Interesting. My RG2 output has been rock steady on 3.322-3.323V since the beginning, and the board is permanently playing music from an NFS server while I'm awake. TP1 to ground doesn't vary either, totally pegged at 4.89V. Temperature-wise, RG1 is at "1 second finger-test" temperature, so not blisteringly hot like the LAN9512.
Looking at a data sheet the top end spec for a LDO1117 type device should be 3.333V and reduces with increased temperature. Heatsinking isn't going to hurt though. I take it that there is nothing connected across from your external psu that would influence things?
Nope, nothing from the external supply, I'm not using a USB psu, it is a direct 5V fee to P1 from a DC-DC converter powered by a bench ps.
I removed the heatsink and stuck a temperature probe on the tab of RG2, and restarted the trend plot. The output of RG2 is showing now about 3.36V
and temperature is about 40C (ambient is about 27C).
It is not a big deal because we are still within spec, and it could have been a measurement anomaly, particularly with LDOs you can have some
variance when the rest of the circuit draws more or less current. I noticed that with the keyboard and HDMI connected the output of RG1 is
slighlty lower, I have no means how to measure the current draw out of it.
Another test I'm thinking of is to remove RG2 and feed the 3.3V with an external ps to P1 via and ampmeter and get some baseline current draw
profile with the R-pi doing different things and stuff like the HDMI monitor connected or not.
I've no clue if anybody at RPF did this kind of tests or not. Is there any record or discussions when the board was on the early design phase ?
You have looked for instability (HF oscillations) with your scope? These issues can be down to (input /) output capacitors. Maybe reflow the solder on these components as well as the regulator. Looking at the trace your 5V line seems to be trending upwards as well?
I assume that you have double checked your readings against another measuring device? Or against another piece of kit with a similar regulator. Its just that the 3.34V start value is out of spec to start with.
For the cost of these things if you are comfortable doing the job then what's to lose. Looks as though its easy to get at as well.
Still not a definitive explanation but I believe the issue is ambient temperature, I'm cranking up the air conditioner to reduce my office temperature from about 27C to 25C and now the output of RG2 shows 3.36V.
I think that since there are several components generating heat and it gets conducted through the power planes, it may generate sort of a cascading event where the internal temperature of some parts start to trigger side effects.
I'll hookup the rework station at some moment to replace the part and test with other regulators I've believe have in stock with the same package.
While the part number on the schematics is shown as NCP1117, the part installed on at least the two boards I've is numbered SE8117T33, which is not from any well known major semiconductor company but a chinese cheap knock off, it could also be a counterfit part, not uncommon when sourcing some parts from China.
This thermal image taken by "Remy" from Spain shows really where the heat is, if you follow the orange haze you can see how heat is being conducted by the board and not just staying around the hot components.
I wonder how many Raspberry Pi's will get "cooked" when people start to put them in cases.
Interesting image, where was that posted? Any more info, like the temperature of the LAN9512 at the time?
The original post is in spanish. Here is the link ...
The original post is in spanish. Here is the link ...
Excellent, thanks jamodio! As it happens I'm fairly good with languages and I speak Spanish quite reasonably, so the article made great reading without translation.
That's a really good series of tests made by Remy, and I'll paste one of the images from the article here because it includes the temperature scale on its right-hand side:
This is Remy's thermal image captured with a Fluke Ti35 during H.264 video and AC3 audio playback at 1080p and with Ethernet connected, which yielded the highest temperature recorded, peaking at 65.1C. It's the first numeric evidence I've seen of the blisteringly high temperature that I've noted (but not measured) on my LAN9512 device.
It would be very interesting to get Foundation engineers' comments on these figures, as so far I think Gert has only commented on subjective observations of "very hot" or finger tests. Is anyone here not banned yet, and dares post Remy's temperature-labelled image in a hardware thread over there?
Interesting images. I notice that 1.8V regulator RG1 (near GPIO connector) isn't the slightest bit warm. That would tend to confirm your theory that the LAN9512 may be supplying 1.8V to the SoC and DRAM, adding to the LAN9512's power problems.
That's a very interesting observation John, on all the thermal images RG1 looks cold. Unfortunatelly we don't have ANY electrical specs of the SoC, not even for the GPIO pins, to have at least a reference on what current draw to expect from the device and how the internal architecture looks is.
