Tamas Rudnai wrote:
> Yes, you are right Wouter, always forget that with RS232 we do not
> usually use a protocol, just the plain wire :-) So yes, USB uses a huge
> overhead but Olin, you said you use bulk transfer, which is supposed to
> be the fastest as long as there is no other devices interrupting the
> channel. You use full speed, so it's 12Mbps (even low speed with
> 1.2MBps is 10 times faster than 115Kbps)

No, it's not.  There are considerable restrictions on low speed USB devices,
including the maximum packet size and the minimum guaranteed time between
these packets.  In the end you don't get anywhere near 115.2Kbaud with low
speed.

However, low speed USB was a bad idea in my opinion, and isn't relevant to
this discussion anyway.

> so full speed is 100x faster

Yes, **when actually sending data**.

> than that which means even with the large overhead with bulk transfer
> you can download largish blocks from the PC so that the overhead become
> relatively smaller I suppose

You should try reading the USB spec instead of supposing.  USB doesn't work
the way you think, and you are reaching incorrect conclusions from your
incorrect assumptions.  For one thing bulk transfers are limited to 64 bytes
maximum per "block".  But the real issue is not the speed that data will be
sent once a block is being transferred, but the various latencies to get a
block started.

> -- I just could not understand why is that
> slower and would like to know to learn how to make faster USB devices?

Start by reading the USB spec.  Basically to go fast on USB you want to make
sure you have at least 64 bytes to send at a time, and design the protocol
to avoid a lot of little questions asked by the host of the device.
Unfortunately this is exactly what my programmer protocol specified, since
this is no problem with RS-232.  I have meanwhile revised the programmer
protocol to allow for queing of requests and for requesting larger chunks of
data at a time, but readback from small PICs is still a little slower than
for RS-232.  Overall programming and verify times for dsPICs are faster when
using USB because the higher bandwidth of USB becomes a advantage.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000.
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist