[Coco] Re: atari USB device

[jdaggett at gate.net]

Now that I can agree on.

james

On 22 Nov 2004 at 12:47, Dave Gacke wrote:

From:           	"Dave Gacke" <dgacke at ektarion.com>
To:             	"'CoCoList for Color Computer Enthusiasts'" 
<coco at maltedmedia.com>
Subject:        	RE: [Coco] Re: atari USB device
Date sent:      	Mon, 22 Nov 2004 12:47:28 -0600
Send reply to:  	CoCoList for Color Computer Enthusiasts 
<coco at maltedmedia.com>
	<mailto:coco-
request at maltedmedia.com?subject=unsubscribe>
	<mailto:coco-
request at maltedmedia.com?subject=subscribe>


> I guess what I'm getting at is this.

> 

> At 1.5mbps (Low Speed USB) your poll intervals come at 10ms fixed.

> 

> The packet sizes are 8 or up to 64 bytes, depending on which type of

> device it is.

> 

> So it's either 800 bytes per second (8*100 polls/sec) or 6400 bytes

> per second. 8 byte report size is for interrupt device, 64 byte for

> isochronous or bulk.

> 

> Both packet types will fit inside your allocated frame time.

> 

> 

> 

> 

> -----Original Message-----

> From: coco-bounces at maltedmedia.com

> [mailto:coco-bounces at maltedmedia.com] On Behalf Of jdaggett at gate.net

> Sent: Monday, November 22, 2004 10:20 AM To: CoCoList for Color

> Computer Enthusiasts Subject: RE: [Coco] Re: atari USB device

> 

> Dave 

> 

> Where I was definitely wrong on the Isochronous mode is that the limit

> is 1023 bytes per pipe, not packet. 

> 

> My understanding is at 1.5 Mbps, one bit time is 667 nS. So in 1 mS

> one can transmit 1500 bits or ~187 bytes theoretically. Add in the

> overhead for polling and the actual data bytes transmitted is markably

> reduced. It seems realistic that 8 bytes to 64 bytes of actual data

> would be a limiting amount in what can be sent in the 1 mS time slot. 

> 

> Considering that there initially be few devices on the bus,

> peripherals like the keyboard or mouse can be allocated more

> bandwidth. 

> 

> james

> 

> 

> On 22 Nov 2004 at 8:51, Dave Gacke wrote:

> 

> From:           	"Dave Gacke" <dgacke at ektarion.com>

> To:             	"'CoCoList for Color Computer Enthusiasts'" 

> <coco at maltedmedia.com>

> Subject:        	RE: [Coco] Re: atari USB device

> Date sent:      	Mon, 22 Nov 2004 08:51:29 -0600

> Send reply to:  	CoCoList for Color Computer Enthusiasts 

> <coco at maltedmedia.com>

>  <mailto:coco-

> request at maltedmedia.com?subject=unsubscribe>

>  <mailto:coco-

> request at maltedmedia.com?subject=subscribe>

> 

> > Hi James,

> > 

> > Unfortunately your math is a little off.

> > 

> > The part that everyone seems to miss about USB is the poll time.

> > Since it is a polled device bus, at low speed, your polls come at

> > 10ms intervals, and full speed they are at 1ms, and I forget what

> > they are at high speed.

> > 

> > Anyway, you are severely limited by the rate at which you are

> > polled. Per poll you can pass 1 packet, so its easy to figure out

> > the throughput from that.

> > 

> > 

> > Dave

> > 

> > -----Original Message-----

> > From: coco-bounces at maltedmedia.com

> > [mailto:coco-bounces at maltedmedia.com] On Behalf Of jdaggett at gate.net

> > Sent: Sunday, November 21, 2004 7:34 PM To:

> > msmcdoug at optushome.com.au; CoCoList for Color Computer Enthusiasts

> > Subject: Re: [Coco] Re: atari USB device

> > 

> > Mark

> > 

> > In handshake mode at high speed the data payload is either 8, 16,

> > 32, or 64 bytes. In Low speed mode the max data payload is 8 bytes.

> > Well capable for the Coco.

> > 

> > Isochronous mode, without hondshakes,  the data payload is limited

> > to 1023 bytes.  

> > 

> > This is per the copy of the USB 1.1 Specs that I have. 

> > 

> > I personally think that Isochronous mode should never need to be

> > used.

> > 

> > 

> > What Torsten was thinking of was that a sector size for a Coco disk

> > is 256 bytes. The SL811HS has the upper sixteen bytes of the buffer

> > reserved for registers. This would leave the largest packet data

> > size of 240 bytes. That ends up being 16 bytes shy of a disk sector.

> > 

> > 

> > The only way to read a complete sector would be to break the 

> > sector up into 4 64 byte packets. Each packet would take 42.7 

> > microseconds to transfer and about 512 machine cycles to do a 

> > load and store for the Coco. At  high speed that is about 286.2

> > micro seconds. Essential about 1.5 milliseconds to transfer data

> > from a disk into the Coco. 

> > 

> > Not sure if we gain that much speed advantage. But maybe some. The

> > Disk drive itself is really the slow element of the whole process.

> > 

> > 

> > james

> > 

> > 

> > On 22 Nov 2004 at 9:42, Mark McDougall wrote:

> > 

> > Date sent:      	Mon, 22 Nov 2004 09:42:27 +1100

> > From:           	Mark McDougall 

> > <msmcdoug at optushome.com.au>

> > Organization:   	Technetium Development Pty Ltd

> > To:             	CoCoList for Color Computer Enthusiasts 

> > <coco at maltedmedia.com>

> > Subject:        	Re: [Coco] Re: atari USB device

> > Send reply to:  	msmcdoug at optushome.com.au,

> >  CoCoList for Color Computer Enthusiasts 

> > <coco at maltedmedia.com>

> >  <mailto:coco-

> > request at maltedmedia.com?subject=unsubscribe>

> >  <mailto:coco-

> > request at maltedmedia.com?subject=subscribe>

> > 

> > > USB data packets (at least v1.x) are only 64 bytes in length. So

> > > I'm not sure what the 240 bytes refers to in your post. With the

> > > USB protocol handling the handshaking the CoCo is left to emptying

> > > the 64-byte packets as fast (or as slow) as it likes. I really

> > > don't see an issue here.

> > 

> > 

> > 

> > -- 

> > Coco mailing list

> > Coco at maltedmedia.com

> > http://five.pairlist.net/mailman/listinfo/coco

> > 

> > 

> > 

> > -- 

> > Coco mailing list

> > Coco at maltedmedia.com

> > http://five.pairlist.net/mailman/listinfo/coco

> 

> 

> 

> -- 

> Coco mailing list

> Coco at maltedmedia.com

> http://five.pairlist.net/mailman/listinfo/coco

> 

> 

> 

> -- 

> Coco mailing list

> Coco at maltedmedia.com

> http://five.pairlist.net/mailman/listinfo/coco