provided that it is not too close to the noise floor. For example, if there were 150 oscillators outputting 20mV each, that would leave you with a 3V output to work with summed up - plenty sufficient for a line signal (it could be attenuated if it is too high for the input stage of the amplifier, for example).

So, because of these requirements, it seems to me that an IC chip could fit the bill because of the sheer number of oscillators I need, and the fact that they don't need to produce a very strong signal.

The grand purpose of this is an esoteric piece of test equipment. I am open to alternatives to an array of sine wave oscillators, but I am for the most part convinced that this is the only way to do what I need to do.

What do you guys think, does an IC like this exist? Advice?
If you need a large quantity of something, it helps to simplify that something. There's this very common three-terminal IC that will work as the active component of a sine-wave oscillator, and you don't need any extra power and ground pins - I count four resistors and three capacitors in addition to this three-terminal device in the oscillator circuit:
1 - 200 Transistor Circuits
The circuit has "reasonable" performance depending on how good a sine wave you want. Googling transistor phase shift oscillator will find many more such circuits, with varying performance (as this is bottom of the barrel, others should have both better performance and more components).

The summing thing is interesting - if you run a bunch of 20mV outputs into an algebraic summer (say, an inverting opamp with 10k feedback resistor, and a 10k resistor from each oscillator to the negative input, with the opamp's output going through another gain-of-1 opamp in inverting configuration), then yes, the possible peak output is the sum of all the peak outputs from each oscillator. 100 oscillators generating 20mV peak sine waves will generate an output with a 2V peak when all the oscillators are at their peak. But this won't happen very often, especially as they're not in any way phase locked.

But with passive summing you should actually have as high an amplitude as you can reasonably have, as resistive summing with a large number of inputs will substantially lower the amplitude of each oscillator.

As for something else that is a "near-equivalent," a white noise source is like a near-infinite number of sine wave oscillators.
 
Sorry, what do you mean by "pulling each other"? Do you mean some kind of inductive crosstalk? Also, is this something that a PLL would solve?

edit: You and I posted at the same time, benb. Could you characterize the poor performance of this transistor oscillator you linked me to? It seems interesting, but I would like low distortion and low noise - if these are the poor characteristics of the circuit, then I would want to steer away from it. You are absolutely right about simplicity though.
 
Last edited: