Mixer created harmonics please

I know some things about electronics but not enough for this, could you help me with choosing a mixer configuration for this task? I want a broad spectrum frequency signal using a mixer feedback loop

I know that mixers are non-linear and give f1 + f2 and |f1- f2|. And other forms such as 2f1 + f2 or 3f1+f2 ect...
I want all these harmonics. So no filtering or balancing mixer. Plus I would like to create a feedback into the mixer to expand the range of harmonics and frequencies..

Using just the basic f1 + f2 and excluding the other frequencies.. I want the feedback to look like this:

f1 + f2 = f3
f2 + f3 = f4
f3 + f4 = f5
f4 + f5 = f6
....

where f3 is the mixed signal and its component harmonics, then a new mix of f3 and f2 is made to get f4 and so on.. where in the end I have a very broad range mix of harmonics starting from the two input frequencies..

I'm just getting started with this question so any refinement you can teach me to bettering my understanding of attaining this would help. Even help better writing my question would help. thanks

Edit:

It's kind of hard to imagine the spectrum analysis of this broad spectrum signal. Do you know of anything that would let me see that this feedback loop circuit would give? Will adding f5 plus f4 give me all the harmonics located in f3 and so on.. So will the spectrum analysis (Fourier transform) shift away from all previous harmonics or mixing of signals as we add more, or will it contain the harmonics inside of all previous frequencies? Giving me one big higher and lower frequency containing range of mixer created harmonics?

Are you planning on using this sort of mixer feedback loop with this type of mixer or have I totally got hold of the wrong end of the stick? They're not one and the same thing. I suppose you could simulate feeding the output of an analogue multiplier back into its inputs, but your output signal would be limited by the voltage rails in practice, making your (RF multiplying) mixer more nonlinear than its design. Plus you have to factor in that real (RF) mixers have bandwidth limitations, etc.

ermine said:

Are you planning on using this sort of mixer feedback loop with this type of mixer or have I totally got hold of the wrong end of the stick? They're not one and the same thing. I suppose you could simulate feeding the output of an analogue multiplier back into its inputs, but your output signal would be limited by the voltage rails in practice, making your (RF multiplying) mixer more nonlinear than its design. Plus you have to factor in that real (RF) mixers have bandwidth limitations, etc.

Click to expand...

Yea the links seem right! I'm not sure what that would look like in circuit.. but that's just me. You are right I had thought about the voltage rail limits and bandwidth limits. I was looking to get between 1hz and 1Mhz. And in actuality I was looking for ~45v and as high as current as possible.. Because I would be using a transformer to get extremely high voltages out of it. But that is besides the point of this because I'm curious just at what the circuit would look like at any voltages. Yea, the non linearity would go through the roof, but I also do not mind it in this case. Just looking for some starter points to make the crazy thing. Which you have helped, thanks.

The audio mixer feedback is a very different beast from the RF mixer - the audio feedback is simply the delays and shifts in the audio paths, combined with limiting in the channels, then feeding it to another audio mixer and getting general positive feedback. Sounds dire to me, but each to their own creative muse

With the RF mixer/multiplier if you need more power to drive the transformer you'd be better amplifying after the multiplier stage. You get ICs that do the mixer function - gilbert cells like NE612, the MC1496 are lowish power. Higher powers are diode mixers like the Mini Circuits SBL-1 but there's not much that will give you 45V. Amplify after the mixer.

But first you have to bottom out what you are trying to do with your mixer, the f1+f2 and difference frequencies sound like the RF mixer, which is in essence something that multiplies two inputs. You could pick off the sum and difference in the IF output and feed it to the RF input, along with f2, keeping f1 on the LO input but you will end up exploring the limits of the output swing very quickly.

What is the overall thing you are trying to achieve with this?

ermine said:

The audio mixer feedback is a very different beast from the RF mixer - the audio feedback is simply the delays and shifts in the audio paths, combined with limiting in the channels, then feeding it to another audio mixer and getting general positive feedback. Sounds dire to me, but each to their own creative muse

With the RF mixer/multiplier if you need more power to drive the transformer you'd be better amplifying after the multiplier stage. You get ICs that do the mixer function - gilbert cells like NE612, the MC1496 are lowish power. Higher powers are diode mixers like the Mini Circuits SBL-1 but there's not much that will give you 45V. Amplify after the mixer.

