My previous column introduced readers to a new parameter, third-order intermodulation (IM3), which is the sideband splatter from a DTV transmitter into both adjacent channels.
This splatter is limited by the DTV RF mask required of all DTV transmitters. What the RF mask filter cannot remove is the splatter close to the DTV signal channel. This splatter is what causes interference into adjacent channels, not poor selectivity of the victim receiver. The maximum IM3, radiated in each adjacent channel, is 44.5 dB below the power radiated in the DTV channel. I define this as the co-channel noise afflicting reception of an adjacent channel, calling it (I):
I = U -46.5 dB.
Did I just contradict myself? No. The maximum power radiated by a DTV transmitter is U -44.5 dB, but a DTV receiver tuned to either adjacent channel has some selectivity that discriminates against noise just inside the desired channel; hence, the bottom line is I = U -46.5 dB. I'll try to pick up a dB of interference rejection anywhere I can find it.
Suppose the D/U ratio at your home is -30 dB. The co-channel noise due to IM3 radiated by one adjacent DTV channel transmitter is D -16.5 dB. This means that a 1.3 dB fade or a little echo will put your signal-to-noise + interference (S/N+I) at threshold, 15.2 dB. If two DTV adjacent-channel signals are at the same level, I = U -43.5 dB, so your S/(N+I) falls to 13.5 dB, and reception fails. Why?
But that was just the first part of the story. Even at a moderate U signal level, there may be additional IM3 generated in your DTV tuner. The ATTC tests demonstrated this at a U level of -25 dBm. Receiver-generated IM3 adds directly to the IM3 that accompanies the adjacent-channel DTV signal into your tuner; receiver-generated IM3 attacks your desired signal, unattenuated. You might have had reliable reception until a DTV transmitter signed on at its maximum authorized power. Fortunately, if you put a 3 dB attenuator in the downlead right at the receiver, you will attenuate the receiver-generated IM3 by 9 dB! With only a 3 dB loss in desired signal power, that may take care of the problem. Suitable 75-ohm, 3 dB attenuators are readily available and inexpensive. That simple remedy may fix your reception problem, but what will happen when your viewers encounter such problems? Perhaps broadcasters should make a brochure available to help viewers with DTV reception problems. They might look to the NAB for such help.
THIRD-ORDER INTERCEPT POWER
This is a figure of merit that tells you how well a receiver, or an amplifier, avoids generating IM3. Readers were introduced to this in the June 23, 2004 installment of Digital TV. A receiver with a +20 dBm third-order intercept power (IP3) can withstand an undesired signal of -20 dBm as the IM3 would-100 dBm well below the receiver-generated noise level and hence, not detectable. In contrast, a receiver with 0 dBm IP3 would generate IM3 at -60 dBm, so the minimum usable desired signal would have to exceed -45 dBm. Ask yourself or your consultant how much of your DTV coverage area has higher than -45-dBm power available to receivers.
Third-order intercept power is not the kind of information consumers ask about, nor is it what manufacturers are willing to talk about. I believe this crucial interference abatement parameter should be talked about and published. A minimum value > +15 dBm should be mandated by the FCC.
And here is why.
Some readers will recall reading about the interference caused by undesired signals on certain pairs of channels near the desired channel. Those pairs include: n-6 and n-3; n-4 and n-2; n-2 and n-1, and n+1 and n+2; n+2 and n+4; n+3 and n+6 and perhaps others. The IM3 generated by signals on all these channel pairs (in a tuner), all land on victim channel n.
There are cities with more than one such channel pair, in which case, the noise in channel n is the sum of the multiple IM3 being generated by those pairs of undesired signals. In the case of both the n-1 and n-2 pair, and the n+1 and n+2 pair, these IM3 add directly to the sideband splatter that accompanies the undesired signal on n+/-1. But is this additional IM3 significant or trivial?
That was the question in my mind until recently. What has changed? I designed and built my own DTV RF test bed, which is shown in Fig. 1. It has two of Prof. Kraus' 8-VSB exciters in it. Each generates an 8-VSB signal from data stored on the hard drives. Therefore, these 8-VSB signals are not coherent. They are at an intermediate frequency of 63 MHz (Channel 3), which is the IF output, and they are upconverted to 611 MHz (Channel 37) for the RF output. The 63 MHz IF signals are upconverted to any pair of UHF channels, passed through HP wideband attenuators, then combined and fed to a master HP undesired signal attenuator. After that, these undesired signals are combined with a desired signal on Channel 37. This test signal can be set to any desired level and fed to a device under test. The entire RF test bed fits in my SUV, including the RF signal generators, HP 432A power meter and frequency counter. The exciters are each on 3 5/8-by-6-inch PC board, and they require 12 V at 300 Ma. They are shown in the Euro card cage above the HP attenuators in Fig. 1.
