[opendtv] Re: Seeing Ghosts on a Single Frequency Network

From: "Manfredi, Albert E" <albert.e.manfredi@xxxxxxxxxx>
To: "opendtv@xxxxxxxxxxxxx" <opendtv@xxxxxxxxxxxxx>
Date: Fri, 11 Feb 2011 16:36:52 -0600
John Shutt wrote:

> Charlie's tests (as described in the article referred to in
> the subject line,) was for a single 10 uS echo 0.5 dB down
> from the main signal.
>
> According to Charlie, a 10 uS echo equates to a 6 mile
> difference in signal path length between the two sources in
> an SFN.  So you have to be 6 miles closer to one transmitter
> than the other transmitter to see a 10 uS echo. If you are
> equidistant to both towers, the echo delay would be 0 uS.
> Since from that equidistant point you could go 6 miles
> closer to antenna A, or you could go 6 miles closer to
> antenna B, a 10 uS echo tolerance gives you 12 miles of
> overlap between transmitters in your SFN to play with.

All due respect to Charles Rhodes, when I read that, I got a really 
uncomfortable feeling that this was way too optimistic and that readers would 
misinterpret what it means about the viability of SFNs. But in the end, his 
essential message was, even with these optimistic models, it still won't work 
reliably. Of course, now the dangling bit left out is, is there a magic bullet?

Here's a simpler way to look at it, and way more real-world.

The speed of light is 3E8 meters/sec in free space, slightly slower through the 
atmosphere. Therefore, light requires 3.33 usec to travel 1 Km, or 5.36 usec to 
travel one mile.

Therefore, if two towers are 12 miles apart, *and* you time-align the 
transmission of symbols from the two towers (which would not be required in a 
MFN, btw), the signal from tower A reaches tower B 64.3 usec delayed, compared 
with the symbol just now being transmitted by Tower B.

If you DO NOT time-align the transmission of symbols from the towers, and Tower 
B is a passive DOCR, then the symbol from Tower B will reach back to Tower A 
with a delay of 128.6 usec, + any additonal delay in the DOCR electronics, 
compared to the time when Tower A originally transmitted its symbol.

What does this mean in practice? It means that in an urban area, where 
buildings exist and signals to the closest tower may very well be attenuated, 
if you are hoping for reliable reception *at all*, you have to have receivers 
that can accommodate those echo delays.

Same applies to more rural areas, by the way, if there are hills or other 
obstructions that may attenuate signals from the closest tower, while not 
attenuating as much the signal from a more distant tower. And the insistence 
that these SFN towers be small and low (e.g. low on the rooftop of downtown 
high rise buildings, or low towers similar to cellular towers in rural 
settings) only makes the problem that much more acute. The closest tower would 
then be much more likely to be attenuated.

COFDM receivers have low tech equalizers, so you have to rely on the GI to 
prevent inter-symbol interference. In COFDM 8K mode, 6 MHz channels, this means 
that towers 12 miles apart can be accommodated safely with a 1/16 GI if 
time-aligned, or 1/8 GI if not time aligned. Check out the Euro SFNs, and you 
will see that this is how they are set up. That's why Berlin uses 1/8 GI, and 
Rome uses 1/4 GI.

With 8-VSB, and towers 12 miles apart, you MUST time-align the towers, and even 
then, the Samsung Gemini is the only receiver that can hope for decent 
reception, assuming only these two towers exist and that the closest tower is 
not attenuated. (No 8-VSB receiver I've seen measured to date works well with 
60 usec of stong pre-echo, not even the Gemini.)

> With two big stick transmitters of equal power, spaced 50
> miles apart, the overlap area would be from 19 miles away
> from transmitter A to 31 miles away from transmitter A,
> which would correspond to the inverse distances from
> transmitter B.

Now, surely you know that is unrealistic, right? The overlap area will change 
according to obstructions, according also to weather conditions, according to 
tree limbs swaying in the breeze, and so forth. That's why the model Rhodes 
used was unrealistically optimistic.

> Oh, and Charlie conveniently left out multiple echoes and
> complex echoes in his ATSC testing.  The more complex the
> echo ensemble, the better a multi-carrier scheme with long
> guard interval performs compared to a single carrier scheme
> with a complex receiver equalizer.

As I expressed to Mark Aitken, I think, the article may have been interpreted 
by readers that there's a magic bullet, aka COFDM, that would solve the 
problems. There isn't. COFDM does not solve the problem, as proven conclusively 
by Qualcomm MediFLO, and also by the Ibiquity testing of the two-tower radio 
SFN in Baltimore.

> And if the ultimate goal of this entire exercise was not to
> provide seamless mobile service, but instead to simply provide
> uniform coverage to fixed receivers, then I agree with you
> that an SFN isn't necessary.

John, a MFN avoids all of the problems we discussed here, even for mobile 
service. And there's another aspect that hasn't been mentioned, and is long 
overdue. You CANNOT keep increasing the COFDM subcarriers, in the hope of 
creating longer GIs, if you want to optimize this for mobile service. Why? 
Because you cannot slow down the symbols indefinitely. A moving vehicle cannot 
see these slow symbols becoming warped while they are being received, and 
expect good reception. So in fact, your idea of 12 MHz is a good idea, because 
it would allow doubling the number of subcarriers WITHOUT slowing down the 
symbols, compared to 6 MHz, so you'd see longer duration GIs and decent mobile 
reception. But, of course, now you have to ask yourself, why not just go MFN.

Bert
 
 
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:

- Using the UNSUBSCRIBE command in your user configuration settings at 
FreeLists.org 

- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.
Follow-Ups:
- [opendtv] Re: Seeing Ghosts on a Single Frequency Network
  - From: Allen Le Roy Limberg
[opendtv] Re: Seeing Ghosts on a Single Frequency Network

Other related posts:
