[FADCA] 9.6KB Throughput vs 1.2KB

[Chuck Hast]

On Sun, 21 Nov 2004 18:42:23 -0500, bud thompson <budt at cfl.rr.com> wrote:
> The radio baud rate of 9.6kb is 8X as fast as 1.2kb...

I am going to sound like a politician here, but it depends on how your link is
set up, and how clean the path (including radios ) is 

> 
> We all know the throughput rate is not 8X as fast.

With large frame sizes, and a clean path indeed it will be 8X. The reason you
do no see the 8X is because of the voice radios which are far from nice clean
data boxes. If you try it on a wire you will see the difference and it will be
quite impressive. If you try it on a true data radio again you will
see that data
flows quite rapidly and you will find that it will be 8X+ because your data box
will have very short TXD times which most radios used in 1k2 service will not
have. Now if you do the 1k2 and 9k6 on the same data boxes you will indeed
see that the speed difference is 8X. This was all the more brought home to me
when I started working with the Tait radios, in fact I link to the Tampa switch
from the Linux FPAC test switch, I will pull say a heard list or a
traffic list off of
the switch, if I am using the kenwood in the car I can hear both the 440 link
and of course the bearer channel off of the fpac switch. And the TX time delta
is just about 1/8 of that of the 1k2 channel. It is QUITE noticable.

> What I don't know is this:
> 
> What is a reasonable comparative throughput rate of 9.6KB compared to 1.2kb?

With reasonable radios and all of your timers set right you should get
from 60 - 70%
of the 8X, if you use data radios and use directional antennas to kill
the multipath
you will get your 8X. 


> Premise:  While we may not make a 'perfect' 12kb test, we are used to 1.2kb
> and will accept almost anything that works.  So, if we observe a 1.2kb 10k
> file transfer with few, if any, retries, we will consider that our
> baseline...
> 
> Now we move to 9.6kb... where we have much less experience...
> 
> All I'm concerned about is encoding/decoding in the absence of noise.  Let's
> assume the RF path/link has virtually no influence.. this is a bench test.
> 
> If the 10k file can be received in 3 minutes using 1.2kb...
> 
> What can we expect to do using 9.6kb?
> 
> Half the time (1.5 minutes?)
> 
> Give me a range - I'm not interested in technical details regarding AX.25
> parameters, etc... just your experience on what we can expect.
> 
> I see no reason to go to 9.6kb if it takes 80% of the time that 1.2KB
> takes.. hardly worth it.
> 
> Where is the break over?  - 50%...40%

On the bench you will get 8X I have done that with TEKK radios, too bad they
quite making the KS-900, it could be whooped up to have a TXD of < 10ms
and with the proper filters in it, it would pass a text book eye.

The killer is the use of radios that are not designed for data and are
nothing more
than a compromise between the two in most cases. Even the D-7 which I have does
a good job decoding 9k6 but the TXD is a bit high needed about 100ms if I recall
correctly.

I should have a tait radio at the hamfest on the switch so you can see
the speed with
which data will come in when the path is good and you have a clean TX and RX.


I know I got long on this, but a little TLC on these links and you
will get the 8X you
are looking for.