Clock vs. Crystal Oscillator

[kodabear]

I am designing a low-power embedded system using a 186. I and trying to decide on a clock vs. a crystal for the processor and an external DUART I am using. Both can use a crystal or external clock. I have used both before and know that generally a clock is easier, you don't have to mess with balancing capacitances. I would be using a standard fundamental mode crystal circuit if I went that route. The processor needs 32MHz and the UART needs 3.6864MHz. Clocks I have found run at around 15mA at 3.3V. That is a third of the processors total current draw. Would a crystal draw less current? I am only making one of these so cost is not a constraint. What other things should I consider? Space is an issue but I have found clocks that are the same size as crystals. Is one more stable than the other, the board will be subjected to significant shock and vibration. Does anyone have any experience or knowledge on this subject?

 

[melone]

How do you plan on generating your clock? How accurate does your clock have to be?

 

[kodabear]

Sorry, I should have been more clear. I mean a "canned" clock, like a "clock in the box". They are usually sold by the same companies that sell crystals. They typically have PWR, GND, Enable/Disable, and an Output. This is what I was saying requires ~15mA. They sell these SMT and thru-hole, ceramic or metal cased, for quieter EMI.

 

[steveowens]

XTAL will be more accurate and stable over temperatue. Resonator will be cheaper. I think that they will both draw about the same amount of current, although, it is likely that if you built your own it would be less than most off-the-self osc. Statek used to sell some real low power osc, not sure if they still do. Most downhole (read high shock and vibration) use Bliley xtals. Remember, power is a function of speed. Want less current, then slow it down.

http://www.bliley.com/

Steve Owens
Finish Line Product Development Services
Steve.Owens@FinishLinePDS.com

http://www.FinishLinePDS.com

 

[automatic2]

Check your processor and uart specs for ratings for crystal against clock. You'll gain on space using sm crystal/caps.

 

[nbucska]

<

http://www.linear.com>

makes a cheap 3mm SMD oscillator programmable with one R up to 20MHz. <nbucska@pcperipherals.com>

 

[Lewish]

My argument for a &quot;canned clock&quot; is you don't have to mess around with tweaking the capacitors to make sure the oscillator always starts when using a crystal. This was a real annoyance with the older 186 chips. The little SMT clocks are quite cheap.

 

[nbucska]

-- cont --
LINEAR'r LTC6900 gives better than 2% accuracy with 1% R. <nbucska@pcperipherals.com>

 

[melone]

As long as he is not trying to perform any communication (for example CAN protocol), then a resonator could work. KODABEAR, please define exactly how much tolerance and jitter your application can handle.

Also, see if you can PLL up the 3.6864MHz (crystal or resonator) to the 32MHz. You don't have to place another device.

 

[nbucska]

The PLL would require extra/more expensive hardware.
<nbucska@pcperipherals.com>