I have a purely analog circuit operating at up to 1MHz, though it is not inconceivable that at some point in the future it will be modified to work at 10MHz. [Just future proofing here - I already saw one project where Marketing asked if they could stick a 0 on the end of the spec. This is better than some places I've worked where they didn't ask!]
Signal levels are in the mV region at the start of the circuit, rising to a couple of volts at the output.
I intend using stripline to impedance-match the output to the cable it drives. My question is, for the earlier stages, is it better to try and use series resistors and impedance-match the tracks between stages, or simply to use big fat tracks for minimal inductance? The IC's will be about an inch apart (much less if I can manage it). The IC's involved have significantly higher GBW's than the signal.
if impedance matching is necessary depends not only on the frequency, it depends on the length of the cable measured in wavelength at the given frequency. If the cable is very short compared to the quarter of a wavelength on the cable at the given frequency, the matching ist not necessary.
Example: at 1 MHz, the vacuum wavelength is 300 m, the wavelength on the cable is about 200 m, a quarter is 50 m, a cable shorter than about 5 m should be no problem without impedance matching. At 10 MHz, the quarter wavelength is only 5 m, the cable should be shorter than 0.5 m. The same is true for striplines instead of cables.
I'm something of a novice here, but at those frequencies and spacing, my opinion is that you don't have to worry about impedance matching.
Speaking generally, you worry about impedance matching when...
1) As others have mentioned, you're driving something at the end of a cable (transmission line) that's long enough that the load termination will be transformed (via the length of the cable) to something you "weren't expecting." E.g., a 500ohm load at the end of a cable that's a quarter-wavelength long at the frequency of interest presents itself as a 5ohm load to your driver -- which is most likely "unexpected." :-) (...but at
10MHz, in coax, a quarter-wave is in the ballpark of 5m -- huge compared to your "serveral cm" IC spacing).
2) You're trying to extract the maximum possible power from a sensor (like an antenna) -- typically because the maximum available power is so tiny (micro- or nanowatts) to begin with. This is the case that seems to be (over-) emphasized in school -- if you have a ~50ohm antenna, surely you want the input of your LNA to be ~50ohms as well, right? Well, in lieu of any other information, sure... but when you're talking about such tiny power levels, you really need to start worrying about noise as well, and sometimes it turns out that your sensor already has so much noise on its output that driving even pretty large mismatches doesn't affect the system's overall noise figure significantly. This is why you often see, e.g., shortwave radio antennas that get fed into (the very high-impedance of) MOSFET gates, where the MOSFET is acting like a pre-amplifier: There's already so much noise on the shortwave bands that even if you did build a proper matching network for the MOSFET, it wouldn't change the system noise figure appreciable -- and it could actually make it worse if it caused you to move the bias point of the MOSET.
Basically: if you're not at the point where you need to worry about the noise of your op-amps and their associated feedback resistors, you likely don't have to worry about this.
3) You're planning to connect a bunch of circuits together and just want a well-defined standard impedance for the sake of convenience: Pretty much all test equipment expects 50ohms (or 75ohms), audio gear expects 600ohms, etc., so why not make your life a little easier and design to these standards anyway if it doesn't add must circuit cost or complexity? This one is pretty self-evident, of course.
You need to provide a lot more information in order to get a satisfactory answer.
In the rf world, devices can be damaged and/or needed power can be lost. In the digital world, data setup and hold time can be diminished and clock rise/fall edge fidelity can be compromised. In the high fidelity audio world, it doesn't matter because nobody can hear the difference.
What happens in your circuit if a certain percentage of the signal gets reflected back and forth?
Bob
--
== All google group posts are automatically deleted due to spam ==
I think it's going to be more like the hi fi audio world. The driving IC's can source /sink about 50mA, allegedly, the signal levels are low and frequencies well under the ICs' GBW limits. The important thing is, the other answers have given me confidence that the best FIRST approach is wide, low-inductance tracks, rather than try to control their impedance, and keep the IC's as close as possible. It'll work well enough for the circuit to get going, and I can probe around and add series resistors if it experiences unexplained problems, and re lay out the PCB if that makes a measurable improvement.
A cable (like twisted-pair 110 ohm wiring in Cat5 cable) has some stray capacitance (so some signal current is lost in transit), and some inductance (so some signal voltage is 'bucked' by the wire). If you drive the cable with high impedance (over 110 ohm, in this case) the stray capacitance causes frequency dependence. If you drive with low impedance (under
110 ohm) the inductance causes frequency dependence. And if you keep exactly at 110 ohm, frequency response is flat (but you still lose power to the terminating resistance).
So, the answer is in your error budget and in the stray and circuit capacitances and inductances upstream of the cable connection.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.