N2+ USB ports no longer working

[servili007]

I have a unit ordered in August of 2020 straight from HK. The USB power rail (VBUS) seems to have totally left the building, as I can't even get a USB powered LED light to respond.

5V rail at the GPIO is fine.

lsusb only contains the linux foundation root hubs, no genesys devices. A powered hub doesn't help the situation.

@odroid this seems like a dead USB hub controller IC?

Code: Select all

test@odroid:~$ sudo lsusb -t
/:  Bus 02.Port 1: Dev 1, Class=root_hub, Driver=xhci-hcd/1p, 5000M
/:  Bus 01.Port 1: Dev 1, Class=root_hub, Driver=xhci-hcd/2p, 480M
test@odroid:~$ sudo lsusb
Bus 002 Device 001: ID 1d6b:0003 Linux Foundation 3.0 root hub
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub

[odroid]

Right. The USB hub controller IC GL3523 on the board was dead.

Since our warranty period is 16 weeks, you need to pay two-way shipping cost to repair your near three years old board. Sorry about that.
viewtopic.php?f=53&t=661

Or, if your application can live with a single USB 2.0 port, consider using the micro-USB port with an OTG-to-Host cable something like this.
https://www.hardkernel.com/shop/usb2-0-otg-cable/

[servili007]

Right. The USB hub controller IC GL3523 on the board was dead.

Since our warranty period is 16 weeks, you need to pay two-way shipping cost to repair your near three years old board. Sorry about that.
viewtopic.php?f=53&t=661

Or, if your application can live with a single USB 2.0 port, consider using the micro-USB port with an OTG-to-Host cable something like this.
https://www.hardkernel.com/shop/usb2-0-otg-cable/

I was not making a comment on warranty length or board reliability, I just wanted to understand the issue I was surprised by the failure since this unit is a headless server with a single usb device connected and sits on a shelf where it isn't touched.

A few follow-up questions:

What is your engineer's opinion on the difficulty of replacing the GL3523 via hot air rework if I'd like to attempt it myself?

Is there an additional repair cost in addition to the two way shipping if I do send it back?

There were some N2 board revisions to address stress to the GL3523. Are any mods performed to repaired boards to approximate these changes?

Thank you

[mctom]

If you have some experience of replacing QFN package ICs, it must be easy.

Our labor cost of replacing the GL3523 IC is a few dollars including a new IC.
But if the USB root hub in the SoC(S922X) is damaged, it will be a totally different story.
Because the SoC is glued with epoxy, we can't replace it.
Therefore, in worst case, we can't repair your board.

The PCB revision of 3.3Volt power-rail isolation was applied from early '2021.
But the power-rail trace exists in inner layers, it seems to be almost impossible to implement a simple mod.
8-layer PCB modding is quite difficult and very expensive.

We've shipper over 200,000 N2 boards in the past few years.
There were about 0.1% of USB failure issues user reported before applying the PCB design change.
After changing, the failure rate seemed to dropped to 0.05%. But it is still existing.

BTW, what is your production date number of your board label showing?
https://wiki.odroid.com/odroid-n2/hardw ... lot_number

[servili007]

BTW, what is your production date number of your board label showing?
https://wiki.odroid.com/odroid-n2/hardw ... lot_number

Label shows 2006

[odroid]

It was made in June 2020.

[servili007]

It was not a beginner level job given the density of the pins, but the Genesys chip has been replaced, and I now have functional USB ports again.

Code: Select all

test@odroid:~$ lsusb
Bus 002 Device 002: ID 05e3:0620 Genesys Logic, Inc. USB3.1 Hub
Bus 002 Device 001: ID 1d6b:0003 Linux Foundation 3.0 root hub
Bus 001 Device 002: ID 05e3:0610 Genesys Logic, Inc. Hub
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
test@odroid:~$ lsusb -t
/:  Bus 02.Port 1: Dev 1, Class=root_hub, Driver=xhci-hcd/1p, 5000M
    |__ Port 1: Dev 2, If 0, Class=Hub, Driver=hub/4p, 5000M
/:  Bus 01.Port 1: Dev 1, Class=root_hub, Driver=xhci-hcd/2p, 480M
    |__ Port 1: Dev 2, If 0, Class=Hub, Driver=hub/4p, 480M

[zyssai]

...

