Homebrew Corner and DIY Studio

Amateur Radio Club OV SØ6 Dresden-Land        


  AD9859 mit vielen Beinchen   Smith-Diagramm   Feldst__rkeindikator.jpg    VV Platine    auch N-Stecker werden gefaßt     Aufbügeln mit gleichmäßigem Druck

RF 2-tone generator

14 MHz 2-tone generator with selectable spacing 5 kHz / 20 kHz and an intrinsic intermodulation distance IMD > 85 dB for IP3-measurement.

To determine the IP3 (Intercept Point of 3rd order Intermodulation Products) a 2-tone RF generator is required for providing two signals of equal amplitude with low frequency spacing. Typically, in many measuring devices a signal spacing of 20 kHz is used. In reality, however, the signals from radio stations are much closer together and their unwanted intermodulation products (IMP) fall in the filter bandwidth. To adjust measurement conditions accordingly, the 2-tone generator can be switched to 5 kHz spacing.

What is the IP3?


Nonlinear components cause mixture of two signals which produce numerous mixing products below and above their frequency with different frequencies and amplitudes (intermodulation products).




The intermodulation products of third order (IMP3) are unfortunately in the very near vicinity of the two signals - only in a distance corresponding to their frequency difference (F1_IP3 = F1 (F2-F1) and F2_IP3 = F2 + (F2-F1) for F2> F1).


Due to this close proximity, the intermodulation products IMP3 for signals with narrow frequency spacing can not be removed by filters and add up to the desired signal as noise.



Since the amplitudes of IMP3 with increasing output power does not increase linearly with the desired signals, but in the 3. Power, they would reach at a specific power output the amplitude of the desired signals. This fictional output power is called IP3 (OIP3 as - O: Output) .


The graph on a logarithmic coordinate system between input and output power (power data in dBm on a linear scale corresponding to a logarithmic representation), the IP3 is represented by the intersection of the straight lines of the main signal and intermodulation product IMP3. The slope of the IMP3 with 3. Power relative to the output  signal in a logarithmic coordinate system corresponds with rise of the lines in a 1:3 ratio (see chart).


The imaginary intersection corresponds to the IP3 (interception point 3rd order). In practice, however, an amplifier droves above a certain power output into saturation, so that not only the gain curve, but also the curve of the intermodulation products increasingly flattened. Consequently, the two curves intersect never - according to this, the IP3 is also never reached and is therefore not measurable. It must be calculated from measured values on two lines. However, because the slope of the lines is known as a ratio of 1:3, the IP3 can be calculated with a single value pair.


Under real conditions many signals with varying size and frequency are mixing. To ever make compareable statements about the intermodulation behavior of electronic modules, a standardized test method with only two signals is required. The two-tone generator provides this and produce signals with equal amplitude and a defined frequency spacing. From the measured output power of the amplifier and the IMP3, the IP3 can be calculated (IP3: imaginary output power, when the 3rd order intermodulation products are of equal size). The IP3 is therefore a fictitious value to characterize the intermodulation behavior under test conditions - no more, no less.




Figure of IP3

  Figure of IP3


MDS: Minimal Detectable Signal

DRim: Intermodulation free Dynamic Range

DRlin: Linear Dynamic Range

The IP3 is calculated from the measured output power and power of the 3rd order intermodulation product according to the formula: IP3 = Pout + 0.5 * IMP3. From the chart above, this calculation can be derived. For clearer presentation, the geometric relationships are mapped isolated. The proof is based on the similarity of the yellow and blue triangle.

Circuit description

For the generation of the signals with 5 kHz and 20 kHz spacing oscillators operate at the frequencies 14.310 MHz, 14.325 MHz and 14.330 MHz in the middle of the SW-bands. Given the relatively high costs of individual crystals frequencies are generated by pulling the crystal oscillators, using low-cost 14.318 MHz crystals. The signals are merged with a 6 dB power combiner. The output delivers a two-tone signal of -10 dBm each.

The circuit of the 2-tone RF generator was designed with the objective of a low intrinsic intermodulation to determine even high IP3 values. Contribute to this highest possible decoupling of signals and low noise of oscillators is required. Oscillator noise was reduced significantly through the use of low noise AF-transistors in comparison              

to common RF-transistors (e.g. BF199), without loss of output power. A subsequent crystal filter reduces additionally the phase noise.

