Estimated reading time: 3 minutes

When Is a 10ghz Transmission Line Not a 10ghz Transmission Line?

A simple question surely? But there could be a whole host of answers, and by now you will probably be anticipating my answer… yes, “It depends.” Just as in life, in electronics the only certainty is uncertainty. -- John Allen Paulos

Surely 10 GHz is 10 GHz? Well, let’s take a look at marketing “spin” first. Often, chipset documents and PCB specs refer to the data handling capability of a link, but in some cases, for example a 22GB/s link may be a pair of 11 GB/s channels, or even four parallel 5.5 GB/s channels, and if you take into account that data can be clocked on the rising and falling edges the clock rate could actually be 2.25GHz. Suddenly, that brings what seemed like a fantastically high frequency down to “real world” levels.

So as a fabricator you really now need to dig under the headlines and work with the designer to find the true nature of the channel under test.

A frequently asked question at Polar is – “Can I check my 1 - 2 - 5 - 10 - 20 GHz PCBs with a TDR?”, and the answer is “It depends”. The “it depends” part is to ask whether the line under test exhibits significant losses at the frequency under test or not. If the line length and operating frequency, along with the laminate type when modelled, predict insignificant losses, then the impedance of the line is relatively frequency independent above a few megahertz. Provided a suitable length test trace is available the impedance can be tested with a TDR with a risetime anywhere between 15ps and 250ps the only caution being that the faster risetime, in general, will require more care over ESD protection.

However, if the line in question, when modeled, exhibits significant dBs of loss, then a simple TDR test (regardless of risetime) may prove inadequate. In this case a fuller analysis of line behavior is required, including examining the losses with respect to frequency. This can either be performed with a vector network analyser (VNA) or with a TDR with post processing software required to extract the frequency domain information, along with the line losses. A variety of techniques are proposed in IPC TM-650 Test Methods Manual¹.

Often PCB fabricators are confused when first confronted with impedance, “Surely this coupon has an open circuit at the end – so the impedance must be infinite?” and at DC this is true; however, for any transmission line the impedance rapidly falls from infinity to a steady state defined by the equation: 

This is clearly shown by the following model which extends the impedance solving of the line from 10KHz up to 10 GHz.

From these two graphs we can see the instantaneous impedance remains roughly the same from 10MHz to 100GHz, so what else is happening as the frequency increases?

Whilst the instantaneous impedance stays the same, the increasing frequency leads to increasing losses. Below 2GHz, even on FR4 most lines of 4 to 5 mil width on ½ oz. copper may be considered “lossless”; i.e. the losses incurred are small enough to be ignored and these lines may be measured on conventional TDR equipment. Above 2GHz, losses may become more significant – this can be down to the Cu dimensions, Cu P-P roughness, and the type of dielectric material employed. Again, modelling can be employed to check if the transmission line in question is “in harm’s way” when losses are considered.

The following graphs illustrate the conductor, dielectric and total losses for 3 and 6 inch lines on FR-4.

The graph of the same structure charts the conductor and total attenuation losses when the copper surface roughness is taken into account.

In conclusion, when looking at high frequency test requirements you should always look “behind the specs” and ask the designer, “Does the speed refer to a single channel? Does the speed refer to a bit rate or a single frequency?” And when you know the answers to those questions you can then investigate further with the designer as to whether the measurements required are those of simple lossless impedance or if additional information on transmission line insertion losses are also required.

All of the above confirms the answer to the question “When is a 10GHz transmission line not a 10GHz transmission line?” – It depends...

References

1. Propagation Loss Test Methods Task Group (D-24b), May 2009
IPC-TM-650 Test Methods Manual, Number 2.5.5.12