The Pi attenuator calculator above designs a symmetrical resistive pad: a three-resistor network that reduces a signal by a chosen number of decibels while presenting the system impedance at both ports. That second property is the whole point. A potential divider also reduces a signal, but it wrecks the impedance match at both ends, and in a transmission-line system a bad match produces reflections that are often more damaging than the level was.
Arb Digital publishes free engineering calculators that give exact figures and then say what a real build does to them. This page returns the exact values from the standard published equations, snaps them to the E24 or E96 series, and works out how much power each element has to absorb, because that last number is what usually decides whether a design is buildable.
What This Pi Attenuator Calculator Does
A Pi pad has one series resistor between input and output and one shunt resistor to ground at each port. A T pad has one series resistor at each port and one shunt to ground in the middle. Both are symmetrical, both are bidirectional, and both present the design impedance looking in from either side when correctly terminated at the other.
The calculator returns the resistor values for whichever topology you select, the values for the other one in the note, the closest available standard parts, and the worst-case dissipation in any single element. It also reports the exact voltage ratio and the output power, which are the practical quantities when you are setting up a measurement.
This page designs a fixed resistive network. It is a different job from computing loss that accumulates along a cable, which the attenuation calculator handles with a per-unit-length coefficient, and from matching two unequal impedances, which needs a reactive network or an unequal-impedance pad and belongs on the impedance matching calculator.
How to Use It
- Enter attenuation in decibels of power. Pads are always specified in power decibels, so 20 dB is a hundredfold power reduction and a tenfold voltage reduction. Getting the factor of two wrong here is the classic error.
- Set the system impedance to match your line. A 50 Ω pad in a 75 Ω system does not simply attenuate slightly differently — it reflects, and the return loss is often worse than the level error.
- Choose the topology that gives sane part values. At high attenuation the Pi shunt resistors approach the system impedance while the series resistor grows large; at low attenuation the T shunt grows very large. Pick whichever puts both values in a comfortable range.
- Check the worst-case dissipation, not the total. The input element takes the largest share, and the imbalance grows with attenuation. On a 20 dB pad the first resistor absorbs most of the power the pad removes.
- Use the nearest standard values to sanity-check feasibility. If the closest E24 part is several per cent away, either move to E96 or reconsider the attenuation figure, because a small resistor error becomes both a level error and a match error.
The Formulas and a Worked Example
Everything is driven by the voltage ratio K = 10A/20, where A is the attenuation in decibels. For a Pi pad the two shunt resistors are Rsh = Z₀(K + 1) ÷ (K − 1) and the series resistor is Rse = Z₀(K² − 1) ÷ (2K). For a T pad the two series resistors are Rse = Z₀(K − 1) ÷ (K + 1) and the shunt is Rsh = 2Z₀K ÷ (K² − 1).
Work the default through. A 10 dB pad in 50 Ω has K = 100.5 = 3.1623. The Pi shunt resistors are 50 × 4.1623 ÷ 2.1623 = 96.25 Ω each, and the series resistor is 50 × 9 ÷ 6.3246 = 71.15 Ω. Those are the values printed in every pad table for a 10 dB 50-ohm Pi. The equivalent T pad has series resistors of 50 × 2.1623 ÷ 4.1623 = 25.97 Ω and a shunt of 100 × 3.1623 ÷ 9 = 35.14 Ω.
Now the power. Feed 1 W into the Pi version and the input voltage across 50 Ω is √(1 × 50) = 7.071 V. The input shunt absorbs 7.071² ÷ 96.25 = 0.520 W. The output is 0.1 W at 2.236 V, so the output shunt absorbs only 2.236² ÷ 96.25 = 0.052 W. The series resistor carries the difference: the voltage across it is 4.835 V, giving 4.835² ÷ 71.15 = 0.329 W. Those three sum to 0.9 W, which is exactly the 1 W in minus the 0.1 W out, as they must.
Notice how uneven that split is. The input shunt alone takes 58 % of everything the pad dissipates. Specify all three resistors at a third of the total and the first one fails first.
Why a Pad Is Not a Potential Divider
Both reduce a signal, and that is where the similarity stops. A two-resistor divider has an input impedance that depends on what is connected to its output and an output impedance that depends on what drives it. Change the load and the attenuation changes with it. In a system where the source and load are both trying to see a specific impedance, that is a fault.
A matched pad presents the design impedance at both ports simultaneously, provided the far port is terminated in that impedance. This means it can be inserted anywhere in a chain without disturbing anything else, which is why attenuators are the standard tool for taming reflections in test setups. Putting a 6 dB pad in front of a badly matched instrument improves the return loss seen by the source by 12 dB, because the reflected wave passes through the pad twice. Signal level is a cheap thing to trade for that.
Pads are also bidirectional and phase-flat, unlike reactive matching networks. They introduce essentially no phase shift across a wide band, they cannot ring, and they work down to DC. The price is that the removed energy becomes heat and cannot be recovered, and that the pad adds thermal noise, which matters in a receiver front end where the decibel calculator and a noise budget decide where a pad may safely go.
What Limits a Real Pad at High Frequency
The equations above are pure resistance and have no frequency in them, which implies a pad works equally well at any frequency. Real components disagree in three ways.
Every resistor has series inductance and parallel capacitance. A leaded resistor's lead inductance alone is enough to lift its impedance noticeably above a few hundred megahertz, which is why RF pads are built from thin-film chip resistors in small packages. Layout adds its own parasitics: the shunt resistors need a low-inductance path to a solid ground plane, and a single via with a millimetre of trace can dominate the network's behaviour in the gigahertz range.
