Signal Distortion And Interference

Distortion comes in two basic classes: Linear distortion, which is not an oxymoron; and nonlinear distortion, which includes all other distortions except noise. Noise is quite another matter and is not covered here.

Linear distortion is not always bad. There are two kinds of linear distortions that are frequency-dependent changes in amplitude and delay. They are linear in that gain and delay can be measured at any signal level with the same result.

Nonlinear distortions come in many orders-second, third, fourth and so on. None may be detected when the signal is small, but they may harm or prevent reception when the desired signal is too strong, or accompanied by stronger undesired signals on other frequencies.

Put simply, nonlinear distortions result from overloading of active devices such as transistors. This is not quite accurate, as that model assumes there is a linear operating range and that only above it is the transfer characteristic nonlinear.

This is a pragmatic approach, so let's use it, but in reality the transfer characteristic of active devices is curved slightly in their so-called "linear range," and more curved in the "nonlinear range."

AUDIO DISTORTION

Carried to the limit, active devices saturate, and no more power output results from further increases in input signal level. Nonlinear distortion of audio signals means harmonic and intermodulation distortion. Nonlinearity in active devices generates harmonics of the input signal frequencies and these usually appear in the output signal.

The ear is a poor detector of harmonics. Musical instruments generate harmonics of their fundamental tone and it is these harmonics that provide the instruments' unique timbre. Nothing sounds more boring than a sine wave!

With only one input frequency present (a pure sine wave), the only evidence of nonlinearity is the appearance of harmonic frequencies of this sine wave in the output. With two or more sine waves simultaneously present, we may get intermodulation products. These are new frequency components that are not harmonically related to the input frequencies. Perhaps this is why intermodulation distortion is so offensive when it is present in music.

Before we leave the subject of nonlinear distortion of audio, we should note that when two (or more) sine waves of equal power are present, at times their peak voltages coincide. This means we have voltage addition-not power addition-and this means we have transient peaks that can overload the amplifier and generate audible distortion.

Because intermodulation distortion produces such havoc in audio, simply measuring the harmonic distortion is not sufficient; we should also measure intermodulation distortion. I believe there is no need to measure harmonic distortion, except in transmitters. Intermodulation is a much more important consideration and it determines whether we will have interference or not, which is the real subject of this column.

VIDEO SIGNALDISTORTION

Intermodulation distortion was not a problem for monochrome TV transmission. It became a problem when chrominance signals were added to the luminance signal within the baseband video signal to transmit pictures in "living color."

The NTSC system adds color information to a monochrome television signal in a way that causes certain intermodulation distortions that the IEEE named "differential gain" and "differential phase." Since these nonlinear distortions could wreak havoc with the reproduction of colors, special measuring instruments were devised.

At Tektronix, I designed vectorscopes to measure the differential gain and differential phase of NTSC signals. When I learned that Dr. Walter Bruch, inventor of the PAL variant of NTSC, found it useful to have a PAL vectorscope, we designed our first solid-state vectorscope to measure both PAL and NTSC signals. This was easy because PAL and NTSC are similar.

What was going on in video amplifiers, transmitters and radio relays and VTRs was that the gain of the amplifier was varying with the amplitude of the luminance signal. That is, the chroma signal centered at the subcarrier frequency was being amplitude and phase modulated by the luminance signal, which is larger than the chroma signal most of the time.

Today, the state of the art has advanced to the point that transmission facilities for NTSC and PAL are so linear that these nonlinear distortions no longer cause the color errors we once had. (Early color broadcasts were plagued with greenish and purple people, for example.) NTSC no longer stands for "Never Twice the Same Color."

However, while we have these problems of analog TV transmission under control, we now have interference between DTV signals on different frequencies, which results from nonlinear distortion.

In my last few columns, I've written about interference to DTV reception caused by intermodulation distortion.

First, I showed that DTV signals on certain pairs of channels could produce co-channel interference (noise) that could decrease the desired DTV signal-to-noise power ratio to its threshold of 15.2 dB with DTV reception failing. I also reported that in many major markets, we have channel allocations in the UHF band that may result in interference that could cause reception to fail.

