Intel® Quartus® Prime Software
Intel® Quartus® Prime Design Software, Design Entry, Synthesis, Simulation, Verification, Timing Analysis, System Design (Platform Designer, formerly Qsys)
17237 Discussions

Stratix IV GXB Refclk IO standard settings

Altera_Forum
Honored Contributor II
1,855 Views

Hi there, 

 

I am trying to set the IO standard for a refclk input for a GXB bank in Stratix IV EP4SGX230KF40F1517 by using the following constraint: 

 

set_instance_assignment -name IO_STANDARD "DIFFERENTIAL LVPECL" -to clk_155mhz 

 

I got the following fitting error: 

 

Error: Differential I/O input pin clk_155mhz is assigned to a non differential location J38. However, it must be assigned to a differential input location 

 

This is compiled for the Stratix IV Development board and the LVPECL is suggested on the schematics. I also tried the other IO Standard and nothing works. The error message also doesn't make sense because the pin J38 and J39 are clearly differential pins. 

 

Could anyone shed some light on this? 

 

Thanks, 

Hua
0 Kudos
10 Replies
Altera_Forum
Honored Contributor II
1,063 Views

BTW, I am using QII 11.0

0 Kudos
Altera_Forum
Honored Contributor II
1,063 Views

The error message is misleading. The real issue is discussed here: 

 

http://www.altera.com/support/kdb/solutions/rd06252010_443.html 

 

Cheers, 

Dave
0 Kudos
Altera_Forum
Honored Contributor II
1,063 Views

Thank you Dave. This makes sense.

0 Kudos
Altera_Forum
Honored Contributor II
1,063 Views

Hi Dave, 

 

I would like to use the refclk input as the clock source for a pll in FPGA fabric and use one of the derived clock as a reference clock of a transceiver. Do you think that will be a problem? 

 

Stratix IV handbook seems to suggest that this is feasible. On vol 2, section I, chapter 2, p.g. 2-5, it seems to have both the clock path to fabric and the clock path back to the transceiver. 

 

In p.g. 2-8, ITB section, the last sentence from paragraph 1 confirms the clock path to fabric. 

 

In p.g. 2-9, the first paragraph confirms the clock path to transceiver.  

 

Am I reading this right? or is it that I have to do something to enable all these? 

 

Thanks, 

Hua
0 Kudos
Altera_Forum
Honored Contributor II
1,063 Views

 

--- Quote Start ---  

 

I would like to use the refclk input as the clock source for a pll in FPGA fabric and use one of the derived clock as a reference clock of a transceiver. Do you think that will be a problem? 

 

--- Quote End ---  

 

 

This works fine. In some cases you'll get a warning about the possibility of increased jitter (I think it was due to cascading of PLLs). 

 

Try it on the board and see how you go. 

 

Cheers, 

Dave
0 Kudos
Altera_Forum
Honored Contributor II
1,063 Views

 

--- Quote Start ---  

This works fine. In some cases you'll get a warning about the possibility of increased jitter (I think it was due to cascading of PLLs). 

 

Try it on the board and see how you go. 

 

Cheers, 

Dave 

--- Quote End ---  

 

 

That's what I did in the compiled project and got those fitting errors. Is there any other constraints I need to add to make it work? 

 

Thanks, 

Hua
0 Kudos
Altera_Forum
Honored Contributor II
1,063 Views

From the Example 2 on p.g. 2-14, I suspect that I have to add location constraints for specify which PLL to use...

0 Kudos
Altera_Forum
Honored Contributor II
1,063 Views

 

--- Quote Start ---  

From the Example 2 on p.g. 2-14, I suspect that I have to add location constraints for specify which PLL to use... 

--- Quote End ---  

 

 

I didn't have to provide a location constraint. Though I am also routing the reference clock to a transceiver block - perhaps that makes a difference. 

 

Cheers, 

Dave
0 Kudos
Altera_Forum
Honored Contributor II
1,063 Views

Hi Dave, Hi Hua, 

 

In my current design (Stx IV EP4SGX230 as well), I'm using a PLL derived clock as reference clock for some transceiver blocks and it works fine. 

Nevertheless, unlike what you would like to do, my PLL ref clock source for this derived clock is provided by a PLL dedicated clock input pin (Left PLL or Right PLL according to the GXB location side). This simply confirm the handbook p.g. 2-9, first paragraph (PLL output clock path to transceiver). 

 

Regarding the PLL reference clock source, the handbook doesn't say much about the dedicated transceiver refclock pin connectivity to the the global clock network. The p.g. 2-8, ITB section, last sentence from paragraph 1 is a bit ambiguous to me. What does exactly mean "the clock logic in the FPGA fabric" ? If the term "clock logic" designates the global clock network, it's fine and your GXB clocking scheme is possible (and I guess it is). 

 

Anyway, I don't know all your board constraints but I would simply recommend you to use a path as "straight" as possible from your ref clock input pin to your GXB RX CDRs or CMU PLLs ref clock input to avoid jitter amplification (refer to handbook p.g. 2-3, table2.2 "Input Reference Clock Source"). 

 

A last word about the location constraint: If your design uses transceiver channels on both sides of your FPGA, you have to duplicate the transceiver reference clock source on your board and to separately route to the reference clock input on each side. Indeed, there is no internal path through the device from one side to the other side. 

 

Regards 

Oliver
0 Kudos
Altera_Forum
Honored Contributor II
1,063 Views

Thank you Oliver. I know my clocking scheme is a little bit odd but we are using Altera dev board to prototype a design and the clock I want to use (with the right clock frequency) is unfortunately tied to a refclk to a GXB. I have to use this clock to generate clocks for FPGA fabric and a refclk for another GXB.... Once we got our own board build this situation will go away. 

 

Cheers, 

Hua
0 Kudos
Reply