RF
I currently working on a RF design, with frequencies up to about 2GHz.
What is the "rule of thumb" for when to stop using lumped components (Capacitors, Inductors, Resistors), and start using microstrips ect. ?
P.S. I can see that lumped components are available up to about 10GHz, but they are pretty expensive.
Edit: What about decoupling capacitors, on supply lines to pll's ect., should they always be associated with RF Chokes to work probably (at 2GHz) ?
There are two methods I am aware of, three if you replace light with RF.
1: A simple circuit with an IR emitter, transmitting a short columnated pulse at a slight angle to the centreline of the measurement device. This refects of the remote object and comes back slightly offset. Capture the reflected light, measure this distance from the centreline, then the distance is simply geometry.
2: Uses a local oscillator, generating a sawtooth wave at a few tens of MHz. We transmit a laser pulse at the bottom of the triangle and trigger a sample and hold to capture the voltage when you get a reflection which is detected with a transimpedance amplifier and a photodiode.
The voltage + pulse count is directly proportianal to the distance.
Light takes 50ns (approx) to travel 15m and 1000ns (approx) to travel 300m. The detector needs a local oscillator of say 2/50ns = 40MHz We need 15cm accuracy, so lets make this 1 bit. We need to encode each 15m into one sawtooth and there are 100 x 15cm of these, we need a 7bit ADC that can sample at 40Mhz.
The counter needs to be able to count to 300/15 = 20 at 40MHz to achieve the full distance. The actual distance is limited by receiver front end sensitivity, output power (safety concerns) and the problems with carry chains on fast(ish) binary counters.
Each 15cm time interval is seperated in both time and voltage, so capturing it should not pose a problem.
The last component is a sample and hold. The circuit would require calibration to remove the error caused by the sample and hold's trigger delay. Other than that none of the components are expensive.
It is just as possible to use a very fast counter instead of the sawtooth, (and these do exist) but would be much more expensive.
3: You can substitude an RF transmitter for the laser and a directional antenna + RF front end for the reciever, otherwise the circuit is the same. (radio and light travel at the same speed)
4: By using RF you get another method, called CWFM, where you use the sawtooth to FM modulate the transmitted signal. The received signal is mixed with the transmitted one, the output of the mixer is a hetrodyne (frequency shifted representation) of the distance, an FM demodulator can turn this into a meaningful signal.
Generally, your assumptions are correct. When connecting a microstrip to a component, how closely you are able to match the physical size and shape of that microtrip plays the biggest role. Package size, termination style, and trim method for a series resistor are what matters.
Your other assumption that you can pretty much ignore transmission line effects for short runs is also correct. Often you'll see 1/4th of a wavelength used as the line between a 'short' transmission line (too short to matter) or 'long' (matters) in regards to the signal propagating through it, but I wouldn't suggest that as a rule of thumb. 1/10th the wavelength and below is where even phase delay becomes inconsequential and you can sigh with relief - you can just ignore transmission line theory completely and you probably won't go to engineer hell or otherwise be a bad engineer for doing so.
An easier way to think about this is with light. @mkeith gets credit for his comment where he mentions 'resolving' stuff. Take a lesson from optical microscopes: they can resolve smaller details if you use violet light, but there is just a limit where everything is too small to be resolved, and that's because it's too small relative to the wavelength to interact with the wave in a meaningful way. This applies to discontinuities for the most part - if its much smaller than the wave, then the wave isn't going to care.
Note: below, I am going to give more general tips on microstripping, but it will apply more or less depending on the wavelength at hand.
Now, back to the first part part, my recommendation on how to connect a 50Ω characteristic microstrip to an 0402 is simply not to. There are two things you have to think about whenever you must cause a discontinuity, reflection and parasitics.
Reflection is easy - keep the instantaneous (characteristic) impedance of a transmission line as close to the same for every step a wave propagating down it has to take, and make sure the other end is terminated with a matched load impedance, and all is well. And the moment you have to put any component in series, that happy dream is screwed. When connecting and layout this stuff, its best to view it in terms of damage control.
