Re: [SI-LIST] : LVCML vs. SLVS

About this list Date view Thread view Subject view Author view

From: D. C. Sessions ([email protected])
Date: Sat May 12 2001 - 06:04:09 PDT


On Thursday 10 May 2001 08:13, Franz Smith wrote:
> Dear D.C.,
> thanks for your comments.
> I saw you didn't add any comment on LVCML, then ...
>
> Vcc(ECL)=0V ; Vee (ECL)=-5.2V
>
> Vcc(CML) = Vcc(PECL) = 5.0V; Vee=0V
>
> Vcc(LV-CML) = Vcc(LV-PECL) = 3.3V; Vee=0V
>
> at least these are the most common definitions I have found!
>
> CML or LV-CML (3.3V) interfaces are mainly used to connect a SerDes
> block with the optical driver/receiver, and there are several
> components on the market using it (MAXIM for istance).

Oh, I'm rather more familiar with LVPECL than I'd like to be, having
designed a few LVPECL I/Os. Nasty stuff if you're concerned about
speed or power.

> From: "D. C. Sessions" <[email protected]>
> >Reply-To: "D. C. Sessions" <[email protected]>
> >To: [email protected]
> >Subject: Re: [SI-LIST] : LVCML vs. SLVS
> >Date: Fri, 4 May 2001 08:25:53 -0700
> >
> >I'm the author of SLVS, and chair of JEDEC's JC-16 committee
> >which is considering it as a standard.
> >
> Is the SLVS-400 a standard or not yet?
> If not, when do you expect a final resolution?

SLVS-400 has not yet been adopted as a standard by JEDEC.
Even as the committee chair, I really shouldn't be saying much more
than that, except that personally I'm quite confident that it will be
adopted by the end of the year.

> >That said, I'm not sure that I'd advise using either of these for a cabled
> >connection. Both have quite narrow common-mode ranges and depend
> >on high-quality ground integrity between the driver and the receiver.
> >Which is Not Guaranteed in a cabled environment. For all of its faults,
> >LVDS really is pretty well-suited to a cabled environment.

> I agree!
> I'm also considering the possibility of using LVDS and a coax cable
> @622Mbps, but I have to double the number of cables!
> Using a AC-coupled solution I should avoid the Vgpd problem, shouldn't I?

Doubling the number of cables would be a pretty poor choice, due to
differential skew issues. You could, of course, simply float the cable.
Not, frankly, a choice that I'd be comfortable with.

At 622, you could also use transformers. As long as you're dealing with
a dc-balanced code such as 8b10b (probably necessary for cabled
connections anyway) this should be OK.

> In the backplane case, I will have to put a constraint on Vgpd
> (for instance 50mV max).
> In this case I could use either SLVS or LVCML.
> >....We're looking into hot-plug
> >issues for SLVS and although we don't want to promise anything it looks
> >good so far.
> Please, let us know, as soon as you have more infos about.

Taking off my JEDEC hat and putting on my Philips hat, we'll be
implementing SLVS-400 at moderate speeds (up to about 2 Gb/s)
in our 0.18-micron library and at high speed (>3 GB/s) for the
next generation (0.12 micron).

For a moderate-length backplane (not needing serious equalization)
I would certainly advise SLVS, not least on the grounds that it will
interoperate gracefully across process generations.

-- 
| I'm old enough that I don't have to pretend to be grown up.|
+----------- D. C. Sessions <[email protected]> ----------+

**** To unsubscribe from si-list or si-list-digest: send e-mail to [email protected]. In the BODY of message put: UNSUBSCRIBE si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu ****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Thu Jun 21 2001 - 10:11:56 PDT