ODROID XU4 Supply Voltage Limit

[BobbyB]

I would like to know the maximum and minimum operating voltages for the ODROID XU4.

I know all of the documentation lists 5V/4A for the power supply, but what is the tolerance on this spec? Assuming 5V nominal, there needs to be a min and max value.

The XU4 schematics show NCP372 Protection ICs on the power input. Datasheet shows an over-voltage threshold range of 6V to 6.6V. Power IC appears to be a Samsung S2MPS11B, but I can't find any input voltage specs for that part.

So has anyone operated an ODROID XU4 at a voltage higher than 5V, say somewhere between 5.5 and 6V, on a consistent basis?

Thanks!

[rooted]

I operate at 5.25v, 5.5v is okay but I'm not 100% sure about 6v...gut instinct tells me it's okay. Wait for odroid or someone more knowledgeable to verify.

I do know a couple of people who mistakenly tried 24v or 12v (can't remember), one board survived one didn't.

[odroid]

Please don't try higher than 5.3Volt. The PMIC's absolute maximum input voltage is 5.5Volt.
Recommended input voltage is 4.8V~5.2Volt.

[BobbyB]

Thanks for the Reply. That answers my question. Will be sure to limit input voltage to that range.
I've been unable to find a datasheet for the PMIC. Do you know where I could find one?

[odroid]

Unfortunately, Samsung didn't release their datasheet publicly.

[pat2]

hi Odroid and forum participants,

I'm becoming crazy. The issue happens only when using an SSD (with filesystem on SSD or filesystem on SD or filesystem on eMMC).
If I don't use the SSD, i don't have issue.

I tried everything: no more tweaks, low CPU temperature (changed thermo conductive paste+tall heatsink+ fan), filesystem installed from scratch, changed the SATAtoSSD connector, changed the voltage supplier (currently a Meanwell 5V, 10 A)...
Randomically (after 10 days, 2 days, 1 week, randomically) when running with the SSD the odroid turns off. Without SSD the system runs forever.

The only 3 remaining ideas are:

- increase the Voltage of voltage supplier,
- change the SSD (currently a Crucial MX300),
- change the OS (currently Ubuntu 16.04 from HK).

With the same OS/filesystem the Odroid works fine without the SSD.

I increased the supplier voltage from 5,14 to 5,3 and testing...
Please let me know if you have reccomendation to avoid the issue.
thanks

[mctom]

Sounds like completely unrelated to power supply if it happens after a few days of operation.

[odroid]

I agree with mctom.

But it is worth to check the power path quality.
If you have an oscilloscope, keep measuring the voltage on the 5V power rail of the 30pin header while heavily accessing the SSD.
If you see a short period of voltage drop like 4.8V while your PSU output is 5.2V, there might be voltage drop between the PSU and the DC barrel jack due to some amount of resistance.
In that case, you might need to increase the diameter of DC cable something like this AWG18.
https://www.hardkernel.com/shop/dc-plug ... bly-5-5mm/
This thread might be helpful too.
viewtopic.php?f=93&t=33365

[pat2]

Sounds like completely unrelated to power supply if it happens after a few days of operation.

I agree with mctom.

But it is worth to check the power path quality.
If you have an oscilloscope, keep measuring the voltage on the 5V power rail of the 30pin header while heavily accessing the SSD.
If you see a short period of voltage drop like 4.8V while your PSU output is 5.2V, there might be voltage drop between the PSU and the DC barrel jack due to some amount of resistance.
In that case, you might need to increase the diameter of DC cable something like this AWG18.
https://www.hardkernel.com/shop/dc-plug ... bly-5-5mm/
This thread might be helpful too.
viewtopic.php?f=93&t=33365

hi guys, thanks for the fast answer. I think too not related to power supply quality. Both power suppliers have good quality: the first coming from HK, the second from Meanwell.

