Re: DTV Reception Maps

N9WOS wrote:
And the second factor, primary to reflected signal ratio. With 8VSB it is an
issue. When you have hilly terrain and poor line of sight signal path, then
a good portion of the signal starts becoming comprised of reflected, and
scattered waves. In some places, the reflected/scattered waves outweigh the
primary signal strength by a large margin.
However if either the primary "or reflected" is consistent and
coherent (not phasing in and out) the reception will be fine. The
problem comes from reflected signals phasing in and out with the
primary. One only has to be a few db stronger than the other with
digital and it will give good reception. OTOH if they are changing in
strength it takes time for the receiver to figure out which one to
lock onto and by then the other may be strong enough to cause a
problem.


Um... The current top of the line adaptive equalizer can (technically) deal with a 0db (100%) reflection. It can even deal with leading reflections. (reflections that arrive earlier than the primary signal, if that is physically possible) It's all technical, because there is a technical difference between a 0db reflection in a lab and a 0db reflection in the real world. The compounding factors are the modulation system and the affect the adaptive EQ has on the noise floor. Dynamic reflection paths is one of them, but it isn't the primary confounding factor.

In the real world, adaptive EQ brings up more problems than it solves.

Lets keep something in mind. There is a duality between frequency and time. When you have a system that adjust the time dependent behavior of a receiver, you also adjust the frequency dependent behavior of that unit. Adjusting the time depending (impulse) response is the same thing as adjusting the frequency dependent response.

In an adaptive EQ, you send out a training pulse. Just like a person making a loud noise in a canyon so they can hear the echo. The receiver listens for echoes of the training pulse. It will store any echoes that it hears as a time dependent map. Lets say it hears a echo at 2mS that is 75% of the original signal strength and another one at 5mS that is 24% of original signal strength. It stores that as a map that it inverts and applies to an incoming signal.

When it sees a transition in an incoming signal, then it applies 75% of the inverse of that transition to the incoming signal at 2mS and 24% of the inverse of that transition to the incoming signal at 5ms.

More technicalities.
Technically, you can adapt to a 0db reflection. You can even adapt to >0db reflection. But there is a problem. You have to be able to lock onto the signal and synch with it to be able to determine what the training pulse is, before you can adapt to it. Your receiver can not lock onto the signal when it is has a 0db reflection. That is because it can't set the adaptive equalizer which needs to know when the training pulse happens, so it can train it's self. But it needs the adaptive EQ to cancel out the reflections so it can detect the signal and synch with it in the first place. So that is basically a chicken or egg problem.

If you have a receiver that is locked into a signal, and the signal has a reflection that grows to 0db, then it can handle it. But if it is a static reflection, then the receiver will never lock in.

Now, the frequency/time problem. If people don't understand what EQ ringing is, then they should take a crash course in it pronto.


When you adjust the time response to negative at 5mS, you basically increase the frequency response at 100khz 300khz 500khz 700khz 900khz.. on and on. (odd order) All the way through 5.900mhz. It also reduces frequency response at 200khz 400khz, 600khz.. on and on. (even order) When you have a 0db reflection at 5mS, then you basically set the inverse response at 100%. That sets the gain on the first set of frequencies to infinity. Any interference in the pass band on those frequencies will be amplified infinitely. It will keep adding onto it's self to infinitum as the receiver passes though the frames between the training pulses. The gain on the second set of frequencies is dropped to 50%

Basically, when you reach 0db reflection, the EQ sets it's self up as a feedback loop with zero db loop loss. That means that it will ring at any number of harmonics. The longer the delay, the higher the number of harmonics. A 1mS reflection will give 30 harmonic frequencies in a 6Mhz channel. A 10mS reflection will give 300harmonics in a 6Mhz channel.

In a noiseless environment, harmonic gain causes no problems because there is no noise for it to amplify. In the real world, when you push up harmonic gain, then you also amplify the noise.

If you have a -3db (50%)reflection that the adaptive EQ compensates for, then you basicly increase the noise response on a set of frequencies to 200%. So a 100 to 1 S/N ratio drops to 50 to 1 instantly.

When ever you compensate for reflections with an EQ, Noise always gains the upper hand.


That is why the earlier generations of receivers limited the adaptation to 10ms. When you get past that time limit, the number of harmonics grows to staggering proportions.

You can have an adaptive EQ that will compensate for reflections approaching 0db, but the noise gain grows to such a large level that any noise will completely kill it. That is because there is no loop loss, so noise adds to noise adds to noise... On and on..

