Re: Pseudo-ECL output models

From: Stephen Peters <sjpeters@ichips.intel.com>
Date: Mon Jan 12 1998 - 10:39:28 PST

Hello Prasad:

    Interesting circuit you have there! I assume that the 'current source' you
show at the output nodes is used to extablish a current flow thru the
termination
resistors even if both of the upper xsistors are off. The upper xsistor then
supply current to change the logic level -- correct? Anyway, my answers are
below, prefefixed with >>>.

                  Regards,
                  Stephen Peters
                  Intel Corp.

>
> Hi, all.
>
> For a communications serializer-deserializer high speed(1Gb/s) serial
> output IC, with the following configuration,(ignore the *star* line till
> you get to question 3)
>
>
> *
> |------|-----*---------------------
> > > * | |
> < < * | |/
> | |___C_*______|___________|
> | | * | |\
> | | * | |
> | | * |/ A____
> |______|___D_*_____| | |
> | | * |\ | <
> | | * | | >
> |/ \| * | | |
> -----| | * | | GND
> | | * B____________|________
> |\ /| * Current Current |
> | | * source Source |
> --------- * | | <
> | * | | >
> | * | | |
> Current * GND GND GND
> Source *
> |
> |
> GND
>
> Where A and B are the IC buffer boundary and the two resistors to
> ground are 150 ohms each from A and B,
>
> I am trying to generate an IBIS model. I have the following questions-
>
> 1. For the pull-up and pull-down tables I get negative currents. I
> understand that for standard ECL outputs,currents are positive. Will this
> cause any simulation difficulties? ( For simulation, the buffer shown above
> is ac coupled to a transmission line and terminated differentially in 2*Z0)
>>> Actually, for standard ECL outputs the current direction is
>>> 'negative' (assuming positive current flows from + to -
>>> (hole flow)). I don't think that should cause simulators
>>> difficulty, as long as all tables are consistent.

>>> By the way, could you please explain what you mean by 'ac
coupling'
>>> to a transmission line? To me, ac coupling means putting a
>>> capacitor is series with the output to block DC current and
>>> voltage -- which doesn't make much sense. Please enlighten.
>
> 2. For generating the IBIS pullup/pulldown table, I bias the buffer in such
> a way that one of the outputs is "high"/"low" and the other is
> "low"/"high", with the 150 ohm resistors in place. Then on the output that
> is "high" a voltage is swept in steps while monitoring the current. Is this
> the correct way of generating the IBIS table?( Note that one leg of the
> differential outputs has only the 150 ohm resistor to ground while a
> voltage is connected to the other leg and is being swept while currents are
> being measured)
>>> From what you have described, it sounds correct. I emphasis
that
>>> the leg you apply the voltage to and measure the current on
does
>>> NOT have the 150 termination resistor. Also, remember to
remove
>>> any package model -- you want just the DC output
characteristics
>>> of the output node itself.
>
> 3. For an output buffer structure that is shown on the left hand side of
> the vertical *star* line, with C and D as the differential output
> terminals, has any one come up with an IBIS model? This output structure
> is used for high speed buffers and the traditional way of generating the
> IBIS tables do not work.
> (This kind of structure is very useful in that in provides back termination
> for transmission lines without any external resistors and the biasing of
> the internal transistors is automatic) Again, typically, this buffer is ac
> coupled to a transmission line that is terminated differentially in 2*Z0
> for simulation.
>>> Fundementally, the C and D outputs are standard
open-collector
>>> outputs, which the IBIS spec will handle just fine. The only
>>> trick is that the two ouputs have a 'differential'
relationship
>>> but again, you should be able to specify that in IBIS.
>
>
> Answers, suggestions and questions are welcome. If this information exists
> somewhere please direct me to it. (I did not find it in the cookbook)
>
> Thanks.

 
Received on Mon Jan 12 10:48:51 1998

This archive was generated by hypermail 2.1.8 : Fri Jun 03 2011 - 09:53:46 PDT