1. OCAU Merchandise now available! Check out our 20th Anniversary Mugs, Classic Logo Shirts and much more! Discussion here.
    Dismiss Notice

5V "ideal diode" / FET switch problem...?

Discussion in 'Electronics & Electrics' started by Menthu_Rae, Feb 3, 2017.

  1. Menthu_Rae

    Menthu_Rae Member

    Joined:
    Mar 19, 2002
    Messages:
    6,874
    Location:
    Northern Beaches, Sydney
    Hey all - I need some help with a project.

    I'm powering a Raspberry Pi through the 40-pin header, this allows you to put in 5V to run the Pi, but, there is no input protection.

    So - I thought I would copy down to the exact ICs the input protection on the micro-USB connector side on the Pi.

    Here are the Raspberry Pi schematics:

    https://www.raspberrypi.org/documen.../schematics/RPI-3B-V1_2-SCHEMATIC-REDUCED.pdf

    Here's the schematic bit I copied from:


    Click to view full size!


    Here is my schematic in KiCAD:


    Click to view full size!


    Note: I am using a slightly larger polyfuse since my project is very intensive in terms of Pi power requirements... only 50mA higher though, so not a lot...

    Here's my board layout (NOT copied from the Pi):


    Click to view full size!


    with ground plane ON:


    Click to view full size!


    My issue is as soon as I apply 5V at the input and turn it on - the IC "U14" / "U0" / BCM857BS shorts out and goes "puff" in a plume of smoke and flame. :upset:

    Here's the datasheet for that part: http://www.nxp.com/documents/data_sheet/BCM857BV_BS_DS.pdf

    I've de-soldered all components and re-soldered, checked that no pins on the IC (it is TINY!) were shorting... alas nothing has fixed it from that... so it must be an error somewhere with the schematic or mapping ICs to footprints??? :confused:

    If anyone has super keen eyes or if I have missed something fundamental electrically - please do tell! :lol: :p

    EDIT: Some info on how the circuit works here:

    https://birds-are-nice.me/ipfs/QmXt1JSG1Hv3pfxeKFktJ8FgZuwd4jwUqe18GTXxjn5QTb/

    and here:

    https://electronics.stackexchange.c...e-from-a-p-channel-mosfet-and-pnp-transistors
     
    Last edited: Feb 3, 2017
  2. rainwulf

    rainwulf Member

    Joined:
    Jan 20, 2002
    Messages:
    4,227
    Location:
    bris.qld.aus
    Found another link to what you need.

    https://www.google.com.au/url?sa=i&...Yo3325wLeHdrGh1VtGyrJ9Yg&ust=1486195349483374

    looks the same as yours though.

    https://www.google.com.au/imgres?im..._PRAhWMWrwKHXC6BIAQMwgjKAgwCA&iact=mrc&uact=8

    Same thing but with less circuitry. uses a P channel mosfet. try that.

    I dont know why they need the matched PNPs.

    http://electronicdesign.com/site-fi...ctronicdesign.com/files/29/9945/figure_01.gif

    I think this is your best bet. http://www.ozqrp.com/rev_pol.html

    As for why your circuit is blowing up... i honestly cant say. There isnt enough current through the device to blow it in the proper configuration. The left most transistor in this case is acting just like a diode, the right transistor is switching though. Its turning on when pin 5 is dropped to zero which wouldn't happen unless you put 5 volts on the right.


    With 5 volts on the left, pin 6, 2 and 5 go to 5 volts. This turns OFF the rightmost transistor, which leaves pin 3 dragged low, thus turning on the mosfet.

    The only way for that to blow would be if the mosfet wasnt turning on, which even if it was the case the body diode would conduct, saving the matched pair.

    The only way for U14 to puff would be if Q3 was backwards or the wrong type. You would get reverse bias across the base emitter of the right diode potentially frying it.

    Here is a question. when U14 puffs, which transistor dies? That's if you can still test it. If its a matched pair they will be on the same silicon wafer.

    The other way U14 would puff would be if pins 6, 2 and 5 were directly on earth. Not through R2.

    rebuild the circuit, leave U14 out. Measure the current across pin 1 to 6 with a multimeter. It should be minuscule. Do the same with 4 and 3.



    btw, your board looks ok to my untrained eye.
     
    Last edited: Feb 3, 2017
  3. RobRoySyd

    RobRoySyd Member

    Joined:
    Jan 24, 2008
    Messages:
    8,204
    Location:
    Sydney
    I cannot see pin 2 of R2 connected to GND on the PCB.

    Agree with Rainwulf, pull the thing that's smoking and measure voltages.
    Often best to populate the board in stages, first with passives and check everything reads sensible values, then add actives with some plan and check.