Let's look at the TV allotments in the Greater Miami area given in Table 1.


(click thumbnail)Table 1. TV allotments in the greater Miami area.I also wrote that the sideband splatter from a DTV transmitter into each adjacent channel is at least 46.5 dB below the power radiated in its assigned channel. This radiated sideband splatter results in co-channel noise in addition to receiver noise.

OLD ATTC TEST DATA REVISITED

Recently, I revisited the ATTC Test Results (1995) for adjacent channel interference between DTV signals on adjacent channels. At D = -68 dBm, the ATTC reported the D/U ratio at threshold was -43 dB. This is the D/U ratio you see in Bulletin OET No. 69. At D -53 dBm, the reported D/U ratio was -39 dB. This confirmed that the D/U ratio is some function of the desired (D) level. While at D -28 dBm, the ATTC found the D/U threshold level was less than -18 dB, quite a change from -43 dB.

If the interference from a DTV signal on an adjacent channel was due to the finite IF selectivity of the one DTV receiver available in 1995, then the D/U ratio would have been the same over the range of desired levels tested.

It was not.

So, what was the interference mechanism of DTV/DTV interference between adjacent channels? At D = -68 dBm, the threshold undesired (U) level turned out to be the noise floor of the ATTC RF test bed, approximately 58 dB below the undesired level.

Our RF test bed was linear. Harris Broadcast built it to the specifications approved by the ATSC, and those required that it be linear. It did not produce any detectable sideband splatter in either adjacent channel.

At D = -53 dB, the D/U ratio had changed, so there had to be an additional source of noise present. Gary Sgrignoli first identified it as third-order intermodulation (IM3) in his March 2003 IEEE paper. I did the numbers and calculated that the IM3 being generated when the undesired signal was -25 dBm was about -70 dBm.

THIRD-ORDER INTERCEPT POWER

Is this good or bad? Well, I think that Zenith's prototype 8-VSB receiver was state-of-the-art then and now.

How good was it? To find out, I calculated the third-order intercept power of that tuner, which is the parameter usually published for broadband RF amplifiers, and it applies to the two-tone test signal. As the power of the input test signal increased, the signal power at the output port increased 1 dB per dB increase of input power until compression was noticeable. That power rating is logically called the 1 dB compression power rating of the device.

At this 1 dB compression power, there was considerable third-order IM present in the output and this increased at 3 dB per dB in the two-tone test signal. At some higher signal level, the signal and the IM3 would be equal. This is the third-order intercept power (IP3) shown in Fig. 1.

(click thumbnail)
(click thumbnail)
Note that no one would operate an amplifier at or even near its IP3 power rating, as it serves as a benchmark for design engineers only. I believe that 8-VSB signals should not be allowed to reach the 1 dB compression power.

A DTV signal can be thought of as a multitone test signal. It has many sidebands all the same power. When it suffers nonlinear distortion, these sidebands generate third-order IM3.

When a DTV signal passes through an overloaded amplifier (or tuner), it generates third-order intermodulation products (IM3) that fall into the adjacent channels. This is exactly what happens in DTV transmitters.

So let's figure out the third-order intercept power rating of the tuner we tested. Such parameters are never specified for consumer products as they are considered to be trade secrets, but they can be measured.

In Fig. 1, I've shown the undesired power to be -14 dB and calculated the IM3 to be -70 dBm. The difference between these is 56 dB. Taking one-half this difference and adding it to the undesired level increases the undesired level by 28 dB to +14 dBm. This is the device's third-order intercept power (IP3) level.

Why did I do that? The IM3 was -70 dBm. This is 84 dB below the IP3 level just determined. IM3 increases 3 dB per dB increase in the signal level, so dividing the 84 dB change in IM3 by 3, we get 28 dB-exactly the change in the undesired level from the test value (-14 dBm) to the calculated third-order intercept power (+14 dBm). This confirms the prediction of IM3.

You may wonder why I ignored the desired power in calculating IP3. The undesired signal was so much stronger that the desired signal can be ignored.

IS +14 dBm GOOD OR BAD?

For that, I calculated the minimum desired DTV received power required by the IM3 that will be generated for a range of IP3. This relationship is shown in Fig. 2.

