Linearity Requirements for MMICs for 16 Through 128 QAM ETSI Compliant Radios

Special Report

Linearity Requirements for MMICs for 16 Through 128 QAM ETSI Compliant Radios

Jim Harvey
Mimix Broadband
Webster, TX

This article highlights some of the regulatory and operational issues that influence linearity requirements and frequency planning aspects of high frequency microwave radios used for point-to-point or point-to-multipoint systems. While most regulations in the US relate to limiting interference to other users, much of the international market follows tighter European regulations, which constrain receiver and system performance. Even in an unregulated market, optimum utilization of the spectrum results from receivers with high resilience to interference.

The European Telecommunications Standard Institute (ETSI) and the UK Ministry of Post and Telecommunications (MPT) have a number of requirements which need to be met to allow the sale of radio equipment in prescribed countries, and many non-US administrations follow the ETSI and UK MPT recommendations. These include:

• Transmit spectral mask
• Receiver thresholds, co-channel and adjacent channel carrier-to-interference ratio (C/I)
• Receiver dynamic range
• Receiver susceptibility to high level interferers

Linearity

There are several constraints on third- and fifth-order intermodulations and one known constraint on second-order intermodulation.

• The transmitter must meet the spectral mask requirements, in particular adjacent channel power constraints due to spectral re-growth arising from third- and fifth-order distortion. (This is a regulatory limit.)
• The transmitter third- and fifth-order distortions will also generate in-band distortion products, and these must not degrade the receiver performance beyond acceptable limits (customer determined).
• The receiver third- and fifth-order distortions at high received signal levels will generate in-band distortion products, and these must not degrade the receiver performance beyond acceptable limits (customer determined).
• Some standards set a limit to the third-order distortion at threshold by requiring tones relative to signal at +19 dBr, spaced to put a third-order intermodulation product (IM3) at the receive frequency, to have a negligible effect on threshold. (This is a regulatory limit.)
• The transmitter power may mix with a high level CW interferer to generate second-order products at the receive IF. The impact of this is dependent on the spacing between transmit and receive frequencies, and the first receive IF. It also depends on diplexer isolation. (This is a regulatory issue.)

Transmitter Linearity

The typical mask for a quadrature amplitude modulation (QAM) signal, specifically 16 QAM in the 24.25 to 29.5 GHz band as defined in ETS 300 431, is shown in Figure 1 . The signal level is relative to the nominal level measured near the carrier, and the +1 dB limit accommodates the leaked carrier as well as minor ripple in the pass band. Spectrum analyzer settings are specified, in particular the IF bandwidth is 100 kHz, and the video bandwidth is 300 Hz. The limit is specified out to 2.5 times the nominal bandwidth of 56 MHz, that is, 140 MHz from the nominal carrier, where another emission specification takes precedence. This sets the levels for spurious rather than intermodulation products. Limits for 128 QAM in 28 MHz are not yet determined, but will probably be close to 50 percent scaling of the mask shown for 56 MHz. ETSI EN 300 430 V1.3.1 (2001-02) defines the mask requirements for Class 5 modulation in 27.5 MHz channels, but this standard is directed at N+1 protected systems using adjacent channel alternate polarization or adjacent channel co-polarization systems, and are more severe than probably required for stand-alone single channel systems.

There is no rigorous way of predicting spectral growth from CW output third-order intermodulation intercept point (OIP3) measurements. A simplistic approach sets the drive level at a point where the two intermodulation tones meet the mask, and this depends on the frequency spacing. Taking two tones at ±11 MHz relative to the carrier gives IM3 tones at ±33 MHz from the carrier, where each tone is required to be -30 dB relative to the signal. The third-order products from two frequencies f₁ and f₂ fall at 2(f₁ - f₂ ) (below the stimulus) and 2(f₂ - f₁ ) (above the stimulus). The spacing of the intermodulation signals equals the spacing between the input tones. If a 20 dBm total output power is needed, each signal tone is assumed to be at 17 dBm, and the IM3 tones must be -13 dBm, that is -30 dBc. Then the OIP3 are required to be 15 dB above 17 dBm or +32 dBm.

Figure 2 shows the relationship used to determine these levels, and assumes a power amplifier with an arbitrary 20 dB gain and an OIP3 of 32 dBm. The gain line is imaginary above the linear range of the amplifier, as this amplifier would be expected to have a P1dB (1 dB single tone compression point) at approximately 22 dBm. In practice, the determination of the intercept point requires careful measurement of the tones at a number of levels to ensure the products move with the correct slope relative to the input signal levels.

Experimental determination of spectral shape for a real 16 QAM signal has been accomplished to confirm this approximate method. For an amplifier with 32 to 34 dBm OIP3 (frequency dependent) and a roll-off factor for a linear modulation process, (alpha) value varying between 0.3 and 0.5, output power levels of 20 to 22 dBm were achieved for a 30 dB suppression at the corner point of the mask.

