Subject: Re: usb 1.1 vs 2.0
From: Christopher E. Brown (cbrown@woods.net)
Date: Tue Jan 07 2003 - 15:53:20 AKST
On 7 Jan 2003, Jim Gribbin wrote:
> My scans were of a portion of a color slide. .91in X .91in X 3200 dpi.
>
> My main issue is that I expected the connection speed to effect the
> scanning speed. The scans I attempted were the same slide at the same
> resolution on the same scanner with the same software on the same box.
> The only change being the usb interface used.
>
> I didn't expect a straight 40x change in scanning speed (the amount of
> theoretical bandwidth change), but I did expect a significant increase.
> My stopwatch says there is no difference. The 90 to 95 sec times that
> you found for a negative implies 2x what I'm getting.
>
> I'll put W98se on a spare drive in the next day or two and see if there
> is a difference with SIIG and Epson's windoze drivers. It would be
> interesting to try a mobo with usb 2.0 built in as well.
>
> I have run across information in researching this that implies there are
> very few devices actually capable of more than 20MB/sec in the real
> world and that the chipset used by SIIG (NEC) has a bottleneck at about
> 28MB/sec, 160Mb/sec and 224Mb/sec respectively. The NEC chipset seems to
> be the main one the the linux drivers are centered on.
>
> 480Mb/sec - Yeah-Right.
>
> Jim Gribbin
I don't see that you going to get much faster than you are getting
now with a 3200dpi scan.
My FS4000US in 4000dpi outputs just under 6000x4000 (uncropped) from a
35mm negative. Scan time for a *single* pass is about 1.5 min, to do
a color pass + infrared pass (for scratch cleanup) outputting 64bit
RGBI is just under 3 minutes, and a normal multipass+infrared for best
quality takes about 5 min per frame. This is on a 20MB/sec SCSI bus.
The timing is not from the transfer, but the scanner mech.
Film scanners are much slower than flatbeds, as a simple matter of
calibration and accuracy. The CCD head positioning accuracy must be
kept to within +/- 1 10,000th/in at worst for a.
How many color bits does the scanner output, and what are you feeding
the computer? How many passes? Are you using any form of infrared
scratch/dust cleanup? (I know the Nikon, Polaroid and Canon scanners
do it, don't know about the Epsons)
What is the raw image size?
SCSI and Firewire are the better interfaces, USB (even USB 2.0) has
efficiency issues similar to those of IDE that limit its
realworld performance.
-- I route, therefore you are.--------- To unsubscribe, send email to <aklug-request@aklug.org> with 'unsubscribe' in the message body.
This archive was generated by hypermail 2a23 : Tue Jan 07 2003 - 15:51:20 AKST