Setup to measure transmitter spectral purityThese measurements were made with Linrad running on a PentiumIII 600MHz computer with a modified Delta44 soundcard. The 144 MHz signals were converted to audio by use of the WSE converters RX144, RX70, RX10700 and RX2500. Only the RX2500 was a regular unit, the others were prototypes. The data on IC706MK2G, IC202mod and FT221 was measured with identical setup after I returned home .The measurement system has much better dynamic range than required for these measurements so no corrections have to be made for the noise contribution from the measurement system.
Spectral purity of a continuous carrier.The unit under test was connected to a 20dB 100W attenuator followed by an ordinary step attenuator which was connected to the RX144 unit. The step attenuator was adjusted to place the signal approximately 1.5 dB below A/D saturation. The transmitter frequency was set to 144.150 MHz and the Linrad baseband filter was set to rectangular shape with a bandwidth of 1kHz. By clicking the mouse on the carrier and at 5, 10 and 20 kHz frequency separation the sideband noise at these separations was obtained directly with high accuracy. For 100, 500 and 1000 kHz separation, the Linrad frequency control was used to shift the center frequency of the Linrad visible window to allow the mouse to be clicked on the noise floor at the correct frequencies.The RX144 unit has a band pass filter at the input so the frequency response is not perfectly flat. With the unit used in Gavelstad, the correction which is applied in all results are as shown in table 1. |
Frequency Correction (MHz) (dB) 144.150 0 144.250 -0.3 144.650 -0.3 145.150 -0.6Table 1.Correction for non-flat frequency response of the rx144 unit. The number given in the table is added to the value given by the Linrad S-meter to obtain the correct signal level at the variuos frequencies used for the tx purity measurements.
|
|
A screen dump from the Linrad display was saved for each unit
showing the spectrum centered 25kHz above the carrier.
To see these spectra, click on the links below:
IC202 mod (SM5BSZ)
The spectral purity of a continuous carrier is listed in table 2. In this table spurious signals are avoided. For spurs, look at the screen dumps. Note that the mirror image that is produced because RX2500 is a direct conversion receiver is located at about 144.200, 50 kHz above the signal itself. This receiver/Linrad spur is suppressed by only 60 dB in these measurements.
|
Model and owner Noise floor in -dBc/Hz
5kHz 10kHz 20kHz 50kHz 100kHz 500kHz 1MHz
IC202 mod (SM5BSZ) 134.9 138.7 140.8 143.1 143.8 145.4 146.8
IC706 (LA0BY) 91.0 99.9 108.3 118.4 125.4 137.1 139.5
IC706MK2G (SM7UFW) 103.8 111.5 117.2 122.9 125.0 126.0 126.2
IC821H (LA0BY) 95.8 105.1 113.1 122.1 127.7 130.9 136.9
IC970H (LA3FV) 100.1 111.9 121.6 130.0 132.0 133.8 134.1
FT100 (SM7GVF) 107.6 112.8 119.0 126.6 129.4 124.0 130.0
FT221 mod (SM5BSZ) 115.5 122.2 130.4 137.1 140.0 148.3 150.0
FT817 (LB9VE) 101.3 109.2 117.2 126.2 130.4 132.6 133.3
FT847 (LA9CM) 96.0 107.0 115.0 123.9 130.4 140.8 141.7
TM255E (LA6MV) 116.2 120.8 122.3 123.7 125.5 141.1 145.8
TR9130 mod (LA6LCA) 124.1 130.2 135.2 139.0 141.4 147.2 151.9
TS850S+conv(LA6MV) 113.9 122.0 129.2 132.5 133.8 137.3 140.6
Table 2. Noise floor at different frequency separations
from a carrier at 144.150MHz.
