Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

The least crystal freq the better?

Status
Not open for further replies.

bremenpl

Member level 3
Member level 3
Joined
Jan 5, 2013
Messages
63
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,286
Activity points
1,827
Hello there,
I have a question regarding crystal oscillators used to clock various MCUs. In general, the higher the signal frequency is, the better PCB enviroment one has to provide for proper operation. Following this, should one always choose the slowest possible crystal to achieve desired MCU clock frequency by using internal PPL's? For example defaultly most of the circuits I see use 8 MHz crystal and then internal PPL is used to get ie. 72 MHz. But wouldnt it be better to use 4 MHz crystal if the MCU still supports it? Wouldnt that mean one could make longer crystal traces on the PCB etc? Are there any cons of this concept?
I would appreciate all help.
 

imo you are kind of right there but the advantages you get from doing this are not significant, reducing frequency to the half does not necessarily provide 2x longer crystal traces or even anything close to that specially at the range of less than 20MHz crystals.

MCUs usually dont require high frequencies, and even if it supports 4MHz you might face slight reductions in performance so you should read the MCU datasheet carefully before committing this.
 
Power consumption also decreases with frequency. As a matter of fact, it is a common strategy employed on devices that must run from batteries for long periods of time.
EMI is also lower.

So, if your device supports it and your application allows it, that is a good strategy.

What I like to do on microcontrollers on which you can switch prescalers or PLLs on the fly, is to use maximum frequency only for the tasks that demand it and lower the frequency afterwards.
Again, this is very application dependent.
 

What I was wondering more about is something like this: Lets say I need 16 Mhz clock. Whats more "stable"? 8 MHz crystal + PLL x2, or 4 Mhz crystal + PLL x4?
 

Hi,

the higher the signal frequency is, the better PCB enviroment one has to provide for proper operation.
I´d say it is somehow different.

For proper operation you always need a good layout. Because EMC doesn´t care about the frequency you use on your PCB.
The external signal may influence your operation even if you don´t have any clock on your PCB. Long messy wiring an a non solid GND plane will always be problematic.
That´s why a circuit maybe runs for a long time witout problems, then you neighbor buys a new handheld telephone and your circuit fails. ---EMC

But EMI, which means: "your PCB is acting like an RF transmitter" depend on clock frequency. But clock frequency is only one item of many.
Trace length, voltage levels, switching speed (rise time, fall time), how many lines are swtching at the same time, coupling to other wires, copper pour and, and, and... will have a lot of influence , too.

I often use a 48MHz Xtal with a short wire to an CPLD or FPGA then I use divided clocks (6MHz, 8MHz, 12MHz...) for my microcontrollers or periferals. I am a friend of only one XTAL clock and all others are generated from this clock. Especially when analog signals, ADC or DACs are on the PCB, this makes noise handling more easy.


Klaus

- - - Updated - - -

added:
PLL and stablity:

What do you mean with "more stable":
* low jitter: this sometimes is not wanted. This causes sharp, high peaks at EMI measurement. Some add jitter to pass the EMI test. It is called "spread spectrum".
* more reliable: I don´t think that one solution is more reliable that the other (of your given two)

Klaus
 

Thank you for answer. I was thinking more of a following situation. Sometimes (nearly always from my experience) ICs have the most passive components required on the same side of the chip as the osc pins are. In that case I always struggle between smaller passives, passives on the other side of pcb and osc crystal traces length. I was wondering either I could add some extra space between the crystal and mcu if i choose a slower frequency crystal.
 

Hi,

8MHz: lambda/4 = about 4.6m
I don't think your xtal wiring is that long.

You are far away from a good antenna. With decreasing wire length your "antenna" becomes worse, and that's good in case of EMI.

For sure you will have some overtones, but with a passive xtal the oscillation shape is more sine than square.

So it doesn't matter if your wiring is 3mm or 12mm...

Klaus
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top