3 ms·

If I understand it right, in the example of clamping a serial line under a monitor, the wires will exhibit higher resistance at higher frequencies, thereby less

by povik 6y ago

If I understand it right, in the example of clamping a serial line under a monitor, the wires will exhibit higher resistance at higher frequencies, thereby less current flows at higher frequencies. Less high-frequency current in the serial line means less radiation and less interfering current induced on wires inside the monitor. Wouldn’t that be to detriment of signal integrity on receiving end of the serial line? Is it so that only higher harmonics are affected, which is okay? Why did I hear putting ferrite on USB cable will improve signal integrity? Is that nonsense?

joshvm 6y ago
Perhaps it reduces interference in the same cable if the USB Vbus is noisy? It might also be just to suppress radiation from the cable which could affect other devices. Though I'm not sure you really need it on USB2. I've built custom totally unshielded USB2 cables that have been put on drones. The important thing is that the differential pair is twisted, otherwise it doesn't work at all. We never had any issue with EMI from the motors, which can be quite bad on quadcopters. This obviously isn't something that might pass CE testing, but my point is differential cabling is fairly noise tolerant. I'm fairly sure you could run USB2 through a coat hanger and it'd be OK. Most USB application notes also specify a small ferrite bead, so there is normally something on the PCB for filtering as well. Wurth have a nice app note on this (search for ANP024). USB3 though, is particularly bad for radiating with low quality cables. My logitech mouse dongle will drop if I connect a high bandwidth camera over a poorly shielded cable.
povik 6y ago
It seems like I was missing the fact that the resistance caused by a ferrite bead is imposed on the sum of currents going through the bead. That means opposite currents on a pair of wires are unaffected, as long as both the wires go through the bead.
kazinator 6y ago
> Wouldn’t that be to detriment of signal integrity on receiving end of the serial line? It was only 110 kbps. Serial consists of binary pulses. A framing bit comes first which starts a clock. The signal after that just has to be in the right range for indicating a 0 or 1 at each subsequent clocking. This is probably self-correcting to some extent, if the threshold for a level change is the same as the threshold for starting the clock. The inputs are likely Schmitt-triggered, which essentially recovers the square wave.

by povik 6y ago

joshvm 6y ago

Perhaps it reduces interference in the same cable if the USB Vbus is noisy? It might also be just to suppress radiation from the cable which could affect other devices. Though I'm not sure you really need it on USB2. I've built custom totally unshielded USB2 cables that have been put on drones. The important thing is that the differential pair is twisted, otherwise it doesn't work at all. We never had any issue with EMI from the motors, which can be quite bad on quadcopters. This obviously isn't something that might pass CE testing, but my point is differential cabling is fairly noise tolerant. I'm fairly sure you could run USB2 through a coat hanger and it'd be OK. Most USB application notes also specify a small ferrite bead, so there is normally something on the PCB for filtering as well. Wurth have a nice app note on this (search for ANP024). USB3 though, is particularly bad for radiating with low quality cables. My logitech mouse dongle will drop if I connect a high bandwidth camera over a poorly shielded cable.

povik 6y ago

It seems like I was missing the fact that the resistance caused by a ferrite bead is imposed on the sum of currents going through the bead. That means opposite currents on a pair of wires are unaffected, as long as both the wires go through the bead.

kazinator 6y ago

> Wouldn’t that be to detriment of signal integrity on receiving end of the serial line? It was only 110 kbps. Serial consists of binary pulses. A framing bit comes first which starts a clock. The signal after that just has to be in the right range for indicating a 0 or 1 at each subsequent clocking. This is probably self-correcting to some extent, if the threshold for a level change is the same as the threshold for starting the clock. The inputs are likely Schmitt-triggered, which essentially recovers the square wave.