linrad support: Testing and tuning the RX70.
(May 21 2004)

Smoke test

Connect the coaxial cables from the output of the RX70 unit to the RX10700 unit and make sure that the RX10700 is connected to the RX2500 which in turn should be connected to the computer and to the power supply. Do not connect the d-sub connectors to the RX70 or RX10700 units unless you are sure the boxes are properly connected to the computer chassis and the power supply zero. The parallel port interface is an open drain, a 74HC03, and it may be damaged by static electricity. Making sure that ground is connected before the cable to a computer parallel port is connected is always a good idea. Replacing the 74HC03 which is mounted in a socket is trivial, but if the computer parallel port is damaged, repair is not easy.

Connect the 9-pin d-sub connectors to the RX70 and RX10700 units to supply DC voltages and connection to the computer parallel port. Look here for details: Controlling WSE units from within Linrad

In case the current differs much from the nominal 0.65A something is wrong. There is not yet any experience on faulty boards so I can not give detailed hints for troubleshooting. The currents through the RF and IF amplifiers are conveniently checked by the voltage across the 22 ohm resistors in their power supply lines. It may be a good idea to have current limited power supplies to avoid burning the 1A fuses. In case some error causes an RF mosfet to have near zero resistance, the power in the 22 ohm resistors could become 8W and since these resistors are rated 3W only, they may become damaged if the fail condition lasts for many minutes.

Tuning the crystal oscillator

Set C25 for maximum capacitance to make sure the drain circuit of Q6 is not in resonance. This makes it easier to monitor the Q6 source voltage.

If the board is new and has not been tuned before, set C4, C7, C15, C24 and C27 for maximum capacitance. Set C49 for about 30% capacitance. Start Linrad and enter the weak signal CW mode by pressing A. In case Linrad was already running, you have to exit and re-enter by pressing X, then B. This way you clock data into the 74HC174 flip-flops and send current to one of the oscillator transistors. Set the frequency in the Linrad frequency control window to 70.2 MHz. The voltage on R6 (and C6) should now be -4.5 V.

Monitor the voltage on the source of Q6 by connecting a voltmeter to R27. When the LO is not oscillating, the voltage on the source of Q6 is about +0.7V with respect to ground. Tune C26. When oscillations start, the voltage will grow. Set C26 for maximum, then tune C49 for maximum and finally tune C26 again. The source should be somewhere around +1 V with respect to ground now.

While still monitoring the Q6 source voltage, tune C25 for minimum. Q6 is a buffer amplifier, when the drain comes into resonance the transistor will saturate and the current decrease. The Q6 source should be at about +0.5 V with respect to ground when C25 is tuned for minimum.

The crystal frequencies are 59.3, 59.4, 59.5, 59.6 and 59.7 MHz. In case you are coarse tuning the board outside its box, you may use a frequency counter to set the frequencies about 200 Hz above the nominal frequencies.

The temperature coefficient of the crystals is not equal. Tuning cold for a frequency error of 200Hz will typically lead to frequency errors within +/- 200 Hz for the warm unit.

Set C59 and C116 for 50% capacitance. Tune C56 for maximum LO level as measured on the test points close to R33. Then alternate between C59 and C56 for maximum LO power. Do not be too careful, these capacitors will be fine tuned at a later stage. The voltage on the test points should be around -5V. Then tune C117 for maximum LO level as measured on the test points close to R58. Alternate between C117 and C116 for maximum LO power. Do not be too careful here either.

Coarse tuning the RF filters

Connect a signal generator to both RF inputs with a T-connector and two 10 dB attenuators. Set the frequency to a few kHz from 70.2 MHz. Anything between 1kHz and 25 khz will be fine. A suitable power level is -25 dBm, but the level is not critical.

Select I=SOUNDCARD TEST MODE on the main menu to display the audio signal with the oscilloscope mode. (In early versions of Linrad this function was called F=HARDWARE TEST MODE.) The keys "1" and "2" are used to select the channel and "+" and "-" can be used to set the oscilloscope gain. Note that the crystal has to be selected before the test mode is entered. If you follow the procedure outlined above it will be automatic, the crystal is selected by setting the frequency to 70.2 in the A=WEAK SIGNAL CW mode.