Residual BER Due to Transmitter Intermodulation

The signal-to-noise ratio (SNR) for a 10^-6 bit error rate (BER) in square 16 QAM is 20 dB; for 128 QAM it is typically 29 to 30 dB. Third-order distortion is expected to generate noise in the adjacent channels. However, it also produces noise products that fall in-band, and which then degrade the threshold.

The determination of how much of the third-order distortion falls in-band is not simple, and depends on the shape of the signal, its alpha and the demodulator characteristics. Simplistically, it can be assumed that the IM3 power is at -30 dBc as required to meet the mask, and with 16 QAM a 20 dB SNR is required to meet a 10^-6 BER. The addition of more noise at a level 10 dB down is known to raise the total noise level by 0.4 dB (assuming both noise sources are Gaussian), so, on this basis, it can be assumed that the received signal level required to restore 10^-6 BER would increase by 0.4 dB before error correction. The change after error correction will be much less than 0.4 dB.

For 128 QAM, the level of noise is comparable to the SNR required to meet 10^-6 BER, so it might increase the effective noise level by 3 dB, which would have a small but noticeable impact on receiver threshold after error correction. In that case, an increase in the transmitter's OIP3 point might be justified, over the level needed to meet the spectral mask. To have less than 1dB increase in threshold due to this source of noise, a -37 dBc interference total is required, or -40 dBc per tone, assuming that all the interference falls in-band. This would require an OIP3 20 dB above each wanted tone, or 17 dB above the total output power, that is 37 dBm for the above example of 20 dBm total output power. This is conservative on several grounds, for much of the interference falls out of band and typical forward error correction (FEC) circuits will correct 10^-2 uncorrected BER to 10^-6 corrected BER. Anecdotal evidence indicates that meeting the mask is the more severe of these constraints, but this needs to be further investigated.

Receiver Linearity

Some ETS documents call for a direct third-order test at threshold, requiring two signals at +19 dB relative to threshold, close to the received signal, with one of the IM3 tones falling on the received carrier. The 10^-6 BER is not to be degraded beyond 10^-5 , corresponding to a 1 dB increase in thermal noise (uncorrected, FEC will relax this constraint).

For 16 QAM, assuming a worst case threshold of -67 dBm, the interferers are each at -48 dBm (+19 dBc). For a 20 dB SNR, the intermodulations must be -27 dBc total that is -30 dBc per tone, corresponding to -97 dBm per tone, or -49 dB relative to the interferers. This implies an IIP3 of (-48 + 24.5) = -23.5 dBm when the received signal level is at threshold, that is 43.5 dB above threshold.

For 128 QAM with a 29 dB SNR, the threshold will be 6 dB higher or approximately -61 dBm (9 dB from SNR but the bandwidth is halved), while the intermodulations must be -36 dBc total or -39 dBc per tone, corresponding to -100 dBm per tone. The interferers are now -42 dBm, so the intermodulations are -58 dB relative to the interferers. This implies an IIP3 of (-42+29) = -13 dBm when the received signal is at threshold, that is 48 dB above threshold.

This requirement is not called up by the September 1996 issue of ETS 300 431 (24 to 29 GHz), but is called up by more recent ETS for 23 and 38GHz, EN 300 198 and EN 300 197, and may flow through to the 24 to 29 GHz standards.

More generally, the receiver must have an input referred third-order intercept point high enough so that at maximum received signal level intermodulations generated in the receiver (generally in the first mixer) do not prevent the radio from operating satisfactorily at maximum received signal level (RSL). ETSI specify the maximum RSL at -15 dBm, at which the corrected BER must be better than 10^-3 . In practice, the uncorrected BER must be better than 10^-3 to 10^-1 to ensure that the error corrector or demodulator do not misbehave. A distortion level of -17 dBc for 16 QAM and -27 dBc for 128 QAM would probably be close to this point, so an IIP3 of 8.5 dB (16 QAM) or 13.5 dB (128 QAM) above the per-tone level corresponding to -15 dBm (ETSI) is required. Assuming 1 dB branching loss, this -15 dBm RSL corresponds to -19 dBm per tone, so the IIP3 at the MMIC input must be greater than -10.5 dBm for 16 QAM or -5.5 dBm for 128 QAM at maximum RSL. Note the noise figure achieved at threshold is not required to be maintained at maximum RSL.