There are other pins from the schematics that would be interesting to know what are they and how they are supposed to be connected, for example SLDO_1V8_1 and SLDO_1V8_2 on the BCM2835 which seems to have some internal power supply block that generates several voltages, I'm pretty sure there is an internal switched power supply for VDD_OUT given the existance of L5 and a feedback loop.
Then is SLDO_1V8 an internal LDO to feed 1.8V to the PoP SDRAM ? if so why it is connected to the +1V8 plane ?
(Notice the comment on the sch "Used to be called SDRAM_1V8 on previous issues".)
Another question is, why all VDD_BATx pins are connected directly to +5V0, or why the VDDIO1_x pins are fed from what looks as an internal VDDIO power supply and VDDIO2_x and the other peripheral blocks like HDMI are fed from the +3V3 plane.
I'll put out a warning out there to be carefull on what you put on the GPIO ports and if you use the +3V3 line to supply power to any external circuitry.
IMHO the entire power architecture on this design is a big mistery and after reading Pete's May 8th article in newselectronics now is pretty clear that the Raspberry Pi has not been a design from scratch but started as "a proof of concept" (the "alpha" board) from Broadcom which Pete canibalized for cost and size.
Hard to say from just looking at the picture of the "alpha" board, besides having a decent DC power jack and a polarity protection diode, the circuitry around U1 and U3 look to me like a switching DC-DC converters and there is an ON Semiconductor 1.8V LDO (bigger package than RG1.)
Anyway, yesterday arrived the little heatsink that I ordered via eBay for the SoC:
Next week I'll replace RG2 by an original part from ON Semi, and since I'll fire up the rework station also replace the polyfuses on the USB lines by 0 ohm resistors.
Is this 1.8V SoC pin issue similar to the floating 1.8V SoC pin issue with the beta boards?
It sounds to me like maybe they had 2 related problems, but only fixed one.
>I'll put out a warning out there to be carefull on what you put on the GPIO ports and if you use the +3V3 line to supply power to any external circuitry.
I don't know if it's at all related, but abisher reported a mysterious
-3.3V measured between a GPIO pin set to be an output, and ground.
And perhaps related, Gert explained that a GPIO pin set to be an output
can source as well as sink, resolving a mystery.
Interesting, I've not seen that comment on the blog about the problem found on the beta boards.
For sure nobody was paying much attention or reviewing the schematics before sending the design to the pcb fab.
Not with the intention to be rude, but I'd not pay much attention at what abisher has to say, in few exchanges I had with him it came out pretty clear that most of the time he does not have much of a clue of what he is talking about, just repeat the message like a parrot, but is fast to aplogize when he realizes he screwed up.
Gert is correct, many GPIO modules on modern microcontrollers and processors can be configured as open collector/drain, which is very handy because you can drive almost anything regardless the power supply to the MCU, but you have to keep in account that whatver current circulates through the load will also circulate or "sink" through the output circuitry of the MCU, so there are current limitations and like any other current circulating through it, it may genearte heat. This is a typical situation with ethernet PHY circuits like the one on the LAN9512, those types of chips are designed to operate somehow "hot" because the differential output drivers sink current that flows through the windings of the ethernet transformer.
But there are ways to use the open drain combined with external driving circuits to handle larger loads without having the MCU to sink excesive currents.
On the other hand if it is not open drain or open collector, you are sourcing the current from the power supply pins to the load, again generating heat on the chip.
In any case, you still need power if you want to put something on the GPIO pins of the R-pi, then where you get that from ?
Depending on the project, I'd strongly recommend to use a separate 3.3V regulator, you can feed it from the main 5V line (asuming that whatever you put won't trigger the
How are you?
Please tell me what are the characteristics of the two sheatsinks used.
I think I found another problem I didn't notice before.
I have the R-pi hooked up in the prototyping gig I described on another thread (http://www.element14.com/community/thread/18981?tstart=0),
currently no keyboard, HDMI monitor, just network and access to console via the serial port.
I've been running a trend plot monitoring the +5V feed to the R-pi and the output of the 3.3V on board regulator. I noticed that gradually
the voltage out of the regulator has increased from 3.30V to almost 3.39V, it is actually not doing much, just responding to ping packtes
(flood mode) I'm sending from another machine and serving the serial console I use once in a while to run procinfo and network stats.
I guess the voltage increase is probably not noticeable when there is more load and current draw, the only thing I can think of is self
heating of the LDO. I just placed one of the little heatsinks I've got for the LAN9512 on the LDO and seems that output voltage is starting
to slowly drop. I'll keep this configuration to see if with the heatsink the voltage keeps droping.