during testing of the ADM3053 as a replacement for our current implementation of the CAN interface, I came across with a serious problem. Also a FAE of AD could not help me with this problem. So now I try this way to contact the developers of AD.
In our testing we noticed that the ADM3053 exposed sporadically. The symptom is that a CAN message is not properly confirmed with an ACK by the ADM3053. The CAN controller behind the ADM3053 tries to send a dominant level, but the ADM3053 stays recessive on the CAN bus. An Active Error Frame is the consequence and the transmitter repeats the CAN message. In the second try, the level dominate for the ACK is correctly interpreted by the ADM3053 and sent on the CAN bus.
In other words: The ADM3053 does not reflect the logic level at his input as the correct differential signal at his output! The occurrence of this bahaviour varies from one error per hour up to one error per a few seconds.
The bad thing is that this behavior occurs on two independently developed hardware designs. In addition, we observed the same behavior on the ADM3053 evaluation board. We have tested several batches of this chip. All have showed the same problem.
CAN baud rate = 1 Mbps
Vcc = 5V
Vio = 3.3V
Rs = 0 Ohms
TX input is fed by a push pull output of the CAN controller. So clean logic LOW and HIGH levels are given.
The supply voltages are stable and do not sag down on failure. We have triggered our oscilloscope for an active error frame conditions, because the "lost" ACK will result in this error frame. We could not trigger a lost bit in a datastream through the ADM3053, but in my opinion if the ACK bit is lost, then also every other bit can get lost. Lost mean, that a logic LOW (must be a dominate level on CAN bus) will result in a recessive level on the CAN bus.
We would like to use the ADM3053 in our designs because of the built-in galvanic isolation and voltage supply. Since we are working in the field of industrial automation, we can not tolerate such a fault because of safty issues.
I can give more technical information, if someone answer this post.