The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> High-Speed I/O Design >> Regarding Serial Link standrds
https://designers-guide.org/forum/YaBB.pl?num=1385062814

Message started by raja.cedt on Nov 21^st, 2013, 11:40am

Title: Regarding Serial Link standrds
Post by raja.cedt on Nov 21^st, 2013, 11:40am

Dear All,

1.Why some standards specify external AC coupling (between TX and RX), some others use Direct coupling, i know from RX design Ac coupling will be useful but i guess there will be a strong reason from Board point of view.

2.Why always TX eye mask will be different than RX mask (actually i don't know how to describe here, TX mask will rectangular,RX will be like rambus shape)

Thanks,
Raj.

Title: Re: Regarding Serial Link standrds
Post by harpoon on Nov 25^th, 2013, 7:09am

raja.cedt wrote on Nov 21^st, 2013, 11:40am:

2.Why always TX eye mask will be different than RX mask (actually i don't know how to describe here, TX mask will rectangular,RX will be like rambus shape)

I have designed a transceiver for USB in the past and the TX is rectangular because it needs to be as perfect as it can be before going into the "channel" (in my case, pins, board, connectors and a long USB cable with reflections and mismatches).

The RX pattern that the chip needs to be able to detect will therefore be non-ideal (i.e. rambus shaped), i.e. a very small eye opening. This is due to the imperfections in the "channel" mentioned above.

Depending on where you measure/define your "channel" / system, the TX/RX eye templates will be different (getting worse going from TX to RX)

Was that what you were asking ?

Title: Re: Regarding Serial Link standrds
Post by hxwang on Apr 1^st, 2015, 9:45pm

DC or AC couple?
If with DC couple, the TX sets the CM voltage. Then the CM range may be large for the RX considering the nonideal effects induced by the channel/cable etc.
If with AC couple, this problem is non-exist. But it throws some requirments on the data.

Title: Re: Regarding Serial Link standrds
Post by loose-electron on Nov 16^th, 2015, 4:42pm

hxwang wrote on Apr 1^st, 2015, 9:45pm:

Agreed - AC coupling dictates some form of run length limited coding structure so that the BW range of the signal is limited on the LF end. DC coupling can tolerate simple binary data but then has to signal process situatiosn with long strings of 1 or 0 which can be difficult.

Title: Re: Regarding Serial Link standrds
Post by Venn on Nov 18^th, 2015, 7:45pm

Hi,

1) When we have a channel that is ac coupled, TX won't set the common mode voltage (or the dc bias) at the RX. RX can set its own common mode voltage. This helps us in several ways. If you want an NMOS input pair for the diff amp at RX, you would rather want a high common mode voltage so that all transistors in the diff amp are in proper saturation. Similarly, if you are planning for a PMOS input pair for the diff amp, you would rather want a low common mode voltage. If TX places some constraints on this common mode level, you won't have the freedom to choose the type of transistor. There are several other reasons in addition to the above mentioned one.

2) For the second question, I believe what you are talking about is the reduction in eye opening (both horizontal and vertical) at the RX. This has something to do with Inter-symbol Interference (ISI) that occurs because of increase in channel attenuation with frequency, and you also have other jitter components in addition. There are lots of materials on ISI online.

Hope this helps.

Title: Re: Regarding Serial Link standrds
Post by raja.cedt on Nov 19^th, 2015, 12:33am

Dear all---thanks for your replies. Few additional points---

1. Dc-oupling mandates same ground potential always at Tx and RX, which is not easy always (especially in case of long trace or server based applications in which each rack PCB separated by long vertical trace)

2. Some times, TX is cmos and RX could be bipolar (in case of 100G optical chips), so TX side 1V and RX side 3.3V(typical). In this case Ac coupling is the only one solution.

Thanks,
Raj.