Simple tests for fake SN65HVD230 CAN tranceivers

[e100]

Datasheet https://www.ti.com/lit/ds/symlink/sn65hvd232.pdf

I bought some 3.3v CAN transceiver breakout boards from Aliexpress and so far I've been unable to make sense of the behavior.
I'm driving one from an ESP32 and there are no wires connected to CANL or CANH.

(Not using the CAN controller) Putting in a 100 kHz square wave into the Data pin gave a weak looking square wave with 1v differential on the CANL/CANH pins, but I only saw that once for a few seconds. The rest of the time CANL/CANH are firmly anchored at zero volts and show no hint of wanting to be at the expected 2.3v recessive voltage.
Holding CANL to ground and putting 3.3v  into CANH didn't move the Receive pin from it's high state.
The Rs pin is at about 0.7v which should be low enough to keep it out of sleep mode.

Are there any other tests I can do?

[rooppoorali]

If you do not get responses here, you can post your problem to the Texas instruments forum. I have seen that the TI officials are quite active in replying.

[AndyC_772]

Is your CAN bus properly terminated?

[e100]

Is your CAN bus properly terminated?

Yes the breakout board has a 120 Ohm resistor linking CANH and CANL

I tried another board in isolation and powered it with 5v and CANH/CANL rose to about 2.8 v, so perhaps these are relabeled TJA1050 chips.

[e100]

I tested another breakout board I purchased last year and it works at 3.3v, although the signal levels are strange.

CANH and CANL idle at 1.5v. The datasheets typically say this should be around 2.3v (to help with compatibility with 5v transceivers).

In the dominant state the CANL and CANH movements are asymmetric.
CANL drops to 1v and CANH rises to 2.3v.

[AndyC_772]

Logically those levels are fine, it's CANH-CANL that matters, not the absolute voltage of each wire individually. I've seen this behaviour before, where CANH-CANL = 0 during recessive periods but the absolute levels drift as per your scope trace; it's not a problem.

What's connected to Rs? Is it connected to GND, or pulled to GND with a well defined and valid resistance?

[e100]

Logically those levels are fine, it's CANH-CANL that matters, not the absolute voltage of each wire individually. I've seen this behaviour before, where CANH-CANL = 0 during recessive periods but the absolute levels drift as per your scope trace; it's not a problem.

What's connected to Rs? Is it connected to GND, or pulled to GND with a well defined and valid resistance?

I measured 10k to ground on the non-working board and 6.8k to ground on the older working board. The data sheet says to use 10k for medium speed.

[AndyC_772]

That's interesting, then; you have something that measures differently between a good vs a bad board.

I suggest swapping the CAN chips between the two, and see if the fault moves with the chip or stays with the board.

I wouldn't be surprised to find the chip is dead, given what demand is like for CAN transceivers right now, and the junk that's likely to have entered the supply chain as a result.