# How to pick a scope and probe to measure a 12.5Gbps signaling frequency?

I fully realize that this signal frequency range is high enough that we do not have scopes and probes today to actually measure the rising edge and see it. This question is a hypothetical question to understand how to pick a scope and probe.

This is my understanding / answer / math for the problem, is it correct?

Signal Frequency, Fsig = 12.5 Gbps Time Period, Tsig = 1/12.5 Gbps = 80e-15 (80 femto second) Rise time, Trise = Tsig/2 (?) = 40e-15 (40 femto second) Knee Frequency, Fknee= 0.5/Trise = 0.5/40e-15 = 3.125e+13 Hz

And we do not have probes or scopes with that high of a bandwidth (3.125e+13 Hz), right?

I want to know if my calculation is correct?

• A 12.5 GHz sinewave actually has a period of 80 ps (80 E-12), not 80 fs. Also there is no need to calculate the period, as Tony says you will probably need at least 5x that 12.5 GHz so 60 GHz is is a minimum. Nov 13, 2016 at 18:22
• Usually DSO BW is (0.35 to 0.45 )/Tr and should sample at least 5xf for quality, thus >60GHz. Nov 13, 2016 at 18:22
• Are there probes for that high frequency though? Nov 13, 2016 at 18:23
• LeCroy has a 100 GHz scope: teledynelecroy.com/100ghz At these frequencies you cannot use a probe in the same way you can use a probe with low frequencies. You need to think RF so power splitters and properly teminated transmission lines/coax cables are needed. That would be a 50 ohm system, not 1 Mohm like a normal low frequency scope probe. So no: there are no probes for that as these would be unusable anyway. Nov 13, 2016 at 18:29
• Have you looked on the websites of LeCroy, Keysight, Tek? What did you find? Nov 13, 2016 at 18:50

To begin, 12.5 Gbps signaling link has the main frequency of 6.25 GHz. The unit interval is 80ps, true, but the period of signal is 160ps.

Now, all modern designs of communication links are more concerned with reducing parasitic emissions, so the signals are designed to have as low edge rate as possible, just enough to deliver a decent eye opening. So 1/10th of the signal rate is plenty, typically the edge transition makes about 1/3 of the Unit Interval (or about 25ps for 12,5 Gbps link). And silicon designers must work really hard to get edges faster than 20ps. So a 40GHz real-time scope should be fine.

If a non-monotonicty of egdes are still of concern, sampling scopes can do a better job, see this presentation from Agilent/Keysight. This was 10 years ago. You can find a used Agilent 86100C for just $20,000 -$30,000.

Since 12.5 Gps optical links use delay compensation to optimizing timing margin or eye-pattern from ISI, the signal uses more bandwidth than 1 Hz /baud on the signal chain. This is why 5x is a minimum for optimizing BER.

There are different categories; scopes, adaptive EQ scopes, adaptive scopes with Bit Error Rate Testing, BERT only with parameters, TDR included with s-parms

Since timing margin, Q factor and BERT are all related a scope can help isolate margin loss for a given budget to each factor that affects BER in terms that can be directly related to BER.

I have not used equipment at this rate but used many at lower rates. The most useful I found measure timing margin histograms such that dependencies for loss of margin can easily be separated by pattern, line loss, asymmetry, skew, path dispersion, spectral effects, compensation EQ effects.

So look for 2 instruments, if you are serious, a good all-in-one tester from Tek and test before you buy.