DAN> 7. At 28,800bps, (244+7) = 9ms and (128+7) = 4.8ms. (I assume that
DAN> the 244 or 128 values are for the data octets only, not the packet
DAN> extras.)  On a intra-city connection I get 2ms roundtrip delay
DAN> reported so I presume that a short (less than 9ms) burst of noise
DAN> occuring at the beginning of a packet would produce a NAK (or
DAN> whathever LAPM uses) at my end within 41ms (9ms+2ms+30ms
DAN> double-pass through the data pumps (source for 30ms figure:
DAN> 20-50ms datapump range mentioned in "What is Selective Reject" at
DAN> http://www.geocities.com/SiliconValley/Park/2432/modem.htm)).
DAVID> At 28,800, you'll be sending 3600 bytes/second so the time
DAVID> for the (244+7) reduces to 7ms vs 3.8 ms for (128+7).
DAVID> Similar to synchronous protocols, no start/stop bits are
DAVID> used so divide by 8 rather than 10.  I haven't bothered
DAVID> with the zero bit insertion here either (a 0 bit is
DAVID> inserted every 60 bits on the average). This would make
DAVID> the equation change to (packet size*8*61)/(DCE speed*60).
Ooops, what a *clown* I've been. I didn't multiply by 8 to go
from (244+7)bits to bytes/octets. Correct timings should be
69.7ms for a (244+7) frame at 28,800bps and 37.5ms for (128+7).
To get these new time values I used: 1/DCE_rate*8*(244+7|128+7).
Could you tell me more about the zero-bit insertions. I haven't
heard about this before. In your equation above shouldn't 61 be
the denominator (the one on the bottom) and 60 would be the
numinator (the one on top) since the presence of an occasional
extra non-data bit would reduce the throughput?
Assuming that there was an extra 1 in every 60 bits in the data
flow, the max cps figures at 28,800bps would be 3,442cps (244+7)
and 3,357cps (128+7). I used:  DCE_rate/8/(244+7)*(244)/61*60.
This would lead to max. theoretical efficiency rates of 119.5%
(244+7) and 116.6% (128+7). I used: cps*10/DCE_rate.
Do you know at what the additional overheads are for Zmodem16,
Zmodem32 and Ymodem-G. I seem to remember someone saying that
Ymodem-G's overhead was 0.5%.
I think the highest overall efficiency I've ever seen (at
14,400bps) was with a 4MB file using MNP4 and Ymodem-G where I
saw 1705cps, or was it 1710cps? (118.4%|118.75%)
DAN> This would be received during the transmission of the 5th
DAN> 244 frame or 8th 128 frame.
Using the new timing figures, assuming an intra-city roundtrip
delay of 2ms, 30ms for a double-pass through the data pumps,
near-instantaneous error-detection and transmission of the NAK
from the remote modem, the NAK would arrive at the tranmitter
either 100ms (244+7) or 70ms (128+7) after the start of
pre-transmission processing of the affected frame. I used:
frame_time + roundtrip_delay + data-pump_double-pass_time.
This arrival would occur during the transmission of the second
frame. So the only difference with/without SREJ would be the
retranmission of frame 1 vs frames 1 & 2. This would lead to a
with/without SREJ throughput degeneration of 7%/14% (244+7) or
3.75%/7.5% (128+7) for every whole-number increment in the
errors/sec rate. This analysis assumes: short round_trip delays
(i.e. not long distance); errors affecting only one frame at a
time (i.e. no overlap); quick NAK response after the data has
cleared the remote data-pump.
Thanks again for your assistance. Cheers, Dan Bridges, Brisbug PCUG.
___
 X SLMR 2.1a X 
--- Maximus/2 3.01
---------------