# Compression and Cross-Modulation

This column has been dealing with DTV-DTV interference and has almost completed this discussion. Readers know, and I hope understand, that most DTV-DTV interference (the exceptions being co-channel and image frequency interference) is due to third-order nonlinearity.

In the case of adjacent-channel interference, nonlinearity in transmitters results in sideband splatter into both adjacent channels, and you know that this splatter is noise within the adjacent channel.

In taboo-channel interference and in adjacent-channel interference, nonlinearity--primarily in the tuner (mostly in the mixer)--results in generation of IM3 (third-order intermodulation) products. We have talked enough about IM3, but have we talked enough about two other IM3 products--compression and X-M (cross-modulation)?

Dr. Oded Bendov's 2004 IEEE Broadcast Symposium paper went into these aspects of third-order nonlinearity quite thoroughly. By the time you read this, the IEEE Transactions on Broadcasting will have published this paper, which is bound to become a classic. So, what did he report to IEEE last October?

LURKING UNDERNEATH

He showed that in addition to IM3 that falls in the adjacent channels, there is much more IM3 (call it noise) which lies inside the channel, underneath the sidebands of the DTV signal. The IM3 within the DTV channel reduces the signal-to-noise ratio of the desired signal. His calculations show that the in-channel IM3 (which you can't see) is 9.8 dB above that which you can see in both adjacent channels; this will also be true of IM3 generated in the transmitter.

However, if the out-of-channel sideband splatter is 44.5 dB below the DTV signal in-channel, then the in-channel splatter is 34.7 dB below the DTV signal, or a signal-to-noise ratio of 34.7 dB, which is excellent performance. Let's see why.

Given a signal power to a spectrum analyzer from the Tx (transmit) output, say at -30 dBm, the noise in the DTV signal is -64.7 dBm. If the signal power at a receiver input is -80 dBm, the Tx noise within the DTV channel is -114.7 dBm, well below the receiver-generated noise (about -92 dBm) so noise from the Tx is inconsequential. That's the good news.

The bad news is that in addition to IM3 noise, there is also signal compression as another result of third-order nonlinearity in the Tx. Ah-ha, you say, but my Tx has adaptive pre-correction so there is no bad news for me.

Ah-ha, I say, but just as third-order nonlinearity in the Tx causes signal compression, it also causes signal compression in overloaded tuners of receivers that have no adaptive pre-correction. Again, the culprit is usually the mixer. Why is it that the mixer always gets blamed for nonlinearity in receivers? A mixer is an inherently nonlinear device. It has a very useful second-order nonlinearity, which is why it can change the desired RF frequency to the IF frequency.

LOCAL OSCILLATOR LEVEL

But to do that without distorting the signal, the LO (local oscillator) power injected into the mixer must be well above the desired signal power at the mixer input. The LO power must be >10 dB above the desired signal power to avoid third-order distortion, which results in compression. LO power is expensive; the obvious cost is the need for a buffer amplifier if the LO cannot provide enough power.

In my RF test bed, I chose extremely linear mixers that require +17 dBm of LO power. That took two stages of buffer amplifiers to get the LO power up to +17 dBm. Those buffers aren't cheap and I didn't have to pass FCC Part 15. The less obvious cost is due to FCC Part 15, which limits LO radiation by consumer electronic devices. I'm talking about the costs of shielding, filtering and decoupling, to say nothing of compliance testing.

The effect of signal compression is to reduce the highest power symbols, making them decode incorrectly. By attenuating the DTV signal, compression ceases, and DTV reception may resume. It usually doesn't take much attenuation; try one or two 6 dB attenuators.

So mixers are going to produce IM3 and compression and cross-modulation. At high RF levels near the transmitter, Dr. Bendov showed that X-M is about the same as the IM3 within the channel. X-M is the transfer of information from one channel to another. What happens is that the desired signal is amplitude modulated by the undesired signal. In case you missed this subtle point, differential gain is cross-modulation from the luminance to the chrominance channel of NTSC and PAL signals.

AM has been around for a hundred years. A strong undesired AM radio station anywhere on the radio dial might be heard while listening to a weak station, especially during those pauses in the modulation of the desired carrier. This led to the invention of special RF amplifier tubes around 1930 that didn't generate X-M.

(click thumbnail)I wanted to see X-M on a spectrum analyzer. This is not possible in the case of DTV-DTV, but it sure is for DTV into NTSC. This is because when the desired signal is DTV, the desired signal sidebands mask the smaller X-M sidebands. The best NTSC signal for this experiment has the least sidebands, so I simply switched off the sidebands, leaving just the unmodulated visual carrier. I created a special version of a two-tone test signal in which there is a "U tone,"--an 8-VSB modulated DTV signal on Channel 33, which I call Fa in Fig. 1, and an unmodulated carrier at the visual carrier frequency of the desired Channel 37 (609.25 MHz), labeled Fb in Fig. 1.

Fig. 1 shows these two "tones" undistorted. Note that there is no IM3 clustered around the DTV signal, or around the NTSC visual carrier either. My RF test bed has a spurious signal denoted by the market symbol, which is of no consequence, but you might have been wondering what it is.

Fig. 2 shows the output of a solid-state amplifier overloaded by this special test signal. Surrounding the visual carrier frequency (Fb) is cross-modulation from Fa. How do I know that? Simple; I turned off Fb, and the X-M vanished, although the IM3 still surrounds the 8-VSB signal. You cannot cross-modulate what is not there! X-M also increases the noise hidden under your DTV signal.

(click thumbnail)Perhaps you have noticed in Fig. 2 a cluster of noise marked "2Fb-Fa" between 628.8 and 634.2 MHz. This is the IM3 generated by the interaction of the two elements of this special two-tone test signal. This is readily demonstrated by turning off Fb. It vanishes. You would have also noted the other cluster of noise marked 2Fa-Fb. This IM3 spectrum is 10.8 MHz wide because it is the difference between the second harmonic of Fa and the fundamental of Fb. The spectrum of the X-M is double sideband amplitude modulation of Fb by the sidebands of Fa, which are at one-half the symbol rate of 8-VSB (5.38 MHz). It vanishes when Fb is switched off.