Forum Discussion

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

Driving terminated controlled impedance trace single ended

Hi,

I'm sure this comes up a lot but I couldn't find an exact match to my query.

Consider a single ended 3.3V LVTTL Cyclone IV output, driving into a 50 ohm trace with far end parallel termination. Say we are driving at 150MHz with a rise time of 1nS.

Given that the Cyclone IV can't source or sink more than 4mA, is this not a problem? 3.3V/50 ohms = 66mA

Is it only valid to drive traces at high speeds that are so short that they don't need a far end termination?

I read (http://www.altera.com/technology/signal/board-design-guidelines/sgl-bdg-index.html (http://www.altera.com/technology/signal/board-design-guidelines/sgl-bdg-index.html)) that if the trace is longer than the wavelength/10 then a termination is required. For a 1nS edge this would be about 1.6".

Regards,

6 Replies

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

    Yes, this is a problem. You can't drive a 50 ohm doubly terminated trace with an LVTTL output. You need 3.3V/100 = 33mA and you only have 4mA.

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

    Oh, and the other problem is that you end up with only ~3.3V/2 signal amplitude at the receiving end.

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

    You can use series termination (20 to 50 ohms in series with the driver output) to gain some extra trace length. The resistor damps the reflected wave when it returns to the source.

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

    The resistor won’t damp the signal when it comes back in my opinion but you can try using parallel resistance. The current signal you have is far less than what’s required, I doubt it’ll work even after putting parallel resistances.

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

    --- Quote Start ---

    The resistor won’t damp the signal when it comes back in my opinion but you can try using parallel resistance.

    --- Quote End ---

    I don't believe that you have seriously studied transmission line termination schemes. It turns out, that for termination of I/O standards like LVTTL or CMOS, that require to achieve a full voltage swing at the receiver side, series source termination as suggested by gj_leeson is the perfect solution for most point-to-point connections. With correctly adjusted series resistor, it achieves almost reflection free waveforms. A small, usually neglectable deviation from ideal behaviour is caused by the input capacitance of receiver.

    It should be also noted, that the 4 mA I/O specification does not mean absolute maximum current. It's the maximum static current to keep the specified output level of the respective IO-standard. In so far it applies to parallel load R termination, but not e.g. to RC load termination or series R termination.

    As the main disadvantage of series source termination, it's not suited for tapped transmission lines, because the intended square waveform is only achieved at the transmission line end.