Hi.
Is there a way to know what IC is damaged?
I can buy an N2 with dead USB ports for a few bucks, I just need to know if GL3523 is faulty.
To confirm, is it U26 ?

[zyssai]

If USB 3.0 host ports are not working, U26 or U12 might be damaged.
If U12 output voltage is steady 5Volt, U26 GL3523 is faulty very likely.

[zyssai]

If USB 3.0 host ports are not working, U26 or U12 might be damaged.
If U12 output voltage is steady 5Volt, U26 GL3523 is faulty very likely.

Hi odroid.
It seems U26 is damaged.
I found QFN76 GL3523-OTY10 and OTY30.
Cannot find difference between them.
Do you know which one is it?

[mctom]

OTY is their symbol of QFN76 package, and I'm just guessing that "30" is an updated version of "10". The latter is removed from the manufacturer's website, so you probably want "30". But I'm not completely sure, @odroid will know for sure.

[odroid]

We can check the suffix early next week since it is already Friday evening hear Korea.

[zyssai]

OTY is their symbol of QFN76 package, and I'm just guessing that "30" is an updated version of "10". The latter is removed from the manufacturer's website, so you probably want "30". But I'm not completely sure, @odroid will know for sure.

Thanks. I don't think this really matter but it's better to ask
It seems there is majority of OTY30 so you're probably right.

We can check the suffix early next week since it is already Friday evening hear Korea.

Thanks a lot. I scratched marking while removing it and forgot to take picture before.

[mctom]

Here is someone else's picture. Looks like there is "30" in the marking, but could mean anything.

viewtopic.php?p=339071#p339071

[zyssai]

Here is someone else's picture. Looks like there is "30" in the marking, but could mean anything.

viewtopic.php?p=339071#p339071

Thanks that's interesting, as I've seen that OTY10 have marking code ending with 10.
This N2 is rev 0.5 so I don't know if previous revisions used OTY10 or 30. I'll check my others N2 this weekend to compare.

[ruppi]

It's a minor chip revision change with some specifications related to USB 3.1, so I believe it's okay to use anything whether it's 10 or 30.

[zyssai]

Thanks. I managed to read marking code of the one I removed on rev 0.5, it seems to be OTY20.
I ordered OTY30. Fingers crossed.

[zyssai]

So, I managed to replace U26 GL3523 IC.
The most difficult part was to remove it, I heated a lot and hope I did not damage something else.
Resoldering was pretty easy.

But, there was some colateral damage




No way to resolder it. I think pads are damaged, I need to reorder somes, but cannot read silk part number.

I tested anyway, So USB devices are powered, but there seems to have no data transfer (keyboard not responding, usb flash drives lights are on but no blinking,...)
Is it possible that I damaged U27? Close to where I heated a lot

[mctom]

U27 appears to be some kind of storage (EEPROM or FLASH), most probably with configuration data for GL3523. If it stops working then the USB hub will operate with default settings, which are no good if the configuration storage is present in the first place.
Could you have damaged it? Probably yes, but for now let's hope not.

It is very hard to tell what the marked component was. Almost certainly a capacitor, but which one?
Have you tried measuring the capacitance of that speck of dust?
I am guessing this is one of C244- C246, as it is close to DVDD 1.2V input pin.
There is a high chance the hub should work without it, perhaps become a little less stable. However if you damaged the trace reaching pin 47 in the chip, then it surely will not work. It should have 1.2V.

