Chip-to-module CAUI-4 CTLE
peaking tolerancing
Piers Dawe
Mellanox Technologies
Problem statement
• Comment 67 (at back of this slide pack):
• The draft assumes that if the host provides an
eye that's open within 1 dB of where the host
says it's open, then it will be open where the
module would use it.
• It seems this is true if the host makes a
reasonable recommendation, but the host
could make an unhelpful recommendation
and still be compliant
• This presentation illustrates the problem and
proposes a remedy
IEEE P802.3bm, July 2014,
San Diego
Chip-to-module CAUI-4 CTLE peaking
tolerancing
2
Simulations based on published
channel model
Example channel vs. OIF VSR insertion loss budget
0
Response (dB)
-5
Chip-to-module
CAUI-4 insertion
loss budget
-15
-20
-25
•
•
CEI-28G-VSR
insertion loss
budget
-10
0
5
10
Frequency (GHz)
15
20
Based on oif2010_407 – there is an equivalent dataset in the 100GE
channels page
A little less loss than worst recommendation. No package model
IEEE P802.3bm, July 2014,
San Diego
Chip-to-module CAUI-4 CTLE peaking
tolerancing
3
Eye opening for
maximum allowed
Vpkpk
Eye at 1e-15, 28 GBd
0.6
0.5
Dashed lines are spec limits
0.4
In this benign example, host could
recommend any value from 1 to 9
0.3
0.2
0.1
0
1
2
3
-11
7
8
9
•
-11.5
-12
dB(EH15) (dBV)
4
5
6
CTLE peaking (dB)
-12.5
-13
This shape
illustrates a rule of
no more than 1 dB
down, one step
either way
-13.5
-14
-14.5
-15
-15.5
1
2
3
4
5
6
7
CTLE peaking value (dB)
IEEE P802.3bm, July 2014,
San Diego
8
9
Change in EH15 for 1 dB change in CTLE setting (dB)
Eye opening (UI bk, V bl, UI*V red)
0.7
All simulations are without
emphasis, with Vpkpk at maximum
1.4
1.2
1
0.8
0.6
0.4
0.2
0
1
2
Chip-to-module CAUI-4 CTLE peaking
tolerancing
3
4
5
6
7
CTLE peaking value (dB)
8
9
4
Eye at 1e-15, 28 GBd
Slower...
0.6
0.5
0.4
0.3
0.2
0.1
0
1
2
3
-12
7
8
9
•
-12.5
dB(EH15) (dBV)
4
5
6
CTLE peaking (dB)
-13
-13.5
-14
-14.5
-15
-15.5
-16
-16.5
1
2
3
4
5
6
7
CTLE peaking value (dB)
IEEE P802.3bm, July 2014,
San Diego
8
9
Change in EH15 for 1 dB change in CTLE setting (dB)
Eye opening (UI bk, V bl, UI*V red)
0.7
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
1
2
Chip-to-module CAUI-4 CTLE peaking
tolerancing
3
4
5
6
7
CTLE peaking value (dB)
8
9
5
Slower...