Isolation sets a second limit. A pad's stated attenuation assumes the only path from input to output is through the resistors. Capacitive coupling across the package, across the board or through a shared ground provides another path, and beyond roughly 30 to 40 dB in a single stage that leakage rather than the resistors sets the real figure. High-value attenuators are therefore built as several cascaded stages with shielding between them, not as one very lossy network.
Power handling is the third. A resistor's rating is a thermal figure at a stated temperature with a stated mounting, and the pulse rating can be very different from the continuous one. The circuit theory behind these lumped networks, and the point at which lumped analysis stops applying, is developed in MIT OpenCourseWare's 6.002 Circuits and Electronics course.
Choosing Between Pi and T
The two topologies are electrically equivalent, so the choice comes down to component values and layout. As attenuation rises, the Pi shunt resistors converge on the system impedance while the series resistor grows without bound; as attenuation falls toward zero, the Pi shunts grow without bound while the series resistor tends to zero. The T network behaves in the mirror image.
In practice that means Pi is generally preferred at moderate to high attenuation because its two shunt values stay near a comfortable 50 to 300 Ω range, while T is often easier below about 3 dB where the Pi shunts would need to be inconveniently large. At high frequency there is a further argument for Pi: its shunt elements can absorb the parasitic capacitance of the layout into their own value, whereas a T network's series elements sit directly in the signal path where their inductance is harder to compensate.
For unequal source and load impedances neither symmetrical form applies, and you need an unequal-impedance pad or a minimum-loss pad, which has a fixed attenuation set by the impedance ratio rather than a value you choose. The cable impedance calculator establishes what impedances you are actually working between, and Georgia State University's HyperPhysics page on impedance covers the underlying circuit concept. If you are attenuating on the transmit side of a radio, remember that transmitting on most frequencies requires a licence in most jurisdictions, and the EIRP calculator is the page for the radiated-power limits that come with it.
Arb Digital builds free tools like this one because useful pages earn attention. If you want tools, calculators or content built for your own audience, we can help.
Browse All Free Tools Talk to Arb DigitalCommon Mistakes to Avoid
- Using the voltage decibel formula — attenuator specifications are in power decibels, so a 20 dB pad divides voltage by ten and power by a hundred.
- Sizing all three resistors from the average dissipation — the input element takes the largest share and the imbalance grows with attenuation, so the first resistor fails first.
- Using a 50 Ω pad in a 75 Ω system — the mismatch produces reflections whose effect on a measurement is usually worse than the level error.
- Expecting one stage to give 40 dB or more — coupling across the package and the board bypasses the resistors, so high-value attenuators are cascaded and shielded rather than built as a single network.
- Ignoring the noise a pad adds — a resistive attenuator degrades noise figure by its own loss, which makes it a poor choice ahead of a sensitive receiver's first amplifier.
Related Free Tools From Arb Digital
For loss that accumulates along a length of cable rather than in a lumped network, use the attenuation calculator, and for ratio-to-decibel conversions in general the decibel calculator. The impedance matching calculator handles unequal impedances that a symmetrical pad cannot. On the component side, the resistor combination calculator builds an awkward value from parts you have, the resistor color code calculator reads the bands, and the power dissipation calculator and Ohm's law calculator check the thermal and electrical margins. The voltage divider calculator covers the unmatched case where impedance does not matter. Everything is on the free online tools hub.
Frequently Asked Questions
They are electrically equivalent and give identical attenuation and matching. A Pi pad has two shunt resistors to ground and one series resistor between them; a T pad has two series resistors with one shunt to ground in the middle. The choice is practical: Pi keeps values convenient at moderate to high attenuation, T at low attenuation.
Because a divider's input impedance depends on its load and its output impedance depends on its source, so it destroys the match at both ends and its attenuation changes when the load changes. A matched pad presents the design impedance at both ports, so it can be inserted anywhere in a chain without disturbing the rest of it.
Not an equal share. The element nearest the input sees the full input voltage or current and absorbs the largest fraction, and that imbalance grows with attenuation. On a 10 dB 50-ohm Pi pad fed with one watt, the input shunt takes about 0.52 W of the 0.9 W dissipated. Sizing all three from the average will burn out the first one.
Power decibels, always. A 20 dB pad reduces power by a factor of one hundred and voltage by a factor of ten. The design equations use the voltage ratio K equal to ten raised to the attenuation divided by twenty, which is where that factor of two enters.
Because a wave reflected from a badly matched load has to travel back through the pad, so it is attenuated twice. Inserting a 6 dB pad improves the return loss the source sees by 12 dB. That is often a very good trade in a test setup, where signal level is cheap and reflections corrupt the measurement.
Not reliably. Beyond roughly 30 to 40 dB the coupling across the package, across the board and through shared ground provides a leakage path that bypasses the resistors, and that leakage rather than the design sets the real isolation. High-value attenuators are built as several cascaded stages with shielding between them.
Yes. A resistive pad degrades noise figure by an amount equal to its own loss, because it attenuates the signal while contributing its own thermal noise. Ahead of a receiver's first amplifier that is a direct loss of sensitivity, which is why pads normally go after the low-noise stage rather than before it.
This tool is provided for educational and preliminary design use. It computes ideal resistive networks with no parasitic inductance, capacitance, coupling or thermal derating, and assumes both ports are terminated in the stated impedance. Verify component power ratings, layout parasitics and isolation against measurement before relying on a design.
