Lpc2000

I have an Olimex LPC2106 header board.  I was not having any luck 
getting this to communicate with an LPT port debugger, so I did a 
few tests.  

According to the LPC2106 datasheet (Section 18 - EmbeddedICE logic), 
both the DBGSEL and RTCK pins must be held high on reset to enter 
debugging mode.  However, the document with the header board from 
Olimex says that DBGSEL should be held low during power up to enable 
debug mode, and it doesn't mention RTCK at all.  I tried all 4 
combinations of logic on these two pins, using both reset and 
removing and re-applying power.  Regardless, I always found that TDO 
is floating.  If I'm understanding correctly, it should be either 
high or low at all times in debug mode since it is the JTAG data 
output.  

To eliminate possible parallel port issues, I used another 
microcontroller to attempt to communicate with the JTAG port for 
testing purposes.  As expected, I found that TDO remains floating 
throughout all attempts at communication.  Next I tried configuring 
the secondary JTAG pins via a short program.  When I do this, I find 
that TDO (using the secondary port) assumes a strong high or low 
state, which changes during communication.  Looking at a JTAG state 
diagram and doing a bit of bit twiddling, I found I was able to get 
into Shift-DR mode on the JTAG port.  I was able to read the device 
ID (I got 0x4F1F0F0F).  However, attempts to write the IR did not 
work.  Regardless what I wrote to IR I would get the device ID back 
when I read data in from DR.

I know my circuit is good, given that I can communicate manually via 
the secondary JTAG port (in a limited fashion, which may be due my 
incomplete understanding of the JTAG protocol).  Why won't the 
primary JTAG port work at all?

I remember once reading that some people had problems getting the 
primary JTAG port to work with a different LPC2* chip (I don't 
remember which chip but it was not the 2106).  The Olimex board has 
a 14.7456MHz crystal (which I can't change).  Could that be part of 
my problem?

Incidentally, I noted that a sample "LED flash" program continued 
running happily throughout all my tests (even where I was 
communicating actively with the secondary JTAG port).  Is this 
normal?  

Thanks for any help.

Frank