Traditional measurementsThere are basically two types of measurements:IMD, intermodulation measurements quantify the non-linearities by measurement of the mixing products produced when two strong signals are fed to the receiver simultaneously. BDR, blocking measurements quantify how strong a single undesired signal may be before it starts to degrade the desired signal. IMD measurements are usually quantified by a single number, the intercept point. The concepts of IP2 and IP3 are well described in many places so there is no reason to repeat the basics here. IP2 and IP3 are usually given as power levels in dBm. Such numbers are meaningless if not the noise floor is specified at the same time. A mediocre radio with an IP3 of -20dBm could seemingly be converted to a good one with an IP3 of +10dBm just by adding a 30dB attenuator in front of it if the effect on the noise floor was not taken into account. Similarly, BDR values, often specified as dB, are meaningless unless the bandwidth is specified.
Comparing IP3 valuesTo make a fair comparison between different IP3 values, it is reasonable to assume that a good preamplifier is present as the first stage. The gain of the first stage is a matter of taste. As long as the noise from the first stage makes a small contribution to the total noise floor it is ok to neglect it. With modern technology there is no reason for the first amplifier to have a noise figure above 2dB which means that receivers with a noise figure above 6dB can be compared directly if IP3 is expressed as dBc/Hz.Another way of saying the same thing is to say that a receiver with a noise figure of 6dB and above should be completely dominated by the noise of stages that come after the non-linearity that limits the dynamic range. Such receivers can therefore be compared using just a single number, the distance from the maximum output level from the non-linear component (usually a mixer) to the noise floor of the amplifier that comes after it. If the noise figure is very low, the dominating contribution should be the first preamplifier. If only 0.1dB of system S/N loss is accepted, 97.5% of the noise must come from the preamp which means that the noise floor is lifted by 16 dB and IP3 values will be 16dB lower when expressed as dBc/Hz compared to the same radio without the preamp. IP2 and IP3 values should be specified as how many dB above the noise floor in one Hz each of the two equally strong signals would be to be equally strong as the intermodulation products when extrapolation is made from low to high levels. Specified this way, IP values will not change if an attenuator is inserted. IP3 levels in dBc/Hz give a fair comparison of receivers with noise figures of 6dB and above. VHF/UHF receivers with much lower noise figures loose some dynamic range due to the noise of the first amplifier. Rather than specifying the intercept point in dBc/Hz one can specify the intermodulation free dynamic range by extrapolating the intermodulation level down to the noise floor. The two tone, third order IMD dynamic range is (2/3) * IP3 while the second order IMD dynamic range is (1/2) * IP2
Comparing BDR valuesBDR values that are measured at different bandwidths can be presented at a common standardised bandwidth of 1Hz. To do this one simply adds the dB value, 10*log10(Bw) so measurements done at for example 500Hz bandwidth become adjusted by addition of 27dB.Unfortunately the comparison of BDR values is not very accurate because the specified bandwidth is often not accurate at all. Particularly measurements made at SSB bandwidths are influenced by the audio frequency response of the receiver under test. The audio response may be far from flat for good reasons. It is for example perfectly acceptable for a radio to compensate poor response at low audio tones for a loudspeaker in a small box with more audio gain at the corresponding audio frequencies. Audio response is a matter of taste. It may slope by 10 dB one way or the other and that does not affect the real performance of the radio at all. The procedure to measure BDR by first locating the frequency at which the strongest signal is obtained, then finding what signal level is required for twice as much power as the noise floor itself should be related to the actual noise bandwidth including the audio response and not the nominal bandwidth of the IF filter. (It is essential to use a true RMS voltmeter here) This measurement is the MDS (minimum discernible signal) and it gives the signal level that produces the same power as the noise floor. (the sum is at +3dB) The noise bandwidth is the integral of the spectral response in linear power scale over all frequencies. For an ideal rectangular response it becomes equal to the bandwidth. For filters that slope by 10dB over the audio range, which is not uncommon at all, the noise bandwidth is much narrower than the nominal IF bandwidth and the BDR values incorrectly become better than they really are if the nominal bandwidth is used for the comparison. Measuring the noise floor as the MDS value is misleading in case the audio response is not flat and the bandwidth associated with the measurement is the IF filter bandwidth. By presenting the MDS value as dBm/Hz, the person doing the measurement takes the responsibility for correcting for any such effects of the audio amplifier. The MDS value expressed in dBm/Hz is directly related to the noise figure. Any room temperature resistor, regardless of its resistance value, will deliver -174dBm/Hz to a resistor of the same resistance that is held at the absolute zero temperature 0 K. Two resistors at the same temperature will deliver the same power to each other because they are in thermal equilibrium. A receiver that has a noise figure of 6dB will deliver 4 times more noise compared to an ideal, noise free receiver when connected to a room temperature resistor. This is the definition of of noise figure. Obviously the MDS value in dBm/Hz converts to NF by the addition of 174dBm/Hz. Particularly at VHF and microwave frequencies one measures noise figures directly with a noise figure meter rather than evaluates the correct bandwidth and measures the MDS value. The blocking signal level is the level of a signal outside the passband that reduces the S/N of the desired signal by a small amount. If "a small amount" is defined as 3dB, the BDR values in dBc/Hz become directly comparable to the noise sideband measurements that conventionally are made on transmitters. For close frequency separations the BDR values should be noise limited and equal to the transmitted noise at the same frequency separation for a modern transceiver.
