Electronics > Projects, Designs, and Technical Stuff
Material selection and manufacturing for high frequency designs
(1/1)
Torpedoes:
Hey everyone,

I've been doing a lot of research lately on high frequency board design and I think I have a design that will work, unfortunately, I've been stuck on manufacturing and the information I find is confusing.

I'm looking to manufacture two daughter boards of different thicknesses. One for PCIe Gen 3 devices (8 GHz) and one for USB 3.1 devices (5 GHz). I've contacted three companies in China and they gave me three wildly different quotes ranging from $200 to $1200. I'm not sure if $1200 is normal for these types of designs.

I've read that FR-4 is not a good material to use so I've looked at other materials such as ROGERS 4350B, however, they seem to be incredibly expensive and I have no idea if it's even worth investing in. I've seen materials with various glass transition temperatures and dielectric constants and I know they play a part in the design but I have several questions:

1) Does the dielectric constant really matter? Is a lower value really needed or if I have the choice between 3.66, 4.05 and 4.6 and I calculate the required differential trace width and spacing for the impedance I'm looking for, can I get away with using a 4.6 board? If not, is there a target value I should be going for to support 8GHz signals?

2) What role does the glass transition temperature play, is this something I need to worry about? Should I try to go for the highest temperature I can find?

3) Do you think I can get away with using an "impedance-controlled" FR-4 from a random manufacturer assuming the design yields the correct impedance according to their stackup?

If anyone has any experience manufacturing small-batch HF designs, I'd love to hear any tips you would consider traps for young players like myself. I feel like I'm making a ton of mistakes but I don't know where to begin.

Regards,
DaJMasta:
PCIe gen 3 motherboards are always FR4, so I certainly assume it's manageable.  Part of what helps is that the traces tend to be pretty short, so even if you're losing some eye opening because of losses in the board itself, you still have enough to be received correctly.

Impedance controlled is really the more important part, because you need to make sure your manufacturer is using materials consistent enough so that your designs don't perform on some batches and fail on others because the dielectric constant isn't consistent, and you want to be sure that the solder resist isn't going to muck everything up either, which should be part of what's controlled in a controlled-impedance fabrication process.


Don't think glass transition temp is going to matter to you, unless your PCIe lanes are going to be routed under power MOSFETs or something, but it can be important in high temperature environments.


If you have any equipment capable of measuring at this frequency, or if you have access to it, you can always verify your work on a test piece.  Make a few stripline and surface microstrip connections on the fabrication process you're considering and determine what the losses/crosstalk really are and if that's controlled enough for your application.  If your trace lengths start to get long, or otherwise start getting towards the limit of the PCIe spec, though, then your dielectric losses could become much more of an issue.
Navigation
Message Index
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod