[PD] Alternate Controllers

[Hans-Christoph Steiner]

On Mar 18, 2008, at 8:56 PM, Charles Henry wrote:

> On Tue, Mar 18, 2008 at 7:34 PM, Derek Holzer <derek at umatic.nl> wrote:
>> What's interesting is that the HID protocol is orders of magnitude
>> faster than using USB-serial to access an Arduino. We figured that  
>> out
>> last week preparing for the gamepad hacking workshop at Pickled Feet
>> here in Berlin.
>
> This is something I've been wondering about... where does that latency
> come from on the usb to serial on the arduino.  I am using debian
> Linux with 2.6.17 kernal on a computer that accesses the arduino
> through /dev/ttyUSB0
>
> The arduino is doing serial communication with a couple of IC's.  Even
> with 24 bits of information sent to an IC using 3-wire interface, the
> process only takes about a millisecond.  The computer interfaces with
> /dev/ttyUSB0 and sends a command, then waits for information to
> return.
>
> I poll the device with select().  The round trip time is around 15-16
> ms.  Only 2-3 ms are actually accounted for in the communication
> routines, so where does the other 12 ms come from?
>
> (Actually, I have thought about different methods, such as triggering
> an interrupt from the actual serial port via another pin on the
> arduino, to indicate that data has been sent.  Would it help?)
>
> I think that the linux kernel doesn't check the file very frequently,
> which gives me this kind of latency, but I don't really know.  I
> wonder, how is the HID interface different?  It gets registered as a
> character device, also, right?
> Any tricks?  fundamental concepts I'm missing?

USB-serial drivers are usually optimized for maximum thruput rather  
than low latency and jitter.  Basically, they were designed to hook  
up modems and the like.  USB HID was designed around low latency and  
jitter.  Most USB HID devices deliver data every 10ms with decently  
low jitter.

The best I've been able to get with Arduino/Firmata is 20ms.  The  
FTDI driver/chip on has a huge and crappy buffer, so lowering the  
poll time would actually make it slower with massive jitter.  With  
the Arduino, you can tie a pin of the Atmel to the FTDI USB chip then  
you can flip that pin to force the FTDI to send it's data  
(circumventing the buffer).  Then you can get reliable 10ms latency,  
perhaps lower.  Check the arduino forums for docs on that.

.hc



>
> Chuck
>
> _______________________________________________
> PD-list at iem.at mailing list
> UNSUBSCRIBE and account-management -> http://lists.puredata.info/ 
> listinfo/pd-list



------------------------------------------------------------------------ 
----

Using ReBirth is like trying to play an 808 with a long stick.    - 
David Zicarelli