Second-order intermodulation is normally not an issue in radio receivers or transmitters. In the case of a frequency division duplexing (FDD) terminal, there is leakage of the local transmitter into the receiver, typically at a level of -40 to -50 dBm (60 to 70 dB isolation with a local transmit power of 20 dBm). There is also an ETSI requirement that a CW interferer at a level 30 dB above threshold will not degrade the threshold, typically by more than 1 dB. Threshold levels for 155 Mbps/56 MHz systems are typically better than -70 dBm, so the interferer level is typically -40 dBm, comparable to the transmit leakage. Careful selection of internal intermediate frequencies relative to the transmit/receive (T/R) frequency spacing can lower the relevant interfering level by rejection in the receive waveguide filter, at the cost of a fixed tuning bandwidth. To determine the worst case, first assume no protection. If the interference level generated by two tones at -40 dBm is desired to be -100 dBm, then the second-order intercept must be 60 dB above the average of the two interferer levels (in dBm). This level is +20 dBm (referred to the receiver input, that is IIP2). Note that this level is required at the threshold, where a good noise figure is also required, so for most receivers some degree of protection is required. This can be done by obtaining more isolation from the transmitter, or by selecting the frequency plan so that some attenuation of the CW interferer is provided at the critical frequency where the interferer to transmitter signal spacing equals the receiver first IF. This requires setting the IF well away from the T/R spacing. Typical T/R spacings are 700 MHz to 1.3 GHz at 23 GHz to 40 GHz. Choosing an IF in the 2 to 3 GHz region will generally suffice, depending on the required receiver tuning range.

Linearity Summary

To meet the transmit mask requirements, the transmitter OIP3 must be 12 dB above the required output power level, that is 32 dBm for a 20 dBm output power.

To avoid degrading the threshold at 128 QAM, the worst case requires OIP3 17 dB above the required output power level, although this may be overly conservative.

To meet the 19 dBc two-tone test at threshold, the IIP3 at threshold should be 45 to 48 dB above the receiver threshold, typically -22 dBm to -13 dBm for 16 QAM and 128 QAM, respectively.

To meet the dynamic range requirement in the receiver, the IIP3 must be -10 dBm or -5 dBm for 16 QAM and 128 QAM. This is at high received signal level, so the noise figure may be degraded to meet this requirement (by a MMIC attenuator before the first mixer, for example), and the use of an adaptive transmitter power control (ATPC) may also relax the requirement.

Filtering Requirements

In addition to the transmit-to-receive isolation requirement, there are a number of filtering constraints required to meet the high level CW interferer previously discussed. In addition to the second-order effects of the high level CW interferer at +30 dBc, two channels away from the carrier (112 MHz for 155 Mbps/56 MHz 16 QAM), the high level interferer must not cause desensitization or blocking in the receiver. Similarly, a "like" modulated interferer in the adjacent channel, as high as 11 dB above the wanted signal at threshold, must not worsen the 10-6 threshold by more than 1 dB. In general, there will be no protection from front-end waveguide filters for signals this close to the wanted signal, so the first receiver IF filter and the linearity of the IF chain must take these into account.

Possible Relaxations

The specification in question does not specifically address the impact of forward error correction, which is generally applied to current radios. The impact of FEC is to improve the threshold by several decibels, and to steepen the typical waterfall curve of BER plotted against RSL. Both of these effects lessen the impact of the in-band distortion previously discussed. The exact nature of the FEC algorithm used must be considered to determine the resulting requirement. Some of the constraints arise from the high dynamic range required, from -70 dBm or so to -15 dBm. This upper limit is relaxed in later specifications for 23 GHz radios to 50 dB above threshold, or -30 dBm, and when ATPC is implemented, the dynamic range is relaxed to 40 dB.

Conclusion

The most obvious concerns regarding linearity relate to the occupied spectrum of the transmitted signal, which must not radiate significant power into the adjacent channel used by another link. This was not an issue in the constant envelope modulation formats, such as 4FSK, which was almost universally used in the previous generation of millimeter radios with saturated output stages, including directly modulated Gunn diode oscillators. The need for better utilization of the spectrum led to 16 QAM, with double the Bits/Hz spectral efficiency of 4FSK or QPSK/4QAM. A further doubling of capacity to 128 QAM may require even better transmitter linearity, not necessarily to maintain adjacent channel power levels, but also to keep in-band distortion low enough not to degrade the link performance.

In unregulated environments, receivers were often allowed to operate "wide open," some with negligible image rejection or adjacent channel immunity. As the value of spectrum increased, regulated administrations began to specify the ability of the receiver to operate in a crowded environment, allowing more licenses to be issued and more revenue obtained. The ETSI and MPT standards led the way in this regard, and have been adopted by many non-European countries. Most users now appreciate the value of having microwave radio links with good electromagnetic compatibility, as it allows them to extract maximum use from the available spectrum.

References
The radio designer is urged to obtain the latest revision of applicable specifications, noting that significant changes in the technology are being reflected in changed parameters in the requirements. Many administrations impose further restrictions over the generally accepted ETSI and MPT documents. Table 1 lists the specifications which may be relevant to high frequency microwave radios used for point-to-point or multipoint systems.

The ETSI specifications can be obtained by visiting the ETSI web site (www.etsi.org). MPT requirements are generally being aligned with ETSI specifications.

Jim Harvey is the chief technology officer of Mimix Broadband, a designer and supplier of MMICs to the high frequency microwave industry. Dr. Harvey has a long history of system design for telecommunications, and for the last eight years has participated in the evolution of millimeter radios from Gunn diode or multiplier-based FSK technology, through GaAs pHEMT MMIC-based linear radios.