Testing short traces with fast rise time TDRs the real costs
Application Note AP8174 

Testing on short traces

It’s a call heard since the early days of impedance controlled PCBs – “I need to test on short traces” – “Faster rise time is essential” and in the correct situations this can be true, but in the headlong push for speed, often OEMs and PCB fabricators overlook the hidden running costs in the rush for risetime.

Why the need for faster risetime?

Step back for a moment and ask why there is a need for a faster rise time. Perhaps to shorten the test coupon and save space – perhaps to test the actual trace on the PCB? Maybe in the view that a faster rise time will give a better indication of the impedance at a higher frequency?

Let's break these down and look at them one by one, taking the shorter coupon route first; yes, agreed – a faster risetime will permit the use of a shorter coupon, but only provided extra care is taken with probing and launch structure of the coupon. IPC specs call for a 150mm / 6” coupon for a reason – to minimise the effects of the trace launch and end of trace on the measurement – regardless of the risetime used. So while a faster pulse risetime could reduce the trace from, say, 6 inches down to 4, making accurate production measurements on traces of less than 3 inches regardless of system risetime may reduce the measurement accuracy and repeatability.

“But surely a faster rise time will give a better indication of performance at higher operating speeds?”

The answer again is “it depends”; for example, if your trace has mimimal losses then the transmission line should be broadly frequency independent, the only changing parameter being the dielectric constant of the substrate, and even then Er will only influence the impedance in a 1/Sq root Er relationship, so the change with frequency is typically minimal when compared with other drivers of impedance.

“I need to test at a specific frequency”

TDR testing is inherently broadband in nature and while, yes, in a 35ps TDR pulse there is more high frequency energy than in a 200ps pulse, the higher frequency components are a tiny fraction of the overall energy. It is possible to extract more detailed frequency information; however, to do this the reflection must be heavily post processed and converted to the frequency domain in order to extract the s-parameters which will accurately describe the impedance vs frequency characteristics, including the transmission line losses. Simply measuring the reflection with a faster rise time on a lossless transmission line will return the same result as with a slower rise time.

If loss measurement or measurement at a given frequency is a requirement, then either VNA, or Lab TDR with s-parameter extraction (for example, a 35ps lab TDR equipped with Polar’s upcoming Atlas GHz test software) becomes a necessity; however, with this comes the overhead of higher running costs through high frequency probe and cable assemblies along with the likely increased incidence of TDR damage or degradation because of either gradual or catastrophic ESD damage, especially in heavy production use. There is certainly a necessity for this type of test, especially when producing boards using chipsets with ultra high speed differential signaling, and the increased cost of ownership is just one part of the required investment for higher speed devices.

However if a PCB is simply operating high speed traces in the upper MHz or low (i.e. 2 to 3) GHz frequency range and the design employs traces with insignificant losses, then the deployment of lab TDRs in the production environment could simply be costing money with no measurable gains in either measurement information. The case then becomes economic, - does the saving in coupon real estate outweigh the increased cost of ownership of using a lab TDR for lossless controlled impedance traces in a PCB manufacturing facility? Some of our customers report up to 10x the incidence of TDR failure when unprotected lab TDRs are deployed compared with testing on slower TDRs with inbuilt protection such as the Polar CITS900.

Examples of measurement correlation between a TDR set for 35ps rise time and 200ps risetime may be seen in application note AP168

In summary:

  • If you need to test lossless PCB impedance on 4 to 6 inch traces a 200ps or 35ps TDR is equally capable, however the 35ps system is likely to have significantly higher running costs.
  • Testing impedance on traces less than 3” long may result in a loss of R&R and measurement accuracy, regardless of TDR risetime.
  • Simply testing impedance with a faster risetime will not give any more information about the impedance at a given frequency.
  • If you need to extract information on trace LOSSES, then there is a definite requirement to use a fast risetime, but the resultant TDR measurements will need post processing into the frequency domain (for example with the upcoming Polar Atlas GHz test software package) and…
  • The techniques to extract valid measurements for loss require traces which exhibit adequate losses and these will be in the region of 4 to 10” long depending on materials in the stackup.
  • The economics of testing with fast TDRs on lossless controlled impedance traces (typically those up to 2 to 3GHz and a few inches long) are dependant on calculating the tradeoffs in additional running costs of Lab TDRs vs the saving of a couple of inches of PCB real estate per panel.