[SI-LIST] Re: PECL termination technique?
- From: "john lipsius" <johnlipsius@xxxxxxxxx>
- To: <hariharan@xxxxxxxxxxx>, "'SI-LIST'" <si-list@xxxxxxxxxxxxx>
- Date: Sun, 16 Feb 2003 15:24:57 -0800
Hariharan,
I overlooked that the MC100 device is a 5V only chip. So is the
Cypress part. Therefore, the Cyp appnote isn't applicable and
you don't need to shift Vcm to a 3.3V receiver. That is, the
pullups to 5V shouldn't be used for that function.
However, the lack of dc pulldown to Vtt = Vcc-2 = 3V is still a question.
The appnote circuit does provide this, so we know it's explicitly called
for on the board, regardless of the receiver Vcc.
-----------------------------------------------------------------
Separate issue: I've further cleaned up the 5V to 5V termination
details.
Y-termination for differential PECL to PECL,
with 5V as an example (driver and receiver)
In the previous circuit I gave, as follows,
---Rs-------------------Vin (+)
|
R
|
gnd--220pF-- (Vtt) --- Rt ----gnd
|
R The 220pF cap is for further common mode
filtering.
|
---Rs-------------------Vin (-)
if Rs is used then the Ioh and Iol equations I gave must include
it:
From dc mesh current analysis (Kirchoff's Voltage Law),
Ioh = [(Rt/[R+Rs])*(Voh - Vol) + Voh] / ( [R+Rs] + 2Rt )
Iol = [(Rt/[R+Rs])*(Vol - Voh) + Vol] / ( [R+Rs] + 2Rt )
but these eq. aren't needed here.
Node currents, Kirchoff current law, is an easier method to get
Rs, R, Rt (with Rs included, as well):
Voh+Vol = Vtt(2+[(R+Rs)/Rt])
or
((Voh+Vol)/Vtt) - 2 = (R+Rs)/Rt
Using Vcc=5, Vtt=(Vcc-2) , Voh=4.1 , Vol=3.3 , R=50 , Rs = 33 ,
then Rt = 119 ohms.
Check input common mode voltage:
Vcm(in) = (Vcm_driver - Vtt)(R/(R+Rs)) + Vtt = 3.42 V.,
It's attenuated from the Vcm=3.7 at driver but it's still ok for 5V
receiver. In general, check Vcm(in) range for the receiver.
Also, check swing at receiver:
Vih = (Voh - Vtt)(R/(R+Rs)) + Vtt = 3.66
Vil = (Vol - Vtt)(R/(R+Rs)) + Vtt = 3.18
Vdiff = Vih - Vil = 480 mV
Vdiff,pk-pk = Vin(1) - Vin(0)
= ( Vih-Vil ) - ( Vil-Vih ) = 480 - (-480) = 960 mV
= 2 * Vdiff
Resistor Rs must be placed next to driver. If not, its purpose is
defeated: to absorb back reflections from common mode (assuming
perfect Zdiff match at load). The closer to the load (R, Rt, and recvr
pin) it's placed, the greater the mismatch to Zdiff and the more
differential reflection occurs, as well. That is, the diff. termination
would look like 100 + 2*33 = 166, not 100.
=========================
----- Original Message -----
From: john lipsius
To: hariharan@xxxxxxxxxxx ; 'SI-LIST'
Sent: Saturday, February 15, 2003 3:54 PM
Subject: Re: [SI-LIST] Re: PECL termination technique?
Hariharan,
The Cypress appnote (link below) does *NOT* match your original
email. Your email doesn't show a ground tap and doesn't have the
220pF common mode filter.
I can only assume that your product works because the MC100's internal
pullups to 3.3v do the dc biasing for the Cypress drivers. But, those
pullups are large, if they exist, so the driver dc current is small --- not
good. They are near cutoff.
Also, lack of the common mode filter is just unwise. It simply improves
your jitter or noise budget, which may be sufficient now, but only
marginally so. Ie: it could fail if you change from the MC100 to another
device.
The appnote also relies on series R of 56ohm to damp the common mode,
and the R_odd that results is (56+75)//82 = 50.4 ohms.
The drivers are dc loaded with 131 ohms, which is in range.
Very ingeniously, their termination uses just its own 5V supply to also
establish the normal 3.3V comm. mode at the load:
Vcm load = Vcm * 75 / (56 + 75)
= (5-1.33) * 75 / (56 + 75) = 2.1 V
In your design:
Vcm load = 3.67 * 60 / (33 + 60) = 2.3 V
The Cy appnote shows this:
-->>----------56----------------
| |
82 75
| |
Cypress +5V-- 220pF --|--gnd MC100 input
driver | |
82 75
| |
-->>---------56----------------
I would just ask Cypress about using their scheme vs. yours with the
MC100 device.