In case the card is never tuned before, set C81, C91, C125 and C139 for 25% capacitance. Set C67,C92, C108 and C126 for 50% capacitance. Tune for maximum signal in the following order for channel 1: C136, C137, then C136 again, C130 and finally C138. For channel two the order is: C76, C77, C76, C80 and finally C78.

Fine tuning the RF filters

The fine tuning procedure uses a pulse generator. Set the pulse repetition frequency to something like 200 Hz.

To have the pulses injected into both channels simultaneously, use a hybrid followed by two 10 dB attenuators. The hybrid will isolate one input from the other only if the output impedance of the pulse generator is 50 ohms. Without the attenuators, the tuning of one channel will affect the other channel if the pulse generator is a parallelled 74AC74 or similar. The filter tuning has some influence on the input impedance.

Connect the RX70, the RX10700, the RX2500 and the computer as described above and enter the J=ANALOG HARDWARE TUNE mode. Then select RX70. With a correctly tuned RX70 unit you should see two spectra side by side as shown in fig. 1.



Fig.1. The screen in ANALOG HARDWARE TUNE mode. The distance between the horizontal lines is 1 dB. The spectra can be moved up or down with the '+' and the '-' keys. This unit is correctly tuned.



The ANALOG HARDWARE TUNE mode measures the five spectra that can be obtained with the RX70, combines them to a spectrum that is about 500 kHz wide.

The RX2500 has to be calibrated and the fft1 bandwidth must be set to at least 100Hz.

Another way to tune the RX10700 RF filters is to use a network analyzer in conversion loss mode. Then tune for a flat response from 69.9 to 70.5 MHz. In this case it makes no difference which crystal has been selected.

The 70 MHz RF filter uses several LC circuits. The tuning points do interact and tuning capacitors one by one at random may give a slow convergency. Here is a way that gives rapid convergency:

1a) Tune C136 and C137 simultaneously with one screwdriver in each hand. These two tuning points interact a little. Find the combination that gives the largest average amplitude, do not worry much if there is some slope - but prefer a horisontal spectrum over a sloping one if the average level is similar. 2a) Tune C138 and C130 simultaneously. These tuning points have a marginal interaction, you can tune them one by one. 3a) Tune C125 and C130 simultaneously. These trimmers interact strongly. Set C125 at different positions and tune C130. Look for the optimum combination. 4a) Tune C138 and C139 simultaneously. These trimmers interact strongly. Set C139 at different positions and tune C138. Look for the optimum combination. 5a) Tune C117 and C116 simultaneously. These trimmers interact strongly. Set C116 at different positions and tune C117. Look for the optimum combination. 6a) Tune C126 and C108 simultaneously. These tuning points have a marginal interaction, you can tune them one by one. 7a) Check the conversion gain of the unit. When tuning several RX70 units, you can calibrate the level shown on the Linrad screen by use of attenuators or by use of the '+' and '-' keys. You can also change the default level, the yzer variable in tune.c of the Linrad source code.

In case the gain deviates from the correct value (see below), the attenuation of the 70 MHz filters can be changed. The filters use near critically coupled LC circuits. They are coupled to each other with 2.2 pF capacitors. To increase the coupling and reduce the losses, squeeze a coil for higher inductance, then retune the capacitor across the modified coil. When the LC ratio is increased, the impedance at resonance becomes higher and then the coupling increases and the losses go down. The best coil to squeeze is the one that had the sharpest maximum when you tuned its capacitor.

In case the gain needed more than 0.7 dB change, repeat from 1a. Do not change the gain by more than 1 dB without going through the tuning 1a to 4a. When the gain is changed much, the coupling capacitors C139 and C125 have to be retuned.

1b) Tune C76 and C77 simultaneously with one screwdriver in each hand. These two tuning points interact a little. Find the combination that gives the largest average amplitude, do not worry much if there is some slope - but prefer a horisontal spectrum over a sloping one if the average level is similar. 2b) Tune C78 and C82 simultaneously. These tuning points have a marginal interaction, you can tune them one by one. 3b) Tune C91 and C82 simultaneously. These trimmers interact strongly. Set C91 at different positions and tune C82. Look for the optimum combination. 4b) Tune C78 and C81 simultaneously. These trimmers interact strongly. Set C81 at different positions and tune C78. Look for the optimum combination. 5b) Tune C56 and C59 simultaneously. These trimmers interact strongly. Set C59 at different positions and tune C56. Look for the optimum combination. 6b) Tune C92 and C67 simultaneously. These tuning points have a marginal interaction, you can tune them one by one. 7b) Check the conversion gain of the unit. When tuning several RX70 units, you can calibrate the level shown on the Linrad screen by use of attenuators or by use of the '+' and '-' keys. You can also change the default level, the yzer variable in tune.c of the Linrad source code.

