[email protected] wrote:
> As discussed on IRC, glitches can easily wreck the current clock
> recovery system. I'll try adding a deglitching circuit on the input
> signal path in the FPGA.

Here's more information on the errors: I finally succeded in
upgrading to the 8k SoC+BIOS, and was promptly greeted by an
endless stream of complaints about IN responses from my
full-speed keyboard.

In a run of 151 seconds, softusb reported a total of 3877
unwelcome patterns. That's about 25 per second or 2.5% of
all IN packets. The corruption rate was as follows:

Pattern Frequency
------- ---------
0xb6    2242
0xda    1580
0xba      55

It also reported a CC 5A once.

Here's what the patterns look like NRZI-encoded, compared
to IN (0x5a):

Byte    Bits (LSB->MSB) D+
------- --------------- ---------
IN 5a   0101 1010       _--___--_
b6      0110 1101       _---___--

Byte    Bits (LSB->MSB) D+
------- --------------- ---------
IN 5a   0101 1010       _--___--_
da      0101 1011       _--___---

Byte    Bits (LSB->MSB) D+
------- --------------- ---------
IN 5a   0101 1010       _--___--_
ba      0101 1101       _--____--

So it does indeed seem that the sample clock is running a bit
fast for some reason.

- Werner
_______________________________________________
http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org
IRC: #milkymist@Freenode

Reply via email to