See:
http://www.cypress.com/cfuploads/support/app_notes/sst33.pdf
----- Original Message -----
From: hariharan
To: 'john lipsius' ; hariharan@xxxxxxxxxxx ; 'SI-LIST'
Sent: Saturday, February 15, 2003 2:51 AM
Subject: [SI-LIST] Re: PECL termination technique?
Hello John,
Thanks for the mail. But the same termination seems have worked for the
previous version of the present product. It has the same kind of PECL
termination between CY7B951-SC (Cypress) to Motorolla's (MC100ELT23D).
That's the reason I'm so curious how it would work?
The present termination what I was describing looks like a standards LVDS
(Differential) termination.
Let me know if you have an answer.
regards
Hariharan
-----Original Message-----
From: john lipsius [mailto:johnlipsius@xxxxxxxxx]
Sent: Friday, February 14, 2003 5:57 PM
To: hariharan@xxxxxxxxxxx; SI-LIST
Subject: Re: [SI-LIST] Re: PECL termination technique?
Corrected eqns. ... bad parentheses
Ioh = [(Rt/R)*(Voh - Vol) + Voh] / ( R + 2Rt )
Iol = [(Rt/R)*(Vol - Voh) + Vol] / ( R + 2Rt )
----- Original Message -----
From: john lipsius
To: hariharan@xxxxxxxxxxx ; SI-LIST
Sent: Friday, February 14, 2003 4:21 AM
Subject: [SI-LIST] Re: PECL termination technique?
Hariharan,
I agree something's missing.
Perhaps your source outputs are pulled down close to the
driver on another schematic sheet. That's the only way it might
work. The MC100ELT23D may have internal input pulldowns, but
that's just to assure safe input levels when undriven.
Perhaps the termination was just 'borrowed' from another
design and is missing the biasing. The single diff. termination
at the load is typical of long links, like a across a backplane, where
grounds can be different. The driver end would have the
pulldowns for biasing in that case.
The 33 and 120 ohm resistors appear to be series damping and
Zdiff termination, but no biasing appears to be provided.
Recap....
1. Somewhere, these are required on each wire:
* establish dc driver current for linear operation
* bias to Vcm=Vcc-1.3 for example (typ. PECL)
2. Another function is Zdiff matching at the load
---Rs-------------------
|
R
|
Vtt ---Rt--gnd Y termination with optional Rs
| for attenuation of reflections
R or for EMC
|
---Rs-------------------
The Y termination can provide biasing & impedance matching
by choosing R and Rt and setting Vtt=Vcc-2. Then the drivers
will operate near Vcm.
Assuming Zdiff=100, then R=50 since Vtt is a 'virtual' gnd.
Applying node current analysis, Voh+Vol = Vtt(2+[R/Rt]).
Then R and Rt pop out.
For Vcc=3.3V, Rt=R but for Vcc=5V, Rt = 2.13*R , using typical PECL
Voh, Vol.
Tolerances:
Just make sure you don't choose R and Rt such that, at VOLmin,
a driver cuts off. R should be 1% tol. or better.
Ioh = (Rt/R)*(Voh - Vol) + Voh) / ( R + 2Rt )
Iol = (Rt/R)*(Vol - Voh) + Vol) / ( R + 2Rt )
Plugging in the Vol(min) and Voh(max) at worst case temperatures,
Rt max., and Vcc(min) will give a min. Iol, which must be greater than
some vendor minimum such as 0 mA (cutoff). If one driver gets in cutoff
it'll be slow and you'll develop a common mode in driven diff. signal
which will skew your duty cycle, create comm mode reflections, worsen
emc, etc. The Rs helps attenuate the inevitable common mode noise
even with a good termination.
-John
----- Original Message -----
From: hariharan
To: si-list@xxxxxxxxxxxxx
Sent: Thursday, February 13, 2003 8:51 PM
Subject: [SI-LIST] PECL termination technique?
Content-Type: multipart/alternative;
boundary="----=_NextPart_002_0005_01C2D412.E14105F0"
------=_NextPart_002_0005_01C2D412.E14105F0
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Blank
Hi,
This is regarding a PECL termination. I'm presently reviewing a
hardware
design, in that the designer has used a PECL to TTL converter for a
clock
Oscillator's PECL output connecting to a CPLD via a PECL to TTL
translator.
Here the designer has used a termination ie PECL - PECL
(Differential), say
he has used a 120 ohm RES connecting the positive and negative of the
differnential pair.
(POS) +------33 ohms---------------------------+ (POS)
>
120 ohms
>
(NEG) - ------33 ohms---------------------------- -
(Neg)
the devices are CY7B951-SC (Cypress) to Motorolla's (MC100ELT23D).
Here in the above case I understand the designer has opted for a power
consumption less termination. But the usual termination we go in for
is
a Y termiantion ( with ref. to ONSEMI PECL design - Application note).