    I don't understand the need for this circuit on the board anyway. The cct on the Pi is there in case the Pi is being powered from external power and the USB port is connected to something. If that something is unpowered then the Pi would be feeding current into it. Unless there's some other way 5V gets to the board other than through the 40 pin header then there's nothing to protect against that I can fathom.
     
    Last edited: Feb 4, 2017
  4. rainwulf

    rainwulf Member

    Joined:
    Jan 20, 2002
    Messages:
    4,227
    Location:
    bris.qld.aus
    its bloody Rainwulf, i have been on this site since Agg was a babby... hahaha

    Seriously though, R2 is connected to ground via the ground plane. I think its either 1, an issue with the matched set of PNPs, or 2, an issue with the mosfet, being either its the wrong mosfet, or its pins are wonky.

    Cant tell though until OP gets back with the troubleshooting requirements that i specified because im some kind of super cool awesome person.

    Really though, electronics becomes like lego when you spend enough time with it.

    What really helps is thinking of transistors as diodes.
     
  5. RobRoySyd

    RobRoySyd Member

    Joined:
    Jan 24, 2008
    Messages:
    8,204
    Location:
    Sydney
    At my age one vowel is as good as another :)



    Yeah well it would have helped knowing it was a 4 layer board :(
    I go back to the days of doing boards with tape. Ground and 5V planes were negative, signal planes positive. Pretty confusing outcome when someone made a batch of PCBs with the Gnd plane as a positive.

    As for the rest, sure I've looked at the transistors as diodes. Even if them diodes are carrying all the 5V to Vcc current it shouldn't be enough power to cause the fireball the OP is describing. Kind of thinking I'll bow out at this juncture. There's many possibilities that maybe the OP has already eliminated or not.
     
  6. rainwulf

    rainwulf Member

    Joined:
    Jan 20, 2002
    Messages:
    4,227
    Location:
    bris.qld.aus
    No response as of yet...
     
  7. OP
    OP
    Menthu_Rae

    Menthu_Rae Member

    Joined:
    Mar 19, 2002
    Messages:
    6,874
    Location:
    Northern Beaches, Sydney
    Sorry for the delay guys - it seems you were all anticipating a response! :D

    Appreciate the feedback and comments - so it appears there's no intrinsic issue with my PCB or schematic. :thumbup: That's good at least.

    Re: all the other items below - in short, I will need some better tools to test the relevant points in the circuit - the pads of the IC (U14) are tiiiiny.

    I will definitely try everything below though.

    Yeah - I double checked the part I ordered - it is definitely the right part number and a P-channel and the footprint / schematic definitely matches the datasheet.

    So it's a bit strange...

    The magic smoke comes out closest to Pin 1 - but I have no way of testing it now unfortunately :upset:

    I am going to need some finer multimeter tips to do that measurement. I have some on order should hopefully have them later this week. :thumbup: I'll let you know how it goes.

    That's good at least! :p

    It's connected to the ground plane on the front layer - and both front and rear ground planes are linked at various points with vias or PTH components.

    Yeah for sure - I will go through this process once some replacement ICs arrive.

    Haha that is true I guess - but with this application it is entirely possible someone could plug in the USB connector at the same time... so it'd be nice to just keep it identical to the reference design.

    It's a two layer board :) Blue is front, red is back.

    Re: the current - even with no load connected - it still wants to short out U14/go puff.
     
  8. rainwulf

    rainwulf Member

    Joined:
    Jan 20, 2002
    Messages:
    4,227
    Location:
    bris.qld.aus
    Yea, and if the ic is correctly wired, thats impossible to happen. There simply isnt enough current through the two resistors to cause the magic smoke.

    Will await the multimeter tests eagerly!
     
  9. OP
    OP
    Menthu_Rae

    Menthu_Rae Member

    Joined:
    Mar 19, 2002
    Messages:
    6,874
    Location:
    Northern Beaches, Sydney
    I did some quick tests yesterday and got some new components from the courier today (DigiKey is FAST! Ordered 5pm Friday, in my hands 12pm Tuesday) :thumbup:

    Anyway from those quick tests I actually believe there is a manufacturing issue with the board.

    With all components removed from the board - I am measuring ~0.1-0.2 ohms across the pads of R2 that should be open-circuit.

    This would explain why U0 (U14) is dying - because R2 is essentially a short to ground.

    Testing the same pads on 2 of the other boards (they give them to you in sets of 3) from the fab house yields open-circuit as expected...

    Will need to investigate further but that is my hypothesis at the moment.

    Annoyingly another (very VERY large [and expensive]) board I got from the fab house also seems to have the same problem... how strange... they were ordered just before Christmas time so I wonder if it's a "meh EOY" QA issue... :rolleyes:
     
  10. rainwulf

    rainwulf Member

    Joined:
    Jan 20, 2002
    Messages:
    4,227
    Location:
    bris.qld.aus
    Well there you go!!! That's where your current is coming from!
     