I thought about drop voltage along the cable and/or crossing the board.
Unfortunately I don't have an oscilloscope, only a multimeter. So I cannot see an eventual transient.
About the cable I realized a self-made DC cable: short lenght and increased diameter. I'm not sure about quality of the barrel.
continuing to test.
I increased the no-load voltage (to 5.3) and changed the SSD too.
I will let you know.
thanks

[mctom]

I used to have similar problems with the first prototypes of PiStackMon, see the picture here: viewtopic.php?p=326978#p326978
The last Pis in a daisy chain often throttled or even reset under the load. So I added a capacitor on each board which solved the problem. There was still a DC voltage drop, but no transients that caused brown-out resets. Pis work well down to 3.8V if you don't use USB and turn off its undervoltage warnings.
Once I saved a movie night by adding a 100uF electrolytic capacitor right onto the GPIO pins, between 5V and GND, otherwise the Pi throttled due to undervoltage and the movie stuttered.
One important thing to note: Do not add a capacitor to an SBC when it's on, otherwise you will surely reset it.
Those were the dark days before I discovered Odroids.

If transient voltage drops are indeed the cause of random shutdowns, you should be able to reproduce it by artificial stress test.
Try firing up stress -c 4 -d 50 to load four big cores and thrash SSD. Note you need to execute this command in a directory that is physically on SSD, otherwise it will thrash some other storage.

[mad_ady]

You can also enable netconsole and send logs to a different system. That way, when the xu4 hangs you might find out what the cause it via its dmesg messages.
Most likely the usb bus resets causing the storage to reappear as sdb. But the rootfs becomes inaccessible, so everything dies.

[mctom]

I understand that XU4 does not hang but turns off completely, but indeed the symptoms could be described in greater detail.

[pat2]

Hi guys,

you are right. I will describe better.
I would not be mistaken with my memory. I will take a picture and a copy of syslog. And post it . I have to wait the phenomenon will appear again.
Currently testing: V=5,2 V (I decreased it a bit), a different SSD, the fan (it's not the HK fan) is currently supplied by an external supplier (before it was supplied by the board, I suspected it could be the cause).
I will try to stress the board and enable netconsole.
thank a lot

Note: mctom, I saw your PiStack! wow!!

[pat2]

yes, XU4 seems a little bit stressed
please, mctom, tell me if I can leave it under stress for days (or not if you see issues on SSD or the board).
temperature has increased 10-14 degrees C. running 7/24 stress and zoneminder (12 cameras, motion detection for 5 of them)
filesystem is on SSD and I tweaked DDR frequency at 933

stress.JPG (257.81 KiB) Viewed 1257 times

[mctom]

I think you can leave it for days but that's just an unnecessary wear to you SSD. If under all that extreme stress the board doesn't turn off, that means it's almost certainly not caused by power supply.

[pat2]

ok, mctom. XU4 passed the test: it didn't turn off and is continuing to run without the stress condition (only Zoneminder 24/7).

temperature decreased. and CPU load too. I will advise if the issue will appear.

[mctom]

Which pretty much defeats the argument that there is anything wrong with power supply voltage. There is no way your system could draw more current than with this synthetic load.
I guess there must be something going on on the software side of things. Hope your logging will reveal something interesting.

[pat2]

Hi,

still running 24/7 without interruption.
At the end, it seems the issue was the f...g fan bought on Amazon.
I'm currently using it but supplied by an external source (not the pins, neither the molex connector on the board).

I will be sure after 15 days of uninterrupted run...I will inform you.

[mctom]

Ah yes, poorly designed fan may consume even 5x the rated current when they start.

In your case, what could have caused random resets was poor supply leads, plus CPU, SSD and a fan all kicking in at the same time - it's not hard to imagine, some periodic operations can do that, like drive indexing, unattended-upgrade or whatever.

Nice detective work, I'd never figure this out as you never mentioned a custom fan.

[pat2]

you are right! I didn't tell you about the fuck.ng fan..I changed it after the original fan from HK died...

I started from sw analysis removing all sw tweaks I did to improve speed: currently the installed configuration is basic: only DLNA server (for music, video sharing), SAMBA server (for SSD sharing), Zoneminder (for surveillance matter). I was sure that such configuration worked without interruption in the past !!

Then I passed to HW: substituting the thermal conductive paste, changing the power supplier and playing with supply voltage.
I changed too the SSD!
About fan, I didn't see it!! It is everytime under my nose and so it became so familiar that I didn't notice much.

Yes, probably - as you told- a poorly designed fan together SSD and heavy 24/7 CPU loads causes the issue under particular condition (drive indexing, unattended-upgrade or whatever). Moreover the fan is an inductive load. Not a beatiful load in terms of electric impedance.

