need help in troubleshooting problem in my layout design with lan8742a

[me.rishabh12]

im facing a very strange issues on my design i soldered LAN8742a Ic which is in QFN24 package which have aground heat pad at the bottom
i soldered the ic using the hot air gun and not the problem is that it work normally for 2 to 3 hour sometime less or more
but after that it simply stop working and when once i reapply heat and solder it start working again but after that there is same issue
i just re soldered ic 3 to 4 times
can anyone explain whats going on my pcb and what is the exact issue

[JS]

Reflowing the IC too many times could be a problem, their are rated for just a few cycles, 1 to 3 usually, IIRC.

Try changing the IC and using a lower temp solder, so you reduce the risk of potential damage to the new IC by overheating it.

JS

[ajb]

How sure are you that the problem is with soldering, and that resoldering the chip is what's 'fixing' it? 

Does leaving the board powered off for a while make it work again?  Does hitting it with some cold spray make it work again? Does lightly heating the board make it fail more rapidly?

Are you absolutely sure that this behavior isn't the result of a software/network issue? 

[me.rishabh12]

no turning off for long not fixing the problem and nor colling either
i need to use hot air gun to reflow solder then it will start working again

[mrpackethead]

Post the schematics and board layouts.  I've used these by the bucketfull without issue, but layout is critical for reliable operation.
When its running what Voltage is at pin 6, ( VDDCR ).. Its very picky about having its gnd plane properly attached.     

[me.rishabh12]

no software is pretty stable i tested the same software on my other design

and both are connected to same network but the other one work fine

[me.rishabh12]

Post the schematics and board layouts.  I've used these by the bucketfull without issue, but layout is critical for reliable operation.
When its running what Voltage is at pin 6, ( VDDCR ).. Its very picky about having its gnd plane properly attached.     

attaching the layout

[mrpackethead]

That pad design under the QFN looks wierd. It shoudl be square.    Check the data sheet for recommended pad shape.    Those vias are large and could be soaking your Solder away.  What type of PCB? ENIG... HASL will never be flat enough to releiably solder QFN on.     Check teh voltage at pin 6, when its running.     and come aback and tell us.

Also, why did you try to path lengh match the TX/RX pairs?  really no need to do that at all, its only 125Mhz..

[me.rishabh12]

That pad design under the QFN looks wierd. It shoudl be square.    Check the data sheet for recommended pad shape.    Those vias are large and could be soaking your Solder away.  What type of PCB? ENIG... HASL will never be flat enough to releiably solder QFN on.     Check teh voltage at pin 6, when its running.     and come aback and tell us.

Also, why did you try to path lengh match the TX/RX pairs?  really no need to do that at all, its only 125Mhz..

the pcb is HASL
and the voltage on pin 6 is 1.3 v approx 
and is it any drawback of length matching the pairs?

[mrpackethead]

At 1.3V you are outside of the operating range of the IC... It should be no more than 1.26V.     A voltage more than this, is a strong indicator that the central pad is not well connected to GND.     Having a look at your PCB design , it appears that the footprint for the QFN is not suitable.  Also when using QFN parts HASL finished PCBS are really problematic, as the HASL is not flat, and you can have all sorts of issues.  ENIG is a much more suitable choice.

Likely you are going to have to look at how your part is soldered.

[bson]

Also, why did you try to path lengh match the TX/RX pairs?  really no need to do that at all, its only 125Mhz..

Agreed, but no good reason not to if there's plenty of space - it does improve common-mode rejection.

And the 100M frequency is only 31.25MHz since it uses MLT-3...

[bson]

How is this chip different from the LAN8740A?  Anything other than WoL support?

I do note you don't seem to have any management interrupt support, but I guess the PHY can be polled for things like link status or speed changes.

Have you made sure it doesn't overheat?  PHYs can be pretty power hungry...

Also, QFN and HASL doesn't go well together in my experience... you should at least clean excess tinning off the pads before reflowing.  It could be a combination of the chip "floating" in solder along with thermal stress that causes cracks in solder joints.  Maybe also the PCB flexing from thermal stress.  I'd make sure it gets a bit of a top/bottom zone to dissipate heat, then pay the extra for ENIG...

Oh, and if you have access to a scope with an active FET probe, check the xtal to make sure the XO is running and the levels look reasonable.  Just a quick and easy check, just to make sure the problem is with the IC and not the crystal or load caps.

Speaking of load caps... I hope they're C0G with a reasonable TC?

[Siwastaja]

and is it any drawback of length matching the pairs?

Yes, the drawback you get is you need to read whining on the EEVblog forum even when your problem is totally unrelated. TLMJW's, or Trace Length Matching Justice Warriors, looking to comment on "unnecessary" trace length matching, is the new trend now after the DGND/AGND dead horse has become too stinky.

I understand the irony in length-matching a 10kHz I2C or a blinking LED IOs, and I tend to have a laugh as well. But really, this is a fairly high-speed impedance controlled design which, while working just fine without it, is not too far from the design territory where the length matching starts to really matter.

But TLMJW's are pissed because now every beginner can use the simple length-matching tool in their PCB packages so it's not their high-speed black art anymore.

[mrpackethead]

I had this exact same issue, when i was developing my first prototypes and did a hand assembly.   When VCCDR is over 1.26V the PHY becomes unreliable.   and its related to how well that central pad is soldered down.

[Apollyon25_]

Another gotcha with these Microchip LAN parts can be the high crystal drive level from the chip, and the crystal's rating. Found this one the hard way.
Agree however that the thermal pad looks insufficient.

 

[ToddP]

What problem did you have with the crystal? I have a 499 Ohm on one leg because the crystal couldn't handle the power. But, we also have oodles of 250MHz radiated noise causing us to fail EMC testing...wondering why the thing is lighting up like a Christmas tree in the test chamber.