|
Spectral purity of voice SSB transmissions.Several of the tested transceivers produce broad SSB signals with a lot of splatter in surrounding channels. Personally I think this is because the transceivers use the ALC to produce speech compression. (My IC202 is an exception, it is modified and I never checked the linearity afterwards since I do not use SSB. My favourite mode is CW.) The rigs were tested in SSB mode by having the owner talk into the microphone with a volume/voice level that produced a reasonable DX clipping as indicated by the peak to average power ratio. Average and peak powers can be read off the Linrad S-meter.A screen dump from the Linrad display was saved for each unit showing the spectrum centered 25kHz above the SSB signal. To see these spectra, click on the links below:
IC202 mod (SM5BSZ)
The peak to average power and the splatter levels are listed in table 3. The splatter level is given in dBc/Hz below the average power. The values are fetched from the screen dumps linked to above.
|
Peak/avg Splatter level at Diff to CW at
Model and owner power 5kHz 10kHz 15kHz 5kHz 10kHz
(dB) (dBc/Hz) (dBc/Hz) (dBc/Hz) (dB) (dB)
IC202 mod (SM5BSZ) 9.4 73 87 99 62 51
IC706 (LA0BY) 14.3 82 98 104 10 1
IC706MK2G (SM7UFW) 7.8 85 105 113 19 7
IC821H (LA0BY) 8.8 84 96 103 12 9
IC970H (LA3FV) 8.4 89 103 108 11 9
FT100 (SM7GVF) 6.2 76 94 106 32 19
FT221 mod (SM5BSZ) 10.1 87 111 123 28 11
FT817 (LB9VE) 6.3 89 104 110 13 5
FT847 (LA9CM) 16.9 87 98 102 9 9
TM255E (LA6MV) 10.8 65 81 90 51 40
TR9130 mod (LA6LCA) 14.4 73 91 101 51 39
TS850S+conv(LA6MV) 11.0 88 104 117 26 18
Table 3. Splatter level in -dBc/Hz at different frequency separations
from an SSB voice signal.
|
|
The splatter measurement is done the following way:
The Linrad screen is used as a spectrum analyzer.
The SSB signal peak level is recorded from the A/D saturation margin
and/or the Linrad S-meter peak level reading.
The level at which the entire spectrum is 10, 20 or 30 kHz wide is recorded and
it is converted to dBc/Hz by a comparision with the CW carrier measurements.
Finally the peak-to-average power is extracted from the S-meter readings
and subtracted.
This procedure corresponds to placing the CW and the SSB spectra from the
links above in the same image with all the curves vertically displaced
for an equal average power.
A new mode for easy testing of transmitters was added in Linrad-01.05 (after the Gavelstad meeting). The tx test mode of Linrad measures the average power as well as the peak power in 2.4 kHz bandwidth and gives a direct indication of splatter and keying clicks on screen. The results presented here are based on average powers only. Various interference peaks that have a low repetition rate are not detected in the average power levels. Phase noise will not be better in SSB compared to CW so for an ideal SSB transmitter, the curves should be equal already a few kHz outside the SSB passband. This test, a real life SSB voice test, does not necessarily correlate to the results of a two tone test. The two tone test is a good test for an individual building block, an amplifier or a mixer, but it may be grossly misleading for a system that relies on various time constants for ALC circuits and possibly other variable gain circuits. To really know what interference a transmitter causes it has to be fed with the real signal.
Receiver sensitivityTo measure the noise figure, a weak signal of known amplitude was sent into the antenna input. Linrad was used to monitor the loudspeaker output from the receiver under test. The signal level was measured by use of a very narrow filter to ensure that the noise floor did not contribute to the signal level. The noise floor was measured with a rectangular filter of known width, 200Hz or 500Hz, which was used to select a flat section of the noise floor at the side of the weak signal. Nothing was changed on the receiver under test so the signal and the noise levels can be used directly to get the S/N ratio in dB/Hz. With an input signal of XdBm, the noise figure becomesNF=174+X-S/N Table 4 shows the results.
|
Model and owner NF Comments
(dB)
IC202 mod (SM5BSZ) 9.9
IC706 (LA0BY) 10.6 Preamp off
IC706MK2G (SM7UFW) 8.5 Preamp off
IC821H (LA0BY) 2.4
IC970H (LA3FV) 7.3
FT100 (SM7GVF) 2.7
FT221 mod (SM5BSZ) 3.0
FT817 (LB9VE) 5.4
FT847 (LA9CM) 6.1 Preamp off
TM255E (LA6MV) 3.3 Unknown if AIP was on or off.