Eye at 1e-15, 28 GBd
EW
limit
0.6
•
0.5
•
0.4
0.3
0.2
0.1
0
1
2
3
-14
4
5
6
CTLE peaking (dB)
7
8
9
dB(EH15) (dBV)
-16
-18
-20
-22
-24
-26
1
2
3
4
5
6
7
CTLE peaking value (dB)
IEEE P802.3bm, July 2014,
San Diego
8
9
Change in EH15 for 1 dB change in CTLE setting (dB)
Eye opening (UI bk, V bl, UI*V red)
0.7
Host could report 2 and be
compliant (because eye meets spec
at 3)
Compliant module might work at 2
(3 dB less Eye Height than at 3, and
failing Eye Width) but might not at
1, where the height is 8 dB less than
at 3, and falling off at 4 dB/step,
and Eye Width is very poor
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
1
2
Chip-to-module CAUI-4 CTLE peaking
tolerancing
3
4
5
6
7
CTLE peaking value (dB)
8
9
6
Eye at 1e-15, 28 GBd
Even slower
EW
limit
0.6
•
0.5
0.4
Proposed 80
mV secondary
limit
0.3
0.2
•
0.1
0
1
2
3
-18
4
5
6
CTLE peaking (dB)
7
8
9
dB(EH15) (dBV)
-20
-22
-24
-26
-28
-30
-32
-34
-36
1
2
3
4
5
6
7
CTLE peaking value (dB)
IEEE P802.3bm, July 2014,
San Diego
8
9
Change in EH15 for 1 dB change in CTLE setting (dB)
Eye opening (UI bk, V bl, UI*V red)
0.7
Host could report 4 and be
compliant (because eye meets
spec at 5)
Compliant module set at 3, gets 3
dB less height than at 4, and
falling off at 6 dB/step, and Eye
Width is very poor
12
10
8
6
4
2
0
1
2
Chip-to-module CAUI-4 CTLE peaking
tolerancing
3
4
5
6
7
CTLE peaking value (dB)
8
9
7
Second-guessing the spec
• The module maker could observe that the rolloff is steeper on the low side and set up
higher than the recommended value
• Or the host maker could observe the same
thing, and recommend higher than actual
• If they both think of it, they will double
compensate, and between them get a bad
result, and/or go beyond 9 dB
IEEE P802.3bm, July 2014,
San Diego
Chip-to-module CAUI-4 CTLE peaking
tolerancing
8
Discussion
• An EH15 minimum that's 1 dB less than the
current one, to be applied at +/-1 dB from
recommended CTLE peaking value, seems to
help
• Comment proposes 80 mV (1.5 dB less than
the 95 mV primary EH15 limit) at
recommendation +/- 1 dB
• A simpler option might be to require 90 mV at
the recommendation, if someone else wants
to do the analysis to show that it fixes the
problem
IEEE P802.3bm, July 2014,
San Diego
Chip-to-module CAUI-4 CTLE peaking
tolerancing
9
Comment 67
83E.3.1
page 166 line 22
According to 83E.3.1.6, the host provides a recommended CTLE peaking value. According to 83E.4.2, the host's
eye must pass the spec at this value or 1 dB higher or lower if within the range 1 to 9. However, simulation
shows that if the host gives a compliant but unhelpful recommendation, EH15 can be up to 8 dB smaller
than the limit in Table 83E-1 within 1 dB of the recommendation (this is allowing 1 dB for the module).
We could impose some accuracy spec on the recommendation, but it is more useful and easier for the host to
simply require that the eye is open at and around the setting the host recommends, by adding a subsidiary
eye height spec. This 80 mV spec is chosen so that a host which gives a reasonable recommendation is not
inconvenienced, and the host can trade off accuracy and eye opening - but the module is protected from a
very unhelpful recommendation.
A simpler option might be to require 90 mV at the recommendation, if someone else wants to do the analysis
to show that it fixes the problem. This would reduce the number of host tests but is less generous to hosts
that give realistic recommendations and may protect the module less well.
The text in 83E.4.2 would benefit from a little wordsmithing anyway.
Insert a second limit below the 95 mV, of 80 mV (1.5 dB lower than 95 mV).
In 83E.4.2, change:
For host compliance, the CTLE peaking in the reference receiver shall be set to one of three values. These are:
a) the recommended CTLE peaking value provided by the host, b) the value 1 dB higher if present in Table
83E-2, c) the value 1 dB lower if present in Table 83E-2. Any of the three CTLE settings that meets both eye
width and eye height defined in Table 83E-1 is acceptable.
to
For host compliance, the CTLE peaking in the reference receiver shall be set to three values. These are: a) the
recommended CTLE peaking value provided by the host, b) the value 1 dB higher if present in Table 83E-2,
c) the value 1 dB lower if present in Table 83E-2. A compliant host passes both the eye width and the
larger eye height limit specified in Table 83E-1 at at least one of the settings, and passes the smaller eye
height limit specified in Table 83E-1 at all of the two or three settings.
IEEE P802.3bm, July 2014,
Chip-to-module CAUI-4 CTLE peaking
10
San Diego
tolerancing