According to the LAN9512 data sheet, the ground pad is the only ground connection, so if it's not connected I can't see how the LAN9512 would work at all. OTOH, there might be enough solder to make decent electrical contact but not enough for good thermal contact. According to the data sheet the LAN9512 dissipates as much as 763 mW which is a lot for such a tiny package. I tried to find the thermal resistance of the QFN package at smsc.com but couldn't.mahjongg wrote:Well, these kind of chips do have a metal pad on their underside that should be soldered to the PCB, if for some reason that isn't done the chip cannot dissipate its heat away effectively to the PCB's ground copper plane, it might explain why its getting hot.
Ah yes indeed, good catch I had not realised that, but you are absolutely right in your assessment and conclusions.johnbeetem wrote: According to the LAN9512 data sheet, the ground pad is the only ground connection, so if it's not connected I can't see how the LAN9512 would work at all.
Re: Temp Sensitive Boot Problem
Postby Jongoleur » 16 May 2012 15:09
In the interest of getting some basic temperature figures from a Pi that appears to be working in a sensible manner, I've waved an (admittedly fairly cheap and inaccurate) infra-red thermometer over the various components of a Pi that has been running for several hours. It has been rebooted several times during this period and restarted flawlessly.
Big voltage reg by large capacitor 38.3C/100.9F
SDHC card 40.8C/105.3F at connector end, 25.6C/77.3F at other end
I measure chip surface temperature in seconds -- specifically, how many seconds I can leave my finger tip on a component before I need to pull it away I need to calibrate this some day.jbeale wrote:In case this is useful, from another thread. Personally, I'd say a 40 C plastic surface is very warm, but not quite too hot to touch. BTW, IR thermometers are as cheap as $25 now.
Can you try running a bunch of Ethernet traffic and see what happens to the USB/Ethernet chip? It may use more current when there's actual traffic, depending on software's use of shutdown modes.spurious wrote:I've just measured my R-Pi temp, which has been on for 5 days
- CPU = 36.5C
Eth Chip = 31.5C
HDMI Port = 26.2C
Eth Port = 23.4C
USB Port = 25.6C
SD Card = 34.3C
PSU Port = 26.9C
Ambient = 21.4C
Just ran a load test for 5 minutes on nginx / MySQL / PHP, from My PCjohnbeetem wrote:Can you try running a bunch of Ethernet traffic and see what happens to the USB/Ethernet chip? It may use more current when there's actual traffic, depending on software's use of shutdown modes.
Something seems very odd here. If that's the normal temperature of the chips when they are working hard, there is no way the Eth chip should be "burning hot" unless the input voltage is way out of spec, or there is an internal or external short-circuit somewhere. You aren't connecting to a PoE port by any chance (power over ethernet)? -and even that, in theory wouldn't be a problem unless the "missing magnetics ethernet jack" problem has returned.CPU = 32.3C
Eth Chip = 28.1C
Ambient = 21.0C
I found a datasheet of another component with the QFN64 package that says 20 degC/W typical without airflow. This assumes the use of 49 (7*7) thermal vias of 0.3mm diameter. At 3/4W the die temperature would rise 15 degC above ambient temperature.johnbeetem wrote: I tried to find the thermal resistance of the QFN package at smsc.com but couldn't.