Digital speed limits of 2 layer FR4

[vvanders]

I've done a bit of digging around but I can't find any solid information. What's the realistic speed limits of a 2 layer PCB, assuming you've got a ground plane on the bottom and signals on the top.

I've heard everything from 5 Mhz to 80 Mhz but was wondering if anyone has some solid numbers.

 

[RCinFLA]

If you are talking about FR4 fiberglass it can be used at GHz's. Just a matter of how much loss you can tolerate.

Mobile phone use FR4 fiberglass and have 50 ohm striplines and microstrips to feed RF filters and antenna.

 

[vvanders]

So it'd probably be possible to use LVDS at 100-200Mhz on a 2 layer board?

 

[crutschow]

As you go to higher frequencies layout and good decoupling becomes an increasingly significant factor in the quality of the signals. At some point you have to start using stripline to control the trance impedances and use line drivers for long trace lengths, terminated in the trace characteristic impedance.

For short traces and a well laid out board, I would expect you could go to 100MHz or more with standard digital circuits. For a sloppy layout you might be doing good to get 5MHz.

Of course with only 2 layers and one of them a ground plane, you're rather limited as to the complexity of the circuit, unless you want to put in a lot of wire jumpers.

 

[vvanders]

I'm working with a FPGA so pinswapping to minimize jumpers shouldn't be a huge issue for me. I've seen a couple of app notes for good PCB layout(no 90deg, track spacing, etc), if you know of any other good resources for that it would be appreciated!

 

[Mikebits]

I have gone as high as 2.4GHz using Micro strip on FR4 material, at 200 MHz it should not be a problem unless your layout is bad. Total ground underneath. You just need to work out the math for trace width. Strip line can't be done on two layer board, so forget about using that.

One more thing. For digital signals at high speed, you should add on the layout, smt pads for a series resistor on the source end, and a shunt RC leg on the load end. These may not be needed and you can keep the shunt end unpopulated and the source end at zero ohms, but these usually help when trying to improve signal quality. I assume you have a high speed scope to make these measurements. If your using large data lines then just use source resistor pads. Use the RC shunt for critical signals like clock and such.

Good starting values for series resistor is like 12-22 ohm and shunt values is 22pf cap and 12-47 ohm resistor. It is somewhat a trial and error exercise.

https://www.electro-tech-online.com/custompdfs/2009/09/Microstrip.pdf

I recommend that you get this book.
https://www.amazon.co.uk/Speed-Digital-Design-Howard-Johnson/dp/0133957241

HJ is the Gordon Ramsay of Digital circuit layout.

 

[vvanders]

Awesome stuff, very much appreciated!

 

[BrownOut]

At some point you have to start using stripline to control the trance impedances

Think he means "trace" impeadances.

 

[Technogumbo]

Mikebits, I came across this post because I was searching for posts related to clock problems. In any case, the book you recommended, High-Speed Digital Design is awesome! I really appreciate it.

In relation to this post, in the book they talk about FR4 boards running in the Ghz range...assuming you layout the board correctly, choose the most appropriate packages, and compensate for adverse effects of running at that high of a speed.

 

[crutschow]

Think he means "trace" impeadances.

Yes, an odd typo. Although sometimes an altered mental state might be useful when designing high speed circuits.

And I assume you mean trace "impedances".

 

[BrownOut]

I didn't point that out just to be picky. To the unintiated, that might look like a real kind of trace, and could be confusing. I cant' tell you how many times I got confused by a typo when I was trying to learn. And, I was well into my 6th year of Engineering before I ever discussed board layout.

 

[crutschow]

I didn't think you were being picky, BrownOut. Certainly typos can be confusing at times. Just though it was kind of funny that a comment about a typo had a typo.

 

[BrownOut]

It wasn't a typo. I'm a poor speller. But I'm a great scrouger