Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: XAUI IO specs




Ali,

By your own statement below, a static Logic 1 is when Vo+ is 400 mV above
Vcm, and Vo- is 400 mV below Vcm, then the differential voltage (Vo+) -
(Vo-) is  800 mV (positive peak).
By your own statement below, a static Logic 0 is when Vo+ is 400 mV below
Vcm, and Vo- is 400 mV above Vcm, then the differential voltage is (Vo+) -
(Vo-) is -800 mV (negative peak).
Then peak to peak is (positive peak - negative peak) = 1600 mV, and not 800
mV as you have in your response.  So, I agree with Mike Dudek, Ed Grivna,
that your statement below is confusing.  I am agreeable to any way of
defining logic levels with or without common mode, single ended or
differential.  I agree with Ed Grivna that Vcm is not necessary to define a
differential signal, and I am only responding to a thread that is using
Vcm.

Given all this confusion, it may be prudent to include pictures of logic
signals, and define the swings in an already standardized way.

May I recommend that future proposals define signals like Figure 3-3 on
page 10 of the IEEE LVDS spec, document # (IEEE Std 1596.3-1996)?  A
picture would eliminate ALL confusion.  In the absence of a picture,
pointing to a standardized way of measurement would also be less confusing.
Perhaps we could use |Vod| = Output Differential Voltage like on page 8 and
9 of the same IEEE document.  Notice the absolute value sign that
eliminates the polarity issue, which eliminates the factor of 2 that is
biting us.  We all understand that |Vod| = 1/2 * (V differential
peak-to-peak).

I actually prefer differential peak-to-peak, but being that there is lots
of confusion, we all need to agree.  Use whatever you like, as long as we
can all understand.


From Ali's Post:

>o. Based on my understanding and everyone who participated during
conference call the Vp-p
>differential max defined as 800 mV.

>    - This means each side of the driver has a swing of 400 mV.

>    - Logic 1 would defined as Vo+ at (VCM+400 mV) and V0- at (VCM-400mV)

>    - Logic 0 would defined as Vo+ at (VCM-400 mV) and V0- at (VCM+400mV)

>    - If the driver output VDp-p <85 mV you got noise.


In your follow up response to this thread, you wrote:
>When I say each side has a swing of 400mV, I refere to each half of the
differential driver.
>Vo(diff)p-p = (Vo+) - (Vo-) = (400) - (-400) = 800 mV.


The two sentences contradict each other.  If you have static voltages, Vo+
= Vcm+400 mV, and Vo- = Vcm-400mV, the difference is 800mV, but is only ONE
peak, and NOT peak to peak.  For any peak-to-peak measurement you need to
know what the other peak is.  On 4/18 Tim Mathey posted an e-mail with
Howard Johnson's recollection of peak differential amplitude vs
differential peak-to-peak.  He even has an ASCII waveform that makes it
crystal clear.


Darn those factors of 2!  :))

Again, if we have pictures you can call it whatever you want and we all
understand.

Kevin Demsky
Mixed Signal and VLSI Development
IBM Corp.
3605 Hwy 52 N
Dept. QXS Bldg. 050-2
Rochester, MN  55901

Internal E-mail:  kdemsky@ibmusm07
External E-mail:  kdemsky@xxxxxxxxxx

Phone:  507-253-5799
Fax:  507-253-4966