In case the gain deviates from the correct value (see below), the attenuation of the 70 MHz filters can be changed. The filters use near critically coupled LC circuits. They are coupled to each other with 2.2 pF capacitors. To increase the coupling and reduce the losses, squeeze a coil for higher inductance, then retune the capacitor across the modified coil. When the LC ratio is increased, the impedance at resonance becomes higher and then the coupling increases and the losses go down. The best coil to squeeze is the one that had the sharpest maximum when you tuned its capacitor.

In case the gain needed more than 0.7 dB change, repeat from 1b. Do not change the gain by more than 1 dB without going through the tuning 1b to 4b. When the gain is changed much, the coupling capacitors C81 and C91 have to be retuned.

Due to small differencies in the layout, the tuning of channels 1 and 2 differs slightly. In case it is not possible to get enough gain in channel 2 by squeezing the coils, one can increase the coupling by replacing the 2.2 pF coupling capacitors C71, C73 and C75 with 2.7 pF capacitors.

Checking gain and intermodulation

Connect two signals, both with a power level of 0dBm to one of the RX70 inputs. Place the signals 5kHz apart at for example 70.205 and 70.210 MHz. Make sure the IM3 level of this test signal is below -80 dB regardless of the load impedance.

Find out what attenuators you need to make the Linrad S-meter show 100 dB for a 0 dBm signal sent into the RX10700. The attenuator needed is typically 40 dB. You may fine tune with the "First FFT amplitude" parameter. The level seen by the Delta44 will be about 13 dB below saturation and IM3 generated within the RX10700, RX2500 and within the Delta44 will not be visible. Once you know that a 0 dBm signal sent through the attenuator corresponds to 100 dB on the Linrad S-meter you can use Linrad as a calibrated spectrum analyzer to measure signal levels. The first production run, 25 units were tested like this with the results listed in table 1.


Unit       Channel 1          Channel 2        
        S(dBm)  IM3(dBm)    S(dBm)  IM3(dBm)
  1      -0.2    -62.1       -0.5    -57.8 
  2       0.0    -57.8        0.0    -57.1
  3      -0.1    -59.7       +0.1    -58.9
  4       0.0    -59.6       -0.1    -58.4
  5      +0.1    -57.3        0.0    -57.4
  6      +0.1    -58.7        0.0    -59.4
  7      +0.4    -58.1       -0.3    -58.8
  8      +0.1    -57.6       +0.3    -58.4
  9      +0.2    -57.7        0.0    -58.1
 10      +0.2    -57.2       -0.3    -57.4
 11       0.0    -57.8        0.0    -57.6
 12      +0.5    -60.7       +0.1    -59.1
 13      +0.3    -57.2       +0.1    -57.7
 14      +0.3    -58.3       +0.2    -57.5
 15      +0.2    -59.4       +0.3    -56.7
 16      +0.2    -59.4       +0.1    -57.1
 17      +0.3    -59.2       +0.1    -57.2
 18      +0.1    -58.6        0.0    -58.5
 19      +0.2    -57.3        0.0    -57.6
 20      +0.1    -60.4       +0.1    -57.8
 21      +0.4    -58.0       +0.2    -58.1
 22      +0.3    -56.1       +0.3    -56.5
 23      +0.5    -57.3       +0.4    -58.3
 24      +0.1    -59.0       +0.2    -58.6
 25      +0.3    -59.4       +0.5    -57.1
Table 1Signal levels and third order intermodulation levels for the first 25 RX70 units when two test signals of 0 dBm each are sent into the unit.