If your microstrip narrows, that will cause a potentially large impedance discontinuity. If your microstrip is 0.1" wide, you never want to do anything that will cause it to narrow or widen, except when you're mitering a corner of course. This means you really really should use an SMD package whose terminals are the same width as your microstrip (or combine parallel packages to simulate this), and one that has a high aspect ratio in the direction of the strip. And also as thin as possible. Basically, you want this thing to seem as if it is just another length of copper microstrip as you can manage. Obviously, a 1210 sized package would be perfect for a 0.1" wide microstrip. It's the same width, and it's aspect ratio is what you want too.
Anyway, the goal is always to minimize all the ways you might be introducing any sort of discontinuity in the characteristic impedance. You're causing damage, but try to do as little as possible. Damage control.
Now, the second issue is parasitics. A passive generally consists of two terminals, and the pads for them. If it is a series passive, you're going to have to create a gap in the microstrip where the passive is placed across. Which means we just created a little series capacitor too! Booooo! If you use a passive wider than the strip, you'll create larger 'plates', and also parasitics between the wider pads and the ground plane, relative to the microstrip. So one series parasitic capacitor with the gap and the two ends of the microstrip and ones to ground at either pad as well. If the pads are not wider, then you mainly just have to worry about that series parasitic capacitance. If the component has a longer aspect ratio, that makes the gap larger, and the larger the gap, the lower the capacitance. So this helps to minimize that.
One final oft overlooked thing (not to say you are doing this, but someone delivered here by the helpful guidance of google and reading this might): When using that 1/10th wavelength rule of thumb, that's 1/10th the wave length in the transmission line medium, not a vacuum. It's a little complex to figure out exactly what this is since a microstrip propagates the wave partially through the FR4 material and partially through air (and soldermask and cat dander or whatever is sitting on top of it), but it's usually within a few % of
$$ V_{p}=\frac{c}{\sqrt{\varepsilon _{re}}} $$
Vp of course being phase velocity, c being the speed of light, and ε_re being the relative dielectric coefficient, which usually is around 4.2 for FR4. Theoretically. Probably. Maybe? In the case of a microstrip, the dielectric coefficient must be corrected since only some of the wave is traveling through the FR4. There are several different ways to go about this using the width of the microstrip to help determine the 'effective' dielectric coefficient. But really, for the uses of figuring out if you even need to worry about any of this or not, its ok to ball park it usually.
Oh, I almost forgot about the antenna! No, the line is never the load impedance. The load impedance is an actual load - the characteristic impedance of a transmission line is the instantaneous impedance (the wave 'sees' 50 ohms impeding it's propagation at any given point along the line. It does not mean there is 50 ohms of impedance between one end and the other end, but that regardless of how far or close from the load the wavefront is, it always seems that same 50 ohm instantaneous impedance). The 50 ohm connector simply maintains this characteristic, but it is not in anyway a load. The antenna is the load, and it will have significant reactive impedance (at least, assuming the antenna is a useful one at your frequency). Anyway, as long as the antenna is a 50 Ω one, you'll be fine. If it's not....you'll need to match the impedance, and this beyond the scope of this answer. And yes, that means if nothing is connected to the antenna jack, you have an unterminated line that is reflecting crap and spraying crap out the end, which is why there are 50Ω termination end caps that too often people don't use but they should! EMC and all that.
Best Answer
The trouble with discrete parts is that they come with parasitics; e.g. an inductor will have a parasitic resistance, parasitic capacitance, and parasitic capacitance to the board. Generally speaking, as you approach 1GHz, parasitics start affecting your frequency response, even on small SMD components. Even if you buy "10 GHz" components, they'll have parasitics. The hard part of RF design is modeling the parasitics accurately (and knowing where they can be ignored). Also, there's a trade-off between precision and cost.
Microstrip elements are more abstract, so you generally have more control over what's going on. As others have mentioned, though, they are bigger.
Modeling components is time-consuming. Keep in mind, though, for a one-off project, you might find that tweaking the circuit is enough to get it to work. Repeatability over hundreds or thousands of units is a different matter. For this spectrum analyzer project, you don't need to worry about the 10.7 MHz IF blocks, but you will have to be careful with the 1-2GHz stuff. The bandpass filter looks like the most difficult part.