A Broadband, Low-Noise Time-Domain System for EMI Measurements through Ka-Band up to 40 GHz

1307 F4 coverIn this article, a time-domain EMI measurement system for the frequency range from 10 Hz to 40 GHz is presented. Signals with a frequency of up to 1.1 GHz are sampled by an ultra-fast floating-point analog-to-digital-converter (ADC) and processed in real-time on a field-programmable-gatearray (FPGA). An ultra-broadband multi-stage down-converter allows for the measurement of signals with frequencies up to 40 GHz. Measurement times can be reduced by several orders of magnitude compared to traditional EMI-receivers that work in frequency-domain.

With preselected integrated low-noise amplifiers, the system offers high sensitivity especially in the Ka-band from 26.5 GHz to 40 GHz. The low system noise figure from 26.5 GHz to 40 GHz yields an average noise floor level of around 12 dBµV using an IF-filter bandwidth of 1 MHz in this range. With a high system dynamic range of more than 70 dB, the system is excellently suited for the measurement of broadband, transient emissions or high-dynamic signals like radar pulses. Non-stationary emissions can be measured via the real-time spectrogram or via the multi-channel amplitude probability distribution (APD) measuring method.


Because of the steadily increasing demand for broadband transmission of information, communication systems and consumer electronics utilize higher and higher frequency bands. In order to protect those systems and services from electromagnetic interference (EMI), the radiated and conducted EMIs have to be measured by dedicated measurement equipment in order to fulfill the requirements of electromagnetic compatibility (EMC) standards like CISPR 16-1-1 [1],
MIL-461F [2] or DO-160F [3].

In comparison to traditional measurement receivers, time-domain EMI measurement systems can significantly speed up EMI measurements, saving time and development and testing costs. In previous work we already have increased the upper frequency limit of time-domain EMI measurement systems to 18 GHz and 26 GHz by broad-band down-conversion of the measurement signals [4],[5]. The presented system allows for the measurement of electromagnetic emissions in the frequency range from 10 Hz to 40 GHz. Measurements that are fully-compliant to the requirements of CISPR 16-1-1 can be done and the system also offers the required frequency range and IF-filters for measurements according to MIL461F and DO-160F. Measurements of the conducted emissions on a PC power supply line in the frequency range from 150 kHz to 30 MHz are presented, that show a reduction in scan time by a factor of 1350 compared to traditional measurement receivers. The measured spectrogram of the radiated emission of a microwave oven in Ka-Band shows the system’s ability to characterize the time-behaviour of non-stationary EMI. Finally a measurement of a frequency-hopping signal in the frequency range from 36 GHz to 37 GHz is presented.


The presented time-domain EMI measurement system consists of an ultra-fast sampler with high-dynamic range in combination with FPGAs for the digital signal-processing and a multi-stage broadband down-converter that enables measurements above 1.1 GHz. The block diagram of the system is shown in Figure 1. The electromagnetic emissions are received e.g. via a broadband antenna for radiated emissions or a lineimpedance-stabilization-network (LISN) for conducted emissions. Signals in the frequency range from 10 Hz to 1.1 GHz are lowpass-filtered to avoid aliasing. A floating-point ADC samples the signal with high resolution as described in [6]. To achieve high dynamic range, the signal is divided in three paths with different gain. The signals in each path are sampled in parallel with three ADCs at a sampling rate of around 2.6 GS/s. The sampled signals are combined, thus yielding a dynamic range of the floating-point ADC of 16 bits.

