# does frequency multiplier work with signal that contain multiple frequency components?

for example, say we have a wide band frequency multiplier that takes the input of 100MHz-200MHz and the output have frequency of 300MHz-600MHz（a frequency tripler).

if the signal contains a 110MHz and also a 150MHz components (to be more realistic, with different power levels), will the output be two components signal that contains 330MHz and 450MHz?

----update below-------

The problem comes from a project where there are two surfaces that reflect MMW wave ($f_0$=50GHz) and one surface is moving, the other one is not. And both reflected signal go into a receiver and were downconverted into RF signal with a carrier frequency of 630MHz (this frequency was chosen to fit the frequency multiplier requirement).

From Doppler Effect the reflected MMW from the moving surface will have different frequency ($\delta f=f'-f_0$). So the intermediate frequency (IF) from the receiver will have two frequency components, one is 630MHz, the other one is 630MHz+$\delta f$.

Next we try to frequency multiply this IF signal with two frequency multipliers (with amps, mixers, filters), as shown in picture.

The frequency multiplier we use are

1 HMC445LP (Hittite, X16, 618.75M-687.5MHz, output 9.9-11GHz, a cascaded X2 multiplier)

2 RMK-5-83+ (Mini-Circuits, X5, 1.0G-1.6G, output 5G-8GHz, a diode quad)

So the question is, we only want the signal that comes from the moving surface, but we cannot eliminate the signal from the not moving surface. If the multipliers work with two (or more) frequency components, this project will probably work. If not, then other solution is needed.

• yes, otherwise it would be useless. any signal that isn't a pure sine wave is built from many frequencies superimposed together. May 1 '18 at 5:10
• Can you provide a link to the type of frequency multiplier your question refers to? I have a feeling that what you may be talking about is confusing you. May 1 '18 at 7:26

It will contain 330 and 450MHz.

But it is not linear so it will also contain

• 110+150
• 330+/-150
• 450+/-110

and all the other possible combinations of the input frequencies and harmonics.

This is called intermodulation.

It will then contain the intermodulation products from those frequencies eg. (450-110)+150

etc, etc ad infinitum these are 2nd,3rd nth order intermods. The higher order intermods fall around the wanted signal, unlike the harmonics, so you can't filter them out.

The whole process in non-linear i.e as the 150MHz gets stronger , the intermodulation mess gets dramatically worse

It depends what sort of 'frequency multiplier' you have. You might have a simple tripler, for instance an amplifier followed by symmetrical limiting, or a step recovery diode. You may have a complicated 'multiplier system' in your back box which is more capable.

The simple amplifier tripler is non-linear. For each input edge it outputs a faster edge. While it's called a tripler, this type generates all odd harmonics, with the 3rd the most powerful (after the fundamental), due to the square wave it outputs. It's only by following it with a bandpass filter that you could remove the fundamental and higher harmonic output.

The step recovery diode is non-linear. It generates a pulse (2 edges) for every input cycle. The ideal spectrum of a series of pulses is flat with all harmonics having equal power, however in practice the amplitude of the harmonics falls as the frequency increases due to the finite width of the pulse. Again it requires a bandpass filter to select the required harmonic.

Both types of simple multiplier output a waveform that is a version of the input, cycle for cycle, changing the waveshape. Therefore anything that distorts the shape of the input wave will change the point in time that the output wave switches.

If we consider a single input signal, then the multiplier will work as advertised. If we allow a very small component of another frequency signal into the input, it will cyclically modify the timing of the output edges, creating phase modulation sidebands on the original signal.

In the case of your two 110+150 MHz signals, let's assume they are phase locked at multiples of 10MHz. Their combination will therefore be cyclic in a 10MHz period, and the output of either multiplier will be harmonics of 10MHz. This is clearly not two outputs, but many, amongst which will surely be 330M and 450MHz.

If you allow a complicated 'mulitplier system', which can separate the input signals, apply each to its own multiplier and then recombine them, then of course you could have something that behaves as you ask.

• I have updated the problem, please give some comments. May 2 '18 at 4:24
• You should find another way of doing it. It's nonsense to throw your composite signal at a frequency multiplier, which then complicates things. Why not analyse the signal at 600MHz, when it's still simple? Trying to do things at 10GHz instead of sub-GHz seems like an exercise in masochism anyway. May 2 '18 at 4:40
• The ultimate goal is to extract the speed of the moving surface, while it is changing very fast (for example, a sudden jump from 0 to 1km/s), but the wavelength is long (6mm), time-frequency analysis can not resolve the sudden speed jump. So we want to increase the time-frequency accuracy by frequency multiplying and downconverting. May 3 '18 at 4:38