Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
15 years ago

Double Rate Error in Arria II GX

I am running several, identical, parallel but asynchronous processes in the Arria II GX. Each process is clocked by an external oscillator, going in on pins defined as clock pins. These processes are all coded to latch on the rising edge of the clock, but some of these processes are occasionally (about once in a million cycles) triggering off the falling edge of the clocks giving double rate for one cycle. This screws up synchronisation of the outputs with external devices causing errors.

The error must be happening close to the clock input as the entire process is affected. I have confirmed with an oscilloscope that the external oscillators are stable.

Has anyone experienced this kind of problem? Any suggestions how the clock might be being affected once inside the chip? Any solutions?

Thanks!

1 Reply

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    15 years ago

    I have found a solution to my own problem.

    The issue seems to have been slow slew rate on the falling edge of the external oscillator signal. With a slow falling edge the voltage spends a relatively long time close to the threshold level of the relatively high speed input pin of the FPGA. In this sensitive crossing period, noise (in this case crosstalk) is able to toggle the threshold of the logic giving a very short pulse as a glitch with additional rising edge where only an isolated falling edge should have been. Given the speed of the Arria this glitch could easily be sub nanosecond in duration and therefore not visible to my oscilloscope probe when I brought the oscillator signal straight out on another pin of the FPGA.

    What I have done is remove some filtering I had on the oscillator output. This has had the effect of reduce the slew time from 2.1ns/V to 1.2ns/V. The faster edges now seem to be immune to the level of cross talk which had been affecting them and there are no more false triggers off the falling edge.