1307 F4 fig1

Figure 1: Block diagram of the time-domain EMI measurement system

To compute the EMI signal spectrum, the digitized EMI signal is transformed by an FFT on an FPGA. For non-stationary signals, a spectrogram can be calculated via the short-timefast-Fourier-transform (STFFT). During the selected dwell-time, a Gaussian window function ω[n], corresponding to the IF-filter of a conventional measurement receiver, is shifted in time with a discrete time-coordinate τ. For every value of τ, the momentary spectrum is calculated via FFT. The short-time spectrum X[τ, k] is calculated by

1307 F4 eq1    (1)

where ω[n − τ] is the shifted window function and x[n] is the discrete input signal. The calculated spectra over time describe a spectrogram. It can be shown that the short-time FFT corresponds to a set of parallel receivers, where the time-domain signal extracted from the spectrogram corresponds to the envelope of the IF-Signal of each receiver [7].

The amplitude spectrum can be calculated using various digital detector modes like the average, peak or quasi-peak mode and is subsequently displayed. In addition, a multiple-frequency amplitude probability distribution measurement method was implemented [8]. Besides the spectrogram, this method allows to evaluate non-stationary emissions by calculating the statistical properties of the signal.


For emission measurements in the frequency range from 1.1 GHz to 40 GHz, a three-stage broadband down-converter is used. Emissions from 1.1 GHz to 6 GHz are bandpass filtered with a single, wideband preselection filter to increase system dynamic range, as can be seen in Figure 2. A wideband low-noise amplifier increases system sensitivity. To maximize the spurious-free dynamic range, the band is subdivided into 16 bands of around 325 MHz bandwidth. Each of those bands is sequentially up-converted to a first high intermediate frequency above 6 GHz by the use of a mixer and a low-noise PLL-synthesizer. An IF-filter is applied and a second mixer converts the subbands to the frequency band below 1.1 GHz, where the signals are fed to the floating-point ADC for sampling [4]. According to Figure 1, the frequency band from 6 GHz to 26.5 GHz is down-converted by the 6-26.5 GHz down-converter. The preselection divides this frequency band into 5 ultra-wide subbands with bandwidths between 3 and 5 GHz. As the assembly is similar to the 26.5-40 GHz down-converter, it is not described in detail. The block diagram of the 26.5-40 GHz down-converter is shown in Figure 3. The input band is divided into three ultra-wide subbands according to Table 1. High-order planar bandpass filters with mid-band insertion losses of 1.5-2.5 dB increase the system dynamic range by attenuating out-of-band emissions and maximize the IF dynamic range by preventing that higher order mixing products are generated in the intermediate frequency band. The switching between the bands is done via SP3T PIN-diode switches with insertion losses of less than 3.5 dB in the frequency range from 26.5 GHz to 40 GHz. Broadband low-noise amplifiers increase the sensitivity of the system. A broadband mixer described in the next section is used to down-convert the bands into the input frequency range of the 6-26.5 GHz down-converter. The local oscillator signals are generated by a low-noise PLL-synthesizer and a frequency multiplier.

1307 F4 fig2

    Figure 2: Block diagram of the 1.1-6 GHz down-converter

Preselection Band Frequency
Band 1 26.5 – 29.25 GHz
Band 2 29.25 – 33 GHz
Band 3 33 – 40 GHz

Table 1: Preselection bands in the frequency range from 26.5 to 40 GHz


1307 F4 fig3

Figure 3: Block diagram of the 26.5-40 GHz down-converter


The components for the down-converters have been realized on glass reinforced hydrocarbon/ceramic and alumina substrates in hybrid assembly. The broadband mixers that have been designed for the down-conversion of the Ka-Band from 26.5 GHz to 40 GHz will be exemplarily described in the following.