So, I hope the issue has been solved (currently DDR frequency is tweaked too at 933 Mhz!!) and my posts could help other friends to do not become crazy like me. A fuck.ng fan...
But better don't claim victory ahead of time! I will keep you informed.

thanks ciao

[pat2]

So, at the end I can definitevely say the issue was related to the not original cooling fan.

I have 2 Odroid XU4Q.
One of them transformed in XU4 through the original HK cooling fan https://www.hardkernel.com/shop/cooling-fan-xu4-blue/ , then died.
When died I bought those 2 cooling fans https://www.amazon.it/GeeekPi-Raspberry ... 07X37NXGX/ from Amazon. And applied to both Odroid XU4Q boards.
The first cooling fan connected to the molex connector dedicated to the original HK fan, the second to pin1/pin2 of CON10 (or CON11).
On both boards I had issues: randomically the boards turn off.

Currently the 2 cooling fans (both) are connected to the meanwell power supplier (50W, 5V 10A) that is serving one board.
The other board is served by the original HK power supplier.

With this configuration the issues disappeared for both boards. Both running 24/7, with filesystem on SSD and DDR frequency tweaked at 933.

XU4s with cooling fan externally supplied.jpg (181.09 KiB) Viewed 1109 times

i know it's a mess

[rooted]

Looks like an active lab to me

[pat2]

Looks like an active lab to me

hahahhaah
it' my game room in the basement.
There is the network star point and, yes, a sort of active lab.

The Lab.jpg (83.23 KiB) Viewed 1091 times

[pat2]

Quick update and probably definitive on the matter - Power supply.

Definitively the issue was related to the not original cooling fan and the heavy load on my Odroid boards: SSD usage, 933 DRAM frequency tweak, GOVERNOR="performance", cores at the highest maximum frequency (2.0 4 big, 1.5 4 little), 24/7 usage, running 24/7 Zonemeninder with movements detection on 5 cameras, my website, DLNA server, samba server.

I have 2 identical odroid boards: one with his original HK power supplier (5V, 6A), the second with a switching supplier by Meanwell (5V, 10A).

I found:

• for original HK supplier: no load voltage = 5,14 V, full load voltage on PIN 1-2 on CON10 = 4,87V

• for meanwell supplier: no load voltage = 5,30 V (fixed by me on the meanwell supplier), full load voltage on PIN 1-2 on CON10 = 5,08V

After I moved the cooling fan power supply to a dedicated supplier (i.e. not coming from the odroid board), I didn't have unexpected crash of the boards, but I didn't imagine a so high voltage drop: 0,27V for the original supplier, 0,22V in the meanwell case.

I imagine the drop voltage due to the high current (4A? I have to measure it) on the cable and the DC5521 barrell.
This post is to highlight the importance of the quality of supplier, but also of cables and barrel.

I found this interesting video on measuring the performance of USB power cables:
https://youtu.be/odu17lhN-vk

A question for Odroid: what's the recommended voltage at pin 1-2 on CON 10? 4,87V in the first case and 5,08V in the second case seem to work well: no board crash for 24/7 and heavy load.

[mctom]

Ahem, I know that wasn't a question for me, BUT...
All ARM SoCs need 3.3V tops to operate - they are designed for mobile devices after all. I tested that long time ago with RasPis 4, and they happily worked under full load at 3.7V input voltage. (a 0.4V margin for the built in Buck converter).

The sole reason why SBCs are powered off 5V, or need 5V on board by any other means, is to provide USB power and HDMI signalling, IIRC. IF you're not interested in any of these then the input voltage doesn't matter that much - it's internally converter to needed values anyway.

But if you use USB (SSD drive?), be aware the USB standard tolerates 5% deviation from the nominal 5V. (4,75 - 5.25V). Every USB compliant device must work properly within this range.

[pat2]

Thanks, mctom. The question was for everyone can add value. You too.
Your answer is really interesting.

First I was not aware about ARM Soc voltage matter, I understand there is a voltage converter on the board to supply properly (3.3V) the ARM SoC.

