The present invention relates to an electrical device comprising analog conversion circuitry for converting signals from a first frequency range to a second frequency range, in particular an analog transmit and/or receiver device, such as for example a direct upconversion transmitter. The invention further relates to a method for deriving characteristics of such a device and precompensating an input signal of such a device.

The direct upconversion (or zero-IF) transmitter is the transmitter architecture which is most probably used in wireless transceivers. In such a transmitter, two mixers are driven by baseband (BB) and local oscillator (LO) signals, which are in quadrature. Ideally, combining the two signal paths as BB_l·LO_l-BB_Q·LO_Q gives a single output frequency ω_LO+ω_BB. However, there are analog imperfections such as DC offsets (δ_BBi,q, δ_Loi,q), LO-to-RF feedthrough (σ_LOi, σ_LOq), and quadrature amplitude (ε_BB, ε_LO) and phase (Δ&phis;_BB, Δ&phis;_LO) errors. This results in an output spectrum which contains also an image and a carrier signal. High-end telecom systems such as WLAN require a high suppression of these spurs. In order to determine the exact origin and contribution of the analog non-idealities to this output spectrum, the amplitude and phase of each spur must be known. In the prior art, this is done by downconverting the RF spectrum back to baseband with a substantially ideal receiver, i.e. an expensive device which has a substantially higher conversion accuracy than the transmitter, which is necessary since otherwise the measurement will be as erroneous as the errors which are to be determined. So this method is not viable for automatic calibration.

Furthermore, up to now only techniques using the amplitude information are known. This amplitude information is obtained by placing a peak detector at the RF output [M. Faulkner, T. Mattsson, and W. Yates, "Automatic adjustment of quadrature modulators", IEE Electronic Letters, vol. 27, no. 3, pp. 214-215, Jan. 1991] or by monitoring the power in the adjacent channel [D. Hilborn, S. Stapleton, and J. Cavers, "An adaptive direct conversion transmitter",IEEE Trans. On Vehicular Technology, vol. 43, no. 2, pp. 223-233, May 1994]. As no phase information on the spurs is available, a time-consuming recursive "trial-and-error" algorithm is always needed in order to determine the optimal baseband corrections which result in the smallest error.

There is thus a need for a direct upconversion transmitter which can be calibrated without needing an expensive ideal receiver and for a method with which the device characteristics or non-idealities can be determined and compensated in a simpler way, avoiding a time-consuming recursive algorithm. Since the non-idealities of a direct upconversion transmitter largely originate from the frequency conversion circuitry which is used and such frequency conversion circuitry is also used in other electrical devices, there is in fact a more general need to provide any electrical device which comprises frequency conversion circuitry with a means for more simply and less expensively deriving the device characteristics, and to provide a simpler method for deriving and compensating the characteristics of such a device.

It is therefore an aim of the invention to provide an electrical device, a method for determining device characteristics and a method for compensating the characteristics which fulfil the above mentioned needs.

This aim is achieved according to the invention with an electrical device according to claim 1, a method for determining device characteristics according to claim 16 and a method for precompensating device characteristics according to claim 21.

The electrical device of the invention comprises analog conversion circuitry having an input and an output, which is essentially provided, i.e. intended for converting a first input signal within a first frequency range applied to its input to a first output signal within a second frequency range different from the first frequency range at its output. The device further comprises signal adding means for adding at least a portion of the first output signal as second input signal to the first input signal. This means that at least a portion of the output of the conversion circuitry, i.e. a portion of the signal within the second frequency range is supplied back to the input of the conversion circuitry. The conversion circuitry is also capable of converting this second input signal, which is within the second frequency range, back to the first frequency range. Finally, characteristic deriving means are provided for deriving at least one characteristic of the electrical device from the frequency converted second input signal, which appears at the output of the conversion circuitry.

By supplying a portion of the output signal back to the input of the conversion circuitry, in the end a signal portion is achieved at the output of the conversion circuitry, namely the frequency converted second input signal, which has been converted from the first frequency range to the second frequency range and back to the first frequency range with the same device. As a result, this signal portion is twice subjected to the same non-idealities. From a comparison with the initial input signal, i.e. a comparison of the frequency converted second input signal with the first input signal, the double influence of these non-idealities on the signal during conversion can be determined and one or more device characteristics can be derived.