In order to realize the ultra-broadband down-conversion of the input signals from 26.5 GHz to 40 GHz into the frequency range from 6 GHz to 13 GHz, a mixer with an exceptionally wide IF-bandwidth is needed. Double-balanced diode designs were implemented, because they provide high port-port isolation and low conversion-loss without the need for an active bias. To achieve high dynamic range, Schottky-diodes with medium to high barrier heights were chosen. Two mixer designs were realized: mixer 1 incorporates an anti-parallel diode pair, mixer 2 uses a quad-diode-ring. As the mixer diodes have to be fed with a balanced signal, balanced-to-unbalanced transformers (baluns) are one of the main elements of the mixer. In [9], broadband Marchand baluns have been described. Such baluns were used for the mixer RF-and LO-ports and have been realized in a planar design on aluminum substrate. The manufactured substrate for mixer 2 can be seen in Figure 4. In contrast to a conventional transformer with center-tap, the common-mode IF-signal is tapped at the radial stubs of the RF-balun, where these taps do not disturb the RF-signal. As the tap-lines have an electrical length of λ/4 at the center frequency of the RF input band, the power divider junction acts as a virtual ground for the odd-mode RF-signal. This yields a low conversion-loss and high RF-IF isolation.

1307 F4 fig4

Figure 4: Picture of the manufactured mixer 2 substrate without diode chip


The manufactured mixers were measured inside their respective housings equipped with 2.92 mm connectors. For each frequency band, a local oscillator signal between 20 GHz and 27 GHz with a power of 15-18 dBm was fed to the LO-port. The IF-signal power level was measured with a precision power meter in the frequency range from 6 GHz to 13 GHz, while a calibrated signal generator fed the RF input signal to the RF-port. Figure 5 shows the measured conversion losses of both mixers. The average conversion loss of mixer 1 is 11.1 dB in the frequency range from 26 GHz to 40 GHz. Mixer 2 exhibits a lower average conversion loss of 8.5 dB because the used diodes have a higher cutoff-frequency than the ones used in mixer 1. Both mixers have very low conversion losses in Ka-band, allowing for a low system noise figure and achieve very wide IF-bandwidths from DC to 14.5 GHz.

1307 F4 fig5

Figure 5: Comparison of the measured conversion losses of the implemented mixers

Port-port isolations are important figures of merit for mixers. A high LO-IF isolation is of special importance for our measurement system, because the strong LO-signal can cause undesired mixing products to be generated at the intermediate frequency of the subsequent mixer. The measured LO-IF isolations of both mixers are shown in Table 2. While mixer 1 achieves a high LO-IF isolation of over 30 dB for local oscillator signals from 20 GHz to 27 GHz, a LO-IF isolation of 8.6-21.4 dB was measured for mixer 2 in the same frequency range. The reason for this behaviour is the additional air-bridge needed to route the LO-signal over the RF-signal for the case of the quad-diode-ring in mixer 2. The diodes in mixer 2 have two internally-crossed diode-pairs, rendering such an external crossing unnecessary. Because of the low-conversion loss, the high IF-bandwidth and the exceptionally high LO-IF isolation, mixer 2 was used in the time-domain measurement system.

ƒLO / GHz LO-IF isol. mixer 1 / dB LO-IF isol. mixer 2 / dB
20 31.6 21.4
23.5 32.7 13.2
27 30.2 8.6


Ka-band radiated emissions commonly exhibit low power levels. Therefore the measurement system should exhibit a low system noise figure to achieve the required sensitivity. Losses in feed lines considerably aggravate this problem.

The noise power at the output of a system, that exhibits the noise figure F can be calculated by

1307 F4 eq2     (2)

k is the Boltzmann constant
T0 is the ambient temperature
BENB is the equivalent noise bandwidth

The system’s theoretical average noise floor level can be calculated using (2). The equivalent noise bandwidth BENB of the IF-filters with Gauss-characteristic is obtained with

1307 F4 eq3    (3)

using the filter’s transfer functions H(f). With the 1 MHz IF-filter and the system input noise figure estimated from the components values of the 26.5-40 GHz down-converter we obtain an estimated average noise floor level of around 11.9 dBµV in the frequency range from 26.5 GHz to 40 GHz.

