Hardware for fixed passband 2.455MHz to 2.545MHz
The A/D boardThe modified Delta 44 provides a dynamic range of 148dBc/Hz with harmonic spurs that are about 85dB below the carrier when two audio channels, I and Q are combined to produce one receiver channel in complex format.
To my knowledge, no conventional mixer, not even CMOS switched mixers can be used for a direct conversion radio which preserves 148dBc/Hz at a spur level of -85dB.
Hopefully, within a few years, the music industry will provide one or two more bits of useful information in the 24 bit audio format so it is highly desireable to make an RF to audio converter that has a dynamic range that is much better than required for the Delta44.
The local oscillatorThe direct conversion principle gives twice the bandwidth and 3dB better dynamic range compared to a design using filters. Good filters are also difficult and expensive.
The direct conversion method puts extreme requirements on the local oscillator since the LO will mix with its own noise sidebands to produce audio noise. Audio noise is within the passband and therefore the direct conversion receiver needs a far better LO than ordinary receivers for which sideband noise is only important when extremely strong undesired signals are present.
The local oscillator is an X-tal running at 10 MHz. The signal frequency is divided twice by a 74AC74 to provide squarewaves at 2.5 MHz and at 5 MHz. These two square waves represent a two bit counter that counts in 90 degree steps. One turn is 2.5 MHz.
Click here for details of the low noise oscillator
The mixerThe AF feedback four phase CMOS switch mixer provides extreme dynamic range. This mixer is near ideal, far more linear than the Delta 44 and the RF amplifier (filter) that provides the input signal.
This mixer is sensitive to signals at LO overtone frequencies. A good filter is required in front of it.
The RF amplifier and filterSince a filter at 2.5MHz with very good attenuation above 4.8MHz is required for the proper operation of the mixer, this filter is also used to suppress signals outside the passband. Even with the built in anti alias filter of the Delta44 plus the sharp filter in the 2.5MHz to audio converter some additional attenuation well outside the passband is desireable.
The RF amplifier is primarily not used to increase the signal level, it is a buffer amplifier that is followed by resistors that give the filter a well defined impedance. This amplifier also presents a 50 ohm load at the input connector.
Click here for details of the 2.5MHz amplifier and filter
The anti-alias filterWithin the desired passband of slightly more than 90kHz, the dynamic range is limited by the soundcard. The mixer saturates about 17dB above the Delta44. It is followed by a very sharp anti-alias filter that allows signals up to 164dBc/Hz just outside the passband without saturation.
Click here for details of the anti alias filter
The complete 2.5MHz to audio converterThe schematic diagrams contained in the links above as well as the DC supply schematics are laid on a double sided PCB. The unit has two channels so there are two RF amplifier/filter blocks, two mixer blocks and four anti alias filters.
All the layout files for the EDWIN CAD system as well as all PCB production files can be down loaded here CAD files for 2.5MHz unit. There are also photos of the assembled unit.
The first audio amplifier, part of the mixer, operates at very low voltages. The noise level is about 1nV in 1Hz bandwidth so already very small AC magnetic fields can induce voltages that show up as spurs in the spectrum. 50 or 60Hz fields can not be supressed easily but higher harmonics of the mains frequency can be well supressed by use of 2mm aluminium for the box. Here are details about magnetic field pick up and the effect of aluminium thickness
The 2.5 MHz to audio converter described here is a free design. All the information on this page is free and may be used by anyone for any purpose
Testing and tuningThe 2.5MHz to audio converter can be tested and tuned by means of simple standard instruments if the unit is connected to a computer running Linrad with a Delta44 board installed. Look here for details: Testing and tuning the 2.5MHz to audio unit
Sensitivity and noise figureThe NF (noise figure) of the RX2500 itself is 7 dB, but when the unit is used together with a modified Delta44 the system noise figure is 14 dB at the 2.5 MHz inputs when the Delta44 is run at minimum gain. Table 1 illustrates how the system noise figure (sensitivity) and inband dynamic range is affected by the A/D converter performance.
A/D converter System NF Noise floor in Saturation Dynamic range (dB) 500Hz(dBm) (dBm) noise floor(dBc/Hz) Ideal 7 -140 -11 -156 Delta44(mod) 9 -138 -24 -141 at max gain Delta44(mod) 14 -133 -12 -148 at min gainTable 1. The system performance depends on the A/D converter used together with the RX2500 unit. This table shows typical data for the dynamic range within the visible passband.
Two signal dynamic rangeA weak and a strong signal are combined in a 3dB hybrid and the result is sent to the RX2500 unit which is connected to a Delta44 soundcard in a computer running Linrad. The weak signal is placed at 2.524 MHz and a notch filter at the same frequency is inserted between the strong signal and the hybrid.
The level of the strong signal is adjusted until the S/N of the weak signal is degraded by 3dB. Table 2 shows the result of this test with the Delta44 board in maximum and in minimum gain mode.