In the device of the invention, the output signal of the conversion circuitry, or at least a portion thereof, is converted back to the initial frequency range by re-using the same device, so without introducing an influence of non-idealities of a second conversion device. As a result, the need for providing an expensive conversion device which is more ideal or more accurate than the conversion circuitry is avoided. Furthermore, however ideal or accurate such a second conversion device may be, it may still further deteriorate the signal, so by converting the output signal back with the same device also such deterioration is avoided, so that the device characteristics can be determined more accurately.

The signal adding means for adding at least a portion of the output signal as second input signal to the first input signal is preferably formed by an electrical connection from the output towards the input of the conversion circuitry. This connection may comprise one or more circuit blocks, such as for example one or more filters for eliminating signals outside the second frequency range from the output signal, or a phase shifter for invoking one of a plurality of predetermined phase shifts to the second input signal before being added to the first input signal, or other circuit blocks.

Analogously, the method of the invention for determining device characteristics comprises the steps of (a) supplying the first input signal to the device, (b) adding at least a portion of the output signal as a second input signal to the first input signal, and (c) deriving at least one characteristic of the electrical device from the frequency converted second input signal. The method of the invention for precompensating an input signal further comprises the step of precompensating the input signal on the basis of the determined device characteristics. This precompensation is preferably performed in the digital domain.

The invention will be further elucidated by means of the following description and the appended figures.

Figure 1 shows a schematic representation of a transmit device comprising frequency conversion circuitry according to the invention.

Figure 2 shows a schematic representation of a receiver device comprising frequency conversion circuitry according to the invention.

Figure 3 shows a direct upconversion transmitter architecture (prior art).

Figure 4 shows a standard circuit diagram for an upconversion mixer (prior art).

Figure 5 shows the typical output spectrum of a direct upconversion mixer.

Figure 6 shows an embodiment of a direct upconversion transmitter architecture according to the invention.

Figure 7 shows a possible circuit implementation of an upconversion mixer according to the invention.

Figure 8 shows a block diagram for baseband calibration of the device of figure 6.

Figure 9 shows a block diagram for time-domain baseband precompensation of the device of figure 6.

Figure 10 shows a possible circuit implementation for baseband calibration. of the upconversion mixer of figure 7.

Figure 11 shows a possible circuit implementation for feedback DC offset calibration of the upconversion mixer of figure 7.

Figure 12 shows a possible circuit implementation for AC-coupling the feedback inputs of the upconversion mixer of figure 7.

The invention is generally applicable to any electrical device having analog conversion circuitry which is essentially provided for performing a frequency conversion on a first input signal within a first frequency range to obtain an output signal within a second frequency range different from the first frequency range. Two such electrical devices are shown in Figs. 1 and 2.

The device of Fig. 1 is an analog transmit device, which has analog conversion circuitry for converting a baseband input signal to an RF output signal. The additional components of the device of Fig. 1 with respect to a known analog transmit device are shown in dotted lines. The baseband input signal is supplied to the conversion circuitry from a baseband section. The output signal of the conversion circuitry is supplied to an RF section for transmittal. The conversion circuitry may comprise one or more conversion steps, each formed by a local oscillator LO and a mixer. At least a portion of the output signal of the conversion circuitry is added onto the baseband input signal by means of signal adding means, which are formed by an electrical connection FB in Fig. 1, but other means may also be provided for this purpose. This has the effect that an RF second input signal is supplied to the input of the conversion circuitry, which is downconverted to a baseband portion in the output signal. This baseband portion is the result of an upconversion and a downconversion by the same conversion circuitry, so by measuring the baseband portion and comparing it with the baseband input signal, characteristics of the conversion circuitry can be determined.

The baseband portion is conveniently extracted from the output signal by means of a low pass filter LPF or alternative means and supplied to a characterisation block, which is provided for deriving the desired device characteristics and supplying signal correction data to the baseband section. The output signal which is supplied to the RF section is also filtered, namely by means of a filter HPF for eliminating signals outside the RF frequency range, so that unwanted components are removed before transmittal.

In the transmit device of Fig. 1, the electrical connection FB also comprises a filters for eliminating signals outside the RF frequency range from the output signal, which is conveniently formed by the filter HPF of the RF section. The connection FB may further comprise a phase shifter (not shown) for invoking one of a plurality of predetermined phase shifts to the second input signal. The phase shifter may conveniently be an RC/CR block, as will appear from the following. Of course the connection FB may comprise further components, but it is preferred to keep their number as low as possible, since each component may introduce further non-idealities.