Figure 6 shows measurements of the system noise floor in the range from 26 GHz to 40 GHz using the average detector and IF bandwidth of 1 MHz and 120 kHz respectively. The system input was matched and the variable input attenuator was set to an attenuation of 0 dB. The time-domain EMI measurement system exhibits a very low noise floor of below 20 dBµV for the 1 MHz IF bandwidth and of below 10 dBµV for the 120 kHz IF bandwidth in this frequency range. The scan time with 1 MHz IF bandwidth and a frequency resolution of 500 kHz was around 30 s, while the scan time using a 120 kHz IF bandwidth and a frequency resolution of 50 kHz was around 90 s.

1307 F4 fig6

Figure 6: System noise floor from 26 -40 GHz


In order to measure broadband transient emissions or general high-dynamic signals like radar pulses, the dynamic range is an important specification for such a system. CISPR 16-1-1 defines broadband pulses for the detector calibration in Band E above 1 GHz and requires an IF dynamic range of at least 40 dB when using a 1 MHz IF-filter. For the characterization of the system IF dynamic range, a pulse generator fed a pulse modulated sinusoidal signal with a frequency of 35 GHz to the system input. The signal pulse width was set to 1 µs and the pulse period to 40 ms. The spectrum was weighted by peak and average detectors and an IF-filter bandwidth of 120 kHz and is shown in Figure 7. With this pulse period, the average detector already shows the system noise floor. The difference in level between the peak and average detector measurements is defined as the IF dynamic range. The measurements in Figure 7 show an IF dynamic range of 62.6 dB. The corresponding value for an IF-filter bandwidth of 1 MHz as specified by CISPR 16-1-1 can be calculated by calculating the change in pulse level ΔAPulse and noise level ΔANoise by

1307 F4 eq4    (4)

1307 F4 eq5    (5)

where Bimp,x and BNoise,x are the equivalent impulse and noise bandwidths of the IF-filters. The measurements indicate an IF dynamic range of 62.6 dB + (18.4 dB -9.2 dB) = 71.8 dB for an IF-bandwidth of 1 MHz, exceeding CISPR 16-1-1 requirements by over 20 dB.

1307 F4 fig7

Figure 7: Measured pulse-modulated signal

The avionic EMC standard DO-160F defines limit lines for the conducted interference signals on supply lines. The presented time-domain EMI measurement system allows to conduct those measurements due to the implementation of the required IF filter bandwidths. The supply lines of a personal computer’s power supply were measured using a current clamp with a bandwidth from 10 kHz to 1 GHz. The measurements are shown in Figure 8. For the measurements, the peak detector was selected with a dwell time of 100 ms. The scan time for the DO-160F scan from 150 kHz to 30 MHz was around 4 s, whereas this measurement would take over 1.5 hours with a traditional heterodyne receiver. The conducted interference currents in the frequency range from 150 kHz to 30 MHz clearly exceed the limit lines defined in DO-160F. The measured power supply would not be suitable for use with sensitive equipment defined in DO-160F.

1307 F4 fig8

Figure 8: Measured spectrum of the conducted interference currents on a PC power supply line

Household appliances can radiate considerable spectral energy densities in the frequency range above 1 GHz. The spectrogram enables the detection of singular or frequency-shifting events in real-time. A microwave oven was measured at a distance of 3 m to a broadband quad-ridged horn antenna with a bandwidth from 1.7 GHz to 20 GHz. To compensate for cable losses and to give the electric field strength of the radiated EMI, the corresponding transducer factors and the antenna factor were applied. Figure 9 shows the time behavior of the magnetron’s 6th harmonic over a time period of 20 s. The magnetron turns on at about 3 s in time and turns off at around 9 s. The free-running oscillator’s output frequency exhibits a frequency drift of around 10 MHz.

1307 F4 fig9

Figure 9: Spectrogram of the 6th harmonic of a microwave oven

Figure 10 shows measurements of a frequency-hopping signal. The spectrum was measured in the frequency range from 36 GHz to 37 GHz using peak and average detectors and an IF-filter bandwidth of 1 MHz. As the detectors are applied on the same sampled data simultaneously, both detectors show the same frequency components, although the signal is non-stationary. The variable attenuator was set to 10 dB. The scan time for those measurements using a frequency resolution of 500 kHz was around 10 s.

