From: briand on
On Tue, 2 Feb 2010 10:10:33 +0000 (UTC)
debian-user-digest-request(a)lists.debian.org wrote:

> You're barking up the wrong tree Brian. That crystal isn't the
> source of your problem. It's well within spec. I'm sure the system
> design tolerance is much greater than 0.0006%.

The crystal was STAMPED 25.0006, so it's actually 600Hz off (24ppm), not
0.0006% off.

> PHYs use PLL circuits
> to clean up the clock anyway, so even if that clock crystal chip was
> running at 25.05 MHz it wouldn't cause problems. BTW, what "lab"
> equipment did you use to "test" this crystal and verify its
> frequency? /laughs

Raises your /laugh with a /chuckle

Know what a spectrum analyzer is ? It's tolerance is about 10ppb.
That's 0.00000001 %, and I know how to use one :-)

The oscillator was running at 25.0006MHz, just like the
crystal said. Well it was actually off a few ppm, but close enough.
Most crystals for mass market stuff are about +/-50ppm, but that's over
temperature. At 25C there going to be pretty much dead-on.

However someone else posted that the spec is 50ppm, which would be
25.00125, so that means that 25.0006 should work, in theory...

It's possible that they did it on purpose _hoping_ that over
temperature it would drift towards 25.0000.

Now that I think about it, when it was closed up in the case and
getting nice and toasty it could have drifted 50ppm high which _would_
put it out of spec, because then it would be 24 + 50 = 74 ppm off.

It's still weird and definitely related to cost reduction in some way.

>
> I'd say you probably just got a defective switch. It happens. And
> probably more frequently with the cheap stuff than the more expensive
> stuff. No surprise there.
>

That's for sure.

> What's interesting in your case is that you're dealing with
> 1000BaseTX. IEEE 802.3ab *requires* auto negotiation for all copper
> gigabit ethernet; manual settings aren't allowed, period.

Well, I am _positive_ that at one point I was able to turn off
auto-negotiation, because that's how I got one of my connections to be
reliable.


Brian


--
To UNSUBSCRIBE, email to debian-user-REQUEST(a)lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster(a)lists.debian.org