But first you have to bottom out what you are trying to do with your mixer, the f1+f2 and difference frequencies sound like the RF mixer, which is in essence something that multiplies two inputs. You could pick off the sum and difference in the IF output and feed it to the RF input, along with f2, keeping f1 on the LO input but you will end up exploring the limits of the output swing very quickly.

What is the overall thing you are trying to achieve with this?

Click to expand...

Thankss
I'm trying to get something going in multisim. But I'm terrible with electronics. I did get a gilbert cell up. I didn't find an integrated one. Unfortunately I am having more trouble because the software is not giving me non-linear results? Don't quote me on this because I may have the actual gilbert cell set up wrong. Oh, ok. I just found out the gilbert cell is double balanced and eliminates those harmonics. I wonder if that was my results. Again, terrible and new at this.

Yes, I had two gilbert cells and was feeding each other. I quickly reached the infinite readings in multisim.

I'm still learning.. So I don't want to say too much without doing experiments to gather information before asking too many questions.

I could not see any resemblance of a feedback loop that I was looking for in my two gilbert cells. Besides infinite voltage. But frequency did not change? Again I wish I wasn't this dumb.

The overall thing I am trying to build is a piece of equipment I seen years ago that created an interesting effect in plasma. Something about the harmonics created self similar embedding and helped with health effects. So I am trying to power a plasma container with these frequency cascades. I want to try an experiment and use it on plants like I heard it was having good results on.

I'm going to play around with mixers tomorrow. I am actually just learning multism has a spectrum analyzer, so I am going to be able to interpret my results better and have a better idea.

I am a hardware on the bench type of guy You can have an ideal mixer by multiplying one input voltage with another, but simple analogue simulation often looks for the DC working point and does a linear simulation around that which isn't going to get you anywhere as far as producing signals that weren't there to start with. You could do a transient analysis and FFT the output, but you need enough lifetime or enough hardware to get the result. Try using a switching mixer - There's some here. The transmission gate they used could be suitable for you, and a lot easier than simulating a Gilbert cell. I observe they did a .tran and then stuffed this into their FFT.

It's not theoretically unreasonable to end up with infinite signal if you are taking a multiplier and feeding the output back to the input, depends how much you feed back. That's not going to happen in the real world because you will get to the supply rail soon enough.

gilbert cell is double balanced and eliminates those harmonics

Click to expand...

People sweated buckets over many many years to try and make the double-balanced mixer approximate as closely as possible the transfer function Vo = Vrf * Vlo. If you make Vrf and Vlo sinewaves, you get the results you described originally, the reason why is described in this PDF. A Gilbert cell is a specific example of a DBM, but as soon as you reach the limits of the device the transfer function goes to pot. You get a lot more than f1 + f2 and |f1-f2|. The whole point of a DBM is to get just f1 + f2 and |f1-f2| or as near as damnit. It's depressingly easy to get 2f1 +/- f2 and all sorts of other products, welcome to the third-order intercept point

If you simply wanted a lot of mush with as many harmonics as you can get you can have life a lot simpler. Add the two sinewave inputs f1 and f2 together linearly and stick the combination through a comparator, which is about as much nonlinearity and you can get. Or square f1 and f2 up which already gives you a hefty set of odd harmonics on each signal and bang the result through an XOR gate, which is a sort of poor man's digital multiplier. If you wanted to do that on the bench then use a CD4046 PLL IC, set the VCO up for your local oscillator, take your F2 in AC coupled via a capacitor to pin 14, wire VCO out on pin3 to comparator in on pin 4 and pick your output from pin 1 - phase comparator 1 out, which happens to be an XOR gate. Lots of nasty noise

The principle can be taken further easier in the digital domain - you can feed the output back to the input easier in the digital domain than your current aims in the analogue domain. f1 and f2 end up generating very noise-like hash - this is the principle behind the pseudo-random binary sequence generator which can be **broken link removed** which probably shows you what you would get in the end - noise.

I'd suggest you back up a bit and ask yourself the more fundamental question - exactly what are you trying to do here?