TR9130 mod (LA6LCA) 5.6
TS850S+conv(LA6MV) 3.7
Table 4. Noise figures
|
Blocking dynamic rangeThe true blocking dynamic range, the one defined in words as "The difference in dB between the weakest useful signal and the strongest out of channel signal that can be present simultaneously without loss of readability for the weak signal". Do not mix this up with BDR as measured by the ARRL lab standards. The tests published in QST give an indication about the level where blocking occurs, but that is usually irrelevant. Readability of weak signals is lost long before due to reciprocal mixing.To be more precise, BDR or two-signal dynamic range, is the difference in dB between the undesired signal that degrades a weak signal's S/N by 3dB and the noise floor. The level of the noise floor will of course depend on the bandwidth and different labs may use different bandwidths. The data presented here is in 1Hz bandwidth. Subtract 27dB to convert to 500Hz. The measurements were made with two generators. One a weak signal, within the passband and another, the strong signal, outside the passband at separations of 5,20,100 and 500kHz. The level of the strong signal P that caused 3dB S/N loss was recorded. S/N was monitored with Linrad at the loudspeaker output. BDR = 174-NF-P
|
Model and owner BDR@5kHz BDR@20kHz BDR@100kHz BDR@500kHz
(dB/Hz) (dB/Hz) (dB/Hz) (dB/Hz)
IC202 mod (SM5BSZ) 138.1 141.1 150.2 150.2
IC706 (LA0BY) 92.6 107.7 125.8 141.3
IC706MK2G (SM7UFW) 106.8 118.7 132.3 139.6
IC821H (LA0BY) 97.8 113.7 129.0 136.5
IC970H (LA3FV) 102.7 123.7 140.7 146.7
FT100 (SM7GVF) 108.6 118.9 130.6 145.1
FT221 mod (SM5BSZ) 116.9 131.2 131.6 147.5
FT817 (LB9VE) 103.3 118.2 132.6 135.3
FT847 (LA9CM) 99.9 116.9 131.9 146.9
TM255E (LA6MV) 128.8 136.9 144.5 151.3
TR9130 mod (LA6LCA) 122.7 135.4 147.8 147.8
TS850S+conv(LA6MV) 112.6 125.6 138.8 149.1
Table 5. BDR, blocking dynamic range in 1Hz.
Subtract 27dB to get BDR in 500Hz bandwidth.
|
Third order intercept pointThe third order intercept point was measured with three signal generators. Two strong signals with equal amplitude plus a third one which was set for a low S/N in the unit under test. On the linrad spectrum display such a signal is well above the noise because of the much narrower bandwidth. The generators were adjusted for IM3 to become equal to the weak signal. The level of the signals at the antenna input give IP3 directly regardless of whether the AGC is in action or not. The strong signal generators were isolated with circulators, combined in a hybrid and finally filtered through a -60dB notch filter at the frequency of measurement. This way the IM3 of the test signal was ensured to be well below the level produced in the units under test and the noise from the strong generators was removed.The results of the IM3 measurements are given in table 6. |
Model and owner IP3@20kHz IP3@100kHz
(dBm) (dB/Hz) (dBm) (dB/Hz)
IC202 mod (SM5BSZ) -10.3 153.8 +2.7 166.8
IC706 (LA0BY) -1.5 161.9 -1.5 161.9
IC706MK2G (SM7UFW) -14.3 151.2 -9.4 156.1
IC821H (LA0BY) -6.7 164.9 -5.2 166.4
IC970H (LA3FV) -5.1 161.6 -0.9 165.8
FT100 (SM7GVF) -19.0 152.3 -16.8 154.5
FT221 mod (SM5BSZ) -45.4 125.6 -17.4 153.6
FT817 (LB9VE) -12.8 155.8 -11.2 157.4
FT847 (LA9CM) Forgot this one -13.7 154.2
TM255E (LA6MV) -12.2 158.5 -12.2 158.5
TR9130 mod (LA6LCA) -21.6 146.8 Forgot this one.
