Hi,
I'm currently working on a PCB design with an Atmega328 / Arduino on it (almost finished).
You often hear that the crystal should be as close as possible to the µC pins, but since PCB space really is an issue here, I moved the crystal to the side (see pic). Is this "too far away", or will I get away with it?
Thanks for your opinions, Phil
Looks OK, however can you not move the RX and +Vcc lines elsewhere?
What about Reset on pin 1? You may be able to squeeze it between the RX and TX lines.
Something like the attached pic.
(sorry done very roughly just to illustrate).
That placement is not in line with Atmel's recommendations in
Atmel AVR042: AVR Hardware Design Considerations. Pg 15
Can you put the crystal under the chip? Or go surface mount?
Well I guess there is your answer PhilPerkins.
I've have made pcb tracks longer then that in the past with no ill effects (@20Mhz), but since it's not compliant with the Atmel recommendations, I would follow Cross Roads recommendation, especially if you are getting the pcb/s made by a company and any required changes in the artwork would cost you money.
I'm lucky enough to be able to make my own pcbs so tend to be a bit more lax when it comes to trying to "push my luck" even if it does not work out, it only costs me a bit of time and some pcb substrate.
Thank you all for your input, I'll re-consider a couple of things ;-).
Without looking at the guidelines, I will note that it is generally considered good practice to have a ground ring enclosing the crystal connections.
There are two considerations; radiation of interference from the oscillator circuit, and if it is of importance, interference with the accuracy of the crystal reference frequency. For both of these reasons as well as mechanical stability, it is also common to solder the crystal cans of larger crystal types to a ground pad. (Some crystals and many resonators provide a ground wire.)
Completely agree, hence my recommendation to move the serial line further away.
25 16 MHz thru-hole mount crystals - not a ground wire in the bunch.
Nor any recommendation from the 5 manufacturers in their datasheets to add one.,
More than likely, the time & temperature needed to heat the steel case up for soldering would damage something too.
(This site's datasheet would not open for me - Citizen Finetech Miyota)
When one uses digital circuits near RF circuits, there is always a chance of the xtal radiating it's fundamental frequency (or one of it's harmonics) to near by RF stages so it's always a good idea to ground the xtals case.
Even though the radiation level maybe very low, depending on the sensitivity and selectivity of the RF stage/s, it may be enough to create problems.
Even though most crystals do not have a ground lead, it's common practice to provide a ground plane under the xtal onto which the xtals case is soldered at one or two points.
This not only minimizes case radiation, but also provides mechanical stability.
Obviously one has to be careful as not to overheat when soldering the case to ground.
How does one do that as in practice with SMD components? Would seem rather awkward?
It is rather akward but do-able and of course if one is using a xtal near RF circuitry, proper selection would be required.
Not all xtals would be useable in that application.
Some smd xtals with the 4 contacts normally have one of the contacts connected internally to ground, example here: http://www.abracon.com/Resonators/abm8.pdf likewise with 4 pin xtal oscillators which give out TTL/CMOS levels.
Crystals should not be exposed to temperatures above their maximum ratings. Exposure to excessive temperatures may damage the crystal, and usually increase the ESR. Crystal "cans" should not be soldered to a PCB. This is sometimes done to ground the case of the crystal. Soldering directly to the case of the crystal usually subjects the unit to excessive temperatures.
--Maxim App. Note 58
The Times They Are A-Changin
-Dylan
Tell that to the thousands of RF products designers and manufactures.
Of course overheating will cause damage - use a proper soldering profile.
Every company that I have worked for solder all xtals to ground.
Between them there must be millions of different products out there and have never seen a huge increase in faulty units due to xtal damage or increased ESR.
Every company that I have worked for solder all xtals to ground.
Yes. But we are talking about amatuer designed hardware and it is my opinion that unless these individuals are proficient electronic technicians, they should avoid soldering crystal cans to the PC board ground plane.
For 4 lead SMD crystals, grounding is not necessary and can adversely affect the load capacitance.
Grounding of the can comes from the early RF days when crystals were prone to multimode oscillations but modern crystal designs are geometrically cut to avoid minimize such problems. RF rules of thumb and ARRL handbooks showed crystals designs and projects with their cans grounded. So, it is entirely correct for these designs to have grounded xtal cans.
Before writing this response, I checked best practices from a number of manufacturers: TI, Tellurian, Maxim and there is no consensus on grounding crystals in digital projects. So, I would say that if the crystal manufacturer states to avoid it, then avoid it. If however the crystal manufacturer , vendor, or engineer (author) specifies grounding, then ground that can but exercise best practices in performing the soldering.
For those fumble-handed among us, you may wish to consider a conductive adhesive.
Ray
Actually, one (not uncommon) approach is to use a wire straddling the can into two through-holes. One can trust the mechanical contact, or if it is necessary to solder the wire to the can in addition, this can be done using a soldering iron with much less heat/ time.
This maybe of interest:
Paul__B:
Actually, one (not uncommon) approach is to use a wire straddling the can into two through-holes. One can trust the mechanical contact, or if it is necessary to solder the wire to the can in addition, this can be done using a soldering iron with much less heat/ time.
That's common in HAM radio equipment, however seen 4-8 crystals mounted vertically as in crystal filters used to set the IF bandwidth in high performance receivers.
http://ecee.colorado.edu/~mathys/ecen2420/psets/ps07/CrystalFilterP1.jpg
Lefty
In a ""Staggered" array of crystals like the one in the picture the 'cans' are soldered together to prevent radiated signals from other places including adjacent crystals from affecting the signals being 'filtered'...
Ideally each crystal would be in it's own shielded enclosure, isolated from each other on both sides of the board. There is a lot of Military equipment that is designed that way.
In practice the cases are soldered together at an "Equidistant point" (for symmetry) and that point is grounded. You might have noticed the ground from the "Center" of the "X" shaped wires straight down to the PCB. this is the equal point and any RF developed across the inductance of the ground wire affects all equally. Remember that these crystals are selected for their resonant frequencies so each will have a slightly different 'sensitivity' to radiation from external sources and the purpose of the ground is to make all equal in RF potential.
Finally the connections on the bottom of the PCB usually are shielded in separate compartments for the receiver filter but may well be unshielded in a transmit filter and just laid down close to the PCB copper.
Edit: I personally have soldered several thousand transceiver crystals to ground without a failure... from 6 to 50 MHz fundamental frequencies. Occasionally one will fail but it is my studied opinion that the failures were defective parts before being soldered to the case. This "Studied" opinion comes from the failure rates which according to some old notes I have dating to that period in my life that show about a 1% failure rate. Digital clock crystals are rarely used in RF applications (Poor relative stability) but when they are used for transceivers they should be grounded, For the reasons I stated and the reasons stated by others in this thread. I hadn't read all the thread before I added my $0.02's worth so a part of my addition is repetition of facts already stated.
Doc
Docedison:
the 'cans' are soldered together to prevent radiated signals from other places including adjacent crystals from affecting the signals being 'filtered'...
Beat me to the punch!
Docedison:
I hadn't read all the thread before I added my $0.02's worth
Hardly surprising, there's quite an effort in keeping up with these discussions. 
Good one, this...thread as it covers a lot of junk... the only thing not covered was the soldering of 32768 Hz crystals... Sometimes it works and sometimes the "Tuning Fork" is in contact with the case and the crystal should be replaced... It's commonly done to keep the crystal isothermal to the chip oscillator mostly... Although occasionally the oscillator has a bit too much gain and then the crystal either won't work or work right as the oscillator is subject to "Jumping Modes" internally.
Crystals are cheap, engineering time isn't.
Doc