ESP32-POE-ISO temperature

Started by Stephane80, November 10, 2019, 07:52:02 pm

Previous topic - Next topic

Thibaut

May 22, 2020, 10:45:34 am #15 Last Edit: May 22, 2020, 10:47:45 am by Thibaut Reason: added temperature from website
Hi,

Thanks for the quick reply.

But the temperature of the board is already above what is marked as safe to operate on the website:
"ESP32-POE-ISO and ESP32-POE-ISO-EA are with commercial temperature range 0-70C"

How hot will the non ISO version get if I would use that one instead?

Thanks

LubOlimex

May 22, 2020, 11:38:53 am #16 Last Edit: May 22, 2020, 11:44:20 am by LubOlimex
0-70 is ambient temperature range around the board, like the air in the room. Board in commercial temperature range allows you to use it in environments that don't go under 0 degrees C and where it doesn't get hotter than 70 degrees C.

Edit: the ISO version would probably get as hot as the non-ISO version
Technical support and documentation manager at Olimex

Thibaut


Huntercover

Hello Olimex-team,

I would like to answer on that old topic because I feel to have the same problem with a non ISO version:
I bought a non ISO version and from my point of view the board is getting too warm as well.

Question: May I connect USB + ethernet together if I am using an external POE splitter directly before the board? I am asking because I have the non ISO version..
To make it clear:
- I am having one ethernet line from a POE switch
- Using a external POE splitter to get a LAN cable and micro USB.

Will the board burn if I plug the 2 connectors in?

Regards,

Andreas

LubOlimex

As long as the PoE power does not go into the ESP32-POE board but only the Ethernet connectivity (as in your case since you get the power off before the board) - it should be safe. Just double check what your splitter does exactly.

If you can keep your finger over the components of the board for more than 10 seconds - it is not too hot.
Technical support and documentation manager at Olimex

gender

Hi,
I have exactly the same problem with my ESP32 POE:

If Im only using POE, the microcontroller is very very hot... so hot that I can not keep my finger over the components for more than 10 seconds.



If Im using Ehternet and USB together (without POE) the problem is gone. There is no hot components at all.


So, it is a workarround for me to use USB-Power but actually I would like to use the ESP32 over POE !!!

Is there any solution for the hot components?

The hottest point by the way is at the black thing close to the letters "olmex":






LubOlimex

It is normal to be hot. Voltage from PoE is above 50V that have to be transformed into 3.3V by a tiny piece of electronics... When you power from USB it is 5V that has to be transformed into 3.3V, so naturally it is much easier.
Technical support and documentation manager at Olimex

Drjaymz

Mine is getting too hot.  I'm measuring 50V and looking at the schematic, that is across R41 which is 4k7. At 50V thats over half a watt.

I am not sure why R41 exists.  But I think its one of a host of fudges to attempt to deal with glitching that this board design seems to have been plagued with looking at the revision changes.  I think its there to discharge C27 and soak up any transients.

Half a watt is easily within that resistor's capability, a hot resistor isn't going to fail.  I just put a thermocouple on mind and its 126C - half of that heat is soaking into the ground plane to which one end is soldered.

Unfortunately, because of the layout it heats up C26 and C27 and will cause them to fail and there's a couple of posts of exactly that type of failure and that will happen as soon as you enclose the design.  105c capacitors or not, remember at 105c their life is 1000hrs.

I will remove R41 and measure the temps again and see how long it is before it glitches.

Drjaymz

Quote from: Drjaymz on November 11, 2022, 11:55:37 amI will remove R41 and measure the temps again and see how long it is before it glitches.

sitrep, with R41 removed its been up and running and has stabilized at about 35C which is a lot better than 90C and I would consider it ok to use in an enclosure now.

Another 'bug' I found was that the PG pin(6) from U2 should be pulled high.  If not then you might find that nearby transients could shutdown the DC-DC converter.

The DC-DC converter is perfectly fine at 0.5A from a 50V source, but its limitation window is a package temperature of 80C according to the datasheet.  Without R41 heating it up it doesn't get that warm at all, it should be better than 75% efficient in this circuit.

LubOlimex

R41 is required to maintain power signature (MPS). Refer to:

https://www.ti.com/lit/ds/symlink/tps2376.pdf?ts=1668174991211

What is the PSE equipment in your setup? Is it compliant with IEEE 802.3af (or newer)?

Technical support and documentation manager at Olimex

Drjaymz

November 11, 2022, 05:44:53 pm #25 Last Edit: November 11, 2022, 07:05:20 pm by Drjaymz
It should be between CLASS (pin 2) and VSS?
Its a loading that is applied to the line to program the power delivery.  The specification says that power delivery will continue as longs as the DET is present - thats R28 on your diagram.
But here R41 is continually on the line which is why its cooking.



