2004-04-10 by Jouni Verronen
To measure RF signals the frequency band must be mixed down to AF range. Figure 1 shows a working circuit (all the ground connections of SRA-1 are not shown). Local oscillator is tuned to the lower end of the 20 kHz slice of an RF band of interest. A balanced diode mixer is very suitable for this kind of application because of its very broad bandwidth and standard 50 ohm impedance level. SRA-1 for example is specified from 0.5 to 500 MHz, but there are newer mixers up to a few GHz.
The lower sideband from the mixer is amplified in the OA circuit to
proper level for SB16 line input.
AD797 is a very low noise, very low distortion
operational amplifier, which works best when driven from low impedance.
So it is most suitable as a diode mixer post amplifier. A discrete
transistor stage, well done, may have lower noise, but distortion
may easily rise higher.
In this setup it is distortion in the soundcard, which limits dynamic range.
There is a diplexer circuit between stages. Its purpose is to terminate the upper sideband from the mixer properly.
Voltage gain of the OA stage was set to 58 dB to see to the noise floor. Signals over about -55 dBm begin to distort with this high gain. The mixer, however, is usable up to around -20 dBm. Gain could be made switch selectable to about 40 dB down to cover whole this range. An RF attenuator can be put in front of the mixer from -20 dBm upwards.
Lower supply voltage may suffice, too. AD797 is specified from +/- 5 V up. There is no specification by Creative Labs of what the line input can take.
A couple of other points are worth notice. There is no attenuation for the image frequency. In many cases this is no major obstacle if there are no signal components either.
Another phenomena may show up, too. There can be an apparent third harmonic on the screen, which should not be. This may then come up so that the RF signal, which you are looking at, has a third harmonic at a reasonable level and then it is mixed down by the third harmonic of the LO. A second harmonic may be there, too, but it is lower as BM has more suppression for even harmonics. In critical cases a low-pass filter can be used in the RF input.
By the way this same mechanism can bring up in-band IMD in ssb/cw detectors, when there is lack of filtering ahead, and third harmonics are produced in IF strip.
These screendumps show some RF measurement cases using the down converter described above and SB_FFTC program.
Fig. 2 demonstrates sensitivity limits. Measured RF signal is 7.005 MHz CW signal from HP8657B at -120 dBm level. As the LO I used my old homebrew grid dipper from seventies which also has +9 dBm RF output. I think this is pretty near the cheapest gear to make meaningful RF measurements at these levels. Admitted, however, that frequency accuracy and stability of a simple GDO are on the borderline for this use. (How about making it with a removable shield cap over the coil and a PIC/LCD frequency counter with a drift correction output.)
We could also estimate the noise floor for the system from the known data:
absolute noise floor -144 dBm / 1 kHz for 68 Hz BW this much lower - 12 dB (Flattop used) ------- makes -156 dBm / 68 Hz then deduct: mixer attenuation 6 dB AD797 input noise figure 10 from unsupressed image 3 ------ 19 dBThis simple calculation gives -137 dBm as the noise floor estimate.
Fig. 3 shows the frequency response of a 9 MHz SSB-generator strip
consisting of an SL640 balanced modulator, an MRF904 isolation amplifier
and a KVG XF-9A SSB-filter.
Modulation frequency was just manually sweeped to 10 kHz with hold
enabled. The center, 10 kHz, corresponds to carrier frequency 8998.5 kHz.
LO to the down converter was a HP8657B, 10 kHz down.
Pass-band ripple may be a bit higher than 1 dB, specified by KVG. There is 32p over one filter termination missing. The filter output was terminated by 470 + 56 ohms to ground, 56 ohm resistor then connected to SRA-1 RF input. It can be seen also that adjacent sideband attenuation is better for USB than for LSB.
Figure 4 shows two tone measurement for the same SSB-strip. Audio tones are 1 kHz and 1.45 kHz. Carrier in the center seems to be over 65 dB down. Also the lower mixing product for 1.45 kHz is shown about 70 down. 1 kHz is not detectable. Inband IMD products cannot be resolved from near signal noise at 60 dB down, perhaps power line harmonics. Some of the garbage are also sidebands from FFT calculation.
Figure 5 shows almost the same measurement plotted on black background. A paper plot is easy to get using one of the various screen dump programs available.
With an RF power attenuator and proper shielding two tone measurements for linear power amplifiers can be done as well. For many cases a simple crystal oscillator can serve as LO.
Figure 6 is still another spectrum measured. It shows an FM signal from HP8657B, center frequency 433 MHz, modulation frequency 1 kHz, deviation 2.4 kHz. Another HP8657B is used as the LO. Zero at the vertical dB scale corresponds to -51 dBm carrier level when unmodulated. The picture shows one of those special points, where the carrier disappears (modulation index 2.4). The measured levels of frequency components accord quite well with theory when Flattop window is used. This phenomenon can be used for calibration of FM generators and modulation meters.
In these screenshots the low frequency noise from power line etc is present. An isolation transformer is not used, as there is no interest in the low end of spectrum, and the transformer could have some effect to the frequency response at the high end.
As the examples show, this simple measurement system can be quite a help
in building radio devices.
Thinking various schools and school projects, a competent professor could also use it to show youngsters much of radio electronics.
In the world there are lot of places with no financial possibilities to latest and most agilent instruments.
Back to SB_FFTC main page
Back to the FrontPage