# Does a BNC splitter introduce defects into the signal?

I'm currently doing some contract work that includes finding and recommending a signal generator model for frequency measurement device verification. The verification scheme requires generating a square signal of sound frequency (20-200 Hz) with small tolerances (10-6 relative frequency error) that goes to two devices - the one being verified and a reference one. The devices then count the impulses over a slice of time. Other properties of the signal specified are slope length and voltage (maximum and mean-square).

The official paper that I'm referring to for the technique uses a two-channel generator for an example setup (but doesn't specify that it's necessary). I'm thus wondering if I can use a single-channel generator with a BNC splitter for the same effect.

That is:

Does a BNC splitter introduce any noticeable defects into the signal - e.g. reflections, distortions etc? The only thing I can readily think of is due to the conservation of energy, the voltage may be two times (or rather, sqrt(2) times) smaller - but that is trivially fixable.

• A Coax cable + BNC connectors is basically a transmission line so start reading here: en.wikipedia.org/wiki/Transmission_line Any change in the transmission line (a connector, bend in the cable, splitting the signal) causes reflections. Question is, how much can your system tolerate ? So the question is not if a BNC splitter will affect the signal because it will. Question is, what can you tolerate. Yes that is difficult to answer. Commented Jun 6, 2017 at 8:21
• "Yes that is difficult to answer" - and that's exactly why I'm asking since I didn't readily find any relevant characteristics for the splitters online. Commented Jun 6, 2017 at 8:23
• If the official paper goes on to say something like "now, adjust the frequency to be verified..." then you're going to need two generators anyway. Commented Jun 6, 2017 at 8:27
• I didn't readily find any relevant characteristics for the splitters online Hmm, but this is not so much a property of the splitter itself but more of how you use it. That makes me agree with Marcus's answer, you should get someone with experience in this field to do this/help you. This is a specialist area, experience is needed. Commented Jun 6, 2017 at 8:39
• 20 to 200Hz is well below any frequency at which a BNC connector would be characterised. If you find values for insertion loss or VSWR or other characteristics, they will normally be given for a specific range of RF frequencies. Low audio frequencies are just really not of interest for that kind of thing. If your audio frequencies have some insanely sharp flanks, you would need to look at the RF characteristics. For any reasonable audio work, it shouldn't matter.
– JRE
Commented Jun 6, 2017 at 9:23

While I don't expect any nonlinear effects that would change the (dominant) signal frequency, the reflections at each interface on your cable will introduce a set of delayed, attenuated copies to your signal.

Whether or not that is bad for your application can't be answered from the data you give.

As a general comment:

I'm currently doing some contract work that includes finding and recommending a signal generator model for frequency measurement device verification.

Doesn't sound like you are the perfect person to do that. Maybe look for a sub-contractor with experience in handling of delicate measurement references. That's not something you want to botch yourself.

• Could you specify as to the order of magnitude of these reflections? While if they're bad for the particular application depends on how tolerant the receiving equipment is (so you can't answer that), the figures themselves only depend on the setup, don't they. Commented Jun 6, 2017 at 10:31
• no, I can't, which you'd know if you had the least understanding of transmission line theory. This project might be over your head. Commented Jun 6, 2017 at 11:47
• FYI, thank you, I completed the work with flying colors. Included 1-channel generator models into the result and just told the client that there are going to be reflections if they are choosing them and let them figure out whether they'd be bad enough or not since doing that wasn't part of the assignment. Commented Apr 17, 2020 at 12:52
• @ivan_pozdeev congratulations! I'm happy you have such a flexible client. Mine want solutions. Commented Apr 17, 2020 at 12:54

The BNC splitter will introduce impedance mismatch when used with two 50-Ohms coaxial cables. But if you would use only sine waveforms in your range (200Hz), the BNC splitter will have no effect.

The effect on sine waves might start at much higher frequencies, and will depend on length of cables. If other ends are poorly terminated, the cable system might experience resonant phenomena, and amplitude of sine (standing) waves might vary by a factor of 2X-3X depending on frequency, depending on wavelength of the wave relative to cable length. With cables of, say 5m, with one-way propagation delay of 30 ns, the effects will start at 15-30 MHz.

However, you did mention a square waveform, and measuring "slope length". In this case the result will depend on initial edge rate of your generated signal (which you didn't disclose). The effect of mismatched impedance will show up as reflections, which will distort edges, the edges will develop a "shoulder", they might become non-monotonic, and the concept of "slope" might be difficult to apply (and therefore measure). If cables are 5m long, reflections will show on 30-ns timescale. However, if your edges are 1us long, there will be no noticeable effect.

• An example reference signal mentioned in the paper is 100Hz pulse rate, 200us pulse time, 40us rise/fall times, 3.8V amplitude. Since this is going to be a test bench, cables aren't likely to be longer than 2-3 metres. Commented Jun 6, 2017 at 17:50
• What I called "slope length" is rise/fall time, I didn't find the proper English term. Commented Jun 6, 2017 at 17:54
• @ivan_pozdeev, if your rize-fall times are ~50us, 2m cables will have no effect, as I explained. But then it becomes unclear what you are trying to achieve having 50-200-50 us timing. Why would the slope matter then? 10-6 out of 100 Hz is 10 ns. Are you trying to time 50us edges with 10ns accuracy in an analog-like setup? Commented Jun 6, 2017 at 18:13

Sounds like you want to detect tiny changes in frequency, by monitoring the output of a high-speed EXOR gate, with a delayline inserted in one of the legs.

Consider the period of 100Hz.....10millisec, 10,000 microsecond, 10,000,000 nanosecond. Thus delay changes of 1nS to 10nS might be your goal?

I'd be more concerned with the two outgoing pieces of coax cable, each 50 ohms thus 25 ohm in combo; Can you buy some 100_ohm coax?

If still applicable after 1.5 year! To solve the problem of obtaining two close-to-identical broad-band signals, keep in mind that common braided coaxcables are highly sensitive to any mechanical deformation like bending. Also, BNC connectors are unstable and susceptible to touching etc. So, despite I do not see what actually you, Ivan_, want to measure (most likely phase variaytions???), to generate the two signals start from a 1-channel generator of 50 Ohm output impedance, split the output using a RESITIVE SYMMETRIC 2-port broad-band splitter (available from many companies at decent prices), add two (preferably semi-rigid) 50 Ohm cables to your two items under test using 50 Ohm terminations at the two feed points. Wile measuring do NOT TOUCH any cable o.t.l! If done in this way phase variations can be reduced to sub-ns level, so well below your 10^-6 level relative to 10 ms (inverse of 100 Hz). Hope this brings any light to the problem.......