Another problem is that noise will mess up an adaptive EQ during the training pulse. When the pulse happens, and there is noise on the channel (pulse or what ever) it will treat the noise as signal reflections. It will overlay a copy of that noise onto the signal every EQ loop. Lets say you have a cyclic noise pulse every 1ms. It will think that there is a reflection every 1mS and apply the inverse of the signal onto it's self ever 1ms. Which basically copies the noise it heard during the training pulse, and replays it over and over again onto the receiver front end many fold.

And repetitive noise causes a whole different problems with loop gain during the adaptive EQ training pulse. If you have repetitive noise, the EQ harmonics of each "apparent" reflection add. That means that you don't need a repetitive noise source that is even close to the signal level of the primary signal to push the EQ past 100 percent loop gain on certain frequencies. Once you go past 100% loop gain you basically have a self starting oscillator. A repetitive noise will cause the adaptive EQ to spontaneously oscillate at specific harmonic frequencies.

And the same thing for actual signal reflections. If you get a reflection at evenly spaced intervals, then they add harmonically in the EQ. The reflections have to be no where close to the primary signal strength, but they harmonically add up to a higher value. When that happens, the EQ spontaneously oscillates.

8VSB works great in an echoless environment. 8VSB with adaptive EQ works great in a noiseless environment. When you get noise and adaptive equalization in the same system, then you have problems.

It is just basically a over engineered regenerative receiver. When you turn the regeneration past 100% then it breaks into oscillation and you are greeted with a nasty squeal.

And even if it don't reach that point, the reflections will still add harmonically to produce numerous frequencies in the pass band that will amplify the noise to extreme levels, even with relatively small amplitude reflections.

It can handle one large reflection close to 100%. Or it can handle two smaller ones, each close to 50%. Or 5 even smaller ones, each close to 20%. Or 100 reflections, each one close to 1% of original signal strength But when ever you have a situation where the total sum of the reflected energy approaches 100% of the original signal, then you will have opportunities for the EQ to brake into harmonic oscillation.

The possible problems in the field that can achieve the necessary conditions to cause malfunctions are mathematically unlimited. So there is no way to redesign the system compensate for specific situation.

An even worse problem is the fact that the noise frequency harmonic response points that is boosted by the adaptive EQ will add harmonically... Yes, we get harmonic adding of harmonics. Noise at 100khz 300khz 500khz 700khz 900khz.. on and on. Will harmonically add together randomly and push the bit error rate to stupefying proportions totally at random.

Basically, when you get a reflection, or reflections that mathematically add to anywhere close to 100% of the original signal, then you boost in band noise amplification many orders of magnitude. That is because harmonics build on harmonics and it all goes to heck in a hand basket. It can easily make 50+db of extra S/N ratio disappear in a heart beat.

If we go back to your dynamic multipart environment, you will see how that just makes thing an order of magnitude more complicated. A magnitude more complicated than insanely complicated. You now have to deal with shifting harmonic response points, that randomly align in the pass band depending on various changing outside conditions. There is not enough computing power in the world to handle some situations that will arise from time to time between all the compounding factors.

You get the same problem when you use adaptive EQ for COFDM.

Time domain/frequency response equalizers are designed for a specific uses under supervised conditions. Canceling out reflections on a received signal in an uncontrolled environment is not one of them.

The only reason engineers are trying to incorrectly apply adaptive EQ it to 8VSB is because 8VSB was incorrectly applied to over the air broadcasting. Stupid begets stupid.


Ugly stupid worst modulation in the world.

And we have been running on stupid in this country for at least the last ten years when it comes to broadcasting.

I sat with more than ten of the top FCC engineers and other knowledgeable types there in 2003. I said to them that we on the outside could only believe that it had to be the POLICY of the FCC to make OTA for broadcasters NOT work well, not work mobile for instance so that the FCC could auction off channels above 51 for a very high price since winners of this auction would then know that they would have NO competition from broadcasters who would be stuck with 8-VSB, a non working, non competitive anchor while they could use state of the art modulations.

One of the FCC top brass in attendance, not a political type, blurted out "We have a policy?" to much laughter from all.

The meeting then went on in a morose fashion as we showed them video of COFDM being used successfully in New York City with a 100 W transmitter set at 400 ft.

What could they do? They were captive of the political process and all were waiting patiently, making sure to keep their jobs, till they could jump to a high paying job in industry.

Politics as practiced in the US today is and has been for some time now way to expensive. It gives us bad peanut butter, waste our spectrum etc, etc and etc.

Bob Miller
.

Loading