Comparing USB dongles. Sideband noise on 144 MHz.
(Jan 9 2014)
Figure 1 is a photo of the four dongles that have been compared in this experiment. The test signal is a combination of three signal sources. A low noise crystal oscillator on 144.050 MHz, a HP8657A on about 144.150 MHz (frequency selected to avoid the worst spurs of the FC0013 tuner.) and a second HP8657A set to emit a frequency modulated signal on 143.170 MHz. The signals are combined and distributed to the four dongles by use of a 7-port resistive combiner. The signal levels at the dongle inputs are shown in table 1.

Frequency    Level
  (MHz)      (dBm)
144.049       -40  "strong signal"
144.142       -96  "weak signal"
143.170       -86  "dither signal"
Table 1.Signal levels at the dongle inputs.

The noise figure of the Logitec dongle is 10.3 dB at the low gain setting required to accomodate a -40 dBm signal at a small frequency separation. This number comes from a NF meter and is not deduced from the screen dumps and the known signal level.

Figure 2 shows all four dongles when two of the signals are switched on and off. All dongles sample at 1.6 MHz and all processing parameters are identical. The E4000 tuner is set in the original gain mode of the osmocom library.

Figure 3 shows the same thing as figure 2, but here the E4000 tuner is set in Linrads sensitivity mode.

Figure 4 shows the screen when the strong signal has been selected in all four dongles. The level is about 118 dB on the S-meters which is about 93 dB above the noise floor in the Logitec/Mirics dongle. This is as expected for a -40 dBm signal in a measurement with a NF of 10 dB in 1 kHz bandwidth.

Table 2 shows signal and noise levels as evaluated from the screen dumps by use of an image manipulation program.

The dongles are arranged the same way in all four images.

    Lifeview LV5TDeluxe
    Manufacturer: Realtek
    Product id: RTL2838UHIDIR
    Tuner: E4000 (Elonics)
    Interface: RTL2832 (Realtek)

    Manufacturer: Logitec
    Product id: none
    Tuner: MSi001 (Mirics)
    Interface: MSi2500 (Mirics)

    ezcap USB 2.0 DVB-T/DAB/FM dongle
    Manufacturer: Realtek
    Product id: RTL2838UHIDIR
    Tuner: R820T (Rafael)
    Interface: RTL2832 (Realtek)

    Terratec NOXON DAB/DAB+ USB dongle (rev 1)
    Manufacturer: NOXON
    Product id: DAB Stick
    Tuner: FC0013 (Fitipower)
    Interface: RTL2832 (Realtek)

Fig. 1. USB dongles for digital TV and FM radio.

Fig. 2. See text.

Fig. 3. See text.

Fig. 4. See text.

The sequence in figures 2 and 3 starts with all three generators running. The first change is that the strong signal is switched off. To a varying degree that affects the level on the S-meter graph. The graph shows S+N in a bandwidth of 1 kHz and when S/N is small, lowering N makes a difference. The second step is to switch off the weak signal. That gives the noise floor which can be expressed as the noise figure (the NF.) The third step is to switch on the strong signal again. The increase in the noise floor is attributed to reciprocal mixing although the mechanism could be something else. Finally the fourth step is to switch on the weak signal.

Table 2 shows the relevant data extractred from the images.

Fig Tuner  Strong   Margin  Noise   Recipr  Weak  S/N    NF   DR   S/W   Sat
            (dB)     (dB)    (dB)    (dB)   (dB)  (dB)  (dB) (dB)  (dB)  (dB)
 2  E4000  (117.3)   2.59    43.5    44.7   60.8  17.3  30.0  77  (55.5)(119.9)
 2  MSi001  118.5    0.81    25.3    33.8   62.4  37.1  10.3  88   56.1  119.3
 2  R820T   117.4    1.80    41.9    42.7   62.2  20.3  27.3  78   55.2  119.2
 2  FC0013  118.8    1.01    40.0    44.0   63.5  23.5  24.1  78   55.3  119.8
 3  E4000   119.0    0.93    37.6    42.4   63.5  25.9  21.7  80   55.5  119.9
 3  MSi001  118.5    1.58    25.5    33.0   63.0  37.5  10.3  89   55.5  120.1
 3  R820T   117.4    2.96    41.9    42.4   62.2  20.3  27.3  80   55.2  120.4
 3  FC0013  118.8    1.89    40.0    43.8   63.5  23.5  24.1  79   55.3  120.7
Table 2.Data extracted from figures 1 to 3. DR is the reciprocal mixing dynamic range at 100 kHz frequency separation normalized to a bandwidth of 500 Hz and a signal 1 dB below saturation.

Each dongle is set to the largest gain possible without saturation. All dongles are run with identical parameters so the signal level plus the margin to saturation should ideally be equal. It is not because the I and Q signal levels differ a little and there is also some DC offset. Table 2 shows that the strong signal is about 55.5 dB above the weak signal. The S/N of the weak signal is on the average 37.3 dB which means that an ideal receiver with NF=0 would give S/N = 47.6 dB.

The level of the strong signal in the E4000 dongle was not measured in the original Osmocom gain mode, but it can be estimated to 117.3 dB from the ratio to the weak signal, 55.5 dB.

Figures 2 and 3 give a good idea about how the noise floor is affected by the strong signal. From R820T and E4000 in the original Osmocom gain mode in which the NF is so high that reciprocal mixing is insignificant to the MSi001 which has a low enough NF to make reciprocal mixing dominate.

The data in table 2 is a bit flawed because special care is needed when studying the dynamic range with inadequate digital resolution. A clear indication about the problem is the upper left part of figure 3. Here the E4000 dongle in sensitivity mode looses 2.5 dB gain when the strong signal is switched off. The reason is inadequate dithering. Note that the noise floor is affected when the weak signal is switched off. That is also an indication that the dither signal is too weak. The FC0013 also shows this behaviour, but to a smaller extent.

By increasing the dither signal by 15 dB to -61 dBm, a little less than 10% of the strong signal amplitude there is enough dithering to make the level of the weak signal nearly independent of whether the strong signal is on or off. At this level it is necessary to switch off FM modulation because all the spurs caused by the dither signalcombine to an increased noise floor. Table 3 shows the RMS power in 1 kHz bandwidth extracted from the screen dumps here:

S0W0 Strong OFF, Weak OFF Frequency = weak signal.
S0W1 Strong OFF, Weak ON Frequency = weak signal.
S1W0 Strong ON, Weak OFF Frequency = weak signal.
S1W1 Strong ON, Weak ON Frequency = weak signal.
S0W0-143.5 Strong OFF, Weak OFF Frequency = 143.5 MHz.
S1W1-143.5 Strong ON, Weak ON Frequency = 143.5 MHz.

          < ----             144.142            ---- >   <---  143.5   --->   
Tuner     S0W0   S0W1   S1W0   S1W1   S/N    NF    DR    S0W0   S1W1    DR 
E4000    38.36  63.13  42.02  63.03  24.77  23.0  79.6   38.26  39.43  82.1 
MSi001   24.96  62.41  32.59  62.42  37.45  10.3  88.7   24.34  29.19  91.5
R820T    41.67  61.75  42.28  61.77  20.08  27.7  79.7   41.67  41.70  80.3
FC0013   40.59  63.16  43.55  63.01  22.52  25.2  78.7   40.38  40.56  81.5
Table 3.Data extracted from measurements with a 15 dB stronger dither signal. DR is the reciprocal mixing dynamic range at 100 and 550 kHz frequency separation normalized to a bandwidth of 500 Hz and a signal 1 dB below saturation.