And yes, with it removed an 802.3af Fortinet PoE is happy to whack 51 Volts up it.
So, with it removed the board consumption drops a lot.  And the environment, the PoE and  and the capacitors can thank  me later.

I get it - you think that 10mA is the minimum load required for MPS?

Drjaymz

November 11, 2022, 06:19:31 pm #26 Last Edit: November 11, 2022, 06:22:20 pm by Drjaymz
https://e2e.ti.com/support/power-management-group/power-management/f/power-management-forum/251393/poe---maintain-power-signature

Here Ti Explain that MPS is pulsed and that the chip switches in the CLASS sense when required.

That will be why mine works fine without the parasitic load.

He says:

 I think I see where you confusion may lie. In TPS2375 DS, Table 1 lists class 0 current (in the 802.3af LIMITS (mA) column) as 0-4mA. This current is not the operating current, but is the current that TPS2375 presents to the PSE during classification

To be fair, the actual datasheet is very waffley and suggests that MPS is a minimum load - its not.

LubOlimex

Your link proves our hardware design was correct!

From your link: "After classification, then 48V is applied and this is when the PD must maintain the 10mA load (see the PD Power (W) column in Table 1)."

Your quote is for the moment before reaching maximum voltage (only between 13V and 21V), and you omitted that second part that clarifies it "In TPS2375 DS, Table 1 lists class 0 current (in the 802.3af LIMITS (mA) column) as 0-4mA. This current is not the operating current, but is the current that TPS2375 presents to the PSE during classification (when the port voltage is between 13V and 21V). "

I don't know how you understand it. It is around 10mA (~47VDC/4.7k).

Furthermore, we used TPS2375EVM's design as a basis and it is connected exactly like that. Check for resistor R11 (controlled by jumper JP7) in schematic on chapter 5-1 in the document here:

https://www.ti.com/lit/ug/slvu108/slvu108.pdf

If it heats for you, you can remove it, but if it heats it is because of your PSE equipment is not made according to the standards, not because our hardware design is bad.
Technical support and documentation manager at Olimex

Drjaymz

QuoteIf it heats for you, you can remove it, but if it heats it is because of your PSE equipment is not made according to the standards, not because our hardware design is bad.

OK, I'll concede if you're going to maintain class zero you'll need >=10mA, although that isn't defined as a DC load (like a resistor), for simple designs it probably is. The evaluation kit you refer to has a parasitic load because it must work standalone without a connected load and that has caught a few people out.  If you have a load connected, then you can remove JP7.  That is basically what the response is saying because none of the reference designs have it.

The resister heats up because its across 50V (I2R losses), regardless of the PSE - you'll always get that heating, don't be blaming the PSE compliance!

Its problematic because the resistor is thermally coupled to ground and heats the everything and it seems a common complaint.  C26 and C27 in particular are not going to like that long term and its not ok to be 90C. But as I said to others its not going to bother a resistor its well within its happy zone.  Mine got to 90C and this room has no heating in it.

But, is it necessary?  Lets have a look at the schematic...

Lets suppose your DC-DC conversion was 80% efficient you'd need to see about 0.4W downstream to ensure a loading of 10mA at 50V.  The ESP32 is nominally about that as it happens, then you have a few other parts like the PoE parts itself etc. In reality, at lower currents the DCDC conversion is more like 65 - 70% as per its datasheet and we don't care about higher loads.

As it happens, experiment is worth a thousand opinions so I measured it - I get 11mA at 51V from the PoE which roughly matches what the PSE says.  And I don't have R41 connected.  So thats why it works.  Another board I have with the resistor still attached measure 20mA which is what you'd expect.

Now, as a board designer, you can't assume that someone will not put the esp32 into sleep mode, but the irony is they won't be saving power.  So thats why you have that load - but I can make a couple of suggestions.

1) thermally isolate R41 from the ground plane better - or use a through hole.
2) provide a jumper just like the eval board - because I think whatever you do it will heat up the board.

LubOlimex

> But, is it necessary?  Lets have a look at the schematic...

Well, you already answered yourself. Yes. In fact we didn't have this resistor when we switched to TPS2375 and weren't aware of the MPS requirement during sleep. This was causing two issues:

1) It turns out most of the customers use these boards on remote locations where sleep and deep sleep are important to conserve power during power downs. Furthermore, most of the customers evaluating which ESP32 board to consider would always measure power consumption in sleep and deep sleep modes. And the boards in sleep would lose PoE power since the PSE would turn off the power supply completely.

2) Same thing naturally happened when using the on-board reset button.

Of course we would consider your advice for future revisions. I've added notices for future revisions to again empirically measure the load and the temperature and to consider better routing of R41 or/and adding a jumper.

Thanks for your constructive feedback.
Technical support and documentation manager at Olimex