5
\$\begingroup\$

USB 2.0 standard uses bi-directional communication over tightly controlled differential transmission line. The signals are originated from a 90-Ohms controlled driver, and are received with 90-Ohm impedance-controlled receiver. The direction is determined by half-duplex protocol, so there is no collision possible. Signal re-drivers/conditioners are not included into USB 2.0 Specification.

To get the signal conditioned, Texas Instruments offers a new class of devices, USB 2.0 re-driver, TUSB211/214 family. In application diagram, the device sits simply on the top on uninterruped differential transmission line, which apparently shorts both in-out ports of the IC, unless they use the miniscule inductance along the 2-mm long trace. The application diagram (and PCB layout) shows this: enter image description here

Could someone explain how do they do this?

\$\endgroup\$
  • \$\begingroup\$ patents.google.com/patent/US20140207984A1 - This patent application may be of interest. It describes the re-driver as a "current injector" \$\endgroup\$ – vofa Jan 25 '18 at 2:49
  • \$\begingroup\$ @vofa, thanks for the link. Still unclear, the signal delay across the 2mm trace (actually it is more like under 1 mm) is about 10 ps, which would require a 100 GHz gate to discriminate. Trace inductance between ports is about 1 nH. What can one do with this? \$\endgroup\$ – Ale..chenski Jan 25 '18 at 3:08
  • \$\begingroup\$ Maybe there is a piece of ferrite inside the chip? Get a magnet out and test it. \$\endgroup\$ – Andy aka Jan 25 '18 at 12:06
  • \$\begingroup\$ I found vague reference to "detectors" that measure slew rate and inject current if it is low on TI's forums. Since this has associated patents I'm sure TI has done everything possible to ensure we cannot find out how it is actually done. Any chance you can decapsulate or xray the IC? \$\endgroup\$ – vofa Jan 25 '18 at 16:35
3
\$\begingroup\$

Okay, after the patent link provided by @vofa and some Googling around, here is the answer.

The standard USB signaling (at high-speed 480 Mbps mode) is a 45 Ohm load driven by 18 mA current source, plus another 45 Ohm load at the receiver. This arrangement makes ~400 mV of signal. The differential eye is therefore +-400 mV. However, over long cables, the signal level gets attenuated, and signal edges slow down as well. This TI circuit is intended to alleviate the problem of signal degradation.

The circuit sits on the top of USB 2.0 transmission line. It has "edge detectors", for positive edge, and for negative edge, sniffing on bus activity. When the detector detects a positive edge (with likely low sense threshold), the other block of the IC injects additional current into the data line. This provides a boost to the edge slope, and to DC level of signal eye as well. When the circuit detects a negative edge, it applies the current into D- wire, boosting the negative transition. The boost is controlled in analog way by external bias resistor, and can be set from 2 mA to 6 mA of extra current, all going into the lump 22.5 Ohm loaded transmission line. Since it is a current source (having theoretically infinitely high impedance), this arrangement doesn't affect transmission line characteristic impedance (aside from parasitic package capacitance/inductance). Also, it doesn't matter for the IC from which side the edge is coming, it boosts it regardless of the direction.

To avoid extra DC shifts in non-HS transmission (LS, FS, suspend, resume, chirp, etc.) modes, the IC has some state machine that follows basic USB protocol and apparently disables the injection in these non-HS regimes. A fairly smart solution indeed.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.