Design Tools - Others

[jwagnerhki]

Hi, a strange issue, 2 x VP5, 1 x XACE (not used) and 1 x V4 in the JTAG chain. Using Xilinx ISE 10.1 and the 2nd USB platform cable.

On one system in question, the chain can be initialized and devices programmed SUCECSSFULLY only when I use the IMPACT GUI. But not from batch.

GUI programming works but log output looks somewhat suspicious with log info voltages 0.00V and temperatures -273C.

Next, I can extract batch commands from the GUI log and place them into a text file. The result is:

setMode -bs
setCable -port auto
Identify
identifyMPM
assignFile -p 1 -file "C:/SysConfig/virt5.bit"
assignFile -p 2 -file "C:/SysConfig/virt5.bit"
assignFile -p 4 -file "C:/SysConfig/virt4.bit"
Program -p 1
Program -p 2
Program -p 4
quit

When I run batch mode "impact -batch cmdfile.cmd" on the above, the first device in the chain is programmed ok. Log voltage shows 3.28V and temperature +273C, ID bits are all '11111111111'. But programming seems to succeed. Programming the second device fails, however, after the first device has been programmed!

And after the first device I get an IMPACT:585 cable error, no ID's can be read back, the cable does not work any more, the FPGA boards must be power cycled.

The USB cable config seem to be identical (per log files) in both the GUI and the batch case. I also tried manual frequency settings of 3 MHz and 6 MHz, to no avail.

Even more strangely, in another system the above batch actually _does_ work fine, without resulting in the cable error. But on this one system, I must use the GUI to get successful programming :-(

Any ideas?

What does the IMPACT GUI cable setup do differently than the command line?

What commands does the IMPACT GUI *NOT* place in the the log file? That is, what commands might I be missing for a successful batch programming run?

Kind regards,

 - Jan

[jwagnerhki]

and it works if instead I use

setMode -bs
setMode -bs
setCable -port auto
Identify
identifyMPM
assignFile -p 1 -file "C:/SysConfig/virt5.bit"
assignFile -p 2 -file "C:/SysConfig/virt5.bit"
assignFile -p 4 -file "C:/SysConfig/virt4.bit"
Program -p 1
ReadIdcode -p 1
Program -p 2
Program -p 4
ReadIdcode -p 4
ReadIdcode -p 2
quit

With the earlier, after programming the 1st device, I get the IMPACT:585 error. After that the JTAG chain can't be accessed any more, I have to power-cycle the FPGA boards. With the new command sequence, programming of all devices works. Why!? :-/

[mcgett]

The symptoms indicate that the JTAG chain is getting broken at some point. 

Does the V-4 design include a PowerPC and is the debug port connected to the FPGA?

Is it possible that the SystemACE device that is part of the chain active at the same time that you are programming through the cable?  This could happen if the control pins are connected to one of the FPGAs in the chain.

[gregmeredith]

Also, to add to this the GUI may be doing some chain management under the hood.  WIth multiple devices in the chain and the DONE pins tied together the startup sequence gets a little tricky via JTAG.  All three .bit files have to be downloaded, and then the start commands have to be issued to each device. 

Here is how this works, when the first device's .bit file is loaded DONE won't be able to go high becuase of the other two.  Then the second devices .bit file is loaded and still DONE will be held low.  Then the third device is loaded, DONE goes high for the third device but the first two are in BYPASS mode and dont' have their startup sequence being clocked.  So the tools will have to issue another START command to each device.  The IDCODE may be doing the same thing by placeing them in ISP mode. 

Try the following which does the chain management for you:

program -p 1 2 4

[jwagnerhki]

Thanks for the explanation! Tried "program -p 1 2 4" (also tried with the option HighZ:True) but the result is that iMPACT thinks it finds too many Unknown devices on the chain and gives up after finding 30 of them.

[jwagnerhki]

Thanks! Yes the V4 is using one embedded PPC, which was connected to a JTAG controller IP. Following your lead, I removed the PPC jtag from the design, but unfortunately it didn't help with the programming issue. Remains the SystemACE... todo..