The decoupling of oscillators is largely determined by the quality of the combiner. For the transformer a double hole core BN43-2402  was wound with 2 × 12 turns bifiliar CuL 0.20mm. The characteristic impedance of the two-wire cable was matched carefully by the degree of twisting to 50 ohms . 6dB-Attenuators at the inputs of the combiner lead to further decoupling. Additionally sheath current filters were inserted, built with RG178 coax wound on a toroidal core with high permeability. A 3 dB-attenuator reduces a possible mismatch of the output. The switching of the oscillator signals is done with a low-reflection RF-relay. The separeted power supply of each oscillator is another factor for a high degree of decoupling.


Mechanical Design

The three oscillator modules and the power combiner were each installed in a separate tin plated housing, sealed with a lid. The PCBs on both sides are completely soldered to the housing. The PCB of the combiner is contacted through additionally at key points. For reduction of stray capacitance the transformer was mounted at a distance of approximately 3 mm above the board on a piece of copper free PCB and fixed with a some hot glue. Finally all modules were assembled into a tin case with the dimensions 160 x 100 x 25 mm.

Platinen-Layout Combiner      Platinen-Layout Oszillator

PCB layout oszillator


The small PCBs were manufactured in a rapid prototyping process. A laser printout of the layout has been transferred with the thermal transfer method, and then the fields were released with a small milling machine. The board allows different circuit variations for pulling the crystals.

PCB layout combiner    


Tips for winding the RF transformer
A 1 m long wire of 0.20 mm CuL was folded in half and twisted slowly with a accu-drill under uniform tension. For this the closed end was hung in a hook bolt and the two free ends are clamped firmly. The twist was finished at about 3 turns/ cm, both ends with uneven twist were cut off. One end soldered to a BNC connector at the other end to a small 100 ohm potentiometer. With the NWA (FA-NWT or VNWA) the impedance of the two-wire line was determined and adjusted by further twisting or untwisting to 50 ohms. It was found that the finished wound transformer at the input and output had a characteristic impedance of 52 ohms. Only with a little bit more twisted wires of 48 ohm impedance the transformer matched exactly to 50 ohm. Presumably, during the winding of the transformer it comes to a low untwisting . The isolation of the combiner inputs increased thereby from -43 dB to over -60 dB.

The effect of the crystal filter - reduction of phase noise
Oszillator ohne Quarzfilter Oszillator mit Quarzfilter
                                         before the crystal filter
                                         after the crystal filter
Oscillator output spectrum measured with a high-impedance HF-FET probe on FA-NWT before and after the crystal filter. The leaner output signals correspond to a lower phase noise.


14,330 MHz crystal oscillator signal without (blue) and with crystal filter (red). The overlay chart shows a much narrower noise and a lowered noise floor by using a crystal filter.


To increase the output of the oscillator for measuring higher IP3 values and to reduce feedback effects and possibly further improve the decoupling, an additional buffer stage would probably be useful.

For different frequency ranges a modular system is required, in which the individual oscillator modules are coupled with the combiner modul. A flexible frequency choice would be possible with a 2-channel DDS. The advantage of the present structure is its compactness, which allows a quick and easy test setup.

The noise of the power supplies could e.g. further reduced by a special circuit technique [Finesse Voltage Regulator Noise].

The sheath current filters at the entrances and exit of the combiner were seen in various schematics. Their effectiveness has not been tested in this configuration.

Some tests of broadband amplifier circuits show so far no significant differences of the IP 3-values between signal spacing of 5 kHz and 20 kHz. If a spacing of 5 kHz make sense will be proved on further measurements of selective receiver circuits.

At a last measured intermodulation distance of more than 85 dB for 5 kHz and 20 kHz spacing no more further optimization is planned currently. The generator should be tested first on various objects.


For the many ideas and great patience I have to thank Lothar DL1DXL and Dieter DL1DRC. Their support has enabled me to realize this project.

Christian, DL9NL

- Website S06 | REDAXO 4.3.1 | Webmaster DL9NL