Close

Lots of ferrites

A project log for Modular differential probe

100 MHz differential probe for oscilloscopes, with modular accessories to suit both high and low voltage use.

petteri-aimonenPetteri Aimonen 08/18/2021 at 11:070 Comments

Every time a point in circuit has both inductance and capacitance without enough damping resistance, there will be a resonance. A common design approach is to try and keep these resonance frequencies away from the signal frequencies you are interested in. But with a wide band design, such as this probe, it is often not possible.

Simplest way to damp resonances is to add a resistor. I've found Okawa RLC design tools especially helpful in quickly trying out different values. But resistors are not frequency selective, so they always affect the normal function of the circuit.

An alternative is to use ferrites, which are lossy inductors. Their frequency response depends on the material used: some are effective starting at 1 MHz, others only above 100 MHz. This allows careful selection to dampen at the resonant frequencies without affecting the signal frequencies too much.

So far I have three main uses for ferrites in this design: power supply filtering, input resonance dampening and common mode resonance dampening.

Power supply filtering

The amplifier circuit requires two-sided supply, in this case ± 6 V. This is generated from +5 V using a 1.4 MHz switching mode boost converter, which necessarily causes some ripple when it switches.

The ± 6 V rails are filtered with two stages of series ferrites and parallel capacitors. Because the fundamental frequency is low, I chose the lowest frequency ferrites I could find, specifically MMZ1608B601, combined with 10 µF ceramic capacitors. These do a fairly good job of filtering out the low frequency ripple and very good job of filtering out the high frequency harmonics. This is important because the opamp power supply rejection ratio gets lower as frequency increases.

But filtering the SMPS output is not enough. The +5 V input goes alongside the signal cable, and usually comes from the same oscilloscope that is doing the measuring. Large current spikes on input supply could couple into the output signal. To avoid this, also the input has two stages of ferrite filtering. But because the input current is twice as high as the output current of each rail, a ferrite with higher saturation current is needed, specifically BLM18KG102. Input from cable goes directly to first ferrite - if there was an initial filter capacitor, it would minimize voltage noise but could cause current noise when connected to a noisy power supply.

Finally, after a bunch of testing, I ended up using a split ground plane for the SMPS portion of the PCB. The power and signal grounds are connected together with a ferrite, and last stage of ± 6 V filtering is against the signal ground. This seems to minimize the noise spikes that get coupled to the amplifier, though I'm not entirely sure of what kind of coupling is at play here.

Input resonance dampening

The input cables have their own inductance, which forms a resonance when combined with the input capacitance. This can be reduced by keeping the cables as short as possible and close together, but practical considerations often get in the way.

A common trick is to add a series resistor to dampen the resonance. This forms a low-pass circuit, which causes the frequency response to droop at much lower frequencies also.

An alternative is to add a ferrite ring around each of the input cables. Because of the frequency selectivity of ferrites, this often has less effect on the lower signal frequencies.

Above, 2x 15 cm input cables are measured. Blue curve has a 50 ohm series resistor, while yellow curve has both a resistor and a ferrite.


Common mode resonance dampening

Similarly to input cables, the output cable between oscilloscope and the amplifier has inductance. Because it is longer, the inductance is higher and the resonance peak ends up lower, at about 40 MHz. The coaxial cable acts as a transmission line, so the inductance does not directly affect the output signal, but it does affect the potential difference between oscilloscope ground and the amplifier ground, i.e. the common mode voltage.

Because the amplifier's common mode rejection is not perfect, the resonance peak shows up as increased or reduced gain at that frequency:

Again, a simple ferrite around the output cable fixes the resonance. Because all conductors go through the ferrite, it only affects the common mode voltage and does not affect the frequency response for the signal. This also improves the common mode rejection at high frequencies:

Discussions