The device of Fig. 2 is an analog receiver device, which also has analog conversion circuitry, but for converting an received RF signal to a baseband signal. The conversion circuitry may comprise one or more conversion steps, each formed by a local oscillator LO and a mixer. The additional components of the device of Fig. 2 with respect to a known analog receiver device are shown in dotted lines. An RF section, which during normal operation receives RF signals, is connected on the input of the conversion circuitry. A baseband section, which is connected to the output of the conversion circuitry, is also provided for supplying a baseband input signal to the input of the conversion circuitry for calibration purposes. This is shown in Fig. 2 by means of the arrow connecting the baseband section to the input (the RF side) of the conversion circuitry. At least a portion of the output signal of the conversion circuitry is added onto the baseband input signal by means of signal adding means, which are formed by an electrical connection FB in Fig. 2, but other means may also be provided for this purpose. Again, this has the effect that an RF second input signal is supplied to the input of the conversion circuitry and added onto the first, baseband input signal. The second input signal is downconverted back to a baseband output signal, which is the result of an upconversion and a downconversion by the same conversion circuitry. So by extracting and measuring the baseband output signal and comparing it with the baseband input signal supplied from the baseband section, characteristics of the conversion circuitry can be determined.

In the device of Fig. 2, the baseband output signal is extracted from the output signal by means of a low pass filter LPF or alternative means, which eliminates the RF components from the output of the conversion circuitry. The baseband output signal is then supplied to a characterisation block, which is provided for deriving the desired device characteristics and supplying signal correction data to the baseband section.

In the device of Fig. 2, the baseband section is also connected to the output of the conversion circuitry via the low pass filter LPF, which is thus conveniently used during calibration as well as during normal operation of the device. During calibration, a switch can for example disconnect the baseband section.

The electrical connection FB may further comprise a phase shifter (not shown) for invoking one of a plurality of predetermined phase shifts to the second input signal. The phase shifter may conveniently be an RC/CR block. Of course the connection FB may comprise further components, but it is preferred to keep their number as low as possible, since each component may introduce further non-idealities.

In the following, the invention is applied to the example of a direct upconversion (or zero-IF) analog transmitter, which is used in most modern integrated transceiver systems. It is understood that the invention can more generally be applied in any electrical device which has analog frequency conversion circuitry.

In the following, the signal adding means, i.e. the connection line or alternative means which add the portion of the output signal of the conversion circuitry onto the input signal, are referred to as feedback circuitry. This terminology is relevant, since the output is (partly) "fed back" to the input, but it should be noted that such terminology is generally used to refer to circuitry which actually measure the output and apply a correction to the source, much like the characterisation block in Figs. 1 and 2.

A known direct upconversion transmitter is shown in Fig. 3. A standard circuit diagram for the mixer block is shown in Fig. 4 and a typical output spectrum of a direct upconversion mixer is shown in Fig. 5. The two mixers are driven by baseband (BB) and local oscillator (LO) signals that are in quadrature. If all circuits are perfectly matched, the RF output signal is given by

Although this architecture is almost ideally suited for this purpose, it has the drawback over heterodyne upconverters of generating some in-band spurs that cannot be eliminated by appropriate RF filtering. The most important spurs are located on the image frequency (due to imperfect image rejection) and on the carrier frequency (due to DC offsets and LO-to-RF feedthrough). The two quadrature paths (I and Q) are not perfectly matched, and real implementations of this circuit will have mismatches in amplitude (ε) and phase (Δ&phis;) and dc offsets (δ). We can describe the complex baseband signal of amplitude A_BB, frequency ω_BB and phase &thetas;_BB with the following equations:

In applying the invention to the direct upconversion transmit architecture, as proposed in Fig. 6, the basic idea is to downconvert the RF signal back to baseband, but doing this by re-using the transmit mixers for the downconversion function. Again, errors will be made in this downconversion, but this time the errors in the up- and downconversion are correlated, so the required measurements and equations to calculate them can be derived. For example, the quadrature error in the downconversion will be the same as in the upconversion, so the actual error will be half of the quadrature error which is measured on the downconverted signal.