  11. RobRoySyd

    RobRoySyd Member

    Joined:
    Jan 24, 2008
    Messages:
    8,204
    Location:
    Sydney
    I have dim memories of a lot of craft involved in PCB design for manufacture.
    I'd add your own goods inwards QA Inspector should have picked up the issue. :)
     
  12. OP
    OP
    Menthu_Rae

    Menthu_Rae Member

    Joined:
    Mar 19, 2002
    Messages:
    6,874
    Location:
    Northern Beaches, Sydney
    OK can confirm it's a board manufacturing issue. I ordered more of the BCM857BS chips (since all mine had fried) and moved all the other components across to a new board (exact same design, my fab house provides 3 boards per order).

    Powered it on... all good! No issues at all. Giving it 5.05V at the input and getting 5.01V at the output. :thumbup:


    Click to view full size!


    So yeah... time to see if the next lot of boards ordered have the same manufacturing defect rate :mad: Pretty disappointing as I haven't had any issues with this fab house before, but I am hoping it comes down to being Christmas time/holiday syndrome. :D
     
    Last edited: Feb 22, 2017
  13. dakiller

    dakiller (Oscillating & Impeding)

    Joined:
    Jun 27, 2001
    Messages:
    8,115
    Location:
    3844
    It is odd to get a short in this situation. The other end of R2 doesn't go very far, you should be able to see where it is shorted.

    Are you pushing the limits of your fab's tolerances?

    3 boards per design, they wouldn't happen to be purple boards?
     
  14. OP
    OP
    Menthu_Rae

    Menthu_Rae Member

    Joined:
    Mar 19, 2002
    Messages:
    6,874
    Location:
    Northern Beaches, Sydney
    Yeah they are perfect purple PCBs :lol:

    Should be well outside of their tolerances - everything has more clearance, larger sizing (vias/tracks) and I am pretty conservative with my routing.

    Now I know where the problem is I might power the other board up and use my thermal camera to see if I can find the short. :thumbup:

    Do you guys have any fab houses that you can recommend preferably with online web uploads + instant pricing/previews?
     
  15. dakiller

    dakiller (Oscillating & Impeding)

    Joined:
    Jun 27, 2001
    Messages:
    8,115
    Location:
    3844
    I've used oshpark quite a bit, never had much problems. They do have multiple sources that they use to do the actual fabbing of them, maybe you just got the rare bad case here.
     
  16. paulvk

    paulvk Member

    Joined:
    Mar 10, 2010
    Messages:
    1,063
    I use KiCad and used Firstpcb to make boards, the first 10 are free (100 x 100mm) as you get a $10 bonus just have to pay for shipping which is about $40 depending on exchange rate but they arrive in about 7 days by courier.
    They were double sided with through hole and surface mount on both sides of the board also had heat sinking on both sides with plated through holes for heat transfer.
    As I did not use the full 100 x 100mm I put two more boards on them which I cut off when they arrived so I actually got 30 boards.
    The main board is an 1284P AVR with Arduino sockets, switching regulator, real time clock, rs232 buffer and 5 headers the two small boards one is a link on batteries for current sensor the other is another AVR in 44 pin PLCC with rs232 buffer and 4 headers.
    So a lot of tracks and through vias.
    Very happy with them!
     
  17. rainwulf

    rainwulf Member

    Joined:
    Jan 20, 2002
    Messages:
    4,227
    Location:
    bris.qld.aus
    Can you see the short?
     
  18. OP
    OP
    Menthu_Rae

    Menthu_Rae Member

    Joined:
    Mar 19, 2002
    Messages:
    6,874
    Location:
    Northern Beaches, Sydney
    Yeah - barely! Thermal camera wasn't a super great help (though it did help me figure out it was directly on the pad). Checking with a USB microscope there is a very tiny link between left pad of R2 and the ground plane:


    Click to view full size!


    On another board - with no issue - all good:


    Click to view full size!


    Mr x-acto to the rescue! :Pirate:


    Click to view full size!


    Beep tested after that and no more problems... Anyway I've asked for store credit from the fab house so let's see what happens. :thumbup:
     
  19. aXLe

    aXLe Member

    Joined:
    Jun 27, 2001
    Messages:
    1,214
    Location:
    Brisbane
    I can recommend Hackvana (hackvana.com) for pcb's - it's run by Mitch Davis in Melbourne (boards are made in China) - prices are good (better than OSHpark) and quality is excellent. Ordering is via email though - just email him gerbers and he gets a no obligation quote from the factory and you can pay via Paypal or CC or DD. The factory can also do stainless steel stencils if you need one.

    Oh - and they e-test the boards (may have found your issue).
     

Share This Page

Advertisement: