Hi Peter,

On Tue, 10 Mar 2020 at 09:03, pgraf--- via Ql-Users <
ql-users@lists.q-v-d.com> wrote:

> > Anyway, it's working now... got about 30Kbit throughput between the Q68
> and
> > a BBQL with GC, which is about the expected rate given the fact that
> QLnet
> > is half-duplex. SERnet between Q68 and PC is about twice as fast at
> 115200
> > baud, but as I said it's easier to do bulk transfers by swapping SDHC
> > cards. At least my BBQLs now have easy access to real mass-storage :-).
> Glad it works for you. Only 30 Kbit sounds a bit low, was that with
> mass storage or ramdisk?

This was by copying a file (unzip) of 111158 bytes from ramdisk on Q68 to
ramdisk on BBQL+GC, which took about 29 seconds from Q68 to BBQL. The other
way round was a bit faster, 22 seconds so about 5000 cps. But copying from
Q68 to BBQL with Trump card took 41 seconds, so processor speed does play a
role. On my Issue 5 (JM) QLs I couldn't get the network to work.

I apparently made a mistake in converting cps to bps, as the bytes on QLnet
seem to be sent with a gap of 5 bits so I should multiply by 13 to get bps,
which yields around 65Kbps. The theoretical limit would be 1bit/11.2us =
90Kbps but the network protocol will have a significant overhead caused by
the small packet size (256 bytes) which means you lose a lot of throughput
since the sender has to wait for an ACK every 256 bytes. Even on SERnet
(which is full-duplex) this is an issue and also the reason why protocols
like XMODEM has a poor performance on high-speed modem lines. On the other
hand, streaming protocols like ZMODEM perform very well if there is proper
flow control on your serial link but fail miserably when there isn't...


*Jan Bredenbeek* | Hilversum, NL | j...@bredenbeek.net
QL-Users Mailing List

Reply via email to