Re: resend the question. (The Criteria of overshoot, undershoot, RingBack for different logic types

Andy Ingraham ()
Wed, 9 Jul 97 08:29:31 EDT

JohnLin wrote:

> I got many responses from people. However, I could not read them due to
> the
> different mailing systems, I thought.
> Please use my FULLY EMAIL ADDRESS , [email protected], for replying
> so that I could read your mails.

I don't understand. I think that if their mail got to you at
all, it should have the message they sent to you. If you have
trouble reading a message, I think it must be due to some other
reason, something wrong on your end.

In what way are they not readable?

By the way, your mail to the maillist comes from
[email protected], and I assume that is what people are
sending their replies to. I did not see any other address they
could have been sending them to. (I assume the capital letters
don't matter?)

Hitting "reply" should work just fine! If you have hard evidence
that it doesn't, I'd like to hear about it.

> Talking about signal quality. Could somebody tell me about the criteria
> of overshoot, undershoot, overshoot ring back, and
> undershoot ring back for TTL(5V) , CMOS(5V), LVTTL(3.3V) logic types?
> Intel has provided the GTL+ (Vtt=1.5V, Vref=1.0) criterion for signal
> quality.
> over/under shoot =0.2*(Vtt) ,
> Ringback= Vref+/-200mv.

In my experience, these criteria are unique to each IC

You may find SOME of this information in their data sheets, in
the form of DC specs, but this does not take into account the
pulsewidth or the energy in the overshoot/undershoot/ringback.
For example, the minimum input voltage spec may be -0.5V, meaning
that you should not apply a constant DC voltage lower than this.
But an undershoot usually may exceed this briefly.

It generally boils down to the energy in the undershoot,
overshoot, or ringback; the integral of the v(t) curve. A vendor
may choose to put this in the form of a limit that the voltage
may not go below (say) -1.5V for more than 10ns.

The same thing is true of ringback. TTL and LVTTL both have
Vinl(max)=0.8V, Vinh(min)=2.0V. Theoretically, ringback should
not cross these voltages. But narrow pulses that cross these
levels might not affect the inputs, depending on its magnitude,
pulsewidth, and/or area. This is a bit riskier, however. A new
process spin may result in faster inputs that do respond to
pulses previously ignored.

Generally you need to contact each IC vendor and ask them
specifically what is acceptable. It may take some pushing to get
an acceptable answer, and some vendors may not give you one.

Andy Ingraham