| Fig. 1 Rhodes' DTV RF test bed, with two of Prof. Kraus' 8-VSB exciters, each generating an 8-VSB signal from data stored on hard drives.|
My initial experiments involved undesired DTV signals on Channels 31 and 34 with respect to a desired signal on Channel 37. Fig. 2 is a plot made with a HP X-Y plotter driven by my HP 141/8554L/8552 spectrum analyzer. This shows the familiar "shoulders" (unfiltered sideband splatter) around the undesired 8-VSB signals on Channels 31 and 34. Due to intermodulation generated in the amplifier under test, which simulates a receiver's tuner, we find in both Channels 28 and 37 the dense spectrum of the IM3 generated in the device under test by the two 8-VSB signals in combination. The spectrum of these IM3 extends over 18 MHz, and is centered on Channels 28 and 37. The spectral power density of the IM3 is shown in Fig. 2 to be maximum in Channels 28 and 37. A rough estimate of the total noise power in Channels 28 and 37 can be made from this plot. Compared to the sideband splatter radiated, we see that this IM is much stronger; it may be around 35 dB below the undesired signals in Channels 28 and 37 that generated it in the receiver's tuner. I want to actually measure the noise power in these victim channels. I'll do that after the custom-made bandpass filters I've ordered are installed. With those filters, I will also be able to plot the spectrum of IM3 due to undesired strong signals on Channels n-4 and n-2; n-2 and n-1 and for n+1 and n+2; n+2 and n+4; and n+3 and n+6. It should be noted that channel n will suffer such interference from every channel pair listed above. Those IM3 add powerwise. Two such channel pairs exist in many major markets.
All these channel pairs are UHF taboo channels relative to channel n. If you read the FCC Office of Engineering Technology Bulletin No. 69, you will not find D/U protection ratios for DTV-DTV interference from taboo channels. This is because the ATTC measured a sampling of UHF taboo channels and found that for DTV-DTV interference, the D/U ratios exceeded 60 dB at weak D-signal levels. Those measurements involved only one undesired signal. UHF taboos n+/-2,3,4 are all ascribed to intermodulation, and it is now a little hard to see why we didn't catch on to the need to test with pairs of undesired signals on pair of taboo channels. Was this topic ever discussed in the Spectrum Utilization Sub-committee of the Advisory Committee? I don't recall such discussions.
| Fig. 2 Spectrum analyzer plot of IM3 generated in an overload amplifier by pairs of undesired DTV signals on Channels 31 and 34.|
There is another nonlinear distortion due to strong undesired signals near the desired channel. This is called cross-modulation (X-M). This does not require two undesired signals, but the undesired signal must be much stronger to cause cross-modulation than would be required of each signal to cause intermodulation. We are all familiar with X-M, even if we didn't know its formal name. X-M from one NTSC signal into another produces a white "Sync Bar," which drifts slowly across the screen. This is the horizontal-sync pulses of the undesired signal downward, modulating the carrier of the desired signal. A single NTSC signal on a UHF taboo channel could possibly cause DTV reception failure due to X-M.
However, our DTV system is quite robust against this sort of problem, as X-M from an NTSC signal will introduce bursty errors during both horizontal- and vertical-sync pulses. Our 8-VSB signal is very rugged against such bursty noise!
Receiver designers, if aware of these problems, may find ways to improve the RF selectivity of DTV receiver tuners. IM, and for that matter, X-M actually take place in the mixer stage of tuners. If there is enough RF selectivity between the antenna port and the mixer, those undesired signals will not be able to overload the mixer and cause reception loss. Receiver designers could also reduce this problem by a wideband RF automatic gain control circuit such as this column has reiterated several times now. If the RF amplifier gain is reduced in the presence of strong undesired signals, IM and X-M would be mitigated. Ideally, and it might come to that, a combination of wideband RF AGC and improved RF selectivity can overcome this interference problem, but are receiver manufacturers listening?
To some manufacturers, if it is not a problem for cable-connected receivers, it just is not a significant problem and won't receive their attention. And this is why the FCC must mandate minimum performance standards for DTV receivers. Just remember, intermodulation is nothing but receiver overload and need not happen, not with good receivers. Will manufacturers build receivers with enough linear dynamic range? It is your future that hangs in the balance.
Next month, I'll report on further experiments with dual UHF taboo signals.