How to select a shunt resistor for high bandwidth current measurements

Selecting the right shunt resistor is critical to achieving accurate, high bandwidth current measurements. From power dissipation and insertion inductance to bandwidth optimisation and compensation techniques, careful shunt selection can dramatically improve signal integrity, dynamic range, and measurement reliability.

By Seamus Brokaw, Application Engineer, Tektronix

Shunt resistors are necessary for many applications requiring higher bandwidth and lower inductance current measurements than a Rogowski or clamp-style current probe can offer. Understanding the resistor’s properties and careful selection from the wide variety of available shunt resistors will ensure accurate, wide-bandwidth current measurements that minimize parasitic inductance and maximize Signal-to-Noise Ratio.

The selection process involves several interconnected considerations:

• Target current range and dynamic range requirements

• Power dissipation limits and thermal management

• Resistance value optimization for signal amplitude and burden voltage

• Form factor selection based on space, power, and bandwidth requirements

• Bandwidth limitations due to parasitic inductance

• Test point implementation strategies

Current range identification
Target Current Analysis

The target current RMS value, peak current, and ideal bandwidth are the first place to start exploring shunt options. These factors will all affect the downstream variables like power dissipation of the shunt, burden Voltage, the resistor value, insertion inductance, and compensation strategy.

In the ideal case, a high value resistor with near-zero insertion inductance has infinite power capability, leading to perfect signal:noise ratio, several GHz bandwidth, and no impact to the device-under-test. In reality, compromises must be made for the shunt’s power capacity and the DUT’s burden Voltage and inductance sensitivity.

For example, identify an input current that averages 8 A RMS with occasional spikes to 12 A and a 40 mA signal-of-interest riding on top of the 8 A. These values will factor into the shunt selection, as we will see later.

Power limits and thermal considerations
With a known target current, browse for current shunts in values that would lead to a power dissipation below 3 W for larger CVRs and 1 W being a safer limit for SMD devices. This power limit will determine the resistance values that you can select from, with higher resistances providing more signal to measure and a lower noise floor at the tradeoff of more power dissipation and a higher burden voltage dropped across the resistor.

This table shows how different shunt values (Column A) can make the same 8 A measurement with varying Power levels (Column D). Higher resistance values lower the noise floor (Column B) at the cost of more power dissipation in the shunt.

For example, an 8 A RMS signal with a 40 mA signal of interest riding on top of it will require that the resistor be able to handle P=I²R power dissipation in its body. With a 1 W shunt, the maximum resistance value will be R = 1 W / (8 A)² = 15.6 mΩ. Sourcing a 3 W shunt raises this limit to 46.9 mΩ and will generate more signal to measure, raising Signal:Noise ratio and making that 40 mA signal-of-interest come further out of the noise floor.

Pulsed vs. continuous current
With a signal that is periodic, rather than continuous, significantly higher current levels can be achieved even with a nominally 1 W resistor. The 5 mΩ Wideband Shunt from Tektronix can reach 200 A square or triangle pulse for a 10 µs pulsewidth where it’d normally be limited to 14 A RMS for a continuous signal.

Form Factor Analysis

SMD resistors
SMD Resistors come in a wide variety of form factors and Ohm values, leading them to be a flexible and affordable option for many applications. The tradeoffs of SMD devices is their limited power dissipation capabilities and uncompensated frequency response.

Shunt Type Comparison

Inductance for SMD resistors can be low to medium, often landing in the few nano-Henry range. The form factor has flexible options with the caveat that as the width of the device increases relative to the length, the resistor starts to look more like a distributed transmission line and less like a lumped element. Impedances vary based on the probing position and therefore currents start being distributed unevenly across the SMD resistor.