Table 1 shows that the gain of the RX70 unit is 0.0 dB plus/minus 0.5 dB. The third order intermodulation is at -57 dBm or below so the input IP3 of the RX70 unit is +28 dBm or above. Two units had intermodulation between -55 and -56 dBm. The reason turned out to be the feed-back transformers on the drain side of the IF amplifiers. The windings were not well enough distributed over the cores causing degraded coupling between the 12 turn bifilar winding on the transistor side and the 22 turn secondary winding. Rewinding these transformers increased gain by about 0.5 dB at the same time as the mixer became better loaded. Retuning the filter to restore unity gain led to a reduction of third order intermodulation by about 2 dB. The unit numbered 22 has the same error in both channels, but the deviation from normal performance is too small for a transformer replacement to be meaningful.

Checking the noise floor

For this test, use some receiver with low noise figure that can measure the noise levels at the output connectors of the RX70 unit.

The Delta 44 is normally operated in is lowest sensitivity mode ("+4dB" with ossmix) to minimize the noise contribution from the internal amplifier in the A/D converter. To check the noise floor of the RX70, by use of RX10700, RX2500 and a Linrad system, set the Delta 44 gain to "-10dB". This way the system noise figure at the input of the RX10700 becomes 10.5 dB, low enough to see modest deviations from normal noise levels at the output of the RX70.

To get an even better visibility for the noise generated by the RX70 a wideband amplifier with a BFR91A was inserted in front of the RX10700 to give a noise figure of 2.5 dB when testing the first 25 units.

Table 2 shows how much the noise floor increases when the output of the 25 first RX70 units are connected to an amplifier with a system noise figure of 2.5 dB.


Unit         Channel 1                Channel 2        
        N(dB)  G(dB)  N-G(dB)    N(dB)  G(dB)  N-G(dB)
  1      6.9   -0.2    7.1        6.6   -0.5    7.1
  2      6.9    0.0    6.9        6.7    0.0    6.7
  3      6.8   -0.1    6.9        6.7   +0.1    6.6
  4      7.0    0.0    7.0        6.8   -0.1    6.9
  5      6.9   +0.1    6.8        6.9    0.0    6.9
  6      6.8   +0.1    6.8        6.7    0.0    6.7
  7      7.0   +0.4    6.6        6.9   -0.3    7.2
  8      7.0   +0.1    6.9        6.9   +0.3    6.6
  9      6.9   +0.2    6.7        6.7    0.0    6.7
 10      7.1   +0.2    6.9        6.8   -0.3    7.1
 11      6.9    0.0    6.9        6.9    0.0    6.9
 12      7.0   +0.5    6.5        6.6   +0.1    6.5
 13      6.9   +0.3    6.6        6.6   +0.1    6.5
 14      7.0   +0.3    6.7        6.7   +0.2    6.5
 15      7.0   +0.2    6.8        6.7   +0.3    6.4
 16      7.1   +0.2    6.9        6.7   +0.1    6.6
 17      7.0   +0.3    6.7        7.0   +0.1    6.9
 18      7.0   +0.1    6.9        6.3    0.0    6.3
 19      6.9   +0.2    6.7        6.6    0.0    6.6
 20      7.0   +0.1    6.9        7.0   +0.1    6.9
 21      7.0   +0.4    6.6        6.9   +0.2    6.7
 22      6.9   +0.3    6.6        7.0   +0.3    6.7
 23      7.1   +0.5    6.6        6.7   +0.4    6.3
 24      7.0   +0.1    6.9        7.0   +0.2    6.8
 25      7.0   +0.3    6.7        7.0   +0.5    6.5
Table 2N is the noise level at the output connector of the RX70 measured by comparision with a 50 ohm dummy load. The test system noise figure is 2.5 dB. G is the gain taken from table 1.



As can be seen from table 2, the noise floor of the RX70 is typically 7 dB above the noise floor of the measurement system when the gain is accounted for. This means that the system noise figure at the input of the RX70 is typically 9.5 dB. Taking the contribution from the 2.5 dB noise figure of the measurement system into account one finds that the noise figure of the RX70 itself is typically 9.1 dB.

Checking close range reciprocal mixing and noise modulation

For this test a low noise crystal oscillator is needed.

Run the Delta 44 in its lowest sensitivity mode ("+4dB" with ossmix). Inject a test signal at 70.010 MHz, 1 dB below saturation into the channels one by one. The noise floor at a frequency separation of 5 kHz should not increase by more than 1.5dB in a single channel.

