About Timing Constraints

[Yamada1]

It would be helpful if you could teach us about the following points with the "Set Input Delay" command.

　1) In the case of the maximum delay setting, does this mean that the delay value is fitted so that the target delay is less than or equal to the set value?

Or does it mean that it is fitted to satisfy the setup time of the receive register given the set value?

　2) In the case of the minimum delay setting, does it mean that the delay value is fitted so that the target delay is greater than or equal to the set value?

Or does it mean that it is fitted to satisfy the hold time of the receive register based on the set value?

I've looked at the materials, but I can't quite understand it, so it would be helpful if you could teach me.

[RichardTanSY_Intel]

As you may already know, the" set_input_delay" is a timing constraint, used to describe a circuit outside of the FPGA. (same as set_output_delay )

The target delay should be equal to the set value that you set, for both min and max setting.

The maximum input delay (-max) is used for clock setup checks or recovery checks and the minimum input delay (-min) is used for clock hold checks or removal checks.

As the -max value increases, the setup requirement gets tighter because the FPGA's internal delays must get smaller in order to meet the setup relationship between clocks.

Likewise, as the -min value decreases, the hold requirement gets tighter, because the FPGA must add more delay in order to meet the hold relationship between clocks.

Reference: https://www.intel.com/content/www/us/en/docs/programmable/683432/22-4/tcl_pkg_sdc_ver_1-5_cmd_set_input_delay.html

Here's a case scenario for better understanding:

Let's say I have a signal coming into the FPGA on port din, which goes through some combinatorial logic and out through dout. To constrain it, I might do something like:

create_clock -period 20.0 -name ext_clk

set_input_delay -clock ext_clk -max 4.0 [get_ports din]

set_output_delay -clock ext_clk -max 7.0 [get_ports dout]

(Note that I did not do -min delays. I am going to ignore hold time analysis for this example, but normally a design should have this too.)

Anyway, the set_input_delay and set_output_delay describe registers outside of the FPGA and states they are clocked by ext_clk.

As such, there is a default setup relationship of 20ns when this clock is the source and destination. This is the lowest priority.

Since 11ns of delay (4+7) are used externally, the FPGA must get its signal from din to dout in 9ns.

A user could then add a multicycles if that is too tight of a requirement:

set_multicycle_path -setup 2 -from [get_ports din] -to [get_ports dout]

This multicycles has priority over the default clock relationship, and makes the setup relationship two clock periods, or 40ns.

Since 11ns are used externally, the FPGA must get its data from din to dout in 29ns. 

Best Regards,

Richard Tan

p/s: If any answer from the community or Intel Support are helpful, please feel free to give best answer or rate 4/5 survey.

[Yamada1]

Thank you for answering.

Your detailed explanation was very helpful.

I understand that "set_input_delay/set_output_delay" is the setting of the conditions of the external circuit of the FPGA, and fitting and timing analysis are performed based on these conditions.

Does that mean that it is correct to use "set_max_delay/set_min_delay" when timing the internal path from the output of the FF to the input of the subsequent FF?

Also, what kind of method is there when there is a combinational circuit between FFs when restricting the timing from the output of FF to the input of the subsequent FF?

I'm sorry to trouble you, but it would be helpful if you could teach me.

that's all

[sstrell]

You can use set_max/min_delay for internal timing paths as you've suggested, but note that it overrides what the Fitter selects on its own, so you really only need to use them if you have specific timing requirements that the Fitter is not meeting, usually referred to as over-constraining.

set_max/min_delay works between registers, so any combinational logic in between is included in the delay value you specify.

[Yamada1]

Thank you for your detailed explanation.

It was very helpful.

I understand that setting "set_max/min_delay" is not normally necessary, and that combinatorial logic delays between paths are included in the set values.

The specified delay value specifies the amount of delay between paths in the same way as "set_input_delay", and does not control the fitting so that the delay between paths is less than the specified value (when "set_max_delay") Am I correct in understanding that it is not?

Sorry for the elementary question, but it would be helpful if you could teach me.

That's all

[sstrell · Accepted solution]

No, as mentioned, set_input/output_delay specifies external delays so the Fitter knows where to place input or output registers to meet timing on the inputs or timing at the downstream device on outputs.  set_max/min_delay is a way of specifically saying to the Fitter "I want this path to have a max delay of x and a min delay of y."  Both sets of commands affect how the Fitter places and routes things.

[Yamada1]

Thank you for your detailed explanation.

It was very helpful.

Thank you for your prompt response, even though it was an elementary question.

[RichardTanSY_Intel]

Thank you for acknowledge the solution provided. 

I’m glad that all your question has been addressed, I now transition this thread to community support. 

Thank you.

Best Regards,

Richard Tan