"Can the Si9000e model Delta-L insertion loss measurements?"

Recently a Polar customer condicting signal integrity analysis asked if the Polar Si9000e was a suitable tool for modeling the s-parameter loss characteristics of a PCB substrate measured with the Delta-L methodology.

Modelling insertion loss with the Si9000e

First, note that Si9000e is suitable for modelling insertion loss on a wide range of PCB structures and stackups and the models are accurate boundary element field solved calculations of insertion loss that will correlate with a variety of insertion loss measurement techniques.

Techniques include Delta-L, SPP, SET2DIL and direct VNA measurements – provided the measurements are performed carefully with probes, cables and well-designed test vehicles. It should also be noted that Si9000e models the pure transmission line loss with the via effects fully de-embedded.

Delta-L measurement technique

One of the benefits of Delta-L is the technique it uses to remove – SI engineers call this de-embed – the effects of the via and test system interconnect, leaving the pure loss of the PCB and its composite materials in the measurement. Delta-L does this by measuring a short and a long transmission line structure and mathematically processing the results; because the short and long line structures both contain almost identical interconnect paths it is possible to "divide out" (though it's more complicated than that!) the interconnect artefacts from the measurement and leave only the losses of the line itself.

The customer asked if the Si9000e would be able to model the intermediate stages of the process – i.e. the loss of the short line and the loss of the long line on their own. The answer to this is "No" as the measurements of the short and long line are intermediate steps to gather raw data that need processing before the finished loss result is mathematically derived.

Using the final result from a Delta-L based measurement system and a correctly configured Si9000e you should be able to establish good correlation between Delta-L and Si9000e, just as has been achieved by previous studies contrasting SPP, SET2DIL and VNA measurements (AP8182)

Notes

There are a few points you need to be aware of, especially if this area is new to you. Delta-L will produce the loss per inch (typically) – this may depend on your system vendor. Si9000e will always produce s-parameters per line length so you will have to set the length of line, LL, to 1 inch in the Si9000e to get a correlating result.

You should also note that the Si9000e presents attenuation (this is probably the easiest graph to use) or 4 port s-parameters or mixed mode s-parameters; again, if you prefer to see the result in dB/inch and s-parameters you should select "Mixed mode" s-parameters as your graph. "Mixed mode" is the Si engineers terminology for "Differential S-parameters".

If you use 4 port in this situation you may see s-parameters that look very wavy; this is normal and the reason for this is outside of the scope of this application note. Simply select Mixed mode or look at the main attenuation graph instead.

If the Si9000e underestimates the loss you should take a look at a cross section of the measured trace and have the surface roughness estimated. This can be factored into the model with the surface roughness capability in the Si9000e.