My customer has started seeing problems with a batch of boards.
I have an opto coupler, MOCD217 because it is cheap, led anode to +12V cathode to 10k. The 10k is pulled down by an open collector elsewhere in the system.
The opto transistor is pulled up to +5V by 36k. The collector goes straight into an input of the FT3150 processor.
This all works and has done for years. 18 months ago the customer started using this a pulse count input. Now he is getting miscounting. It appears to be due to the slow rising edge due to the
36k pull up. The scope shows about 800us rise time which appears monotonic and has less than 50mV of noise on the signal. The processor is specified as TTL levels, 0.8V and 2V, with minimum 175mV of hysteresis.The problem seems to tie in with a change of batch of the MOCD217, 07 to 09 date codes.
Changing the resistors around the opto I got much faster rise times and greatly reduced miscounting.
As far as I can see with 175mV hysteresis on the processor input it shouldn't matter if the rise time was a fortnight or 10us.
My concern is I can use a "trick" (copyright CRU) and make it work. But as yet I don't understand why it should make a difference unless of course the data sheet is wrong and the processor does not have hysteresis on the inputs.