Ultra96 Safety missing

Products Support Ultra96
#1

Hi,

I am looking for a hardware engineer who can help me get the ultra96 board up and running again. I accidentally plugged in the mezzanine board with unaligned LS connector. The board was switched off but was plugged-in. Which resulted into the VCC_5V0 and VSYS_IN input get shorted because of the GND pins very next to them. Ultra96 have a missing design safety feature which could lead to this problem very easily. And I have seen they have rectified it in version 2 with a schottky diode which is good.

VCC.PNG
VCC.PNG1446Ă—494 47.7 KB

My project is in middle and boards are out of stock at avnet. The only options available are either to fix my current board. Or to find someone selling it.

If you know anyone who can fix it please let me know. Or anyone who is selling it, in the USA please let me know. I am in urgent need. can’t wait till June for version 2 to arrive.

Thank you

#2

I’m curious, how the schottky diode in v2 prevents the problem you had?
Btw, did you solved it?

#3

If you see the attached photo diode would prevent the flow of current towards the pin. Which would not lead to the 5VDc plane.

Screenshot_20200223-235927
Screenshot_20200223-2359271080Ă—1920 222 KB

#4

Yeah, I smoked a Dragonboard 410c a few years ago though misalignment. Had a new baby in the house, and was apparently not paying enough attention to what I was doing. Smoke came out of the PMIC.

Now about that diode; take a peek at pages 8 and 9 of the CE specifications;

Power shall be provided to the board by one (and only one) of the following:
...
2. An 8V to 18V power supply from the SYS_DCIN pins on the low speed expansion
connector.
...
The board shall be able to provide the following power to external devices when a
sufficiently rated power supply is connected to the DC Jack:
1. A minimum of 7W to a mezzanine module via the SYS_DCIN line, and
...

To translate; the SYS_DCIN pins on the low speed connector are required to be usable as EITHER power input OR OUTPUT. Which means that that diode violates the specifications.

Further, there are more opportunities to smoke parts by misaligning the pins; for instance, if you misalign in the opposite direction from what you did, you would dump 5v to the 1.8v on the mezzanine, which would kill everything touching the 1.8v. You could also dump SYS_DCIN into GPIO at pin 34, which would certainly kill everything.

The bottom line is simply this; make sure the pins are correctly aligned before powering anything.

1 Like
#5

The diode was provided as a safety measure. There is always additional risk in a system when you allow multiple input voltage sources. We added the diode to protect the most common way of using a mezzanine, which from what we have seen is to plug power into the Mezzanine and have the mezzanine supply power to the baseboard.

You may still deliver power from the Ultra96-V2 to the Mezzanine, but it requires that you desolder the diode and short those pads. A user would have to conscientiously do this in order for it to work, at which point they should likewise increase their caution of the misalignment and board damage issue.

Bryan

1 Like
#6

Well I personally don’t agree with the idea of forcing the user to solder on surface mount components in order to get a product to work according to specifications it is supposed to comply with. I also question the validity of suggesting that a user will be more aware of alignment if they have to do soldering. Every time you give a user an opportunity to solder something, all you really do is give them an opportunity to screw it up.

If it were me, in the VERY LEAST, I’d insist on having a switch in place to bypass that diode.

#7

I appreciate the feedback. There are always design choices to be made. On one hand, the specification requires multiple locations for input DC power. That same specification also states, “If multiple in-specification supplies are connected there shall not be a safety issue and there shall be no damage to the board.” We chose to focus on the “no damage to board” part, even if that meant a little inconvenience for what we thought were a small number of users. Perhaps we didn’t have all the information though, and we welcome the feedback.

Out of curiosity, which mezzanine are you using? I’d like to keep track of which ones do not have the ability to supply power through the LS expansion.

Bryan

#8

There are lots of mezzanine boards that do not have the ability to supply power through the LS expansion! What matters though isn’t whether they supply power, it is whether they consume power from SYS_DCIN (which they are entitled to do)… many mezzanines simply ignore it because the prefer the nice clean 5v line instead :wink:

#9

Exactly.

And it’s about complying with the spec rather than trying to be compatible with specific boards.

One of the boards I designed myself needs to be able to either supply the SBC from the mezzanine (normal use), or be able to supply the mezzanine via power from the SBC (under bench test). Which means that my board requires full compliance with spec.

Besides which, that line in the spec about “no damage to the board” is really more about having sufficiently wide traces between the barrel jack and the LS header to avoid burning it off. The issue the diode deals with is misalignment.

Oh, here is another question for you with regards to spec compliance: what happens if you put an 8v power supply with a tvs diode or other similar functioning over voltage protection on the sbc’s barrel jack and an 18v supply on the mezzanine? They’re both in-spec supplies that should be able to be hooked up without causing damage according to the letter of the specifications.

#10

If you’re offering to make a suggestion on what we should do differently in the future, please feel free. The “no damage to the board” came in the power supply section, not the mezzanine alignment. I think it’s fair to say that it is open to interpretation, but our interpretation was to protect against the scenario where supplies are plugged in at both the host and mezzanine. That in turn caused the conflict between 1) supplying the system from either host or mezzanine while 2) eliminating the possibility of damage to the board during a simultaneous plug-in. We felt the current design was a reasonable compromise. You are welcome to disagree and provide a different suggestion for consideration in the future.
Bryan

#11

Since you seem to be having a hard time understanding what I am suggesting, I’ll be blunt: My suggestion is to remove the diode. It isn’t a compromise, its a violation of specifications.

#12

What you see as my lack of understanding is me disagreeing on your interpretation of the spec. I’d be interested to know other people’s opinions as you have clearly stated yours multiple times now.

#13

It isn’t a question of opinion or agreement.
The spec has also been explained in this thread by @danielt who you have apparently chosen to ignore.

The specifications explain very clearly that the SYS_DCIN pins on the low speed connector are available BOTH for the purpose of powering the SBC from the mezzanine, AS WELL as powering the mezzanine from the SBC. There is no room for interpretation.

#14

I’ve not really formed an opinion yet, in part because I’m not even sure that the spec is implementable: 8v on the barrel jack and 18v via the LS connector without a safety issue and 18v on the barrel jack and 8v via the LS connector, also without satefy issue. Since in both cases there is little constraint on the power supply specificiation (current delivery, internal resistance, etc, etc, etc) I don’t quite see how the baseboard could ensure safety.

If I were asked (right now this is unofficial rambling :wink: ) I think I would recommend resolving the ambiguity in the spec by making it explicit that the barrel jack and LS connector can be on a common power rail (since AFAIK that is what most boards already do). I think the most important safety issue is that having an 18v supply on either the barrel or LS connector must not damage a built-in USB type C power supply.

#15

One of these or similar on the SBC between the LS header and the barrel jack should do the job of ensuring safety;
https://www.littelfuse.com/products/polyswitch-resettable-ptcs/surface-mount.aspx

And perhaps another one at the USB-C power supply, if applicable.

#16

But, isn’t shorting +5V in itself a problem? That is not solved by the diode…