Here are some DC voltage measurements you can try to debug the issue.
All measurements to be done relative to GND.
#1-2 are L10, #3-4 are C243, which are parts of 1.2V Buck converter.
#5 is pin 47, which uses voltage from 1.2V Buck. Be very careful not to short it to neighboring pin with a probe!
#6-#7 are C256, which provides 3.3V to th memory chip.

If your multimeter can measure frequency, try measuring it across capacitor #8 - it should be 25MHz.

[zyssai]

Thanks my friend I will try.
I just took a look at the N2 datasheet and it seems to be R120, connected to pin 46.
But just don't know what means the 'F' letter after R120 20k value

[zyssai]

Here are measurements.

1: 1.21V
2: 1.16V

3: 0V
4: 1.21V

5: 1.21V

6: 0V
7: 3.28V

For 8, I'm not sure but there seems to be no frequency (0Hz).

[mctom]

Excellent job, all voltages look completely fine.
The frequency could be way too high for your multimeter, that's the most reasonable explanation.

I thought it was a capacitor because of a color, but maybe this resistor used to be black before your hot air rampage
"20k F" probably means 20k with 1% tolerance. "F" is a very oldschool way of denoting that.
These days 1% is pretty much a standard tolerance of resistors so I wouldn't worry about that.
Other resistors may be of 5% tolerance to reduce cost, but this one must be precise, so it makes perfect sense to me.

The bad news is, this resistor is required.
If you say the pads are so damaged it's impossible to solder it back, you could try buying a bigger one and solder it somehow..
One end to the chip pin, and the other to the capacitor next to it, maybe? If that capacitor is connected to GND, which you can easily check.

[zyssai]

Regarding the frequency it's possible that I am out of range with my old Metrix MX24

So yes this is a resistor, I could have told that, the picture is not clear enough.
That's the resistor pads that are damaged. The PCB ones seems fine.
I have some 1206 but not sure they will fit
I surely have some smaller ones on out of order spare boards, I'll check that today, thanks mctom for your help as always

[mctom]

Yeah, your multimiter is limited to 500kHz. But I'll be honest, I don't remember any multimeter measuring 25MHz reliably.

For electrical contacts, we usually use the nomenclature that pads are on PCB, and components have pins, even if component contacts don't look like pins. That helps avoiding confusion.

Instead of reusing a component from another board, you could try verifying it is indeed 20k resistor to make sure, and buy a new one, which will cost you shipping alone.
Get yourself 20 pieces so you can try as many times as you want.
I think this is size 0201.

If you have a sacrificial N2, why didn't you use USB hub chip from it?

[zyssai]

It was too late to verify component value, it goes to 9th dimension before I could do something with it

So I found a perfect 20k resistor on an old motherboard, I think size is 0805, couldn't fit in the right place.
So now, take a picture of this piece of art, print it, and put in a frame in your wall



I can't clean it with alcoohol because of wires. But I will put a point of hot glue to save it from bad gestures.
It works!! USB ports are all working, thanks for your precious help as always

[mctom]

@zyssai, this is incredible! Congratulations!

But please clean it! You can just splash it with acetone and leave it to dry.
Hot glue is a good idea, and another reason why this surface must be clean, so it can stick properly.

Odroid Wiki should have a gallery for pieces of art like that!

[zyssai]

It works!! USB ports are all working, thanks for your precious help as always

Congratulations!
It seems like your hardware technology has made amazing progress over the past 2-3 years.

[zyssai]

Thanks guys

I've just made a speed test with HK 5BK Wifi Module to be sure there is no data transfer problem with USB and I think results are convincing.

[mctom]

This resistor is required by the reference current generator inside the chip. With known resistance connected to known voltage, you get known current. by the power of Ohm's law.
It is very hard to make a precise resistor in silicon, so some chips require external resistor instead.

This is one of them. I want to point out that this resistor passes constant current - so its parasitic properties, capacitance and inductance, do not matter at all.
If you did the same with USB data line, I don't think this would have worked. You were both lucky and persistent, which I continuously admire.

[zyssai]

I always love to learn some useful facts like this, you know what you're talking about, it's impressive