Bus bar types
Bus bars have a larger size than SMD devices and therefore can dissipate much more heat. This can be a big advantage when in a high-RMS current measurement such as a 100 A continuous DC current with a smaller, 1 A transient of interest riding on top. The resistor needs to dissipate the power created by the full RMS current, not just the smaller signal of interest. These devices also are made of materials that have stable resistance over temperature and can have inductance values in the 10s of nH range, barring the same transimpedance issues that start to appear on large geometry resistors, similar to the larger SMD resistor devices.

Their larger size means the designer often needs to think ahead and design in these resistors to the PCB layout. If your traction inverter or motor drive application already uses current sense devices for over-current protection or short-circuit detection, building in the bus bar shunt and adding an additional test point for a high performance current measurement is much more feasible than an impromptu test point with the same shunt.

Wideband shunts
One shunt designed to work with the Tektronix Isolated Current Probe and MSO oscilloscopes are the Wideband Shunts. This design interfaces with square pins on the board and features frequency compensation, temperature compensation, and high-performance fuses to protect the test equipment. Values range from 5 mΩ to 5 Ω, providing 250 MHz of bandwidth and easy implementation onto the board.

Note for Wide Bandgap devices, the square pin Wideband shunts add several nH of insertion inductance. If placed in the current loop, that could affect the device performance and make measurements on the lab bench differ from simulation or production devices.

Ultrafast current shunts (and other new entries)
New current shunts from test and measurement companies are launching every year to improve on off-the-shelf resistor performance. One is the Ultrafast Current Shunt (UFCS) from PMK. In testing, the UFCS shows several-hundred MHz bandwidth with a desirable rolloff, indicating it is a frequency-compensated design. It performs well with low insertion inductance and somewhat high 3 W power dissipation capability. The challenges with using the PMK device come from its large form factor. The low insertion inductance requires a wide space on the test board and the shunt is also tall, presenting positioning challenges in crowded enclosures.

This device is well-suited for Wide Bandgap current measurements where the insertion inductance is paramount and compromises can be made to fit the shunt onto the test board. It’s cost is more in line with a measurement device than a mass-produced SMD component, so it’s also more suited for R&D lab benches than production tests.

Bandwidth and compensation
At a first order level, a shunt resistor can be modeled as a single pole RL circuit: a resistor and inductor in series. The inductance is caused by parasitic design of the shunt resistor and has a large effect on the effective bandwidth of the current measurement. This basic model ignores transimpedance and coupling effects which can be significant but are very difficult to characterize and model correctly.

Bandwidth calculation

Corner frequency is correlated with R and inversely correlated with L. That shows how the resistor value selection and form factor will greatly impact the measurement bandwidth. Here is a simulation showing the effect of resistance value and 800 pH of inductance on the frequency response. The corner shifts with increased Resistance.

Uncompensated Frequency Response

Compensation network design
The parasitic inductance can be compensated using an RC network where:

After compensation, frequency response can be flattened and extended for an order of magnitude or more! Notice also how the addition of a steep rolloff element can change the deceiving high frequency amplification in the original chart to a more desirable HF rolloff.

One important note is that although the measurement inductance is improved through compensation, the DUT is still experiencing the same insertion inductance caused by the shunt. It is only in the measurement that the effects of the inductance are compensated.

Resistor value selection

Compensated Frequency Response

Design Trade-offs
The resistance value selection involves balancing several competing factors:

• Power Dissipation: Higher resistance increases power dissipation in the shunt (P = I²∙R)

• Burden Voltage: Higher resistance increases voltage drop across the shunt, lowering final rail. Voltage experienced by the DUT. This factor can matter on low Voltage processors that have a 5% Power Rail tolerance. (e.g. ±40 mV tolerance on a 0.8 V core Voltage power rail)

• Signal Amplitude: Higher resistance provides larger signal amplitude, raising signal:noise ratio

• Measurement Bandwidth: Higher resistance and lower insertion inductance will both improve measurement bandwidth.

Selection Methodology
Now familiarized with the tradeoffs associated with different current shunts, select a device based on what is mechanically possible and then optimize for measurement bandwidth and device insertion inductance.

