Making a phase shifter on FPGA

[gmish27]

I looked all the suggested posts when I typed the topic and so I am asking about this once again.

I wanna produce a constant phase shift of 120 degree at 1-3 GHz. Is there any possible IC for it?? Probably the answer is NO (I am not sure).
So I wanna know how to implement the same on FPGA, i.e I provide an input signal to the programmed FPGA and I get an output signal phase shifted by 120 degree. (NOTE: frequency range is 1 to 3 GHz).

 

[FvM]

The question doesn't sound well considered. The I/O delay of any digital IC can be expected to be a multiple of 120°. At best you can achieve it as a differential delay between two outputs. It's unlikely however, that you'll find a FPGA suited for your purpose. I would rather try with passive LC and distributed circuits.

 

[kalyanasv]

Hi,

If the signal is digital, which I presume it is, you can consider the Rocket I/O's provided by Xilinx App note. This si somethign I have tried but there are new varieties which probably better rates or better signal integrity which is the main concern at such rates more than anything else.

Have fun

 

[gmish27]

but are passive LC circuits gonna provide accurate output SUCH HIGH FREQUENCIES? Also, you said I/O delay of any pin would provide the delay, so how can I use this feature of a digital IC.

---------- Post added at 08:39 ---------- Previous post was at 08:38 ----------

Kalyanasv: sorry I did not mention about it, but I have to produce the phase difference between two sine waves oscillating at 1-3 GHz. The signal is analog

 

[FvM]

but I have to produce the phase difference between two sine waves oscillating at 1-3 GHz. The signal is analog

So why do you think to use digital logic ICs?

Also, you said I/O delay of any pin would provide the delay, so how can I use this feature of a digital IC.

I said it's a multiple - and it will be very inaccurate. It's just a bad idea.

but are passive LC circuits gonna provide accurate output SUCH HIGH FREQUENCIES?

What do you know about RF circuits?

 

[gmish27]

Very less FvM. I am just in my 3rd year where I will be studying the RF devices and circuits for the first time.

 

[gmish27]

FvM can you provide me instructions to do that? Someone suggested me to go for transmission lines (microstrip) to do the same...

 

[FvM]

Transmission line segments can provide exactly defined signal delays according to it's "electrical length". Unfortunately, fixed delay involves frequency dependent phase shift. If you intend a fixed phase shift over a frequency range, it can't be achieved by a simple delay line. A complex filter will be required, the number of elements depends on the frequency intreval and the tolerated phase error.

 

[gmish27]

ok FvM, what if I juts want to produce the phase shift at constant frequency, say 1GHz? Then how should I design my microstrip circuit?

 

[lucbra]

the stripline technique is indeed a good way to go. Unfortunately, this is not for beginners. FvM stated is exactly: the electrical length and frequency are closely related.

 

[FvM]

what if I juts want to produce the phase shift at constant frequency, say 1GHz? Then how should I design my microstrip circuit?

That's actually easy, I think. 120° at 1 GHz corresponds to 0.33 ns, which is about 5 cm stripline length, depending on your stripline design parameters. In addiion, you would want to implement a power divider, e.g. a Wilkinson divider to split the input signal in two pathes. Power dividers and directional couplers - Wikipedia, the free encyclopedia

 

[gmish27]

FvM please tell me how did you find that length. I wanna learn, please.

 

[FvM]

I assumed a propagation delay of 6.6 ns/m, or in other words half lightspeed. For transmission lines with homogenous dielectricum, the lengthening factor is equal to √Er. FR4 has an Er of 4.3 to 4.6, but part of the electrical field is straying in air, so the effective micro strip Er is lower, e.g. around 4, resulting in a lengthening factor of 2.