0
\$\begingroup\$

I am designing a PCB with 10 individual USB hub ICs onboard. They all require a clock input and I am wondering if it is possible to drive all the clocks from a single crystal oscillator?

The hub IC is http://www.ti.com/lit/ds/symlink/tusb4020bi.pdf

\$\endgroup\$
4
\$\begingroup\$

Yes, you can -- take a 24MHz oscillator and feed it into a clock buffer IC such as the CDCLVC1310 (http://www.ti.com/product/cdclvc1310). The buffer IC will allow you to do easy point to point links for each hub's clock. Please read the datasheet for that part in detail -- it has many clock input options as it supports single-ended, LVDS, SSTL, etc. Additionally, the datasheet for your hub says it wants a 1.8V input clock (Section 9.1.2) -- you will need a 1.8V supply (a LDO should suffice) to power the clock buffer's I/O rail (VDDO). You could even choose a single 24MHz crystal and provide it to that part to generate the ten copies of your clock.

A single oscillator has a maximum load on its output that will likely be exceeded by 10 USB Hubs. On top of that, you'd have to route the clock in a fly-by fashion with likely AC termination (since it is a clock this is easy) or Thevenin/parallel termination at the end to prevent reflections. If you imagine your single clock-source having to fan-out (which you identified in the other answer correctly), you can see why drive requirements + branching the signal can be troublesome.

Finally, my gut suggests that running 10 hubs in phase from a signal clock course may contribute to EMC issues and slightly stronger emissions, but I cannot quantify it. It would be 24MHz regardless, but this solution puts every hub in phase with the other. A solution with 10 discrete xtals would be at the same frequency, but likely out of phase with each other.

\$\endgroup\$
  • \$\begingroup\$ Actually, I re-read your USB hub datasheet in detail. It needs a 1.8V clock input (I was suspicious of it taking in a 3.3V clock). I will update my answer with a more suitable part. EDIT: The CDCLVC1310. \$\endgroup\$ – Krunal Desai Dec 28 '15 at 7:25
0
\$\begingroup\$

From the datasheet (emphasis mine):

XI | 38 | I | Crystal input. This terminal is the crystal input for the internal oscillator. The input may alternately be driven by the output of an external oscillator.

You should consider using a clock driver IC though, so that you don't violate the maximum capacitive load for the oscillator in question.

\$\endgroup\$
  • \$\begingroup\$ Ahhh this makes sense. I'm just reading now, what I am referring to is what is meant by the 'Fan-Out', am I right? (en.wikipedia.org/wiki/Fan-out) \$\endgroup\$ – rupertbg Dec 28 '15 at 7:03
  • \$\begingroup\$ @rupertbg: Correct. \$\endgroup\$ – Ignacio Vazquez-Abrams Dec 28 '15 at 7:35
  • \$\begingroup\$ it's possilbe that the crystal output can be wired to the crystal inout on the next few chips, reducing the fan-out frquired from the single oscillator, \$\endgroup\$ – Jasen Dec 28 '15 at 8:44

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.