Calculate Maximum Acceptable Burden Voltage
With good decoupling capacitor selections, the burden Voltage imposed by the shunt will be limited by the average, or RMS, input current.

Example: a 40 mV Voltage budget with 8 A RMS average current allows a 5 mΩ resistor.

Calculate Power Dissipation
The next limiting factor is power dissipation in the shunt.

Example: The 5 mΩ resistor experiencing 8 A RMS current needs to dissipate 1/3 W heat continuously. That Power value is easy to find in SMD devices and trivial for larger form factor bus bars, wideband shunts, and other dedicated current viewing resistors.

Verify Signal Amplitude and Check Noise Floor
After finding some devices that will survive the current measurement, the next task is to optimize the measurement for dynamic range, bandwidth, and insertion inductance.

Dynamic range is the ratio of the largest Current that can be measured to the noise floor of the measurement device, in dB.

Example: The same 8 A RMS current has occasional impulses up to 12 A which need to be measured. On a 5 mΩ shunt, that creates a 60 mV measurement. Using a TICP probe in its ±45 mV range allows this measurement when at least 15 mV of offset is applied. This TICP input range with a 5 mΩ shunt results in a 13.3 mA RMS noise floor.

If this same measurement were performed on a 1 mΩ shunt, the noise floor raises to 66 mA RMS and Dynamic Range falls to 45 dB.

Check Measurement Bandwidth

Resistor value selection affects dynamic range and it also affects another critical measurement specification: bandwidth.

As shown in section Bandwidth and Compensation, both resistance and insertion inductance impact the measurement bandwidth. Compensation helps but the effectiveness is limited by the signal coming off of the shunt in the first place.

For these reasons, a resistor with the largest value that is possible is always recommended for current shunt measurements. Also for bandwidth, Lshunt of the device should be minimized. This can be accomplished through resistor selection, PCB layout, and test point strategies.

SMD resistors by themselves could have several nanoHenries of inductance but try putting four to eight in parallel to drop the inductance. This approach has limits as current stops being distributed evenly at wider form factors, but it does work to lower the insertion inductance.

Newer shunts, designed to optimize insertion inductance, have accomplished more and some designs also include compensation networks to further tune the measurement frequency response.

Shunt Selection Summary
Choose the highest value Resistor that will survive the RMS current while being minimally invasive to the DUT. Then choose a resistor that is designed for lowest insertion inductance.

Test point implementation
Ideally the shunt resistor, compensation network, and test points are all built into the board design. In the second best case, a shunt resistor already exists on the board. And finally, there are cases where no measurement point exists and the entire chain needs to be added.

Implementation Strategies

[1] Integrated PCB Design: Best performance, requires early planning

[2] Existing Shunt Modification: Quick implementation, may degrade signal integrity.

[3] Complete Addition: Maximum flexibility, highest complexity

Connector choices
In addition to the shunt location and compensation, the physical connection to the device will also impact measurements through their sensitivity to radiated emissions. In the examples above, MMCX, square pins, a coaxial cable with MMCX connector, and twisted pair with square pins were shown. These different methods have different levels of radiated emissions sensitivity and care should be taken to choose the best connection that is realistic and possible for each measurement.

Shielded test points using MMCX, SMA, and coaxial cables will perform the best, as the outer braid provides effective shielding to the center conductor. Even better than coaxial is a twinax cable as used with the Tektronix Wideband shunts. Twinax cables put both conductors inside a third, braided shield—protecting both positive and negative conductors from EMI with the added benefit of true differential loading down at the DUT.

Conclusion
Selecting the appropriate shunt resistor for high bandwidth current measurements requires careful consideration of multiple factors including current range, power dissipation, resistance value, form factor, bandwidth limitations, and test point implementation. The selection process involves balancing competing requirements to achieve optimal performance for the specific application.

[1] On this board, the Is_Lower testpoint at MMCX connector J3 is designed into the board, complete with compensating pole at C1, R32, and R20.