On Monday, June 10, 2019 at 4:11:53 PM UTC-7, Eric Jacobsen wrote:
On Mon, 10 Jun 2019 12:51:17 -0700 (PDT), RichD
<[email protected]> wrote:
Recently I attended a coding seminar. The speaker
very briefly reviewed the relevant history;
first Hamming codes, ~1960: BCH codes, ~1970: convolutional,
1995: polar, 2002: LDPC
Clue me in - what was the advance in each case?
I worked on a BCH project once, that seemed fairly
efficient, how is it, or other algorithms, deficient?
(snip)
You'll probably find different histories out there from different
sources, as the lineage of some of these aren't all that clear. It's
like asking when the FFT algorithm was invented...you may get vastly different answers.
It seems to me that part of it is the ability to process such signals
at an appropriate speed.
My favorite in the history of transforms and coding is ICT,
the Integer Cosine Transform. It is similar to DCT, but the
coefficients are optimized for speeding up the forward transform,
at the expense of slowing down the inverse transform:
https://ntrs.nasa.gov/search.jsp?R=19940025116
Specifically, the forward transform is done on the CDP1802,
a microprocessor from the 8080 and 6502 days that doesn't have
hardware multiply.
It seems to me that the math for the more complicated coding
systems might have been around for years, but no ability to
use them until computing hardware caught up.
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)