[opendtv] CRC comparison report

  • From: "Manfredi, Albert E" <albert.e.manfredi@xxxxxxxxxx>
  • To: "OpenDTV (E-mail)" <opendtv@xxxxxxxxxxxxx>
  • Date: Mon, 3 May 2004 17:12:02 -0400

The CRC comparison report I was referring to was published
in the IEEE Transactions on Broadcasting, Vol. 49, No. 3,
dated September 2003. This test is devoted entirely to
performance in high multipath environments. But it was
based on results with a 2nd gen 8-VSB receiver, echo
tolerance of which was only -1 to +38 usec. So, without
the benefit of the much better echo tolerance specs
available with the upcoming Linx and Zenith receivers, and
possibly Broadcom as well (will wait for the complete
test).

It compared 8-VSB with COFDM in four modes, with and without
on-channel repeaters (OCRs). The four COFDM modes ranged
from 64-QAM, 3/4 FEC, 1/16 GI down to QPSK, 1/2 FEC, 1/16
GI (i.e. 19.76 to 4.39 Mb/s).

One interesting sidebar was that with a 0 dB echo, the
COFDM receiver used for the test required a 31 dB C/N,
whereas with an echo 1 dB lower, 22 dB C/N was needed.
Neither of these profiles is as difficult as Brazil E.

But the 8-VSB receiver required 31 dB of C/N for a -2 to
-3 dB echo, so it was not as good as the COFDM receiver.
But flash forward, the newest 8-VSB receivers from Linx
and Zenith can operate in Brazil E profile, with
reacquisition, with about 25 dB C/N, and echo tolerance is
now more like +/- 38 usec or even +/-50 usec. So these are
the improvements that 1999 results can't possibly account
for. These upcoming 8-VSB receivers are better than the
COFDM receivers of 1999 or of 2002 (at the 19+ Mb/s level),
at least in some multipath scenarios, and we just don't
know how they compare with the more current COFDM
receivers.

The executive summary is that without OCRs, the 19+ Mb/s
COFDM signal and the 8-VSB signal were roughly matched
in terms of reception success. Adding the OCRs improved
8-VSB reception about as much as, or more than, reducing
COFDM to 16-QAM. But due to the limited 8-VSB echo
tolerance, adding OCRs in a few cases killed 8-VSB
reception entirely. OCRs were also beneficial to COFDM
reception, and in fact there was never a case where OCRs
reduced COFDM performance (thanks to the much better echo
tolerance).

Below is a quote from the end of the article, which I
think tells the story pretty well.

Bert


---------------------------------------
For 8-VSB, with turning the OCR "ON", the reception sector
in 9 sites (53% of the total 17 sites) became larger than
those obtained by using DVB-T without OCR at the bit rate
of 11.71 Mbps (mode 3) [this would be the 16-QAM mode].

It should be pointed out that as shown by parts of these
tests, the reception sectors for 8-VSB and DVB-T mode 1
[19.76 Mb/s] were almost the same when the repeater was
"OFF".

V. FUTURE WORKS

A new prototype 8-VSB receiver was evaluated in the
laboratory at CRC [3] [the Linx test published October
2002]. This prototype performed much better in the
presence of multi-path than previous receivers. The
equalizer operated in the range between -29 to +40 us,
and in some parts of that range, could handle a zero dB
echo [that was in the range -17 to +17 us]. This
receiver should be evaluated in field tests similar to
the ones presented in this paper to determine if, as
one can expect, similar improvements to those obtained
with DVB-T mode 1 (19.8 Mbps) with OCR, could also be
obtained with 8-VSB.

VI. CONCLUSION

In this study, an on-channel repeater (OCR) was used to
cover a small part of the coverage area of the main
transmitter. The results showed that when the delay
between the signals coming from the 8-VSB main transmitter
and the repeater was very short and within the pre-ghost
capability of the receiver, a significant improvement of
reception sector was obtained in the majority of the sites
in the test area.

In order to have a very low delay situation, the internal
delay of the OCR must be very short, and there should also
be no or very short delay due to path difference between
the signals coming from the two sources to the target
area. These requirements put some constraints on the
configuration of the OCR, as well as on its location with
respect to the main transmitter and the target area.
However, even under these conditions, a few locations
could still be found in which the 8-VSB reception failed
when the repeater was turned "ON". The cause of failure
was believed to be the reflected OCR signal with a delay
outside the pre-ghost range of the receiver.

With an 8-VSB receiver more robust against pre-ghost, the
constraints on the configuration and location of the
repeater could be relaxed to some extent. In this case, it
is expected that even if the OCR signal delay at the
reception points is longer (but still within the delay
range of the equalizer), the OCR can provide reception
improvements similar to those obtained in phase II of this
study, and comparable to those that can be achieved by
more robust modes of transmission.

[Note: this was in fact tested, also in Ottawa, with the
predictable improvements in performance, in the Linx OCR
test of April 8, 2003, results available at the Linx
site.]

It is recommended that broadcasters consider OCR as an
alternative for reception improvement in their technical
planning.
 
 
----------------------------------------------------------------------
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.

Other related posts:

  • » [opendtv] CRC comparison report