The test signal, 1dB below saturation with a power level of -13.5 dBm is at 125.8 dB on the Linrad S-meter. The noise floor with a dummy load at the RX10700 input is at 10.9 dB in 1 kHz bandwidth which means that the system noise figure is 15.6 dB. At a distance of 5 kHz from the test signal, the noise floor is typically at 12.1 dB which is -143.7 dBc/Hz. At the point of saturation, the noise floor is below -144.5 dBc/Hz. Only a small fraction of this noise is due to reciprocal mixing from the RX70 local oscillator.

In the first production batch of 25 units, one of the boards failed this test. The failing unit had a noise floor that was always at least 2 dB, but intermittantly as high as 15 dB above the normal level when the strong signal was switched on. Running the unit overnight with a hope of permanent failure made the unit close to normal, intermittant noise floor increases in the order of 3 dB or less only, but the level was significantly above the 1.2 dB obtained for normal units. Trouble-shooting is not trivial, but by lifting U6 and U7 from their sockets and injecting the LO signal from another unit by means of a coaxial cable coupled with two turn links to L10 of the faulty unit and L33 of the normal unit it was possible to make sure that the buffer Q6 and everything following it was operating properly. Since the problem was visible regardless of whether Q1 or Q2 was oscillating, the faulty component had to be within limiter or some component close to it. Replacing Q7 and Q8 eliminated the problem.

Checking medium range reciprocal mixing and noise modulation

For this test a crystal notch filter is needed.

Inject a test signal at 70.020 MHz with a level of 0 dBm into the channels one by one via a notch filter at 70.040 MHz. Set Linrad to 70.075 MHz so the strong signal will reach the RX2500 at 2.555 MHz causing an audio frequency of 55 kHz after mixing with 2.5 MHz. Run the Delta 44 in high gain mode ("-10dB"). The low test signal level, 0 dBm is because the RX2500 can not accept much more when the audio frequency is as low as 55 kHz, only 7 kHz above the Nyquist frequency for the Delta44.

The noise floor should not rise by more than 1.5 dB.

The noise floor change is measured at the center of the notch filter, 70.040 MHz, which is converted to an audio frequency of 40 kHz. A noise floor increase by 1.5 dB (1.41 times) means that the sideband noise due to the strong signal is 41% of the power associated with the 12.4 dB noise figure, (5046 K). This means that the noise temperature of the sideband noise is below 2069 K or below -165.5 dBm/Hz. Since the test signal level is 0 dBm the test ensures that reciprocal mixing and noise modulation together are below -165.5 dBc/Hz at a frequency separation of 20 kHz when the strong signal is just outside the visible passband.

In the first production batch of 25 units, none of the boards failed this test.

Checking wide range reciprocal mixing and noise modulation

Inject a test signal at 70.140 MHz with a level of +5 dBm into the channels one by one via a notch filter at 70.040 MHz. Set Linrad to 70.025 MHz so the strong signal will reach the RX2500 at 2.385 MHz, well outside the 2.5 MHz RF filter. Measure the noise floor at 70.040 MHz with and without the test signal. Run the Delta 44 in high gain mode ("-10dB").

The noise floor should not rise by more than 2 dB.

This test gives the noise floor at 100 kHz separation for signals outside the visible passband. A noise floor increase by 2 dB (1.59 times) means that the sideband noise due to the strong signal is 59% of the power associated with the 12.4 dB noise figure, (5046 K). This means that the noise temperature of the sideband noise is below 2977 K or below -164 dBm/Hz. Since the test signal level is +5 dBm the test ensures that reciprocal mixing and noise modulation together are below -169 dBc/Hz at a frequency separation of 100 kHz when the strong signal is well outside the visible passband.

In the first production batch of 25 units, two of the boards failed this test by giving an increase of the noise floor by 2.2 and by 2.5 dB respectively. Replacing Q6, the first J310 LO buffer amplifier brought both units below the limit of 2.0 dB.

Note that the sideband noise tests give an upper limit for the performance of the RX70 unit. What is measured is the entire receiver chain: RX70 -> RX10700 -> RX2500 -> Delta44 -> Linrad. At a frequency separation of 5 kHz, reciprocal mixing and noise modulation contains contributions from all five processing blocks. At 20 and 100 kHz separation, the Delta44 is protected from the filters inside the RX2500 unit but still the sideband noise is the sum of the contributions from three units.