Testing a MICSIG CP2100B Current Probe

Background

A colleague of mine bought a MICSIG CP2100B Current Probe and asked me to test it with some adapters I made for a Keysight N7026A 150MHz Current probe as a reference (beware, this one is about 5K€ without VAT).

The CP2100B has a good price performance ratio, is USB powered (no batteries) and measures up to 100A (ranges are 10 and 100A). 3dB Bandwidth is 2.5MHz, and rise/fall time is about 175ns. Its main use is probably the debugging of switching power supplies, PFC circuits, motor controls and the like. The probe is shown here:

Dave Jones has discussed this probe and made a teardown in his EEVBlog Youtube Channel.
Click here to see Dave Jones’ CP2100B review and teardown
Dave also sells this probe (I guess in Australia).

Testing a current probe so you get repeatable measurements is less simple that it might look in the first place. Some of the pitfalls are:

1. At higher frequencies (and steeper signals), you need a “defined” environment, ideally without a lot of unknown stray inductances, capacitances and couplings. This translates to a “test cell” with fixed geometry, not dangling wires. The differences in rise and fall times measured with a defined environment and loose wires can easily sum up to 200-300ns, not what you need for a probe with a 175ns risetime spec.
2. At lower currents, current probes tend to be noisy. You would want to provide some electric (but not magnetic) shield around the head of the probe.
3. A good test for probe rise and fall times is  a high performance square wave generator. Needless to say, generator rise and fall times ought to be considerably shorter than probe rise and fall times. For the MICSIG CP2100B, this is no challenge (standard square wave generators like the Rigol DG1062Z have risetimes of ca. 10ns, the probe has a spec of 100ns, so no issue. For the 150MHz N7026A, we have a different story).
4. The setup should be free of reflections caused by mismatch. I use a coax feed to the test cell (a Hammond box with two BNC sockets and a short straight wire connecting them, just long enogh to attach the probe), and behind the test cell I have an attenuator and a scope with a 50Ohms termination. For the 2.5MHz MICSIG, a 100MHz scope is fine, for the Keysight N7026A we need 1GHz or better, preferably. The scope can verify that the current probe does not impose any excessive distortion of the square wave signal.
5. Although a probe might be designated as “up to 100A” or so, this is a low frequency limit. If you ever try this a RF (inside a transmitter or the like), the limits are much lower and if you measure 10A at some MHz your probe will be dead instantly. Check the datasheet, and dont try this !
6. For low level measurements (low in relation to the maximum current the probe can accomodate), current probes generate a lot of noise. Averaging is the way to get around this. If you are looking for single events, this is not a possibility and you are out of luck (and you have to buy a more expensive probe).

The Test Setup

The idea is to create an “ideal” environment for the CP2100B and the N7026A by passing the same current thru two shielded enclosures with just a bare conductor between two BNS sockets inside. The square wave generator is put to 100kHz, 10Vpp into 50Ohms. Rise and fall times are below 10ns according to the DG1062Z datasheet. The setup is shown here:

We then measure the CP2100B output signal, the N7026 output signal and the passthru signal via an 10:1 50Ohms attenuator and 50Ohm scope input channel. The shielded adapters are shown in more detail below:

The risetimes with this setup are shown below (measured between 20 and 80%, discussed later):

Blue is the waveform in the 50Ohm terminated scope input, green is the N7026A and yellow is the CP2100B.

And the fall times are here:

As expected, the terminated voltage and the N7026A are very close (the N7026A has a risetime of below 3ns, considerably faster than the generator). The CP2100B has a bit above 100ns, which is better than its spec at 175ns. For the fall times we have similar values. The CP2100B has some “lazy” settling that would result in a lot longer rise and fall times if a 10/90% limit rule is applied (more than 200ns). The N7026A shows no settling issues. For a budget probe the CP2100B rise and fall times are acceptable in my view.

 

A More Realistic Practical Setup

Now lets try what happens if we measure current using loose wiring instead of shielded test cells (The same way Dave Jones does it). The measurement environment now looks like this:

This is not exactly an extreme example, if you compare with what Dave Jones is doing in his test. The total unshielded wire length was less than 20cm (BNC connetor to BNC connector), which is not a lot.

The shapes of the signals change significantly. Rise time is here:

… and fall time is here:

Things are no longer pretty now: While the N7026A shown no visible changes, the CP2100 exibits under- and overshoot and has a total risetime in the order of several hundred ns, plus a slow settling. The fall time curves show the same issues. Even with benevolent judgement we have about double the specified times, plus ringing.

 

Lessons Learned

The effect of the measurement environment on waveforms is surprisingly large, even for a low-frequency probe like the CP2100B. What can be learned:

1. If you want precise measurements, use a defined environment (shielding, short wires, …). A good idea could be “current test ports” in your circuit where the current probe just fits over it.
2. If you must measure using open cabling, do not rely on published bandwidths, risetimes, over- and undershoot. Its probably safe to divide bandwidth by a safety factor of ca. 5, and to not believe in signal waveforms shorter than several times the published rise/fall/settling time.
3. The CP2100B is good to measure switching PSU waveforms up to some 10kHz, but not a lot higher. My absolute limit would be ca. 100kHz. Its still a usable probe for this, especially for the price tag of ca. 375€.
4. Noise in the CP2100B is not a big issue, because it can be eliminated by averaging. Measuring 10mA of current works fine (after zeroing, …).

Conclusions

The CP2100B is OK for switching PSU work and similar tasks not involving too high frequencies (> 100kHz fundamental). It does measure DC, by averaging 10mA currents can be displayed without excessive noise. Not a bad bang for the buck. BUT – be aware of what it can do and what not.