1307 F4 fig10

Figure 10: Measured spectrum of a frequency-hopping signal


A time-domain system for EMI measurements from 10 Hz to 40 GHz which allows for measurements according to CISPR 16-1-1, MIL-461F and DO-160F was presented. The system offers high sensitivity due to a low system noise floor and achieves an average noise floor level of around 12 dBµV in Ka-band using an IF bandwidth of 1 MHz and of around 2 dBµV using 120 kHz IF bandwidth. With a dynamic range of more than 70 dB, the system fulfills the requirements of CISPR-16-1-1 and is excellently suited for the measurement of high dynamic signals like radar pulses. An STFFT-based spectrogram and a multiple-frequency amplitude probability distribution measuring method can be used to investigate the properties of non-stationary EMI. Measurements of the conducted emissions on a PC power supply line using DO160 IF-bandwidths, spectrogram measurements of the radiated emissions of a microwave oven in Ka-band and of a frequency-hopping signal in the frequency range from 36 GHz to 37 GHz were presented. favicon


The authors would like to thank the Bayerische Forschungsstiftung for co-funding this project.

© 2012  IEEE. Reprinted, with permission, from the 2012 IEEE International Symposium on Electromagnetic Compatibility proceedings.


  1. CISPR 16-1-1, Ed. 3.0, “Specification for radio disturbance and immunity measuring apparatus and methods Part 1-1: Radio disturbance and immunity measuring apparatus – Measuring apparatus,” International Electrotechnical Commission, 2010.
  2. MIL-STD-461f: Requirements for the Control of Electromagnetic Interference Characteristics of Subsystems and Equipment, Department of Defense Interface Standard, 2010.
  3. DO-160F: Environmental Conditions and Test Procedure for Airborne Equipment, RTCA, Incorporated, 2007.
  4. C. Hoffmann and P. Russer, “A Real-Time Low-Noise Ultra-Broadband Time-Domain EMI Measurement System up to 18 GHz,” IEEE Transactions on Electromagnetic Compatibility, Vol. 53, Issue 4, 2011.
  5. C. Hoffmann and P. Russer, “A Broadband High-Dynamic Time-Domain System for EMI Measurements in K-Band up to 26 GHz,” IEEE Symposium on Electromagnetic Compatibility 2011, Long Beach, USA, pp. 489-492, 2011.
  6. S. Braun and P. Russer, “A low-noise multiresolution high-dynamic ultra-broad-band time-domain EMI Measurement System,” IEEE Transactions on Microwave Theory and Techniques, vol. 53, pp. 3354 -3363, Nov. 2005.
  7. J. Allen, “Short term spectral analysis, synthesis, and modification by discrete Fourier transform,” IEEE Transactions on Acoustics, Speech and Signal Processing, vol. 25, pp. 235 -238, 1977.
  8. H. H. Slim, C. Hoffmann, S. Braun, P. Russer, “A Novel Multichannel Amplitude Probability Distribution for a time-domain EMI Measurement System According to CISPR 16-1-1,” EMC Europe 2011, York, Great Britain, 2011.
  9. N. Marchand, “Transmission Line Conversion Transformer,” Electronics, vol. 17, no. 12, pp. 143-145, 1944.


Christian Hoffmann, GAUSS INSTRUMENTS GmbH, Munich, Germany, choffmann@tdemi.com


Ayoub Sidhom, Technische Universität München, Institute for Nanoelectronics Munich, Germany

Stephan Braun, GAUSS INSTRUMENTS GmbH, Munich, Germany, braun@tdemi.com

Peter Russer, Technische Universität München, Institute for Nanoelectronics Munich, Germany, russer@tum.de