[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[HLL] Big bug !?!

Status: RO
Content-Length: 1886
Lines: 52



> The protocol proposed by Volker assumes that you get 14400 bps out of your
> V.32 modem link. This is not the case unless you also turn on V.42.
> Without V.42, the modem will send one start and one stop bit for each byte
> of data. Consequently, you need 10 bits per byte, leading to a loss of 20%
> of your bandwidth. The effective bandwidth of a 14.4 modem without V.42 is
> 14400 bps/(10 bits/8 bits payload transmitted) = 11200 bps. Which is not
> enough to fit the 13000 bps data stream generated by a GSM full rate
> vocoder.


I'm afraid you are right...

If the modem is "so stupid" to transmit 10bits/byte and waste 20%,
we loose enough bandwidth to be unable to transmit gsm-full-rate.
I thought this would not we the case, even without V42 (Hacko was sure,too).
But a short test this morning showed you are probably right !

To test "real" performance, i sended long patterns of "random" data between
the two 14.4. The effective end-to-end user-bit-rate (14.4 carrier) was:
* without V42 10640 Bits/Sec
* with V42=on 12280 Bits/Sec (more, even though uncompressible data!)
So there is no chance to get 13000 Bits/Sec over the line :-( this way.

There are different solutions, at least:
* Stay with 14.4-modems, try using them in sync-mode
* Use something higher than 14.4
* Use different coding algorithm

Thanks for bringing up this alarming point !

It seems we had a major theoretical error, witch would not
have shown up before first practical tests...


Version: 2.6.3i
Charset: latin1