Some numbers for a modern transceiver.To get some actual numbers for a modern transceiver we can use the measurements on the ICOM IC-756PROII done by the ARRL lab and published in the February 2002 issue of QST. Below the values at 14MHz are used, and the bandwidth 500Hz is assumed to be equal to the true noise bandwidth.Noise floor, MDS, is -131dBm, -139dBm and -141dBm respectively for preamp off, preamp one and preamp two respectively. These values correspond to -158dBm/Hz, -166dBm/Hz and -168dBm/Hz and expressed as noise figures, the values are 16dB, 8dB and 6dB. The third order IMD dynamic ranges are 97/95/91 dB at 20kHz spacing and 76/75/72 dB at 5kHz spacing when referred to 500kHz bandwidth for preamp off/preamp one/preamp two. When referring IMD dynamic ranges to narrower bandwidth, the numbers do not grow by the bandwidth factor. When the noise floor is lowered by X dB due to narrower bandwidth, the two test tones have to be reduced by X/3 dB for the third order IM products to go down to the new noise floor. Therefore the IMD dynamic range changes with two thirds of the bandwidth factor. Referred to unity bandwidth IMD dynamic range values for IC-756 become 115/113/109 dBc/Hz and 94/93/90 dBc/Hz. From these values we can conclude that IP3 is at 172.5/169.5/163.5 dBc/Hz for 20kHz separation and 141/139.5/135 dBc/Hz at 5kHz separation. The measured IP3 values are +20.2/+10.2/-4.1 dBm and -18.8/-28.8/-35.5 dBm. When we subtract the noise floor, the following values are obtained for IP3 in dBc/Hz: 178.2/176.2/163.9 and 139.2/137.2/132.5. The 20kHz values differ by about 6dB from the IP3 values deduced from the third order IMD ranges for unknown reasons. The blocking dynamic range is measured as 118/116/111 dB and 100/97/94 dB for 20kHz and 5kHz respectively. In unity bandwidth these values correspond to 145/143/138 dBc/Hz and 127/124/121 dBc/Hz. When making a comparison with a wideband digital radio it is essential to have some more data than what is supplied in the ARRL measurements. A digital receiver becomes abruptly overloaded with very strong spurious signals as a consequence if the input exceeds the range of the A/D converter. If for example a 100kHz passband is filtered out and sampled by an A/D converter, the instantaneous voltage produced by all signals in the passband occasionaly becomes very high. It must never exceed the A/D saturation limit. For an analog receiver, the corresponding signal level is the 1dB compression point. Analog receivers do not produce anywhere near as sharp limitation as a D/A so such comparison is not quite fair to the analog receiver, but it is far more reasonable than noise limited BDR values. As a rule of thumb one can say that 1dB compression happens 15dB below IP3. This means that 1dB compression for IC-756PROII is at 163dBc/Hz or at 157dBc/Hz depending on which of the above results for IP3 one uses. It would be very interesting to have the 1dB compression point at a larger frequency separation too. It could well be a little better or even much better. In case the BDR of an analog receiver is noise limited at 100kHz separation it is not fair to compare the 1dB saturation point to the saturation point of a digital system. The important number is the BDR, the discussion about 1dB compression was intended to give an indication as to the level where BDR is saturation limited. For IC-756 a reasonable guess is that it happens at a frequency separation of about 100kHz. Comparing analog and digital receiversThe most critical number for a digital radio is the noise level in relation to the saturation level. In the data sheet this is usually expressed as the SNR value (S/N ratio) at full bandwidth.A typical radio A/D, the AD6644 from Analog devices gives typically 74dB SNR at a sampling frequency of 65MHz. When expressed in unity bandwidth, the saturation point (1dB above the test level) becomes 150dBc/Hz in relation to the noise floor. This is about 10dB below the performance of the IC-756. For wider bandwidths it will compare even less favourable. It is possible to achieve similar and even better results by use of top class audio converters but it is really hard (or impossible) to beat the analog radio saturation performance with a wideband digital design. On the other hand, the analog radio is usually noise limited well below the saturation point because of LO sideband noise. This is easily avoided in a wideband digital design. If the standard BDR measurements were done on a wideband design the numbers at 20kHz and at 5kHz would both be close to 150dBc/Hz. At wider separations IC-756 should reach a BDR of 160dBc/Hz or possibly 170dBc/Hz but a wideband digital design will stay at 150dBc/Hz also for large frequency separations. Until wideband A/D converters have improved by about 20dB one will have to add some analog circuitry to do the filtering required in front of the chip. In most cases this means that the A/D conversion is made at some IF frequency and that a fixed frequency filter is used. This is exactly what we see in modern transceivers like the IC-756. Tunable preamplifiers, one for each band is another possibillity. Not difficult to design and excellent BDR performance since no variable local oscillator is used. The A/D converter itself and the digital processing that follows it do not have the normal behaviour with quadratic and cubic growth of second order and third order distortion. The analog circuits that preceed the A/D converter have normal behaviour however, so intermodulation measurements are relevant. The bandwidth of the signal reaching the A/D converter is very important. If a blocking dynamic range of 160dBc/Hz is adequate at a bandwidth of 40kHz, one would need 166dBc/Hz for a bandwidth of 80kHz with the argument that twice as many strong signals will be present when the bandwidth is doubled. They will sum up to twice the amplitude. It will happen less frequently by a factor of two, but for an A/D converter it must not happen at all. Possibly new algorithms can change this requirement in case saturation is caused by a few narrowband signals. Knowing how they add will allow the computer to estimate the missing data |