Hi all,

Looking to posts on this forum and others, most questions are about calculating the caps needed for a given crystal. However, I'm trying to figure out which values mentioned in a chips datasheet are the most important ones when choosing the crystal.

Say I'm creating a circuit for the LAN9512 chip. The datasheet (http://ww1.microchip.com/downloads/en/DeviceDoc/00002304A.pdf) does a crystal specs recommendation on page 43:
25Mhz, CO|Shunt @ Nom 7pF, CI @ Nom 20pF, Drive @ min 300mW and ESR @ max 50Ohm

I'd like to know which values given by the chip manufacturer are the most important ones which should match. Lets take this crystal: MCRS025000F183000RR (datasheet: http://datasheet.octopart.com/MCRS025000F183000RR-Multicomp-datasheet-8454482.pdf)

Could I use this crystal for this chip? Or should I match the CI .. or CO|Shunt, drive, ESR or should even every value match?

I hope I created this question with enough (useful) info. And hopefully I didn't overlooked an existing post. If so, please let me know.

Regards,
Onno.

 

The LAN9512 will drive its crystal at 300 uW, but the crystal to which you l inked is intended to be driven at 50 uW. I do not see the datasheet mention and option t reduce the drive level. For a possible solution check out page 4 of this application note:
www.nxp.com/docs/en/application-note/AN3208.pdf

Plan to use 18 pf as the load capacitance since this is what the crystal is intended to work with. Subtract the 3 pf parasitic capacitance on the oscillator pins on the LAN9512, so if you use very short traces to the crystal and its load capacitance you can us 15 pf load capacitors.

The LAN9512 datasheet recommends designing for ±45 ppm for the lifetime of the equipment. .

The initial accuracy of the crystal with 50 uW of drive is ±30 ppm. Aging rate of the crystal with 50 uW of drive is ±3 ppm/year. That looks like about 5 years before there is a good chance of having errors with ethernet. You might want to use a higher precision crystal if 5 years sounds like too short of a time. While on the subject, use NPO or COG load capacitors so you won't have to worry about them aging.

 

Hi Dick,

Thank you for your extensive answer. According your reply the most important value from the chip datasheet is the drive level, right?
So when that taken in account, one of the following would be OK (random order)?
FOXSLF/250F-20 - http://datasheet.octopart.com/FOXSLF-250F-20-Fox-Electronics-datasheet-87579709.pdf
FOXSDLF/250F-20 (PartID: FC4SDCBMF25.0) - http://datasheet.octopart.com/FOXSDLF-250F-20-Fox-Electronics-datasheet-81464241.pdf
ABLS3-25.000MHZ-D4YF-T - http://datasheet.octopart.com/ABLS3-25.000MHZ-D4YF-T-Abracon-datasheet-10283362.pdf

Than based on the load capacitance from the datasheet of the chosen crystal I calculate the capacitors.
And based on the age rate, the Abracon would be 'worst' because that one will age to 45ppm in about 3 years, since it has a rate of +-5ppm. Correct?

 

Both Fox crystals: Yields may be low because the drive level from the LAN9512 will be 300 uW vs 500 uW maximum required for the Fox crystal, though the Fox crystals typically only need 100 uW to oscillate. Since the drive level will be low, that will probably throw the frequency off by a few parts per million. My opinion is that if you are going to build one or a few then this is probably not going to be an issue. If you are going to be building production lots, it can be a real headache.

The Abricon parts are not good candidates for many reasons.

Incidentally, you should probably take the operating temperature range and its effect on frequency into account. It i nice that the Fox crystals can be had with very tight initial tolerances.

 

"25MHz" is a lot, normally the values of capacitors are about "22pF", I dont know how to calculate them either, I just take the standard value.

Some MCUs support an "RC" oscilator if the accuracy is enough for you.

 

Hi,

Again thanks. So it's better to try to match the Drive level as much as possible!? So with that given: the following crystals will do the job?
Diodes Inc. / Pericom
FL2500213Z - http://www.mouser.com/ds/2/115/FL2500213Z-1129545.pdf
FY2500117 - http://www.mouser.com/ds/2/115/FY2500117-1129613.pdf
FL2500316Z - http://www.mouser.com/ds/2/115/FL2500316Z-1129531.pdf

Or maybe could give an example of a crystal you would use / could be used?

pfiew, these little things can really be hard/take a lot of time in the beginning. But won't give up ;o)

 

I think you understand the important factors.

Which of those looks like the best candidate to you?

 

Hi Dick,

Based on the factors, and the last three mentioned, I would go with the following order:
FL2500213Z : Thicks all boxes from LAN9512 datasheet. 20ppm over 10years. so +/- 2ppm/yr. ESR 30Ohm < 50Ohm
FL2500316Z : same as the first one, accepts only higher temp.
FY2500117 : Thicks almost all boxes. max drive level 300 uW. 5ppm/yr against stability of 20ppm > to reach +/-45ppm = 5yrs

However I'm still struggling with the drive level. :s When I lookup the explanation of drive level, I find:
"Drive level indicates power consumption by crystal unit while oscillation circuit works.
It is important for crystal to keep within drive level specification.
Excessive drive level might cause unexpected change of frequency and equivalent series resistance."

So when looking to the Fox crystals, they state: Drive Level : 0.5 mW (0.1 mW Typical)
As 0.1mW is typical and 0.5mW is max, I understand that they are pushed near the max with 0.3mW. But I noticed that almost all crystals have a drive level of 0.1 mW (typ) but have a max of 0.5 mW. Like one other I found has:
Load Capacitance (CL) 18.00pF
Shunt Capacitance (C0) 7pF max
Drive Level 500μW max
ESR 40Ω max

Since 300uW is below the 500, I would say that one could also be used.

I also noticed that most suppliers don't have a filter option for drive level. Which in my opinion would be strange when it's a that important value.
Eg. when I search a given supplier for crystals with a drive level of 300uW, I only get three results..

 

Very good. It looks like you found a very good candidate in FL2500213Z. The excessive drive will probably not be a major problem, though you can cut down the drive using that series resistor shown in that NXP application note if you feel like going to the extra work.

 

Hi Dick,

Thanks! Good to know I'm finally a bit on the right track. So when a crystal datasheet states typical 0.1mW with max 0.5mW and the chip states 0.3mW, it's still best to cut down the drive level from the chip to 0.1mW...

At least we now know that the drive level is most leading in choosing the right crystal. To bad my program library doesn't have a footprint for this one .. guess I need to make my own.

Thanks for your time! Really appreciated.

 

If you are in production, you should supply the max. If only making a few, go with what is convenient.

 

That clear. Higher level than typical could be used, but should always be less than the crystal max drive level.

For the complementation of this thread, I found two application notes who also shed some more light on the drive level and ways to reduce them (as in: Adding just a series resistor).
https://www.idt.com/document/apn/801-crystal-high-drive-level
https://www.idt.com/document/apn/830-quartz-crystal-drive-level