Then, yes, I use SSD (my rootfs is on SSD). I argue that since 4,75 - 5.25V is the recommended voltage range I'm ok in the first and second case (4,87 and 5,08 voltage Pin1-2 on CON10).
thanks a lot

[mctom]

Actually ARM SoC, like many other sophisticated chips, requires a few voltage levels to operate. The fastest logic (CPU core) operates at about 1.1V. Even Intel CPUs work at very low voltage and crazy high currents - that's what makes them so fast.

3.3V that I mentioned is required by the "slower" parts of your SBC, such as network controller, SoC peripherals, DDR perhaps?

But the takeaway for you is that the most critical parts of SBC require 3.3V to run - or not reset, at least. If you experienced random resets, that must mean whatever power supply you used could not provide enough current, thus voltage dropped. Or couldn't adapt to sudden current change fast enough, which also could be an issue (often solved with "more capacitors").

I'm really glad you chose an industrial PSU, I own a few myself. I operate my Odroid M1 server with spinning HDD off a Mean Well unit, can't complain at all.

[pat2]

Crystal clear, mctom.

thanks a lot.

[pat2]

Anyway, I think there is a too high voltage drop on the board. A board design/realization issue?

I solved with the Meanwell supplier through the over voltage workaround i.e. setting the power supplier's no load voltage to 5.3V and obtaining 5.08 between pin 1-2 on CON 10.

With the HK supplier, since not possible to set a voltage greater than 5.14, sometimes, more rarely without the cooling fan, the board turns off when SSD is connected.

I don't know if possible to solve with:

solution 1: using an SSD hub with an external supplier
solution 2: disconnecting the power pin/ground pin of the board USB port (to avoid to force current into the board) and supplying the SSD directly with a different supplier

Odroid, did other XU4 owners report such experience?
thanks

[odroid]

As far as I remember, there was no such issue with a USB-SSD enclosure.

I have used a generic WD 2.5inch 4TB HDD with my XU4 for near 5 years as a personal home NAS with a stock 5V/4A PSU.
The USB VBUS voltage could be down to 4.8Volt shortly in the booting process but my NAS is still working well 24/7.

What is the minimum working voltage of your USB-SSD?

[mctom]

Anyway, I think there is a too high voltage drop on the board. A board design/realization issue?

No. The difference in voltages that you have measured is not a "voltage drop" in its common understanding. DC voltage sources often have slightly higher value when not loaded.
If you want to measure the actual voltage drop on XU4, you'd have to probe Meanwell output voltage and XU4's USB output voltage, and compare these two. I'm sure there's nothing wrong with that.

But anyway you did the right thing, I also tuned my meanwell's voltage while it was operating. No worries about the higher voltage when it's not loaded - won't hurt anyone in such conditions.

If you manage to hang or reset the XU4 with a USB device plug-in, that means one of the two things:
- that USB device is not compliant with USB standards (that is surprisingly common these days)
- That XU4 is broken (but of course, we can blame hardware for almost anything )

What you can do about it is to keep the SSD connected before you turn on XU4, that can help.
Or, use a USB hub with external power supply to reinforce HDD.
Or, use a USB galvanic isolator and provide 5V power to HDD separately.

One thing for sure, never cut the ground connection in USB cable, or any wire in it. Because reasons...

[pat2]

As far as I remember, there was no such issue with a USB-SSD enclosure.

I have used a generic WD 2.5inch 4TB HDD with my XU4 for near 5 years as a personal home NAS with a stock 5V/4A PSU.
The USB VBUS voltage could be down to 4.8Volt shortly in the booting process but my NAS is still working well 24/7.

What is the minimum working voltage of your USB-SSD?

Hi Odroid,
do you mean a "powered" enclosure?

With original HK supplier (5V, 6A): no load voltage is 5,14 V. The voltage on PIN 1-2 on CON10 is 4,87V (when connected the SSD).
The value is stable and it is the normal run voltage between PIN 1-2 on CON10.
Is it a correct value?
I don't have issue at boot and also during normal run. The only, suddenly during the usage the board turns off. More rarely without external cooling fan, but turns off.

Please let me know if you have suggestion.

[pat2]

Anyway, I think there is a too high voltage drop on the board. A board design/realization issue?

