Spend a full day on an EMC bench and you will move between four or five different ways of expressing the same electrical magnitude. The receiver displays dBμV. The conducted-emission limit line in CISPR 32 is written in dBμV. The radiated-emission limit line is written in dBμV/m. The spectrum analyzer in the next room reads in dBm. The power amplifier datasheet specifies output in watts. The bench-side oscilloscope is in millivolts. None of these are wrong; each one is the right unit for the instrument or document that uses it. But moving between them is where good test reports go bad, because the conversion is not just a multiplier — it depends on the system impedance and on whether the quantity is a voltage, a current, or a power.
Why are EMC limits written in dBμV in the first place?
Conducted-emission limits in CISPR 22, CISPR 32, ANSI C63.4 and MIL-STD-461 are expressed in dBμV (sometimes dBμA, for current-probe measurements) because that is the natural quantity the measurement instrument actually reads. A CISPR-compliant EMI receiver presents a defined 50 Ω input impedance to whatever artificial mains network or coupling/decoupling network is feeding it, and the receiver's front end measures voltage. The receiver does not need to know — and cannot directly measure — the power flowing into its input. It measures the voltage at the receiver's input port, and that voltage is reported in dBμV.
Radiated-emission limits are written in dBμV/m for the same kind of reason: the field-strength receiver measures the voltage induced in the antenna at its receiver port, and then the antenna's calibration factor — the antenna factor, in dB/m — is added to convert from receiver-port voltage to incident field strength. The whole measurement chain stays in voltage units because the receiver is a voltage instrument.
Why does impedance matter for a conversion?
Voltage and power are different physical quantities, related through impedance by P = V² / Z. A voltage of 223.6 mV across a 50 Ω load dissipates 1 mW. The same 223.6 mV across a 75 Ω load dissipates 0.667 mW. The voltage is the same; the power is not. Whenever a conversion crosses the line between voltage units (dBμV, dBmV, mV, V) and power units (dBm, dBW, mW, W), the impedance is what bridges the two, and you must pick a value.
On an EMC bench the default is overwhelmingly 50 Ω, because that is the impedance presented by every standard CISPR network, LISN, current probe shunt, antenna feed, and receiver input. The 75 Ω world exists mainly in cable television, broadcast-distribution, and certain video-instrumentation contexts, and it is a deliberate choice driven by the loss-versus-power-handling tradeoff for long coaxial runs. A conversion that does not state its impedance is incomplete; a conversion that uses the wrong impedance silently produces a wrong answer.
What is the conversion formula?
Start from the canonical quantity — power in watts — and everything else falls out of two identities: P = V² / Z and P = I² · Z. Given any one input, compute the canonical power, and then radiate back to every other unit. Concretely, for an input voltage in dBμV at impedance Z:
V (in volts) = 10^(dBμV / 20) × 10⁻⁶, and P (in watts) = V² / Z. To get from there to dBm, take 10·log₁₀(P / 1 mW). Simplified at 50 Ω, that whole chain collapses to dBm = dBμV − 10·log₁₀(50) − 90, which evaluates to dBm = dBμV − 106.99. People round it to 107 and move on with their lives.
Memorise the 107 figure. It is the single most useful number on an EMC bench. The CISPR 32 Class B conducted-emission limit (quasi-peak) is 56 dBμV at 1 MHz, which is therefore −51 dBm. A receiver reading of 60 dBμV is −47 dBm. A spectrum analyzer noise floor of −100 dBm is 7 dBμV. Translating in your head between the limit line and the spectrum-analyzer display is the difference between knowing a problem when you see one and waiting for the test report to find it for you.
What is the 20·log/10·log trap?
The logarithmic family of EMC units splits cleanly: voltage and current units use 20·log₁₀, power units use 10·log₁₀. The factor of 2 is not arbitrary — it falls directly out of the squared term in P = V² / Z. When power doubles, voltage rises by √2, which is 3 dB in voltage and also 3 dB in power, only because the 20·log on voltage and the 10·log on power are exactly the conversion that keeps a dB always a dB.
The mistake is using the wrong factor when converting between adjacent units inside the same family. Going from dBμV to dBmV is 60 dB of offset, because mV is a thousand times μV, and 20·log₁₀(1000) = 60. Going from dBμV to dBV is 120 dB. Going from dBμV to dBμA at 50 Ω is exactly 20·log₁₀(50) = 33.98 dB — because I = V / Z and the dB-of-the-ratio is the 20·log of the impedance. Get any of these wrong and you are off by a tidy round number that looks like it could be right.
The defence against this is to never convert between dB units by remembered offsets when the impedance is in play. Convert to canonical power, then back out to the target unit. The arithmetic is the same, but the impedance term appears where it belongs and the conversion is auditable.
When is the answer different on a 75 Ω system?
Move the same numeric reading from 50 Ω to 75 Ω and the dB-voltage units stay the same — a dBμV is still a dBμV, because the unit is referenced to 1 μV, not to a power — but every dB-power unit shifts by 10·log₁₀(50 / 75) = −1.76 dB. A 0 dBmV signal is −46.99 dBm at 50 Ω and −48.75 dBm at 75 Ω. That 1.76 dB matters: it is most of an order of magnitude on the power side, even though the voltage reading is identical.
This is the source of nearly every cable-distribution measurement disagreement we have ever seen between an EMC bench (50 Ω) and a customer's broadcast or CATV measurement setup (75 Ω). Both readings are correct; they are reporting the same voltage at different impedances. The disagreement is in the implied power, not in the measurement.
What does field strength have to do with any of this?
Nothing — and that is the point. dBμV/m is not a unit conversion of dBμV. The /m is the antenna factor, in dB per metre, that converts the voltage at the receiver port (in dBμV) into the field strength incident on the antenna (in dBμV/m). The antenna factor is frequency-dependent, antenna-specific, and provided on the calibration certificate. There is no impedance-only formula that gets you from dBμV to dBμV/m; you need the antenna's calibration data at the measurement frequency.
Treating field strength as if it were just another unit conversion is the most expensive arithmetic mistake on a radiated-emission test, because it produces a number that looks like a limit-line comparison but is silently missing the antenna factor entirely. The number can easily be off by 20-30 dB at common test frequencies. That is the difference between a pass and a structural redesign.
What does CTS use internally?
We keep a unit-conversion tool in our internal toolchain that takes any one of the 18 standard EMC amplitude units — dBμV, dBmV, dBV, dBμA, dBmA, dBA, dBpW, dBm, dBW, and their nine linear counterparts — together with the system impedance, and returns the other 17 equivalents. The arithmetic is identical to what we have walked through here: reduce the input to canonical power in watts using P = V² / Z or P = I² · Z, then radiate back to every other unit. We verified the implementation against the published formulas in Cantwell Engineering's RF calculator, and the round-trip stability is at the floating-point noise floor (1 × 10⁻¹² relative error across all 18 units). The point of having it as a single tool, rather than a stack of remembered offsets, is that the impedance is a required parameter — there is no way to do a conversion that crosses the voltage/power boundary without stating the impedance, which is exactly the discipline a compliance lab wants.
An accredited test report needs more than the bare conversion. The measurement-uncertainty budget for an EMC test is built up from the receiver uncertainty, the antenna-factor uncertainty, the site-attenuation uncertainty, the cable-loss correction, and the detector and bandwidth corrections (quasi-peak versus average versus peak, at the resolution bandwidth specified by the standard). A unit-conversion tool is the foundation, not the whole stack. Future briefs will work through the antenna-factor mathematics and the detector/bandwidth corrections in their own right.