Right angle RF tracks - case study


#21

This is amusing. All the usual suspects here who previously lamented neglect of scientific evidence joined in chorus of blindly repeating: " Carthage must be destroyed!" Forget that pesky science that has nothing to do with the real world and all those pesky textbooks which are written by those clueless engineers.

I was wondering how long it would take you to arrive to this final argument. If you think about it, it is really irrelevant to what we are discussing here. Again, the point of the case study is to demonstrate that right angles in transmission lines can have detrimental effect to the circuit performance. I believe I have done it. Can those issues be fixed? Of course! Anything can be fixed! Theoretically you can match any output to any input without worrying about all those pesky 50 Ohm transmission lines, RF layout best practices etc. The question is how much effort and lab time with a network analyzer it would take to do so. Even then it helps to understand how you got certain impedance in the first place, to be able to predict matching elements response. Now imagine those cases when people don’t even worry about tuning the circuit and just copy a reference design and end up with a circuit that doesn’t work.

And in the words of Jayne Cobb: “If wishes were horses we’d all be eating steak.” Enlighten me what is a half wave length at 915Mhz? I thought we were talking about practical things here. How many additional turns you will have to make to accommodate that length? Even at 10Ghz the trace lengths can be prohibitive in a busy layout design, just to add additional trace lengths because of a simple turn.

Now let me save you some time on typing up the counter argument. Something to the effect of : "Well you used half wave length of traces as an assumption in your example and now we can’t use it to solve the problem :frowning: ) Am I right? When I did it, I used it to isolate the problem and show its affects more clearly. I could’ve just as easily added whole bunch of other impedance turns on the Smith chart just to show the effects if all the real transmission line lengths, but why? What you are proposing is a “practical solution” to a real problem. That doesn’t seem very practical to me.


#22

My reading of the disagreement is as follows:

Side 1:
A sharp 90 degree turn has an effect but it’s influence is overpowered by other more important effects until you reach ultra high frequencies.

Side 2:
In high frequency applications you already need to eliminate all other factors, which means the form of the bend is one other influence that you can optimize.


Both sides agree that two 45° bends are much better than a single sharp 90° bend.
You even agree that a curved trace is equally as good as two 45° bends.

At least if i read this comment correctly:


So if i am not totally wrong you guys mostly agree and we can all calm down again and use our time to help kicad users with kicad specific problems.


#23

Man, you’re like two steps ahead of me.

I dare you, do it. Include all your transmission lines with real lengths and corners and show me how you’ll match it with that single inductor. You would have to constrain each line segment to multiple of half wavelength. That doesn’t sound like it would help in layout.

Yes, I agree with that. So if you want we can stop here and you can ignore what I wrote before this.


#24

Then I guess I’m in the “Side 3” or in both at the same time. I’m a lone reed … :slight_smile: My point was to be careful to use categorical statements like :“90 degree corners never have any substantial effect until you reach ultra high frequencies” I just demonstrated that it can have significant effect even at sub-gigahertz frequencies. But then again, I admit that there are many cases when you can easily disregard corner influence altogether. It all depends. As usual you just have to use your head to think to figure out when it is important and when it is not.


#26

Your previous comment would have been such a nice place to stop :wink:


#27

Agreed! (additional symbols)


#28

As usual you just have to use your head to think to figure out when it is important and when it is not.

Or just make traces rounded if software could do it automatically without ever resorting to use less reliable tools like head


#29

I think that since sometimes it matters, not doing right angles became a ‘best practice’. Avoid them and you don’t have to think about it on an individual case by case basis?


#30

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.