Estimated reading time: 3 minutes

The Bare (Board) Truth: Netlist Mismatches Redux

Greetings all. In this month's Bare (Board) Truth I will once again be talking about IPC netlists and common netlist errors that can cause time delays at the fabrication level.And once again, I feel the need to clarify the intent of the netlist compare at the fab level.

This is a comparison between the design criteria and the output Gerber files via the provided IPC-D-356 netlist. This is not to be confused with a 1 up comparison done at the CAM stage by most fabricators. I still get requests from customers to just "generate a netlist" based on their Gerbers. As I have said before, since the intent of a netlist compare is to compare the design criteria against the exported Gerber files, this would never find a mismatch.

Generally, when a customer asks that question, they feel their Gerbers are correct and they assume a 1 up compare in CAM would find things caused by a fabricator, such as etch compensations that cause shorts or other anomalies caused by file manipulation for fabrication.

(This is where the 1 up compare in CAM comes into play, where a comparison between the CAM edited files vs. the original data is done to ensure there are no electrical issues.)

The most common forms of IPC netlists provided by our customers are IPC-D-356, IPC-D-356A or Mentor neutral files. If net list mismatches are found at the fabrication level, it is necessary for the fabricator to contact the customer and share the net compare results. These results can vary from legit netlist issues, such as unconnected features (opens), missing relief pads on planes (shorts), and bridged isolation barriers (shorts) up to net list issues caused by the netlist's definition.

An example: Mounting holes where a post-fab connection is made by the mounting screws themselves. The IPC net list itself may show these as connected, but the Gerber files and drawing call these holes "non-plated" and therefore get designated as a "non-electrical," causing the broken or open net.Net list mismatches can also be caused on the fab end if the layers and netlist are not properly registered. Symmetrical boards are particularly susceptible to this.

Also, if metal features extend past the fab lines on a part and the fab line is not removed before running net compare, the net compare will come up with erroneous "shorts" where the fab line is connected to the overhanging feature.One of the most common netlist mismatches is an analog-to-digital ground short (AGND to DGND). Customers intentionally create this short between analog and digital ground, but it is sometimes not identified on the "read me" drawing note or even in the header of the IPC file. (This kind of identification is very helpful when a board is time-critical and a customer can't afford delays.)

Other common netlist errors are associated with "castellations" or edge-plated features. Here, the netlist defines these points as connected at assembly but they show up as broken or open nets at fabrication, requiring a phone call to the customer to clarify. Again, this call can be avoided by calling out any intentional shorts or opens on a read me file or fab note.

Some customers' Gerber files feature entities such as a metal land area split into two separate nets that - after etch compensation (a compensation to all metal features to account for the known loss at an etcher) - show up as shorted nets. Other times, depending upon the netlist definition, non-plated through-holes can come up as "missings" in a netlist compare.

Whereas these anomalies are not "real" netlist issues, they still must be communicated to the customer for disposition. A good fab shop can usually diagnose the source of the short or open for the customer and can generally get an answer fairly quickly.

So, you ask, "Do I even need to supply an IPC netlist?"

If the job is called out on the drawing, spec or read me file as an IPC-6012 Class 3 part, then the fabricator is obligated to run a netlist compare. And let me add that, from a fabricator's standpoint, we would like all customers to provide IPC netlists, regardless of the respective board class.

You would not believe how many times we have received incomplete data on very densely packed multilayer boards, but we were saved by the netlist the customer provided!

As always, thanks so much for your time. Feel free to contact me with any questions.

Mark Thompson is in engineering support at Prototron Circuits. To contact Mark, click here, or call 425-823-7000.