How to characterize vector response of future broadband instrumentation

Technology News |
By eeNews Europe


Frequency spectrum allocation for wireless communications is outstripped by demand and regulators are tackling this issue by re-releasing ‘white space’ and frequency blocks previously used for analogue services. In turn, the specification standard bodies such as 3GPP [1] define performance metrics and measurement configurations for wireless system. Their objective is to guarantee compliance with frequency allocation and to ensure interoperability of wireless systems as nobody wants products that do not work together or cause interference. The challenge for designers of both RF frontends [2] and digital circuits is to accommodate the wider range of frequency channels and achieve acceptable performance. Measurement is about commerce so the instrument manufacturers that support the development and manufacturing processes face similar issues of performance demands at acceptable cost. Engineering is the art of the achievable and so this is a dynamic process with considerable interaction between these players.

Let us briefly introduce properties of the main measurement instrumentation used in R&D of wireless systems. Vector signal generators (VSG) currently work with carrier frequencies up to 30 GHz or more and have a complex modulation bandwidth (in-phase and quadrature) of up to about 2 GHz. A vector signal analysers (VSA) is the instrument of choice to analyse modern broadband communication standards, such as LTE. Commercially available instruments have demodulation bandwidth of up to 300 MHz operate over a similar frequency range as the VSG.

In order to check instrument’s specifications, it must be sent to an independent or manufacturer’s calibration laboratory where parameters are measured with uncertainty traceable to basic SI units. Traceability means an unbroken chain of comparisons relating an instrument’s measurements to a known standard. This is well established for parameters such as RF power, however, traceability is difficult to achieve for dynamic measurements such as the vector response of a VSA or error parameters of modulated signals, where no accredited method exists yet. This especially holds for error vector magnitude (EVM), which is one of the key parameters used to define the performance of wireless communication equipment and circuits. Manufacturers of measurement instruments use slightly different methods to calibrate their own instruments. Moreover, the calibration usually forms a closed loop, e.g., for calibration of a VSG, one needs a calibrated VSA and vice versa. These methods are clearly not traceable to standards of a higher precision.

In order to overcome these drawbacks, a manufacturer-independent calibration method has been developed by a consortium of national metrology institutes in the framework of the EMRP IND16 “Ultrafast” joint research project [3]. In this article we will describe a method for achieving manufacturer independent traceable characterization of VSA’s vector response and for EVM measurement.

Traceable measurement of EVM

A traceable approach to EVM measurement is to use a well characterized sampling device such as an oscilloscope to sample the signal directly at the carrier frequency or a VSA to sample the signal at the intermediate frequency or in the base band. The second step is a signal processing and EVM calculation using a validated method, which means the use of external software where all computational steps and errors are open to analysis. In the framework of current project [3], freely available software is being developed for the EVM and uncertainty calculation of basic modulation schemes. A digital real-time oscilloscope (DRTO) is attractive for measuring the waveform as it can acquire over long epochs and provides a high degree of oversampling, compared with a VSA. A digital sampling oscilloscope (DSO) can provide the calibration link to the primary standard (electro-optic sampling system, EOS) but is not suitable for measuring communication waveforms for a variety of reasons. However, they can be used with a VSG to calibrate a VSA. The drawbacks of the DRTO are low vertical resolution, typically less than 8 bits, and the volume of data (10 ms at 5 GSa/sec is a 50 Mb file). This is mitigated to some extent by the temporal oversampling. Also, the multiple A/D converters within these instruments give rise to some nonlinear behaviour and a complicated error structure. Measuring at certain frequencies introduces spur components within the measurement bandwidth. In general this can be mitigated by multiple measurements.

Instruments such as a wireless communications test set will contain both a VSG source and a receiver. The source only generates wireless communications signals and the receiver only responds to a wireless communication signal. The problem is that measurement of a single receiver with different sources gives different answers, so which one is correct? Calibration of a modulated source is schematically shown in Fig. 1. The CW source helps to correct for the digital oscilloscope’s timebase error and the attenuators eliminate unwanted reflections that also contribute to the EVM [4]. The comparison of combined measurement results obtained with the WCDMA sources and dedicated receiver together with the corrected results for the source and receiver EVM components is shown in Fig. 2.

Fig. 1 Reference receiver for calibration of a modulated source

Fig. 2 Combined EVM of four WCDMA sources

VSA’s vector response characterization

When characterizing a vector response of a VSA, the situation is similar to a VSG. Well established methods exist, traceable to high-precision standards, to measure frequency or power. The scalar response of the IQ demodulator can be calibrated with a power meter. An oscilloscope impulse response can be measured with a calibrated electrical pulse generator or phase reference (comb generator). The same approach cannot be readily applies to the measurement of the phase response or group-delay dispersion of a VSA because the bandwidth is too low. The problem with a VSA is that every time it triggers the relative timing of the trigger and the RF phase will be different. For example, a time variation of 500 ps and a carrier frequency of 2 GHz means you can get any answer you like for phase. All the frequencies of interest must be present at once. Currently a generator with multitone option installed is used for calibration; however the generator itself must be first characterized. Ideally, the broadband multi-tone signal must have bandwidth comparable or greater than the receiver under test. Signal properties of a multitone signal can be chosen in order to stress different elements of the response (missing tones, random fill etc). The same (multitone) signal is measured using two instruments: a). DSO characterized against EOS, b). VSA under test and to achieve the highest accuracy, correction for impedance match is also necessary. The process is schematically shown in Fig. 3. Use of a DRTO is gives higher uncertainties due to its limited dynamic range.

