Replacing crystal oscillators with a global clock

[smashedProton]

Hello all,

I'm designing a 16 channel usb to UART board using 4 FT4232 usb to 4xuart chips and a TUSB2046B usb hub chip.  The ftdi's want a 12mhz clock and the usb hub chip wants a 48mhz clock.  I'm wondering if I can clock everything using one 12mhz crystal and an IDT ICS512.  (datasheet: http://www.idt.com/document/512-datasheet)  Would I be able to do this?  What should I take into account when designing the circuit?  Right now, I have the pll set to 4x multiplication, the reference clock is connected directly into the crystal input of the ftdi chips, and the pll output clock is going into the 48mhz imput clock of the usb hub chip.  Is my specific circuit correct, or should I add more circuitry to condition the signal to work with the clock input of the ftdi chips?

Thanks!

[T3sl4co1l]

Is that really cheaper than using two separate crystals?

I've used that TUSB thing before; it bothers me that they don't document the "crystal" pins at all (same goes for basically any device, including that PLL).  In my case, it was using two TUSBs on one board, where it would be nice to save a crystal by running the other chip from the first one's clock output.  They don't say that it's a clock output, but as it turns out, the signal level on the CLKOUT pin seems CMOS-ey, so it may very well be practical to do that.

Point being, maybe you'd still need two crystals, one for the FTDI and one for the TUSBs, but only one for them, not one each.

Of course, it would be more applicable if what you were doing involved multiple hubs, rather than multiple UARTs... but maybe the same scheme can be applied to the FTDI chips.  Dunno.  You can try asking them; maybe you'll actually get a useful reply (unsurprisingly, the reply I got from TI's 'help' forum was terse and useless).

Might also look into a hub that uses 12MHz rather than 48.  As I recall, the TUSB has a 6MHz mode, but, that doesn't help very much.  (On the upside: if you have a clean 12MHz, you can simply use a 74HC1G74 to divide it to 6!)

Tim

[smashedProton]

Is that really cheaper than using two separate crystals?

I've used that TUSB thing before; it bothers me that they don't document the "crystal" pins at all (same goes for basically any device, including that PLL).  In my case, it was using two TUSBs on one board, where it would be nice to save a crystal by running the other chip from the first one's clock output.  They don't say that it's a clock output, but as it turns out, the signal level on the CLKOUT pin seems CMOS-ey, so it may very well be practical to do that.

Point being, maybe you'd still need two crystals, one for the FTDI and one for the TUSBs, but only one for them, not one each.

Of course, it would be more applicable if what you were doing involved multiple hubs, rather than multiple UARTs... but maybe the same scheme can be applied to the FTDI chips.  Dunno.  You can try asking them; maybe you'll actually get a useful reply (unsurprisingly, the reply I got from TI's 'help' forum was terse and useless).

Might also look into a hub that uses 12MHz rather than 48.  As I recall, the TUSB has a 6MHz mode, but, that doesn't help very much.  (On the upside: if you have a clean 12MHz, you can simply use a 74HC1G74 to divide it to 6!)

Tim

I'm using 4 FT4232 chips so the crystal + pll  solution would replace 5 crystals.  It wouldn't save a whole lot of space, but I'm trying to cram my circuit into a small space so wouldn't want 5 crystals if I don't need to have them.