TS850S+conv(LA6MV) -4.1 166.1 -0.5 169.8
Table 6. IP3 absolute and relative to the noise floor.
|
What can we learn from comparing numbers?Since both transmitter and receiver is measured in relation to the noise floor in 1Hz bandwidth we can compare tx and rx dynamic range directly. See table 7. |
Model and owner Rx-Tx@5kHz Rx-Tx@20kHz Rx-Tx@100kHz Rx-Tx@500kHz
(dB) (dB) (dB) (dB)
IC202 mod (SM5BSZ) 3.2 0.3 6.4 4.8
IC706 (LA0BY) 1.6 -0.6 0.4 4.2
IC706MK2G (SM7UFW) 3.0 1.5 7.3 13.6
IC821H (LA0BY) 2.0 0.6 1.3 5.6
IC970H (LA3FV) 2.6 2.1 8.7 12.9
FT100 (SM7GVF) 1.0 -0.1 1.2 21.1
FT221 mod (SM5BSZ) 1.4 0.8 -8.4 -0.8
FT817 (LB9VE) 2.0 1.0 2.2 2.7
FT847 (LA9CM) 3.9 1.9 1.5 6.1
TM255E (LA6MV) 12.6 14.6 19.0 10.2
TR9130 mod (LA6LCA) -1.4 0.2 6.4 0.6
TS850S+conv(LA6MV) -1.3 -3.6 5.0 11.8
Table 7. Difference between rx and tx dynamic range.
A negative value indicates blocking or bad filtering in power supply
for rx amplifiers, positive values indicate noise in the transmit
amplifiers.
|
|
When looking at table 7 it immediately becomes clear that the TM255E has
a problem that would be easy to fix.
The first stage after the filter in the transmit chain is noisy
(could be some other stage, but that is not likely).
By increasing the level by 20 dB in front of the filter or by modifying
the amplifier after the filter this transceiver could be converted to
an excellent rig when it comes to tx purity.
It is similar to the other rigs as it is in its original shape, but since
it has an excellent local oscillator in contrast to all other unmodified
units, this transceiver should be easy to make excellent on the transmit
side.
IP3 and BDR can also be compared directly. see table 8.
|
Model and owner IP3-BDR@20kHz IP3-BDR@100kHz
(dB) (dB)
IC202 mod (SM5BSZ) 12.7 16.6
IC706 (LA0BY) 54.2 36.1
IC706MK2G (SM7UFW) 32.5 23.8
IC821H (LA0BY) 51.2 37.4
IC970H (LA3FV) 37.9 25.1
FT100 (SM7GVF) 33.4 23.9
FT221 mod (SM5BSZ) -5.6 22.0
FT817 (LB9VE) 37.6 24.8
FT847 (LA9CM) 35.4 22.3
TM255E (LA6MV) 21.6 14.0
TR9130 mod (LA6LCA) 11.4 Forgot this one.
TS850S+conv(LA6MV) 40.5 31.0
Table 8. Difference between IP3 and BDR, both in dB/Hz.
The negative value for FT221 is because a strong signal reduces gain
without affecting S/N.
A large value means that the receiver is limited by sideband noise,
probably due to reciprocal mixing although bad filtering of supply voltages
could give the same effect.
|
| Large numbers in the comparison tables 7 and 8 or in the last two columns of table 3 may indicate a design error of some kind. The tables are shown here to demonstrate the advantage of having a common reference power level for all dynamic range measurements, for tx as well as for rx. Wheter a large number indicated a correctable error or not must be judged by a comparision of the dynamic range data themselves between the different transceivers. |