Fig. 3 Measurement of the VSA’s transfer function

The frequency response Y of both the analyser and the oscilloscope can be written as the product of the input signal X and the transfer function of the instrument H. The transfer function of the analyser under test, which combines together both the amplitude and phase response, is then calculated as a ratio of the analyser’s response and the input signal. In reality, the influence of a matching attenuator must be taken into account. An example of the measurement of VSA’s amplitude and phase response is shown in Fig. 4. The bandwidth of the multitone signal and of the VSA was 80 MHz, yet the limitation of this technique is only the bandwidth of the DSO and therefore much wider bandwidths can be characterized [5]. Fig. 4a shows the comparison of the amplitude responses of the analyser and a DSO. Fig. 4b shows the final phase response of the analyser. Signals with low peak-to-average power ratio in the time domain are preferred; a good example is a signal with near quadratic phase profile.



Fig. 4 Amplitude (a) and phase (b) response of a 80 MHz bandwidth VSA

Measurement uncertainty

In metrology, measurement uncertainty is a non-negative parameter characterizing the dispersion of the values attributed to a measured quantity. Each measurement instrument is designed in a unique manner, yet they are comprised of certain common function blocks (downconverter, digital stages, amplifiers, RF frontend etc). Each block generally influences the signal due to its imperfections and might contribute to the final EVM. The propagation of errors from the functional blocks to the final EVM value is a challenging task. Typical impairments include nonlinearity, gain/phase mismatches, noise, coupling, impedance mismatch, time-base errors and resolution constraints.

Accurate prediction of device level contribution to the final EVM requires accurate models of RF and analog device impairments based on specifications provided by instrument manufacturers (limited due to IP constraints). The dominant source of error can vary depending on signal properties and method of calibration.

Besides instrument hardware-dependent errors, the sophistication of the demodulation software inside a VSA also contributes to the final EVM value. Since EVM has been introduced to the standards [1], the method to obtain it should also be standardized. The general uncertainty budget exceeds the scope of this overview article, yet one can find more information at the website of the IND16 Ultrafast project


A full vector characterization of modern broadband signal sources and analysers is currently insufficient in metrology laboratories. One could ask – why do we need traceable communication? We believe the argument is that at present the traceability is achieved in industry through a technical audit process and the information is proprietary. One could argue that millions of mobile phones and other wireless devices have already been successfully fabricated using instrumentation which has been characterized using existing techniques, so why should we care about a more precise characterization or uncertainty calculation? Part of the question is: what benefit does the traceability give? The EVM test with several sources answers this. Once the phones are available the QoS is dominated by the air interface. The target here is integrity of the supply chain. Some companies work on the bases that their suppliers must re-create the same test environment (up-front cost). High measurement uncertainty can behave as noise, which contributes directly into EVM. Poor EVM may be equated as a noise component, so it will impact sensitivity and cell size (RF front-end). As the bandwidth and carrier frequencies grow, the influence of test and measurement instrumentation itself to the signal quality becomes more important.


The work has been supported by the EMRP joint research project “IND16 Metrology for ultrafast electronics and high-speed communications”. The EMRP is jointly funded by the EMRP participating countries within EURAMET and the European Union.



[1] 3rd Generation Partnership Project, ‘Universal Mobile Telecommunications System (UMTS); Base Station (BS) conformance testing (FDD),’ 3GPP TS 25.141, version 11.6.0 Release 11, 2013.

[2] RF Front-End Technology Challenges, ICT KTN and Cambridge Wireless, December 2012.

[3] EMRP joint research project “IND16 Metrology for ultrafast electronics and high-speed communications,” available at


[4] D. A. Humphreys and J. Miall, “Traceable measurement of source and receiver EVM using a real-time oscilloscope,” IEEE Trans. Instrum. Meas., vol. 62, no. 6, pp. 1413–1416, Jun. 2013.

[5] D. A. Humphreys, M. R. Harper, and M. Salter, “Traceable characterization of vector signal analyzers,” Proc. of 75th ARFTG Conf. Dig., Anaheim, CA, May 2010, paper 3.2.

About the authors:

Martin Hudlička works for the Czech Metrology Institute, Prague, Czech Republic

David A. Humphreys works for the National Physical Laboratory, Teddington, UK

Mohammed Salhi works at Physikalisch-Technische Bundesanstalt, Braunschweig, Germany

Faisal A. Mubarak works at the Dutch Metrology Institute(VSL), Delft, The Netherlands


Linked Articles
eeNews Europe