Path: utzoo!utgpu!attcan!uunet!husc6!mailrus!uwmcsd1!bbn!gatech!udel!garrett
From: garrett@udel.EDU (Joel Garrett)
Newsgroups: comp.dcom.modems
Subject: Re: Direct modem line from USA to France?
Summary: BYTE's tests were pretty skewed...
Message-ID: <3608@louie.udel.EDU>
Date: 4 Aug 88 17:32:59 GMT
References: <297@amanue.UUCP> <247@westmark.UUCP> <6435@bloom-beacon.MIT.EDU> <791@riddle.UUCP>
Reply-To: garrett@udel.EDU (Joel Garrett)
Organization: University of Delaware
Lines: 28

In article <791@riddle.UUCP> domo@riddle.UUCP (Dominic Dunlop) writes:
[Discussion of "extensive" BYTE modem benchmarks deleted]
>By comparison, the much-discussed Telebit
>Trailblazer Plus did 5568 and 5078 bps respectively in each test.

I think it was discussed here not long after that issue of BYTE came out
that they were not very "real world", but then again, few benchmarks are.
I think the biggest source of incorrect results for the TB+ in these
tests were a result of 1) not fixing the interface speed at 19.2K bps
2) sending "random" data and thus not letting the modem's data compression
stuff to work, and 3) Their definition of line noise for the test was evidently
wide frequency spectrum noise, which would obviously blow away the PEP
method of line compensation.  The other thing that made me suspicious of
the test was the different results obtained from the Ventel modem which was
essentially the same modem in a different box.  I know, same thing as always,
no benchmark is totally accurate, special accomodations must be made sometimes,
but most of the modems that they tested came with instructions to crank up
the interface speed, etc for best performance, but I think they held the
speed back because one or two of the units tested couldn't do it.  Why not
just make these exceptions a separate test and let the ones that can be in
a different group,  as they obviously are?

>-- 
>Dominic Dunlop
>domo@sphinx.co.uk  domo@riddle.uucp

Joel Garrett
garrett@udel.edu