soldering crystal oscillators

yeah, and I`m guessing it shouldn`t be used in Medical equipment either, section 2.VII  "A latch-up phenomenon and a spurious phenomenon might be caused when an excessive outpatient noise is impressed to the power supply and the input terminal..."

Actually, I have found that it's rather unusual nowadays for Japanese documentation to be that poorly translated.  It used to be quite common.  My experience with crystals is that some cannot take the heat of IR reflow soldering, so you have to hand-solder them later in the manufacturing process.  This could very well be the case with RasPi, and people have reported manufacturing defects such as bent through-hole pins.  I know I'll be checking my RasPi carefully before plugging it in.

Here's what Citizen's web site has to say:

When heat is applied across the crystal (such as reflow oven for example) is very small cage, and because the injection method of soldering, it may significantly degrade performance, avoid using Please.

 

http://cfm.citizen.co.jp/english/product/cvo_manual.html

see paragraph 1.II.3.

I don't know if RasPi's cage is very small or solder is being injected.

John,

  You're probably right about the crystals being hand-soldered.

We've seen what appears to be evidence of hand-soldering on

the back side of the board:

http://www.raspberrypi.org/phpBB3/viewtopic.php?f=6&t=6124

and there isn't much else on the back side that looks like a

candidate for hand-soldering.

I thought things like bent pins and stray solder blobs were supposed

to be detected by visual inspection, or automated optical inspection,

at the factory.

I'm saddened that Pete seems to have changed his mind about talking tech with us on Element 14.

The conversation about polyfuses ended up being totally one-sided with a grand total of zero contributed to our understanding by the Foundation, when it would have been far better as an open Foundation-community dialogue about Pi power design.  These crystals are another area of the Pi where this applies.

Wish they'd talk openly.  Secrecy has no place in a project with such worthwhile educational goals.

Morgaine.

coder27 wrote:

 

John,

  You're probably right about the crystals being hand-soldered.

We've seen what appears to be evidence of hand-soldering on

the back side of the board:

http://www.raspberrypi.org/phpBB3/viewtopic.php?f=6&t=6124

and there isn't much else on the back side that looks like a

candidate for hand-soldering.

 

I thought things like bent pins and stray solder blobs were supposed

to be detected by visual inspection, or automated optical inspection,

at the factory.

Through-hole parts are often hand-soldered as well.  It can be hard to wave solder SMT components.  I'm told you have to be careful how you orient resistors and capacitors or you get "tombstoning".

http://www.npiengineer.com/tombstoning.html

although I`m not quite familiar with ethernet, I have done a little work with USB, and that likes divisons or multiples of 12Mhz, that are taken care of on chip by either a divider or PLL.

25MHz seems a strange number to start with (as opposed to a 24MHz for instance), could that be causing the problems?

or can someone tell me how the 12MHz is derived for the USB please.

Thanks

YT2095 wrote:

 

although I`m not quite familiar with ethernet, I have done a little work with USB, and that likes divisons or multiples of 12Mhz, that are taken care of on chip by either a divider or PLL.

25MHz seems a strange number to start with (as opposed to a 24MHz for instance), could that be causing the problems?

or can someone tell me how the 12MHz is derived for the USB please.

25 MHz is what the LAN9512 data sheet specifies for the XI/XO pins.  They probably quadruple this internally to 100 MHz for 100 Mb/s Ethernet.  For USB 2.0 you need 480 Mb/s so they probably need 480 MHz internally.  They probably have a 480 MHz PLL oscillator divided by 96 to get a 5 MHz feedback signal.  This is locked to the 25 MHz crystal reference divided by 5.

Apparently the timebase for 100 Mbps ethernet is actually 125 MHz.

With 100BASE-TX hardware, the raw bits (4 bits wide clocked at 25 MHz at the MII) go through 4B5B binary encoding to generate a series of 0 and 1 symbols clocked at 125 MHz symbol rate. 

-from http://en.wikipedia.org/wiki/Fast_Ethernet

See also: http://www.designspark.com/content/raspberry-pi-waveforms

Thanks, although I am somewhat confused about your assertion of USB2.0 needing 480 MHz to work?

I use the PIC 18F4550 with full USB 2.0 capabilities, and that requires 12MHz to run, made by a 20 MHZ crystal, divided by 5, and then taken up to 48MHz for the CPU clock and 12MHz for the USB.