Thorough investigation of this idea shows that, although at first sight very simple, retrieving the TX errors requires some more operations than this. The main reason is that there are a lot of unknown and uncertain phase shifts in the RF path, which complicate the mathematical formulas. The final circuit topology that allows to successfully recover all errors is shown in Fig. 7. By comparison with the standard circuit topology of Fig. 4, the additional components are immediately clear. The standard upconversion operation passes the LOxBB signal through a high-pass filter to the RF port. In the circuit of Fig. 7, a fraction α of the RF signals is tapped and fed back to the BB input ports of the TX mixers. This will create an RFxLO=BB component in the output spectrum, which passes through a low-pass filter and is measured at the LF ports. This signal can be amplified and converted to the digital domain (e.g. by the receive VGA and ADC already present in the system), where the necessary calculations required for determining the quadrature errors can be performed.

An RC phase shifter generating the FB signal with either 0 or 90 degrees delay is inserted for obtaining two output signals which make it possible to perform all the mathematical operations for retrieving two LO quadrature errors ε_LO and Δ&phis;_LO. All the added blocks are however also not perfectly matched and they have quadrature errors associated with them as indicated in the Fig. 6. A possible circuit implementation of a mixer including the low- and high-pass filtering in the output path and the extra feedback inputs in parallel with the baseband inputs is shown in Fig. 7.

Of course, all the extra circuitry is not free from nonidealities, and will introduce errors in the calibration measurements which are performed. These errors are also indicated in Fig. 6. They include:

• quadrature errors (ε_FB and Δ&phis;_FB) and DC offsets (δ_FBi and (δ_FBq) in the feedback signals FB_l and FB_Q
• quadrature errors (ε_LF and Δ&phis;_LF) and DC offsets (δ_LFi and δ_LFq) in the low-frequency signals LF_I and LF_Q
• amplitude and phase errors (ε_RC and Δ&phis;_RC) in the 90-degree rotation in the feedback path.

In the following calibration procedure, sufficient measurements and mathematics are employed to cancel out the effect of these extra imperfections, and acquire a good estimation for the errors in the baseband and local oscillator signals.

Below it is described how the feedback circuitry can be used to automatically calibrate the transmit spectrum, at regular times before actually transmitting data. Several measurements are performed, making regular use of switches or multiplexers that guide low-frequency signals from one part of the circuit to another. Care must be taken to design these multiplexers such that they do not influence the measurement. Preferably, multiplexing is done in the current domain and simple CMOS pass transistors can be used to switch the signal from one node to another. The calculations presented make use of the FFT function, a block that comes for free in e.g. an OFDM modem since the receiver is not running at this moment. For other applications where such an FFT is not readily available, other mathematic derivations can be analogously developed.

The complete calibration sequence is performed in 6 steps, as set out below.

In this step a sine wave BB signal is applied to the mixer, but the circuit is put in a configuration where it does not perform an upconversion. Instead the baseband signal is transferred directly to the output, where it takes the path through the low-pass filter and is detected at the LF outputs. A block diagram for this is shown in Fig. 8 and a possible circuit implementation for this is shown in Fig. 10. The LO signal can be running, but the DC level of the LO mixer transistors is set to ground to completely eliminate the mixing operation. This way the BB current signal is directed towards the LF outputs. An alternative would be to apply no LO signal, and shift only the DC level of the inner mixer transistors (driven by the signal LOn) to zero, while the outer transistors stay active and conduct the BB current without any mixing operation to the output.

In order to cancel the quadrature error of the LF path (ε_LF and Δ&phis;_LF) (both in the filter circuitry shown and in the following amplifiers and A-to-D converters), two measurements must be done with I and Q signals swapped:

• BB_l signal to LF1_l signal and BB_Q current to LF1_Q signal
• BB_l signal to LF2_Q signal and BB_Q current to LF2_l signal

• -BB_l signal to LF3_l signal and -BB_Q current to LF3_Q signal Since all these switches (only one pair CaIBB is shown in Figs. 8 and 10) are done in the current domain, the influence of imperfect matching in the switches should be negligible.

