Post Go back to editing

ADUM2250 problems

Hello,

I'm an intern at Philips in Aachen Germany. I received a PCB a month ago to troubleshoot/repair. The PCB has major design errors, I was able to repair most of them so far. 8 ADUM2250 are on board. By reading the datasheet, I noticed that the sides of the ADUM2250 differ from one another, i.e. side 1 is I2C compatible while side 2 is I2C compliant. The ADUMs on board are connected in such way that side 2 is the input and side 1 the output. How critical is this?

I managed to connect one of Isolators as described in the datasheet and cut the I2C from the rest, butthe I2C signals coming from a microcontroller seem to get distorted once I connect the I2C bus to the ADUM. The following images should explain what I mean:

I2C SDA when MCU not connected to ADUM

I2C SDA when MCU is connected to ADUM:

ADUM's output SDA looks even worse than the input! Any idea why this is happening?

  • Hello hhachem,

    The ADuM2250 is actually two pairs of channels which are cross coupled.  This unfortunately makes a crude flip-Flop that would catch the first low coming through and stop working.  So we employ a scheme used in some NXP I2C buffers that relies on having a 0.9V VOL on side 1.  This is what you are seeing on the scope, is the part holding the low input from side 2 until the correct state passes through the isolation barrier.  I made a video that discusses this in detail at the link below.

    http://videos.analog.com/video/products/digital-isolators/2013357472001/How-I2C-Isolators-Work/

    If you have any questions after looking at the video, let me know.

    MSCantrell

  • Hello,

    Thanks for the reply. That makes sense. I was unable to read the datasheet in detail due to the time pressure. In the video, you explain, that when side 1 communicates with side 2, the I2C on side 1 looks weird but it should look normal on side 2. However, this is not the case for me, side 2 looks even weirder. At a 0 logic level, the signal oscillates between .4 and .9V . I will post some images on Monday, once I go to work.

    Hachem

  • That is odd.  Please send a schematic of how you are using the part, including the power supply connections, pull-up values, and other parts on the busses.

    Is it possible that you have an I2C buffer or bus extender on the side 2 bus?

    MSCantrell

  • My I2C bus is connected to multiple ADUMs. The power is supplied by isolated DC-DC converters. I will try to draw a schematic from memory tomorrow (as I said, I did not design the PCB).

    The outputs of the ADUMs are connected to I2C compatible ADCs.

    Pull up resistor: 4.7k

    Pull up voltage : 5V

    Hachem

  • Hello Hachem,

    You can only have a single side 1 ADuM125x connected to a bus segment.  If you tie more than one side 1 to a bus segment they will not be able to communicate with eachother.  There are also NXP buffers that have the same restriction.

    MSCantrell

  • Hello again,

    Well that will be shocking news for my boss, as he wanted each ADC's I2C bus to be isolated. So if I understand you correctly, the following does not work:

    ,

    instead, the following should be done:

    I surely hope there's something written about this in the datasheet, so I can use it as an argument when I confront the original designer of the PCB.

    Thank you for your help

    Hachem

    EDIT: here's what I get , when the ADUMs are connected as shown in the 1st picture above:

    1 is side 1 and 2 is side 2. Note that the 1st ADUMs: I2C-> side 1 -> side 2 -> ADC

    The rest: I2C -> side 2 -> side 1 -> ADC

  • Hello Hachem,

    Both of your layouts are possible, you just need to be aware of the orientation of the couplers.  Side 1 can not share a bus segment, so you need to make sure that the common bus segment only links the side 2's of the ADuM2250. Actually a single part could be turned around, but no more than one.

    In the second configuration, the multidrop network can have a fairly large capacitance so it is preferable to point side 2 at the ADC's.  But this one could work either way.

    From your description of the scope screen capture, as I understand it, you have one side 1 and multiple side 2's along with the MCU on the blue segment.  The trace does not look correct. The only time this bus segment looks like it is going high, it is ramping very slowly like there is a large capacitive load.  That may be because the pullups are sized for a side 1 driver, and it is making the bus run slowly.

    I would start by removing the component that has side 1 attached to the MCU bus, and see if it starts working.

    Best Regards,

    MSCantrell

  • Thanks for the reply. I disconnected the above mentioned component, now I get the following:

    So it is obvious, somehow side 1 of the ADUMs is affecting the bus, as the small edges amplitude are multiples of 0.9V. (Notice that the bus is supposed to stay high at the end, due to an I2C-NACK!)

    Also here is the schematic:

    Please note that the pull up resistors are present in the ADC & MCU schematics (i.e. they cannot be seen in the above schematic).

    Regards,

    Hachem

  • Hello,

    At this point, I think we have more than one issue to work through, and it might be faster if you could contact me directly.  I have sent a privare e-mail with my contact information.

    I am sure we can get your issues sorted out.

    Best regards

    MSCantrell

  • Hello,

    actually i have the same problem with this weird step in the beginning of every cycle like hhachem told in the beginning. Have you found a solution for this in the meantime?

    greetings,

    Daniel