But what is the range of undesired DTV received signal power we will experience when all DTV stations are operating at their maximum authorized power? More than 500 DTV stations now operate below their authorized power and significant power in-creases must be in their future.

For a full-power UHF station, the peak visual power is 5 megawatts ERP or 37 dB above 1 kW (37 dBK). The received power can be as high as -15 dBm close-in, using the FCC planning factors for receiving antenna gain and downlead loss.

As the undesired signal is generating interference by a nonlinear process (intermodulation), we should be concerned with its peak power, which is at least 5 dB above its average power. For a maximum-power DTV facility in the UHF band, its average power is 30 dBK or 35 dBK peak.

We see that the peak powers of DTV and NTSC are comparable, so I believe we will encounter undesired signals of -15 dBm in the field.

Looking at Fig. 2, we see that the minimum desired signal that can be received with U = -15 dBm is -70dBm for a receiver with an IM3 of +20 dBm. In my opinion, this would be a very good receiver, and I really doubt that one this good exists. If I'm wrong, please tell me who makes such a product.

Now let's look at the minimum usable received DTV signal power for lower IP3 receivers. An IP3 +15 dBm receiver will require a desired signal received power greater than -60 dBm. We get a 10 dB improvement in minimum usable signal power for a 5 dB increase in IP3.

As Fig. 2 shows, this 2:1 leverage exists for all values-it is not just a special case situation seldom encountered, it is a fact of life.

I am not an experienced designer of TV tuners, but I find it difficult to believe that tuner manufacturers in the fiercely competitive world of consumer electronics would provide more performance than necessary unless such performance is mandated. I'm not knocking the manufacturers-they live in a highly competitive world where nearly all customers are fed by a CATV system or get their signals from a satellite.

In both cases, all signals fed to the receiver are at about the same power and that power is well above the receiver's own noise level, yet well below its overload power. Only off-the-air reception poses the challenge of wide dynamic range to the tuner designers.

WIDER DYNAMIC RANGE?

Is it technically feasible for a tuner to have an IP3 of, say +20 dBm? I believe it is. There are three or more ways to get there:

1) Use smart RF automatic gain control circuits that reduce RF amplifier gain when strong signals on adjacent channels are detected;

2) Increase the DC current in the RF amplifier transistor; or

3) Use a somewhat better transistor in the RF amplifier stage, a better mixer or both.

For reliability reasons, designers operate transistors at the lowest possible power level. Just increasing the operating current can improve the IP3 of the RF amplifier, but there are strict limits to be respected.

As an example, Stanford Microdevices SGA-6486 offers an excellent transistor with an IP3 of +38 dBm. If the maximum undesired signal is taken at 0 dBm, the IM3 would be 76 dB below the 0 dBm undesired maximum power. That puts IM3 down at the noise floor where it belongs.

In the above example, I picked 0 dBm as a possible worst-case situation because, although one undesired signal may be at -15 dBm, there will usually be a number of powerful undesired signals present. Remember that the peak envelope power (PEP) for multiple signals is critical. Look again at the Miami data in Table 1.

You can draw up such a table for your market, or I'll do it for you for a fee (of course).

Remember, the peak envelope power (PEP) is what counts here, not the average power. PEP increases very rapidly with the number of strong signals.

Finally, Fig. 2 demonstrates the effect of the noise generated within the desired channel by one strong undesired signal on one adjacent channel. It does not take into account the co-channel noise received from adjacent-channel undesired signals due to radiated splatter. Therefore, some upward adjustment of the minimum desired signal is justified.

CONCLUSIONS

To close on a positive note, I bought four low-noise preamplifiers. Their IP3 is +40 dBm, while their noise figure is below 1 dB. Another example of what is currently available has an IP3 = +47 dBm and a noise figure of 3.5 dB.

There is no technical reason why adequate interference rejection cannot be provided to consumers purchasing DTV receivers. It's all a question of getting tuner manufacturers to know what is needed-let's hope the FCC will tell them.

Although impressive improvements have recently been made in DTV receivers, tuners are still in need of minimum performance mandates.