To determine the BB quadrature errors, the first two measurements can be combined:

• one at frequency +ω_BB with a complex amplitude A+j.B
• one at frequency -ω_BB with a complex amplitude C+j.D
• one at DC with a complex amplitude E+j.F

In the extreme case for the WLAN OFDM system, a BB signal with 26 carriers at all positive frequencies nx312.5kHz can be applied, and for each component the resulting signal at the negative frequency (given by the FFT component C+j.D) gives the quadrature error information. Care must be taken however that harmonic distortion components from e.g. carrier x do not disturb the measurements at carriers 2x, 3x, etc. Therefore it is preferred that only a limited number of BB carriers are applied, whose frequencies are chosen such that the harmonic distortion components do not fall on top of other fundamental frequencies. The quadrature errors of the other (non-used) carriers can easily be retrieved from interpolation between the known points. Also the phases of the applied carriers must be chosen such that the generation of signals with high crest factors is avoided.

Next the baseband DC offsets are determined by combining the first and the third measurements:

• average of LF_I = G
• average of LF_Q = H

Next the DC offset in the feedback path must be measured. This is necessary because later we will activate them to feedback the RF signal to the LF ports, and the DC signal present at the LF ports will be used as an estimation of the carrier spectrum of the RF output. If however this feedback path inserts also DC offset, a false carrier component will be generated and the actual LO feedthrough will be incorrectly compensated for.

For this, the feedback circuitry is activated, but no RF signal is applied to it. The digital TX block, taking into account the previously estimated DC offset, must generate a zero baseband signal. As for the baseband calibration, the LO transistors of the mixer must be biased at ground level and are short-circuited by a switch leading the FB signal directly towards the LF ports. A possible circuit is shown in Fig. 11.

In order to cancel the DC offsets in the LF path (δ_FBi and δ_FBq), two measurements are required with the sign of the FB signals swapped:

• FB_l signal to LF1_l signal and FB_Q current to LF1_Q signal
• -FB_l signal to LF1_l signal and -FB_Q current to LF2_Q signal

• average of LF_l - LF2_l = G
• average of LF1 _Q - LF2_Q = H

Alternatively (and even preferably) a feedback circuit can be built that does not generate any DC offset, e.g. by simple AC coupling (or high-pass filtering) the feedback connection to the mixer input. An example circuit for a Gilbert-cell upconversion mixer is shown in Fig. 12. In this case it is not needed to calibrate out the feedback DC offset, saving the two measurements in step 2. And even more, also the third measurement of step 1 can be omitted because in the next step the LO-to-RF feedthrough is estimated. If there is a baseband DC offset, it will be combined with the LO-to-RF feedthrough and be compensated for in the same way. Of course, the baseband DC offset can be dependent on the baseband gain setting, so one must be careful in choosing the set of measurements to do.

Because DC offsets in the LO signal pass a fraction of the baseband signal directly to the mixer output, an error will be introduced in the measurements in steps 4 and 5. This error is measured now in order to cancel its contribution later.

For this a single BB tone is applied, preferably the one of a low frequency. It should have no quadrature errors or DC offsets, so the results of step 1 should already be applied now. It should also be generated with zero phase, i.e. the delay through the BB path, the LF measurement and the FFT calculation should be compensated for. This can easily be done by calculating the phase of the BB signal in step 1 (&thetas;_BB = arctan(B/A)) and applying this value.

For the baseband DC offset, as explained before, one is free to compensate them at this point or do a combined estimation in this step of BB DC offsets (δ_BBi,q) and LO-to-RF feedthrough(σ_LOi,q) in steps 4, 5 and 6.

The mixer now operates normally and shifts this baseband signal towards RF frequencies, but also generates some low-frequency signals. Two measurements are performed to cancel out quadrature errors in the LF path:

• output of mixer I to LF1_l signal and output of mixer Q to LF1_Q signal
• output of mixer I to LF2_Q signal and output of mixer Q to LF2_l signal

The same BB signal as in step 3 is applied. The mixer now operates normally and shifts this baseband signal towards RF frequencies.

But now the feedback path FB is also activated with the 0-degree delay setting, which causes the circuit to generate a low-frequency component that will be measured at the LF outputs. The delay setting at 0° is just a relative number, there are other phase shifts in the RF section which are unknown now but which will be cancelled out by the final mathematic formulas.

Again two measurements are done to cancel out the quadrature errors of the LF path.

