USB SIE command timing?

[lpcware]

Content originally posted in LPCWare by gbm on Thu Dec 09 08:03:00 MST 2010
LPC17xx use some cryptic stuff called SIE for USB interface. From the manuals it is not obvious how this secret hardware works. Some important actions of the USB interface are implemented as SIE commands. After the command is written, the SIE executes it and sets the CCEMPTY bit when ready.
My question is: how long does it take for SIE to execute a command? Is there any guaranteed maximum execution time? Is it reasonable to wait in a polling loop for the command to be executed inside an interrupt service routine, like the LPCUSB firmware does?

Either the documents are silent on this or I am too dumb to find the information in the manual and datasheet. NXPers, please help.

[lpcware]

Content originally posted in LPCWare by gbm on Wed Dec 15 14:20:09 MST 2010
Thanks a lot! I'll try to port it to LPCxpresso with my USB engine.

[lpcware]

Content originally posted in LPCWare by drs on Tue Dec 14 16:26:00 MST 2010
Apologies, that one was a 7z renamed to zip. Try this one...

[lpcware]

Content originally posted in LPCWare by gbm on Tue Dec 14 14:32:38 MST 2010
That's exactly what I am looking for, but the .zip file you've attached is corrupted :(
Cold you please try to post it again?

[lpcware]

Content originally posted in LPCWare by drs on Tue Dec 14 13:49:00 MST 2010
I've attached a USB composite device that includes CDC/ACM and MSC. Is this what you were looking for?

[lpcware]

Content originally posted in LPCWare by gbm on Sun Dec 12 04:29:58 MST 2010
Thanks for the response. Somehow I find it strange that the representative of a vendor must take measures in order to determine the timing characteristics of a circuit designed in-house. Why isn't the definitive data available?
And why so many important (for a design engineer) details are missing from NXP documents?

800 ns is a long time - 80 instructions on an LPC17xx running at 100 MHz. That's what I was afraid of. Waiting this long in a polling loop is not a good idea but USB interrupt state machine with 100+ states looks even worse.

I will try to rewrite USB code so that it makes some use of the time between issuing a command and completing it by the SIE.

So far I got my VCOM code working with full handshake traffic, based on LPCUSB with major modifications - no NAK interrupts. May next goal is a composite device with CDC and MSC.

[lpcware]

Content originally posted in LPCWare by NXP_USA on Fri Dec 10 13:08:55 MST 2010
We did some measurements on the LPC1343 and it looks like the following SIE commands take approximately 800 nS:

CMD_VALID_BUF
CMD_SET_EP_STAT(x)
CMD_SEL_EP_CLRI(x)

Timing should be similar on the LPC1768.

-NXP