RE: PC based CNC system architecture

.... further discussion about "black box" and the EPP interface

>
> Date: Sat, 09 Dec 2000 02:37:58 -0500
> From: Matt Shaver <mshaver@...>
> Subject: Re: Re: PC based CNC system architecture
>

<major snip>

>
>At this point we spent quite some time researching how to connect the
>controller to the PC. I read most of the USB spec and concluded that

it

>wasn't feasible since USB devices can only start data transfers at

1ms

>intervals. The data rate is high, but sending a command and receiving

a

>reply would take 2ms minimum due to the way the bus works. I'm still

not

>sure whether this is built in to the USB hardware. If anyone knows

the

>USB interface at the hardware level, I'd like to hear your thoughts

on

>this. Another problem with USB is that someone will inevitably plug
>something else into the bus and goof everything up.
>

<more stuff snipped>


About a year ago when Matt was starting down this path, I had a
discussion with him about possible interfaces including USB. At
the time I shared his misgivings about USB, but I am starting to
come around. I have participated in two USB designs and now have
a lot more faith in it.

The 1 millisecond interval is definitely real, but as I understand it
most current EMC implementations use 1 millisecond updates. While this
is a lower limit to response time on USB, the latency can be made very
stable. We have succeeded in transferring digitized voice over USB.
In the case of an intelligent device such as Ron Ginger is proposing,
the update speed is much less critical.

The goof it up situation is getting better, but will probably always
exist. On the other hand, who in their right mind would be fiddling
with the USB bus on a machine controller. Set it up, test it, and
leave it alone.

Curiosity aroused.... I will investigate further.

Regards,
Steve Stallings