RF - High Frequency Issues in Design/Implementation

[sanchez9457]

Hello!

I'm new to the world of RF! I had a pretty basic question. So I'm currently building (or trying) a Radar system using some off-the-shelf components, two coffee tin can (acting as my antennas) and a teensy for ADC, and MATLAB to process the teensy data. So I'm currently building and testing using 2.5GHz components. But I was just wondering: what are some issues that you run into when you use higher frequency like say 20-30GHz? I know power becomes an issues - components will need more power to operate at higher frequencies - which also creates a cost issue for components that can take higher frequency. But what other issues in RF design and implementation come about with using higher frequencies?

Any discussion/feedback is much appreciated!

[dmills]

Way too broad....

**EVERYTHING** starts to look like a transmission line, including the passive components as they are no longer electrically short. Parasitic inductance and capacitance start to really MATTER, to the point that many networks are just carefully calculated shapes in the copper on a teflon PCB rather then being conventional passives at all.

PCB materials become increasingly lossy, and possibly dispersive, gain becomes increasingly expensive, small impedance discontinuities start to really matter, so connectors start to get really expensive, vias on a PCB start to look like stubs so you get into pain like back drilling vias. 

You get weird skin depth related effects, and signal propagation modes start to really matter, eventually you end up using waveguide rather then conventional traces. 

Small slots start to look like aerials, then waveguides.

Lots of fun playing up there.

Regards, Dan.

[T3sl4co1l]

Overly ambitious in many areas:
- Since you're new, you probably don't know (correct me if I'm wrong here) about signal propagation and losses (Friis' formula, for example), about how much gain is required in a typical radio receiver (a RADAR set is no different), about the methods employed to solve that gain requirement (IF strip + AGC).

This alone makes the receiver considerably more complicated than you are expecting.

Any one of those subcircuits would be a wholesome introduction to radio, particularly in the shortwave band (say around 7MHz).

- I don't see that you'll get anything into a Teensy.  The detected signal won't even twiddle a bit, at most distances.  Nor is the sample rate high enough to resolve reasonable distances: you're looking at more than a football field per sample!

- Circuit design at 30GHz is really cool, but intractable without a deep knowledge and understanding of transmission line theory; that, and a PCB level EM field simulator!  These are not hard to come by, but they are expensive, and well worth the price.

In contrast, at the 100s of MHz, you stand some chance of signals staying in wires (the PCB, and all its traces, may actually be under 1/4 wavelength), and in the 1.8-30MHz range (the shortwave band), components are downright easy to work with (capacitors are common off the shelf, inductors can be bought, or hand-wound with whatever wire is handy).

Tim

[scatha]

Gunnplexers could be a good entry point if you're OK with it being a monostatic rather than a bistatic system.

[sanchez9457]

Way too broad....

**EVERYTHING** starts to look like a transmission line, including the passive components as they are no longer electrically short. Parasitic inductance and capacitance start to really MATTER, to the point that many networks are just carefully calculated shapes in the copper on a teflon PCB rather then being conventional passives at all.

PCB materials become increasingly lossy, and possibly dispersive, gain becomes increasingly expensive, small impedance discontinuities start to really matter, so connectors start to get really expensive, vias on a PCB start to look like stubs so you get into pain like back drilling vias. 

You get weird skin depth related effects, and signal propagation modes start to really matter, eventually you end up using waveguide rather then conventional traces. 

Small slots start to look like aerials, then waveguides.

Lots of fun playing up there.

Regards, Dan.

This sounds pretty cool, actually. I have a few questions though if I may:

Can you explain a little about what you mean when you say that passive components are no longer electrically short? Like do you mean that these components will cause some type of impedance?

Why does gain become expensive?  Would we need to have more amplifiers? If the frequency is high up there in the 30GHz+ range, doesnt that mean attenuation is less frequent so we need less amps?

At high frequency would the system sense small impedances? Or why would the smallest impedance discontinuities matter?

Sorry for the questions and sorry if some of these are really dumb but I'm just really curious. Any discussion would be highly appreciated! Thanks!

[sanchez9457]

Overly ambitious in many areas:
- Since you're new, you probably don't know (correct me if I'm wrong here) about signal propagation and losses (Friis' formula, for example), about how much gain is required in a typical radio receiver (a RADAR set is no different), about the methods employed to solve that gain requirement (IF strip + AGC).

This alone makes the receiver considerably more complicated than you are expecting.

Any one of those subcircuits would be a wholesome introduction to radio, particularly in the shortwave band (say around 7MHz).

- I don't see that you'll get anything into a Teensy.  The detected signal won't even twiddle a bit, at most distances.  Nor is the sample rate high enough to resolve reasonable distances: you're looking at more than a football field per sample!

- Circuit design at 30GHz is really cool, but intractable without a deep knowledge and understanding of transmission line theory; that, and a PCB level EM field simulator!  These are not hard to come by, but they are expensive, and well worth the price.

In contrast, at the 100s of MHz, you stand some chance of signals staying in wires (the PCB, and all its traces, may actually be under 1/4 wavelength), and in the 1.8-30MHz range (the shortwave band), components are downright easy to work with (capacitors are common off the shelf, inductors can be bought, or hand-wound with whatever wire is handy).

