BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING
FIG. 1 is a graphic view of a communication system having multiple channel carriers over a wide system bandwidth of frequencies, each channel having a specified wide bandwidth.
FIG. 2 is a block diagram view of a system using a stepped CW signal for characterizing the frequency response of a device under test containing a frequency translation device according to the present invention.
FIG. 3 is a flow chart view of the process for characterizing the frequency response of a frequency translation device according to the present invention.
FIG. 4 is a block diagram view of a system using a modulated signal for characterizing the frequency response of a device under test containing a frequency translation device according to the present invention.
DETAILED DESCRIPTION OF THE INVENTION
For the following discussion “calibration” refers to the factory calibration of a device using external test equipment, and “normalization” or “alignment” refers to device self-calibration using an internal test signal.
Referring now to FIG. 1 a typical RF system has multiple channels over a wide system bandwidth, such as tens of gigaHertz, each channel having a unique center frequency, ωc, and a specified channel bandwidth, ωb on the order of 100 MHz or greater for state of the art wide band systems. The wide system bandwidth covers a frequency range from a start center frequency, ω1, to a stop center frequency, ω2. A channel frequency response at a device is modeled as follows:
H(ω, ωc)=HR(ω+ωc)HM(ω, ωc)HIF(ω)
where H is the overall channel frequency response, HR is the frequency response preceding a mixer, HM is the RF mixer frequency response, HIF is the combined frequency response of the circuits following the mixer and ω is the frequency offset from the channel center frequency, ωc. The frequency response may be expressed as a complex function that contains both magnitude and phase information. As a result the magnitude of the frequency response may be expressed as an absolute value of the complex frequency response. By recognizing that the mixer and its interacting circuits are essentially not separable, the combined frequency response is characterized and may be used as a correction factor for a transmitter or receiver, or as a figure of merit for the mixer.
As shown in FIG. 2 during initial calibration a signal source 12, such as a stepped continuous wave (CW) signal generator, swept CW signal generator, modulated CW signal (multi-tone) generator or the like, provides an RF signal that is varied across each channel of a device under test (DUT) 14, either by stepping a frequency for the RF signal across the channel bandwidth about a center frequency, sweeping the frequency for the RF signal across the channel bandwidth about the center frequency, modulating the center frequency with a second signal having a frequency equal to or greater than the channel bandwidth, or the like. For ease of understanding the following description assumes that the signal source 12 is a stepped CW signal generator. The RF signal from the signal generator 12 is input to a power splitter 16 that provides a portion of the power from the RF signal to an input port of the DUT 14 and an equal portion of the power to a first power meter 18. A second power meter 20 is coupled to an output port of the DUT 14. The power meters 18, 20 are magnitude measurement systems. The magnitude of the RF signal at each frequency step, as measured by the first power meter 16, is compared (22) with the magnitude of the RF signal output from the DUT 14, as measured by the second power meter 20. The resulting differences in magnitude in dB as measured by the respective power meters 18, 20 represents the frequency response of the DUT 14 for the particular channel. This calibration process occurs for each channel within the wide system bandwidth to provide a separate frequency response for each channel or center frequency.
More specifically as shown in FIG. 3 during the characterization phase the RF signal from the signal generator 12 is used as a stimulus signal, which is split between the first power meter 18 and the DUT 14. The power measurement starts at the start center frequency, ω1, (step 24) with a predefined portion of the signal being applied to an input of the DUT 14 (step 26). For each channel the signal generator 12 is stepped across the channel bandwidth using a predefined frequency step, called an “IF frequency” step. For example if ωb=100 MHz and the IF frequency step is 10 MHz, then the following frequencies are tuned from the center frequency, ωc: ω=−50 MHz, −40 MHz, . . . 40 MHz, 50 MHz where ω=0 at ωc. For each offset frequency the magnitudes of the stimulus signal and device output are measured (steps 28, 30). The difference in dB between the readings from the power meters 18, 20 is stored at each frequency step over the channel bandwidth to produce the frequency response (step 32). The process is repeated (step 34) for each channel center frequency from ω1 to ω2 by incrementing the center frequency by a predefined RF frequency step (step 36).
The IF and RF frequency steps or intervals are selected so that any interpolation error introduced at an arbitrary center frequency and frequency offset is much smaller than any specifications under consideration. While fine steps result in better accuracy, it takes longer to calibrate and requires a larger data store for keeping the calibration data. The phase data may be derived from the magnitude data, if required, by assuming the system is a minimum phase system, as described in co-pending U.S. patent application Ser. No. 11/272,285 filed Nov. 9, 2005 entitled “Filter Equalization Using Magnitude Measurement Data.” Where, as shown in FIG. 4, a signal source 40 provides a modulated signal, the amplitude and phase may be measured directly at the same time at IF frequencies covering the specified channel bandwidth using a measurement device 42 attached to the DUT 14. Similar to the procedure described above for the stepped CW signal, the entire wide system bandwidth is covered by incrementing the center frequency by the RF frequency step.
Prior to saving the calibration results in a data store, the following conditioning (step 38) is preferably performed. These steps help to reduce the amount of calibration data and improve the accuracy of interpolation for determining the frequency response of the DUT 14 at any desired center frequency. The frequency response for each channel is scaled with respect to a reference frequency, which for convenience may be the alignment frequency referenced above:
where ωr is the reference or alignment frequency and the magnitude terms are expressed in dB. Then for the frequency response at each center frequency subtract the frequency response at center (ω=0) from each frequency response:
|H2(ω, ωc)|=|H1(ω, ωc)|−|H1(0, ωr)|
where H1 is the scaled channel frequency response for a particular channel, and H2 is the stored frequency response for that channel.
From the calibration data the frequency response at any arbitrary center frequency may be determined by interpolating the frequency response along the RF frequencies for every IF frequency used in the calibration. If a fine IF frequency resolution is desired, then the calibration IF frequency step is used and the frequency response along the IF frequencies may also be interpolated. In both cases common interpolation methods, such as a cubic spline, may be used.
A source, such as a transmitter, also has a mixer, the frequency response of which may be characterized as indicated above. The frequency response may then be used in the transmitter to predistort the signal output to compensate for the mixer in the transmitter. This is especially useful for a precision arbitrary signal generator which requires the production of an RF signal having known characteristics for testing RF devices.
Also, using a receiver that is corrected for its own frequency response, as described above, any variations in a signal having known characteristics may be considered to be the result of variations in the frequency response of the source. Since such variations may be distinct for each source, such variations may be used as a signature to identify the particular source from which the signal is received.
Thus the present invention provides a technique for characterizing frequency translation devices as a two-dimensional function of a center frequency and an offset frequency to model the frequency response, normalizing the resulting calibration data to a reference frequency to reduce the amount of calibration data for storage, and interpolating from the calibration data a frequency response at any arbitrary center frequency.
Patent Application
20080095271
