EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: ANTALIFE on March 22, 2022, 10:51:24 am
-
Hi All
A question for those familiar with USB 3 (or any other high speed/bandwidth standards). Signal Integrity (SI) wise, does 2-3GHz band (and above) hold much useful information for USB 3?
Reason I ask is that I am going through one of Keith Armstrong's PCB design techniques module (10A) (https://www.emcstandards.co.uk/advanced-pcb-design-for-cost-effective-si-pi-an) and at one point he says that using an ENIG finish actually dampens emissions/signals above 3GHz (due to nickels ferromagnetic properties). This (https://www.signalintegrityjournal.com/articles/25-how-interconnects-work-modeling-conductor-loss-and-dispersion) is the article that he references for those curious
-
First time i hear about that. Id say the effect would be pretty minimal. How you route those lines and do grounding is likely to have a much much bigger effect on EMI.
USB 3.0 is 5Gbit and so to get a nice clean eye diagram you need to have components even higher than 5GHz
If this is true then how does PCI Express work properly? The latest generation of CPUs is using PCIe 4.0 that runs at 16Gbit so we are well into 10s of GHz territory. Yet the LGA pads on the CPU are gold plating on a nickel plating. The spring pins that make contact with the CPU pads are gold plated steel. The pins in the PCIe slot are gold plated steel that make contact with gold plated contacts on a PCIe card going to a GPU chip that likely has ENIG gold pads to make BGA soldering more reliable.
Most of the reason for using ENIG is to provide a nice consistent flat surface for reliable soldering of the modern tiny SMD chips.
-
Depends how long lines are.. Few millimeters won't make a difference, unless it makes transmission line nonuniform a that point, which creates impedance nonuniformities, which create reflections...Which also might be important or not depending on signals risetimes and signal rates..
As you go faster that few GHz, even copper surface roughness will make a large difference. At super high frequencies, current won't go through nickel but gold because conducting depth will be very small.
As author say, that is not necessarily bad, because if bandwidth is sufficient enough to have eye that is open enough, but you suppress higher harmonics, you might even get less radiated EMC and less reflections...
At 10 GBits and up, you start back drilling vias because 0.5mm stub in a via can be a problem...
At 5Gbits, Nyquist frequency (as called in SI, not sampling theory) is 2,5GHz (half the bit rate). Up to 10-12 GHz (4-5x more) will be frequencies that will be needed to define edges. But those can have certain roll off in frequency losses.There is no such thing as brick wall filter. It might even help with pulse response.
Also at those speeds, transceivers already use all kinds of frequency/pulse compensations and pre distortions to compensate for transmission line non linear loses...
-
USB3 puts out a lot of crap in the 2.4GHz band among other places. Its got a radiaton for every band. Splattertastic! Its caused me numerous issues with RF measurements. I turn off the offending device- that where my know how ends.
A while back I read this white paper. It goes in to a lot of detail about the EMI issues associated with USB 3 and some mitigation schemes.
Its on the usb.org site for free, strangely they have not monetised every word. www.usb.org/sites/default/files/327216.pdf (http://www.usb.org/sites/default/files/327216.pdf)
Great advice on high speed PCB design:
www.ti.com/lit/pdf/spraar7?keyMatch=SPRAAR7H (http://www.ti.com/lit/pdf/spraar7?keyMatch=SPRAAR7H)
www.ti.com/lit/pdf/szza009?keyMatch=SPRAAR7H (http://www.ti.com/lit/pdf/szza009?keyMatch=SPRAAR7H)