I have a question regarding best practices for function generator to circuit/scope connection.

I recently had an issue chasing down excessive ringing on square wave signals coming from a function generator set to Hi-Impedance mode (output is 50 Ohms, setting only affects the reading). The connection at the generator was a BNC T-connector, with one side connected via 3ft of pomona coax to an oscilloscope with no termination (just the standard 1M input impedance), and the other end was connected to a 3ft pomona mini-grabber for circuit connection. According to "The Art of Electronics, 3rd. Ed." Appendix H.1.2 and H.2 on transmission lines, terminating the cable at the source (series/backtermination through the built-in 50 Ohm at generator output) should be enough to eliminate wave reflections, and terminating both ends is not required.

After contacting the manufacturer about the excessive ringing, they initially diagnosed this as a faulty generator, but were able to reproduce the same results on other generators with the same setup, so recommended not to use a BNC-T if I want to use HiZ mode, but to just probe the circuit input with a scope probe. I find the more permanent connection via T more convenient.

So is there some kind of best practices here that people follow to assure clean square waves from their generators to the DUT? Is it generally not a good idea to BNC-T the connection from generator to scope? And/or do both ends need to be terminated, contrary to the position of AoE (although it does mention it as a "just to be safe" measure)?

UPDATE: coming back to this after some time and hardware changes (no T-connectors, just single 50 Ohm coax):

Image 1) a 10MHz square wave using 50Ohm coax, terminated with a 50Ohm in-line BNC terminator and connected directly into the scope. The result is decent. enter image description here

Image 2) same as 1, except using a 10M probe to measure the BNC output of the feedthrough resistor. enter image description here

Image 3) same as 2, except with a short bnc-to-mini-grabber adaptor connected to the 50Ohm feedthrough resistor. Scope probe is connected to mini-grabber tips. enter image description here

Image 4) same as 3, except 100kHz square wave instead of 10MHz. enter image description here

Anytime the mini-grabber is connected, I get massive overshoot on square waves at any frequency, even with proper termination. What is the cause of this, and how do I remedy it to get a similar wave as shown in image 1 to appear at the input of my test circuits (without having a board made with a BNC connection)?

  • \$\begingroup\$ Transmission line effects are a complicated beast. Discontinuities in impedance always lead to the field behaving in unwanted ways, mostly reflections. A T in a transmission line is usually such a discontinuity, even just a connector is. \$\endgroup\$
    – PlasmaHH
    Commented Dec 8, 2017 at 19:59
  • \$\begingroup\$ What is your 'feedthrough' resistor? Is it a center-to-shield resistor terminator, or (like some) is it a center-to-center series resistor? If the latter, it is NOT terminating the function generator signal, unless your probe is connected to very LOW impedance points (not usually the case). \$\endgroup\$
    – Whit3rd
    Commented Jan 13, 2018 at 2:15
  • \$\begingroup\$ Center-to-shield, the datasheet is linked in the description above \$\endgroup\$
    – User7251
    Commented Jan 13, 2018 at 4:07

3 Answers 3


The generator has a source impedance of 50 ohms. For best results it should be connected to the device under test with a 50 ohm coax cable which is then terminated with a 50 ohm load. Without the 50 ohm load, the generator is essentially unterminated and the signal will reflect back since it sees an impedance mismatch. The best termination is a 50 ohm feedthru but a T connector will suffice if the feedthru is not available.

  • \$\begingroup\$ According to AoE, the signal will reflect back but will be absorbed by the 50 Ohm resistor at the generator, eliminating the need for a second (load side) termination. They call this "series" or "backtermination". I have a feedthrough 50 Ohm terminator, but sometimes using Hi-Z is more applicable for the larger amplitude/offset capability. \$\endgroup\$
    – User7251
    Commented Dec 8, 2017 at 21:11
  • \$\begingroup\$ Look at the difference in the scope signal when it is not terminated and when it is terminated. You already said there was ringing without any termination. \$\endgroup\$
    – Barry
    Commented Dec 8, 2017 at 22:01

Put to the input of your oscilloscope a T-BNC connector. Extend the cable to your circuit from it. Have a matched 50 ohm load at the circuit. If needed, add a parallel or serial resistor for proper matching at the circuit end. Without it the reflection stays and your oscilloscope can show different voltage than your circuit gets. You can also have an attenuator at the circuit end of the cable.It gives some insulation between the instruments and your circuit. The signal strength is unsure, if there's poor matching.

I personally try to connect the oscilloscope to the circuit with Hi-Z probes, if it's possible.

  • \$\begingroup\$ Interesting, why exactly do you suggest to using the T at the scope instead of the generator? To specifically terminate the ends of the signal chain as opposed to the middle? Is there a difference? \$\endgroup\$
    – User7251
    Commented Dec 8, 2017 at 21:15
  • \$\begingroup\$ @User7251 The 1Megohm input of the oscilloscope is a quiet bystander, only one matched termination is needed for reflectionless operation. \$\endgroup\$
    – user136077
    Commented Dec 8, 2017 at 21:24
  • \$\begingroup\$ Right, but how does that explain the difference? If the termination is in the middle of the cable vs. the end you still have one termination. \$\endgroup\$
    – User7251
    Commented Dec 8, 2017 at 21:34
  • \$\begingroup\$ @User7251 The 1Megohm input of the oscilloscope is not a termination, it watches the happenings in the cable without any substantial role. My answer suggest the proper 50 ohm terminations to be at the ends of the line - one as automatically inside the generator and one is needed to be built at the DUT. My solution is not good if the frequency is in VHF or UHF range or you have nanosecod rise times. The input of the oscilloscope has some capacitance, too and this can cause substantial mismatch in high frequencies. \$\endgroup\$
    – user136077
    Commented Dec 8, 2017 at 21:43

A 3-foot section, out and back, with assumed Er of 4, has electrical length of

3-feet * 2x (out,back) * 2x (sqrt(Er)) 3 * 2 * 2 = 12 nanoseconds.

If your generator rise time is any faster than 5 or 10 times that 12 nanoseconds, you will see degraded pulse flatness (aka overshoot).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.