No. The difference in voltages that you have measured is not a "voltage drop" in its common understanding. DC voltage sources often have slightly higher value when not loaded.
If you want to measure the actual voltage drop on XU4, you'd have to probe Meanwell output voltage and XU4's USB output voltage, and compare these two. I'm sure there's nothing wrong with that.

But anyway you did the right thing, I also tuned my meanwell's voltage while it was operating. No worries about the higher voltage when it's not loaded - won't hurt anyone in such conditions.

If you manage to hang or reset the XU4 with a USB device plug-in, that means one of the two things:
- that USB device is not compliant with USB standards (that is surprisingly common these days)
- That XU4 is broken (but of course, we can blame hardware for almost anything )

What you can do about it is to keep the SSD connected before you turn on XU4, that can help.
Or, use a USB hub with external power supply to reinforce HDD.
Or, use a USB galvanic isolator and provide 5V power to HDD separately.

One thing for sure, never cut the ground connection in USB cable, or any wire in it. Because reasons...

Hi mctom, many conversation starters. I will answer later.
Thanks

[odroid]

Hi Odroid,
do you mean a "powered" enclosure?

With original HK supplier (5V, 6A): no load voltage is 5,14 V. The voltage on PIN 1-2 on CON10 is 4,87V (when connected the SSD).
The value is stable and it is the normal run voltage between PIN 1-2 on CON10.
Is it a correct value?
I don't have issue at boot and also during normal run. The only, suddenly during the usage the board turns off. More rarely without external cooling fan, but turns off.

Please let me know if you have suggestion.

I meant a bus powered enclosure. There is no additional power source.
4.87V should be okay since the standard USB VBUS specification is still 4.75 ~ 5.25Volt.
But, the actual voltage might be lower than your measured value due to a short period (severl millisecond) of in-rush current.
You might need an oscilloscope to capture the short-period of voltage drop.

Do you know the maximum in-rush power consumption of your USB storage device?

A thicker USB cable between your storage device and XU4 might be helpful.

[pat2]

ok, odroid thanks.

It's not a big issue since I don't have issue with the production odroid board (supplied by meanwell and the workaround I told)
Probably for the second odroid board a powered enclosure is enough. Do you have suggestion? on Amazon?

Regarding your second comment, I agree, 4.87 should be ok. And it is ok: no issue at boot, no issue during normal run. Except sudden turns off after 1 week, 10 days. Probably due to an in-rush current for a short period. I don't have such issue whit the Meanwell or when there is only the SD card (or the eMMC card).

I don't know how much is the in-rush current, but I read that It could be till 1 A.

https://www.anandtech.com/show/7594/sam ... b-review/8

Anyway the production is ok, I will try with the powered enclosure.
thanks

[pat2]

No. The difference in voltages that you have measured is not a "voltage drop" in its common understanding. DC voltage sources often have slightly higher value when not loaded.
If you want to measure the actual voltage drop on XU4, you'd have to probe Meanwell output voltage and XU4's USB output voltage, and compare these two. I'm sure there's nothing wrong with that.

But anyway you did the right thing, I also tuned my meanwell's voltage while it was operating. No worries about the higher voltage when it's not loaded - won't hurt anyone in such conditions.

If you manage to hang or reset the XU4 with a USB device plug-in, that means one of the two things:
- that USB device is not compliant with USB standards (that is surprisingly common these days)
- That XU4 is broken (but of course, we can blame hardware for almost anything )

What you can do about it is to keep the SSD connected before you turn on XU4, that can help.
Or, use a USB hub with external power supply to reinforce HDD.
Or, use a USB galvanic isolator and provide 5V power to HDD separately.

One thing for sure, never cut the ground connection in USB cable, or any wire in it. Because reasons...

today also the odroid board in production, supplied by meanwell, turned off.

the usb device is the famous StarTech
https://www.amazon.it/gp/product/B00XLA ... UTF8&psc=1

the SSD is a Samsung 1Tb
https://www.amazon.it/gp/product/B08PC5 ... UTF8&psc=1

the Meanwell output voltage is 5,30 V and XU4's USB output voltage is 5,08V
The issue is not related to SW. Same SW on SD only runs without issue.
I will try with a powered enclosure/USB hub.
if still KO, probably I'm not so good to use properly the board when rootfs is on SSD.
thanks