最近聽到一種說法，認為沒有reference plane的differential pair才是好的設計。筆者一開始並不理解這說法從何而來，直到一位熱心的網友提出來討論，我才知道原因出在一篇IEEE的文章被斷章取義所造成。EMC Considerations for Differential Lines, 2007

首先該文一開始是說：

A full balanced system requires no separate "ground". 接著再說

The presence of any closely coupled conductors to a differential pair is a complicating factor that in many cases will result in the generation of unwanted, common mode current.   看到這裡，難怪有人誤解了

多數人忽略了這裡所指的[full balanced circuit/system]是什麼意思，所以就沒注意該作者接下來所說的：

Unfortunately, in many cases the physical implementations of many differential links are not simple, two wire systems, but rather are comprised of two, independent, single ended circuits.

Full balanced system是一種理想的differential pair，但要做出真正這樣的電路，Tx需要有光耦合隔離電路或中間抽頭變壓器隔離。所以該文作者也說，像Ethernet就是balanced differential pair，而Ethernet就是有Transformer isolator。另外一種常見的differential pair是USB，它就是由兩個反相的single-end所組成的differential pair(更精確的說，USB Tx是由一個電流源，透過開關改變方向切出來兩個反相訊號)，所以USB 2.0 spec.要求cable要有 shielding。

該文作者接著說明，大部分的電路是以產生兩個獨立的反相訊號來實現差分訊號傳輸，此時參考平面對差分線的影響就很重要。若參考平面完整，且差分訊號的兩個輸出設計對稱(驅動能力與相位都對稱)，則其分別的return current會在reference plane大部分抵銷，反之則產生common mode return current. (EMI issue)

剩下筆者就不多說了，各位自己再去看看原文囉。

參考文章

• EMC Considerations for Differential Lines , 2007

• Differential Signals are NOT Immune to EMI/EMC Concerns! , 2010