Post Go back to editing

ADUM1281 gives wrong output

Category: Hardware
Product Number: ADUM1281

We use a ADUM1281 to isolate a CAN bus driver. The driver chip is a 82C251T. It usualy works fine, but at 1Mbit/s about once pr minute we have an error frame.
We have (with some difficulties) been able to catch the situation on an osciloscope. The magenta trace is the input to the isolator, and the yellow trace is the output from the same isolator channel. The yellow trace rises correctly about 30nS after the input. Then the input has a big ringing with the voltage dropping to ca 2V before rising again. 20nS after that, the output goes low. But it does not go high again. For the rest of the pulse, the input is high and the output low, which is clearly wrong. (The cyan trace is the TX signal and is not relevant here). Common mode voltage has been monitored, but is quite silent.

This looks for me like an internal hardware problem in the ADUM1281 isolator that is not documented in the datasheet. Both version A and C seems to have the same problem.

Parents Reply Children
  • Hi jnaso

    The 200nS delay is between the  CPU-TX pin and the CPU-RX pin. And what looks like a long delay is actualy data from the other end of the bus. This is the acknowledge pulse, where all nodes send acknowledge to the current message. The local node terminates its acknowledge pulse 200nS earlier that the node(s) on the far end. This is normal behaviour.

    The VDD2 is 3.3V, and is reasonably stable. There is a 100nF capacitor close to the VDD2 pin, not shown on the schematic. And also close to VDD1. This is a large system with a lot of nodes and close to maximum bus length. It is not a prototype, but a production system. It works ok with optocouplers (HCPL-0720), but not with the ADUM1281 isolators.

    Has anybody else observed this kind of glitch?

  • The ADuM1281 is an edge-encoded architecture. That means when the isolator's input detects an edge transition (high->low) or (low->high) the transmitter sends a series of pulses across the barrier. The decoder on the output side needs a certain period of time between series of pulses so it can properly interpret what it is receiving. There is a glitch filter built into the input side to prevent edges coming in too quickly. 

    It seems the input signal looks like a short high pulse followed by a glitch to the isolator. If the system's bit width was longer we'd see the output state correct, but the correction takes about 1us. 

    You could switch to a different isolator architecture. On-Off-Keying is level encoded, which means it doesn't require the glitch filter. Both the ADuM121N1 and ADuM321N1 are pin-compatible with the ADuM1281. Those parts would eliminate the prolonged low state. However, if TXD gets low enough to trigger the VIL, both parts will reproduce the input signal as a low on the output side. It may be best to work on cleaning up the input signal before changing parts. 

    Regards,

    Jason