Tim

Football field per sample is quite a lot! Thanks so much for the info! I don't actually intend on going into the GHzs when playing around with the antenna just a curiosity of mine to ask about the effects/issues with working with high frequencies.

[RadioNerd]

Antenna design gets very interesting at higher microwave frequencies as it possible to build small and high-gain antennas (veeery interesting for radar) with low effort. Parabolic dishes and horn antennas for 10+ GHz are quite small and portable... And can even be 3D printed if you are able to plate PLA or ABS...

[donmr]

Can you explain a little about what you mean when you say that passive components are no longer electrically short? Like do you mean that these components will cause some type of impedance?

When a wire's length is a significant fraction of a wavelength at the operating frequency then is isn't just a wire anymore, it's a TRANSMISSION LINE, with inductance and capacitance!  And it's an antenna (which receives and transmits)

At 30 GHz a conductor a few mm long has reactance.

[HAL-42b]

Gregory Charvat did one of those precisely as you describe it.

http://mrvacuumtube.blogspot.com/2011/01/mit-iap-11-radar-course-sar-example.html



Dave and Chris interviewed him on the Amphour some time ago.

http://theamphour.com/the-amp-hour-115-watcher-of-wraithlike-walls/

[TheUnnamedNewbie]

Can you explain a little about what you mean when you say that passive components are no longer electrically short? Like do you mean that these components will cause some type of impedance?

Think of it this way: In classical circuit analysis, you assume (gennerally) that any point in a trace is at the same voltage: If I measure 0.5V at the beginning of the trace, it will (at the same moment in time) also be 0.5V at the end of the trace, ignoring any drop due to losses.

If you go up in frequency, the wavelength becomes shorter. At mains frequencies (a few hundred Hz at most), the wavelength is on the order of 1000's of km. If you go to tens/hundreds of MHz, the wavelength is on the order meters. But at 30GHz, the wavelength is just 1cm! That means that if you were to look at a wire of 1cm carrying a 30GHz signal, the voltage on that wire would go from zero to the maximum, to zero, to -maximum, and back to zero again (it is a bit more complicated than this, but you get the idea).

What this means is that you can't really say very sensible things about the voltage over, say, a through hole resistor, because within the length of the resistor, there is more than an entire wavelength! Now how much current is flowing through that resistor if different parts inside of that resistor have a "different" voltage drop across them?! All those simple models start falling apart!

It gets even worse when you start including parasitics. A coil is made up out of a wire. That wire has a resistance. At low frequencies, we can just pretend the resitor is pretty much in series with the inductor. Now remember that the short waves caused a different voltage on all the bits of the resistor example earlier? It does the same in this inductor. Now we can't just pretend it's a resistor and inductor in series, but it's like the resistor and the inductor have "fused" into one strange, complicated component.

And it get's even worse when you start thinking about reflections and so on, where a two cables that are shorted at one end, appear to be an open circuit at high frequencies, or vice versa!

In short, RF design is a mess and complicated bit of wizardry - but that's what makes it fun for those who enjoy the complexity.

[SingedFingers]

Sub 50MHz club here. I liked my stuff lumped

[Mechatrommer]

But I was just wondering: what are some issues that you run into when you use higher frequency like say 20-30GHz?

save all the theories for others to comment. maybe you'll find this interesting... (among the cheapest 2nd hand in town)...
http://www.ebay.com.my/itm/Anritsu-37377C-40-MHz-to-50-GHz-Vector-Network-Analyzer-with-Opts-2-3-6-10-11-/252763504551?hash=item3ad9e0ffa7:g:ln8AAOSwWxNYnQld

[bson]

Can you explain a little about what you mean when you say that passive components are no longer electrically short? Like do you mean that these components will cause some type of impedance?

The shape of a component will matter and the signal will reflect partially off say the front edge of a surface mount resistor.  It will want to creep along the outer skin of the resistor (what is its material?  does it have print on it?), and then has to make it down the other end to the pad and trace.  As a result, at adequately high frequencies the resistor value ceases to be a resistance and it just becomes a kind of bumpy poor transmission line with lots of 3-axis 90 degree angles.  It will still have some resistive properties among others, just not the value printed on it.  Its resistance, whatever it has, will also be a function of frequency (like any transmission line) - R(w) since it depends on the skin depth.

[radar_macgyver]

In the 20-30 GHz region, everything is lossy, especially FR4. If you can build all the high frequency stuff in waveguide (or a pre-made module on high frequency substrate), it's not terribly difficult to build a radar with a Gunn diode, especially if you can find a module that incorporates a Gunn diode, mixer and iris (the aforementioned Gunnplexer module). If you can score one with a varactor, then FMCW radar is quite easy with something like a Teensy, since it only has to deal with the IF (in the audio band), and generate a VCO sweep voltage.

One source I've found for Gunnplexers is shfmicro.com (although they no longer sell them now, they do have a DRO-based replacement at X-band), they also have some inexpensive horn antennas that will easily outperform what you can manage with a coffee can. I've MacGuyver'd a couple of demonstration radars with their X-band Gunn modules (hence my username ).

If you stay in the Ku-band or lower, you can get by using connectorized parts. SMA connectors and cables are relatively cheap for these frequencies.