24GHz Mixer - How to shift the frequency of a 24GHz(100MHz bandwidth) signal by +-4kHz

Why do I want this? When a 24GHz radar signal hits an object that has a relative velocity, the frequency of the signal will shift by a frequency on the order of kilohertz. I want to create this process of frequency shift electronically.

I receive a 24GHz(100MHz bandwidth) signal from the antenna. I want to shift the frequency by something on the order of kilohertz (from -4kHz to +4kHz).

(the carrier signal is 24GHz, I want to turn it into what it would have been, had the carrier frequency been 24.000004Ghz)

I have a few ideas:

(+4kHz here is just an example, it could be anything between -4kHz and +4kHz)

Using this.

I will put the received signal to LO. Put cos(2pi 4kHz t) and -sin(2pi 4kHz t) to IF1 and IF2 ports. Does this have any chance of working? Would it cause problems to put a 100MHz bandwidth signal to a LO port?

Another idea is: I can demodulate it by, say 20GHz. Put 20GHz to the IQ modulator I posted above and shift it above by putting cos(2pi 4kHz t) and -sin(2pi 4kHz t) into IF1 and IF2. And then modulating the signal by 20.000004GHz.

• Move fast? .... – user16324 Oct 27 '15 at 12:24
• Are your signals and oscillators even stable within that frequency range? – PlasmaHH Oct 27 '15 at 12:31
• What problem are you trying to solve? Frequency-shifting a signal by 0.2 ppm seems like an odd thing to do. – Adam Haun Oct 27 '15 at 12:31
• @AdamHaun When talking about radar applications, it would not be odd at all :-) – Bimpelrekkie Oct 27 '15 at 12:47
• @PlasmaHH I need the oscillators to remain coherent for 80ms. I am not concerned by the received signal. I assume the received signal is proper. – grdgfgr Oct 27 '15 at 12:56

Note that this "GaAs MMIC I/Q MIXER" is, to put it bluntly, basically just a couple of fast diodes in a package !

"I will put the received signal to LO.... Does this have any chance of working?" Yes if your received signal is strong enough. Note that on the datasheet, first page above the table it says: LO = + 10 dBm If your received signal is less than that, it might not work.

"Would it cause problems to put a 100MHz bandwidth signal to a LO port?" Not as long as that 100 MHz is within the 21 - 30 GHz range. On page 12 there's a schematic of the LO input, note that it is AC coupled and that there is an inductor. These are tuned to that 21 -30 GHz range so any signal outside that range will not be able to get in as the input is not tuned for that.

You mean that you want to frequency shift your received signal by 4 kHz by applying 4 kHz quadrature signals at the IF inputs, 24 GHz (BW = 100 MHz) at the LO input and then have the resulting 24.000004 GHz at the RF output ? That might work but I would start by applying the Received signal at the RF input instead if the LO. And even then I doubt whether the frequency shift will happen and if it does, if it will have enough conversion gain.

This module looks like a downconversion mixer to me so maybe it is not what you need.

Why don't you look for a module from for example MiniCircuits, Skyworks, Tryquint or Avago which is more suitable for this task ?

• That was very helpful, however I do not fully understand what kind of modules would be more suitable for this. Also the reasonI want to put the received signal to LO is because the only components that allowed the received signal to be +20GHz were demodulators. – grdgfgr Oct 27 '15 at 13:36
• After reading Dave's answer I agree that shifting a 24 GHz, 100 MHz BW signal by 4 kHz is not an endeavor that is very likely to succeed. You mentioned that you are new to this area. Combined with the fact that anything above 1 GHz is not for the uninitiated I suggest you have a look at some literature and see how these things are done. – Bimpelrekkie Oct 27 '15 at 14:03

Trying to shift a signal by a tiny amount that's within its own bandwidth is virtually impossible to do with a single stage. The problem is that mixers are far from perfect, with not only the sum and difference signals appearing at the output, but also a significant amount of the original input signals. In most applications, these unwanted signals are removed by post-filtering, but that can only work if their spectra don't overlap.

In other words, what you're trying to do will almost certainly require at least two stages of conversion, one to shift the signal by more than 100 MHz, and another to shift it back by whatever the first shift was ±4 kHz.

(It's possible that this is what you were saying in the last part of your question, but it wasn't at all clear what you meant.)

• I also had this idea: I can demodulate it by, say 20GHz. Put 20GHz to the IQ modulator I posted above and shift it above by putting cos(2pi 4kHz t) and -sin(2pi 4kHz t) into IF1 and IF2. And then modulating the signal by 20.000004GHz. So I use 2 regular mixers and 1 quadrature mixer? – grdgfgr Oct 27 '15 at 13:36
• You are using standard terms in nonstandard ways. "Demodulate" means to reduce a signal to baseband; "modulate" is the opposite process. What you seem to be describing is more like the double conversion that I was talking about, with a 4 GHz IF. I don't know why you'd need a third mixer, though, unless you're using it to generate the 20.000004 GHz second LO signal from the 20.000000 GHz first LO. – Dave Tweed Oct 27 '15 at 13:42
• I am sorry for my incorrect terminology, unfortunately I am quite inexperienced. I call it 'demodulate' because usually demodulators take in a high frequency signal and high frequency LO. "I don't know why you'd need a third mixer". Yes, I figured that it would be impossible to have two 20GHz signals with 4kHz difference unless one of them were created from the other. What are the standard names for the components I am looking for? Mathematically all I want to do is multiply and add some signals, but with real life components at 24GHz, it gets far more intricate than that. – grdgfgr Oct 27 '15 at 13:48
• Yes, it does get very intricate, and there's probably a better approach. Please add a higher-level description of what you're trying to accomplish overall to your question, and maybe we can offer suggestions that avoid the need for such a strange conversion in the first place. – Dave Tweed Oct 27 '15 at 13:57
• When a 24GHz radar signal hits an object that has a relative velocity, the frequency of the signal will shift by a frequency on the order of kilohertz. I want to create this process of frequency shift electronically. – grdgfgr Oct 27 '15 at 13:59