Differential Pair Routing Phase Lead Black Magic

Making PCB Layouts, Manual routing, Auto-routing, Copper pouring, Updating from Schematic, Manufacturing Output
Post Reply
Message
Author
bixelps
Posts: 12
Joined: 15 Aug 2011, 16:31

Differential Pair Routing Phase Lead Black Magic

#1 Post by bixelps » 13 Mar 2021, 09:40

I am not an expert at differential pair routing as you are about to see. I can understand that the two traces need to be length matched and I can manually add meanders to make that happen.

What is complete black magic to me is the concepts of "Forward Point Phase Lead" and "Backward Point Phase Lead" as shown in the phase tune tool. I have no idea what these concepts mean and it seems that no matter what meanders I add the situation is always un-acceptable to the DRC. I see that DipTrace places a red line with arrows between the pair on the layout which is supposed to help me in some way but I have no idea what it is trying to tell me.

I have read for several hours on this forum, searched youtube, googled and not seen any example of how to resolve phase lead issues in diptrace. Yes I even resorted to reading the manual. I cannot find any discussion on what phase lead is in any of these sources.

So now I am here hoping that someone on this board will be able to explain what phase lead is, why its important, why forward phase lead is different than backwards phase lead. Why at different points on the same trace I get different answers from the phase tune tool and what to do to resolve these issues.

It would be much appreciated.

novarm44
Technical Support
Posts: 418
Joined: 09 Jun 2010, 06:24
Contact:

Re: Differential Pair Routing Phase Lead Black Magic

#2 Post by novarm44 » 13 Mar 2021, 19:29

Differential pair has source of signals and receiver. Ideally all opposite trace segments have similar length from both source and from receiver. Phase lead means some signal has shorter length at the specified area: Forward lead from the start point of pair and backward - from the end point of trace/pair (how it was built). DipTrace can not determine itself where is the source of signal, so checks phase difference in both directions. You can play with phase tune tool to see how it changes when you add meanders and increase length: usually it is necessary to add meander where phase error starts, on the trace which has shorter path before error. Use real-time DRC to show errors at once when you add/edit meanders.

bixelps
Posts: 12
Joined: 15 Aug 2011, 16:31

Re: Differential Pair Routing Phase Lead Black Magic

#3 Post by bixelps » 14 Mar 2021, 15:03

I do appreciate the response to my post but I am afraid I can not make sense of many of the things you wrote. First let me give a little more information. The trace I am working with goes from one chip to another. This is a USB 2.0 data pair. There are no complicated "star" connections or anything, its point to point but does travel across my board on two layers. Since the pair run together everywhere along the way they are already "approximately the same length".

I have already spent many hours "playing with the phase tune tool" to no avail.

Specifically about your post.

1) What is an "opposite trace segment"?
2) Regarding "Phase lead means some signal has shorter length at the specified area: Forward lead from the start point of pair and backward - from the end point of trace/pair (how it was built)". These two sentences make no sense to me. Can you try again? Specifically what is Forward and what is Backward when it comes to these traces?
3) Are all problems like this solved by just adding meanders in the right places?
4) How do I know what tolerances to use in the DRC for my signals?

Alex
Technical Support
Posts: 3536
Joined: 14 Jun 2010, 06:43

Re: Differential Pair Routing Phase Lead Black Magic

#4 Post by Alex » 15 Mar 2021, 10:33

Ideally, diff pair is two coupled traces of the same length. Suppose one end is transmitter and other end is receiver. Differential signal goes from transmitter to receiver with constant speed. They(positive and negative signal) start together. But if diff pair turns to some angle (90, 135, 180 degrees, etc) one signal outpaces the second one. This is phase error. To fix the problem a meander can be added to the signal that goes ahead another.
The software doesn't know which end is transmitter and which is receiver. Therefore it calculates phase error in both directions.

bixelps
Posts: 12
Joined: 15 Aug 2011, 16:31

Re: Differential Pair Routing Phase Lead Black Magic

#5 Post by bixelps » 17 Mar 2021, 12:03

Hi Alex,
Does your response refer to the length of the traces or the phase error? You mention both in your post.

Its pretty easy to get the length of the traces to match with meanders. What is not seemingly possible is to use the phase tune tool to solve issues with phase lead. The tool suggest to add a meander to this or the opposite trace, that's great. But when I do that the tool shows no improvement in the error and often suggest I now add the meander to the other trace. No matter how small I make the meander it will often then suggest I add another meander to the opposite trace. The sign of the two phase errors are usually opposite and there seems to be no way to improve.

Is the position of the meander on the trace important? When the tools is active as I move the mouse along the trace I see different phase errors calculated in different locations along the trace. The tool gives no hint on where to add the meander(s), how big or how many meanders to add and no consistent information on which trace to add the meander to.

When I add a meander should I expect the phase error to become more positive or more negative?
What is the definition of "forward" in this tool?

Post Reply