• output of mixer I to LF1_l signal and output of mixer Q to LF1_Q signal
• output of mixer I to LF2_Q signal and output of mixer Q to LF2_l signal

To estimate the LO quadrature errors and LO-to-RF feedthrough components, the FFT of the complex signal(LF1_l+LF2_Q) + j.(LF1_Q+LF2_l) is taken, which contains three spectral components:

• one at frequency +ω_BB with a complex amplitude A1+j.B1
• one at frequency -ω_BB with a complex amplitude C1+j.D1
• one at DC with a complex amplitude E1+j.F1

This step is a copy of step 4, but now the feedback delay is set to 90 degrees. This phase shifter does not have to be a very good one, because if the phase difference is not exactly 90° or if the amplitude does not remain equal, this can be detected in the LF signal and the final mathematics used in step 5 to estimate the errors will take this into account. So a single RC/CR phase shifter is sufficient for this purpose.

Again, two measurements (to cancel LF quadrature errors), subtraction of the waveforms of step 3 (to cancel LF and LO DC offsets), and the FFT of the average of the two obtained signals will result in three spectral components:

• one at frequency +ω_BB with a complex amplitude A2+j.B2
• one at frequency -ω_BB with a complex amplitude C2+j.D2
• one at DC with a complex amplitude E2+j.F2

The following formulas are able to give a good approximation of the LO quadrature errors:

A correction in the time domain is also possible, provided that the baseband quadrature errors are not (or only minor) frequency-dependent. The following equations apply:

These formulas are of course linear approximations with respect to all other errors in the circuit. They are however only:

• second-order dependent on ε_BB and Δ&phis;_BB.
• third-order dependent on ε_LF and Δ&phis;_LF.
• second-order dependent on ε_FB and Δ&phis;_FB.
• third-order dependent on ε_RC and Δ&phis;_RC.

• Although perfectly valid with ideal mixer circuits, simulations with real-life implementations show a small systematic deviation from these results. E.g. the BB phase &thetas;_BB seems to be not perfectly the same as the compensation needed in the measurements on step 2 and 3. This is however something that can be easily detected during simulations, and the algorithm can be adjusted for it.
• This technique might be expanded further to compensate other transmit non-idealities, the most important of which are of course non-linearities. If the RF feedback signal is taken not directly at the mixer output, but at the PA output just before the antenna, sufficient information should be present to detect and correct the nonlinear behavior of the PA.

In conclusion, the invention provides a.o. a method for measuring and correcting the RF output spectrum of a direct upconversion mixer. Amplitude and phase information of all the spectral components of the output signal is obtained by downconverting the RF signal back to baseband. However, unknown errors in the downconversion operation are avoided by re-using the transmit mixer as a downconverter An automatic calibration procedure is presented that explains all the measurements and calculations to be performed in order to obtain an accurate estimate of both the image rejection and the carrier feedthrough. This procedure could even be extended to include other analog non-idealities such as e.g. intermodulation distortion.

This automatic calibration procedure can be generally described as a method for calibrating a direct upconversion transmitter as defined in claim 11, which comprises one or more of the following calibration steps:

• applying a sine wave baseband signal to both the I and Q branch as input signal and transferring it directly without upconversion to the output of the conversion circuitry and conducting a first measurement with the characteristic deriving means; swapping the input signals of the I and Q branch and conducting a second measurement with the characteristic deriving means; and deriving quadrature errors of the baseband signal from the first and second measurements;
• applying a zero baseband signal to both the I and Q branch, shortcircuiting the conversion circuitry and conducting a third measurement with the characteristic deriving means; swapping the sign of the second input signal and conducting a fourth measurement with the characteristic deriving means; and deriving a DC offset of the signal adding means from the third and fourth measurements;
• applying a single baseband tone to both the I and Q branch while the conversion circuitry is operational and the signal adding means are not operational and conducting a fifth measurement with the characteristic deriving means; swapping the output signals of the I and Q branches and conducting a sixth measurement with the characteristic deriving means; and deriving a DC offset of the conversion circuitry from the fifth and sixth measurements;
• applying a single baseband tone to both the I and Q branch while the conversion circuitry and the signal adding means are operational with a first phase shift on the second input signal, and conducting a seventh measurement with the characteristic deriving means; swapping the output signals of the I and Q branches and conducting an eighth measurement with the characteristic deriving means; applying a second phase shift to the second input signal and conducting a ninth measurement with the characteristic deriving means; swapping the output signals of the I and Q branches and conducting a tenth measurement with the characteristic deriving means; and deriving conversion circuitry quadrature errors and/or carrier feedthrough amplitude and/or carrier feedthrough phase from the seventh to tenth measurements.