Altium Designer 6 multilayer routing problem

[[email protected]]

Hello everybody,

Problem description:
I have a 4 layer + 2 power plane mixed-signal PCB with SMD components
on one side and I would like to route the digital signals on Mid Layer
1+2 and the quiet analogue signals on the Top and Bottom Layers.

Altium version: Altium Designer 6.7

Layer Stackup Buildup (no blind and buried vias)
Top Layer
GND Plane
Mid Layer 1
Mid Layer 2
Power Plane
Bottom Layer

I placed the digital signals in a 'Digital' Net Class to be able to
assign a routing layers rule to these digital signals - routing layers
enabled for this rule are Mid Layer 1 and 2. Priority=1.
The analogue signals where also allocated to an 'Analogue' Net Class
and assigned to another routing layers rule with the Top and Bottom
Layers enabled. Priority=2
I also have a default rule with all the routing layers enabled.
Priority=3.

Behaviour:
For some reason when I run an Auto Route->All the signals in the
digital Net Class are not routed. Signal fan-out vias are generated
but then the connections on the inner layers are not made. It's not a
problem of the router not having enough space to route the signals,
the router simply stops. I could try to route the signals by hand but
there are too many of them and it's clearly not the point of using an
autorouter.

When I try to route the 'Digital' Net Class separately by going to
Auto Route->Net Class... the same problem happens (on a board with no
pre-routes). The autorouter doesn't do anything. Running an Auto Route-

Fan out->Signal Layers also doesn't help.

Strangely enough if I replace the fan-out vias with pads and then I
run Auto Route->Net Class... the connections are made.

Am I missing something, or is this a bug?
Hope somebody out there is able to help!

Best regards,
Patrick

 

[Brad Velander]

Patrick,
I would be suspicious that the rule only allowing digital signals in the
mid-layers is jamming up the autorouter. Because the signals actually
originate on the top or bottom layer, even though you have fanned them out
already. Does that make sense? The vias probably are not seen by the router
as a valid starting point for routing, it is likely looking right back to a
pad to start the autoroute, thus when you changed them to pads then they
seemed to route.
So I would suggest routing it with the fan out vias changed to pads,
afterward change them back to vias and lock those routes or don't reroute
the digital net class again.

 

[pacikk]

Patrick,
I would be suspicious that the rule only allowing digital signals in the
mid-layers is jamming up the autorouter. Because the signals actually
originate on the top or bottom layer, even though you have fanned them out
already. Does that make sense? The vias probably are not seen by the router
as a valid starting point for routing, it is likely looking right back to a
pad to start the autoroute, thus when you changed them to pads then they
seemed to route.
So I would suggest routing it with the fan out vias changed to pads,
afterward change them back to vias and lock those routes or don't reroute
the digital net class again.

Hi Brad,
thanks for the reply. I think you're right about the fact that the
problem is caused because the signals originate in the top or bottom
layer, since when the routing layer on which the signals are
originating is enabled the autorouter finishes successfully. The idea
of switching the vias in question to pads to enable successful routing
is certainly a solution.

Yesterday I did some more experiments and found another method. I
discovered that if I locked these fanned-out vias the autorouter also
routed the nets in the digital class. Maybe the reason for this
behaviour is that once the autorouter sees that these vias have been
locked it shifts the signal origination point to these vias.

I wonder how other board designers are handling the design of mixed-
signal pcbs in Altium. I thought routing noisy digital signals on the
inner layers was the standard way of routing mixed-signal PCBs.

Regards,
Patrick

 

[Brad Velander]

Hi Patrick,
Your methodology is perfectly sound, routing the digital signals on
inner layers. Alternatively though one could argue that properly routed,
impedance matched and terminated digital signals should not be noisy
anyways. At least not where they would cause violations of any EMI
regulations.
Obviously locking the preroutes is the answer to the issue. Prior to
locking those preroutes the autorouter is undoubtedly reasoning it's routing
based upon shifting back to the pads on either the top or bottom layers.