Can anyone help me out to understand how the designer would have
arrived at
those values and the termination technique.
regards
Hariharan
------=_NextPart_002_0005_01C2D412.E14105F0
Content-Type: text/html;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3Diso-8859-1">
<TITLE>Blank</TITLE>
<STYLE>BODY {
COLOR: #000000; FONT-FAMILY: Arial, Helvetica; FONT-SIZE: 10pt; =
MARGIN-LEFT: 25px; MARGIN-TOP: 25px
}
P.msoNormal {
COLOR: #ffffcc; FONT-FAMILY: Helvetica, "Times New Roman"; FONT-SIZE:
=
10pt; MARGIN-LEFT: 0px; MARGIN-TOP: 0px
}
LI.msoNormal {
COLOR: #ffffcc; FONT-FAMILY: Helvetica, "Times New Roman"; FONT-SIZE:
=
10pt; MARGIN-LEFT: 0px; MARGIN-TOP: 0px
}
</STYLE>
<META content=3D"MSHTML 5.00.3315.2870" name=3DGENERATOR></HEAD>
<BODY background=3Dcid:328163704@14022003-0c2a=20
style=3D"COLOR: #000000; FONT-FAMILY: Arial">
<DIV> </DIV>
<DIV><SPAN class=3D328163704-14022003>Hi,</SPAN></DIV>
<DIV><SPAN class=3D328163704-14022003></SPAN> </DIV>
<DIV><SPAN class=3D328163704-14022003></SPAN><SPAN =
class=3D328163704-14022003>This=20
is regarding a PECL termination. I'm presently reviewing a hardware =
design, in=20
that the designer has used a PECL to TTL converter for a clock =
Oscillator's PECL=20
output connecting to a CPLD via a PECL to TTL translator.
</SPAN></DIV>
<DIV><SPAN class=3D328163704-14022003></SPAN><SPAN =
class=3D328163704-14022003>Here=20
the designer has used a termination ie PECL - PECL (Differential), say
=
he has=20
used a 120 ohm RES connecting the positive and negative of the =
differnential=20
pair.</SPAN></DIV>
<DIV><SPAN class=3D328163704-14022003></SPAN> </DIV>
<DIV><SPAN class=3D328163704-14022003></SPAN><SPAN=20
class=3D328163704-14022003> &nbs=
p; (POS)=20
+------33 ohms---------------------------+ (POS)</SPAN></DIV>
<DIV><SPAN=20
class=3D328163704-14022003> &nbs=
p;  =
; =
&=
nbsp; >=20
</SPAN></DIV>
<DIV><SPAN=20
class=3D328163704-14022003> &nbs=
p;  =
; =
&=
nbsp; 120=20
ohms</SPAN></DIV>
<DIV><SPAN=20
class=3D328163704-14022003> &nbs=
p;  =
; =
&=
nbsp; =20
></SPAN></DIV>
<DIV><SPAN=20
class=3D328163704-14022003> &nbs=
p; (NEG) =20
- ------33 ohms---------------------------- - (Neg)</SPAN></DIV>
<DIV><SPAN class=3D328163704-14022003></SPAN> </DIV>
<DIV><SPAN class=3D328163704-14022003>the devices are CY7B951-SC=20
(Cypress) to Motorolla's (MC100ELT23D). </SPAN></DIV>
<DIV><SPAN class=3D328163704-14022003></SPAN> </DIV>
<DIV><SPAN class=3D328163704-14022003>Here in the above case I =
understand the=20
designer has opted for a power consumption less termination. But the =
usual=20
termination we go in for is </SPAN></DIV>
<DIV><SPAN class=3D328163704-14022003>a Y termiantion ( with ref. to =
ONSEMI PECL=20
design - Application note).</SPAN></DIV>
<DIV><SPAN=20
class=3D328163704-14022003> &nbs=
p; =20
</SPAN></DIV>
<DIV> </DIV>
<DIV><SPAN class=3D328163704-14022003>Can anyone help me out to =
understand how the=20
designer would have arrived at those values and the termination=20
technique.</SPAN></DIV>
<DIV><SPAN class=3D328163704-14022003></SPAN> </DIV>
<DIV><SPAN class=3D328163704-14022003>regards</SPAN></DIV>
<DIV><SPAN class=3D328163704-14022003>Hariharan</SPAN></DIV>
<DIV><SPAN class=3D328163704-14022003></SPAN> </DIV>
<DIV><SPAN=20
class=3D328163704-14022003> &nbs=
p;  =
; =
=20
</SPAN></DIV>
<DIV><SPAN class=3D328163704-14022003></SPAN><BR> </DIV>
<P> </P></BODY></HTML>
------=_NextPart_002_0005_01C2D412.E14105F0--
-- Binary/unsupported file stripped by Ecartis --
-- Type: image/gif
-- File: Blank Bkgrd.gif
-- Binary/unsupported file stripped by Ecartis --
-- Type: text/x-vcard
-- File: Ramalingam Hariharan (E-mail).vcf
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts: