RF Design: Guidelines for Driving 125MHz Square Wave w/o Ground Plane

[dink]

My take on RF in electronic design has always been, if the wavelength of the frequency you are working with is less than 10 times the length of the board/device(s), you should be using impedance matched networks for traces and cabling.

Currently I have a problem where the wavelength is the same length as the object I am designing (isn't a PCB), and the cost of RF matched cabling is 10x to 100x the budget I was shooting for. Specifically trying to drive a 125MHz square wave into an off the shelf flat flexible cable approximately 10 inches long, without a ground plane. I am assuming the third taylor expansion and thus designing to 625MHz.

What are your thoughts on these rules? Would significant jitter and signal degradation occur?

Thanks

 

[berkeman]

My take on RF in electronic design has always been, if the wavelength of the frequency you are working with is less than 10 times the length of the board/device(s), you should be using impedance matched networks for traces and cabling.

Currently I have a problem where the wavelength is the same length as the object I am designing (isn't a PCB), and the cost of RF matched cabling is 10x to 100x the budget I was shooting for. Specifically trying to drive a 125MHz square wave into an off the shelf flat flexible cable approximately 10 inches long, without a ground plane. I am assuming the third taylor expansion and thus designing to 625MHz.

What are your thoughts on these rules? Would significant jitter and signal degradation occur?

Thanks

If you match the Zo of the drive and receive to the cable, you shouldn't get much in the way of ringing. You will get a fair amount of crosstalk to other conductors in the cable, though. So it would be best if you put this signal at one end of the flat cable, and surround it with two ground lines.

It would be best to use a separate small diameter coax for the clock connection, though. Shouldn't be all that expensive...

 

[yungman]

Is this a digital clock signal? If so, it is not nearly as critical as analog rf signal. Like Berkman said, use two adjacent lines as ground. Also, you can also try copper tape and tape on one side of the flat cable. Make sure...I repeat make sure you ground the copper tape at the pin of the two ground lines sandwiching the 125MHz signal. Just simply ground in any convenient point is a no no. This is because of the image current. Image current is the surface current induced on the surface of the ground plane from EM boundary condition. It always like to stay close as possible to the signal, or else, it will form a loop and you'll have all sort of noise issue, impedance issue etc. Put the tape onto the flat cable, end the tape to the two ground pins where the cable ends and solder right there. Then you'll have the continuation of the ground return signal. This is the most basic thing to look for as a signal integrity engineer...to control the ground image current.

Just use a scope with probe that has very short ground and look at the signal, you might be able to get away without even putting the copper tape. 125MHz clock is not very fast! Without looking at the cable, if you have two ground one on each side, my wild guess is it is about 90 to 130 ohm characteristic impedance, do put a resistor something like a 100 ohm at the terminal point ( or 200 pull up and 200 pull down so you don't load the output too much. )and see whether it will tame some of the ringing. 10" and 125MHz clock, you almost can get away with not copper tape.

 

[berkeman]

Good point by yungman about the 'scope probe. I mention one version (a Z-lead probe) in this thread:

https://www.physicsforums.com/showthread.php?t=458430

.

 

[Enthalpy]

USB and SATA cables use much worse frequencies and lengths.

Yes, insert ground wires betweens the signals.

Better, ribbons exist that have twisted pairs (for differential signals), with adjacent pairs having different pitches (or it would be useless).

You write "square wave"... If it's data then the main difficulty isn't steep edges but dispersion, that is, unbalanced 1 and 0 will produce a less attenuated mean value arriving later. Answers: modulate a carrier frequency, use NRZ, or represent each byte of data by something like 6 positions at 1 among 12, or add enough parity... Ethernet does such things.

 

[yungman]

USB use shielded twisted pairs, it is very controlled impedance. I designed a 36" flex circuits using Dupont AP material. It is only 0.5" wide, I fitted a USB, Firewire and two of the 2.4GHz RF signal through it. It was tested cross talk of about -27 to -30dBm to the adjacent signals. But I had to use guard grounded trace like Berkman suggested and with ground plane on both top and bottom to achieve the crosstalk spec and characteristic impedance of 90, 110, 50 and 50 resp. It's the Firewire that was hard because of the 110 ohm, the trace ended up to be 4mils width only and the attenuation is quite high. Can't exactly get a perfect eye pattern. Started rounding already.

From my experience, dispersion is not that bad for a short length, at 125MHz, you get about 4nS or high and 4nS of low. The transit time through the 10" cable is only about 1nS or so, dispersion is at pS range for this short length, it is really not too bad. It is more the ringing due to mismatched impedance that matter, you don't want the ring to cross the threshold of the logic level. That's why I suggested using pull up pull down resistors to give some sort of termination to lower the ring a little and use the scope to look at the receiving end to make sure the ringing is under control. But op has to be very careful on the scope probe setup.

 

[berkeman]

USB and SATA cables use much worse frequencies and lengths.

As yungman points out, there is a reason that they work. Uniform Zo.

You write "square wave"... If it's data then the main difficulty isn't steep edges but dispersion, that is, unbalanced 1 and 0 will produce a less attenuated mean value arriving later. Answers: modulate a carrier frequency, use NRZ, or represent each byte of data by something like 6 positions at 1 among 12, or add enough parity... Ethernet does such things.

Dispersion issues are distance related in cables. The OP was about a short cable in overall length.

 

[dink]

Thanks for all the responses. I pitched the copper tape idea using kapton tape layers and foil to impedance control, it was shot down as "arts and crafts." 125MHz isn't very fast, but we have a very critical engineering crew. I think I have a solution. Thanks again.

 

[yungman]

Thanks for all the responses. I pitched the copper tape idea using kapton tape layers and foil to impedance control, it was shot down as "arts and crafts." 125MHz isn't very fast, but we have a very critical engineering crew. I think I have a solution. Thanks again.

Is this a real project at work? Tell them to layout a flex circuit, you can get any length, any stack up you like, all the control impedance, signal integrity requirement you want!

If it is a class project, you don't have to listen to them!

Too "arts and crafts"??! I work close to 15 years in a company consists 75% PHDs in the lab, we defined the meaning of arts and crafts if you see how we experiment new ideas! I remember one scientist was working on laser stuff on the optical table, care to guess how he fixed the lens in place?...chuing gum!...on top of a coke can! That's arts and crafts!