Frequency Level Level Notes [+4dB] / [-10dB] separation [+4dB] [-10dB] (kHz) (dBc/Hz) (dBc/Hz) 10 147 140 A/D saturation 20 146 139 A/D saturation 30 165 169 40 167 172 50 169 174 60 171 177 70 174 180 80 177 180 100 178 181 RF saturation/reciprocal mixing 150 179 182 RF saturation/RF saturation + reciprocal mixing 200 179 184 RF saturationTable 2. Two signal dynamic range of a RX2500 Delta44 combination.
An inspection of table 2 shows that the dynamic range is better by 7dB for offending signals within the visible passband if the Delta44 is operated in minimum gain mode. One can also see that the dynamic range is better by about 5dB for signals outside the visible passband if the Delta44 is operated in minimum gain mode. A better soundcard than the Delta44 may provide even better results than combining the best values of the two columns of table 2.
The RX2500 and Delta44 combination has a mirror image spur which is typically reduced to a level of about -70dB. There are other spurs that correspond to audio frequency overtones. Table 3 shows some typical spur levels for a system that was calibrated several months ago.
Input Main Mirror 2nd 2nd 3rd 3rd power signal image mirror mirror (dBm) (dB) (dB) (dB) (dB) (dB) (dB) -12 127 61 22.1 28.9 7.5 35.3 -15 124 58 14.7 25.8 1.5 21.1 -18 121 55 10.2 20.5 15.0 -21 118 52 5.7 14.9 3.3 -24 115 49 8.9 -30 109 42.4 -40 99 32.4 -50 89 22.3 -60 79 12 -70 69 2 -80 58.9 -90 48.9 -100 38.9 -110 28.8 -120 18.9Table 3. When the audio overtones of the desired signal fall within the visible passband, spurs are generated. The main source of audio non-linearities is the Delta44.
The spur free dynamic range is about 100 dB as can be seen in table 3.
Signals outside the visible passband do not generate any spurs except for harmonics from 1.25 MHz and responses attributable to LO overtones at 5 and 7.5 MHz. The responses at 5 and 7.5 MHz are approximately 120 dB below the main signal. I have not taken the trouble to make the filters needed to measure the 1.25MHz overtone response.
Three signal dynamic rangeFor two signals within the visible passband, the third order intercept point, IP3, is +23dBm while for two signals outside the visible spectrum IP3 is much higher. Table 4 shows the level of the third order intermodulation signal at different signal levels and frequency separations. The received signal or intermodulation product is at 2.524 MHz
Input SAignal IM3(33kHz) IM3(63kHz) IM3(233kHz) level S-meter S-meter S-meter S-meter (dBm) (dB) (dB) (dB) (dB) +12 151 *** 108.6 100.0 +9 148 124 100.7 93.7 +6 145 112.6 91.0 83.9 +3 142 85.6 82.2 74.8 0 139 78.7 73.7 66.0 -3 136 71.0 65.3 57.2 -6 133 63.6 57.0 48.5 -9 130 55.5 47.9 38.9 -12 127.1 48.2 39.4 30.3 -15 123.7 41.0 30.5 21.7 -18 120.8 33.2 22.4 13.1 -21 118.0 24.6 13.2 4.9 -24 114.7 17.4 6.0 1.1 -27 111.8 10.0 1.5 0.0 -30 109.0 3.7 0.0 0.0 -33 106.1 0.0 0.0 0.0 -36 102.8 0.0 0.0 -39 100.0 0.0 -50 89.9 -60 79.0 -70 69.0 -80 58.9 -90 48.9 -100 38.9 -110 28.8 -120 18.9 -130 9.4 -140 2.2 off 0.0Table 4. Signals and third order intermodulation products. The levels are from the Linrad S-meter. Levels above 127.1 are extrapolated.
The third order intercept point can be obtained from the data in table 4:
IP3(33kHz) = +26.5dBm
IP3(63kHz) = +32.0dBm
Use of RX2500 with simple converters.If the RX2500 and Delta44 combination is connected directly to a TUF-1H Schottky mixer, the Delta44 should be run at maximum gain with a system noise figure of 9dB at the 2.5MHz input. Signals within the visible passband are allowed to be 140dB above the noise floor in 1Hz at a level of -28dBm. Signals outside the visible spectrum are allowed to be 40 dB stronger which is a signal level that a TUF-1H can not deliver. The TUF-1H has 1dB compression at +14dBm at the input and the maximum level obtainable at the output is about +8dBm. Without any amplifiers the system noise figure at the RF side of the TUF-1H is about 15dB which is adequate for HF bands. For VHF some amplification is needed.
ModificationsOne of the units in the second production batch showed a slightly higher noise floor in channel 1. (0.5 dB with the Delta 44 in its most sensitive setting) This was traced to be caused by noise from U4, a 7805 voltage stabilizer which is decoupled by a 0.1 microfarad capacitor C164. The 5V line goes close to L23, one of the 2.5 MHz filter inductors of channel 1. Replacing the 7805 brought the noise down to normal, but having noted the phenomenon on an unusually noisy 7805 it was discovered that channel 1 typically is 0.1 dB noisier than channel 2. By changing the value of C164 to 0.47 microfarad the noise level is drastically reduced, enough to make the noise from the noisiest 7805 invisible.