General Category > Discussion - Hardware

About the input delay of FPGA when connecting FT601 with FPGA

(1/1)

StanleyChien:
Hi,

I am using the FT601 chip with Cyclone V GX FPGA.

I have difficulty in setting the input delay of FPGA's timing constraint.

In common definition of input delay, input delay = the time from clock rising to data stable (both clock and data  are from FT601).

In FT601's datasheet,  (https://www.ftdichip.com/Support/Documents/DataSheets/ICs/DS_FT600Q-FT601Q%20IC%20Datasheet.pdf)
In Figure 4.8 and Table 4.2, T1 and T2 give the AC timing of signals from FT601 to FPGA.

It seems that the data is stable between CLK's rising-T1 to CLK's rising+T2.

There's an sdc example in AN_421 FIFO Bus Master for FT60x (https://www.ftdichip.com/Support/Documents/AppNotes/AN_421_FIFO_Bus_Master_For%20FT60x.pdf)


--- Code: ---set_input_delay -clock [get_clocks fifoClk] -max 7.0  [get_ports {RXF_N}]
set_input_delay -clock [get_clocks fifoClk] -max 7.0  [get_ports {BE[*] DATA[*]}]
set_input_delay -clock [get_clocks fifoClk] -min 6.5  [get_ports {RXF_N}]
set_input_delay -clock [get_clocks fifoClk] -min 6.5  [get_ports {BE[*] DATA[*]}]

--- End code ---

The input delay max = CLK period - T1 = 10ns - 3ns = 7ns (for CLK = 100MHz)



I am confused about the input delay min = 6.5ns.

Since T2 is a minimal value, after CLK's rising+T2, the data of next cycle may be stable before the next CLK's rising-T1.

So I think the input delay min is T2 = 3.5ns.

Why the input delay min is 6.5ns in the AN_421?


Thanks,
Stanley

FTDI Community:
Hello,

We are actively working with R&D to review FPGA timing.

Our IC designer has confirmed that the timing we are using in our demo application is a known good reference:

https://ftdichip.com/software-examples/ft600-601-software-examples/

You can also contact your local support team via email as it seems you have already been in contact before:

https://ftdichip.com/technical-support/

Best Regards,
FTDI Community

NorthGuy:
Here's what the example application suggests for Quartus:


--- Code: ---set_input_delay -clock [get_clocks fifoClk] -max 7    [get_ports {BE[*] DATA[*]}]
set_input_delay -clock [get_clocks fifoClk] -min 6.5  [get_ports {BE[*] DATA[*]}]

set_output_delay -clock [get_clocks fifoClk] -max 1.0 [get_ports {BE[*] DATA[*]}]
set_output_delay -clock [get_clocks fifoClk] -min 4.8 [get_ports {BE[*] DATA[*]}]
--- End code ---

Here's what I used for Xilinx 7-series in Vivado, which I believe is correct:


--- Code: ---# max delay is 7.0 (from FTDI docs) + 0.4 data wires - 0.2 clock wire = 7.2
set_input_delay -clock [get_clocks clk_ftdi] -max 7.2 [get_ports {ftdi_data_*}]
# min delay is 3.5 (from FTDI docs) + 0.2 data wires - 0.2 clock wire = 3.5
set_input_delay -clock [get_clocks clk_ftdi] -min 3.5 [get_ports {ftdi_data_*}]

# max delay is 1.0 (from FTDI docs) + 0.4 data wires + 0.2 clock wire = 1.6
set_output_delay -clock [get_clocks clk_ftdi] -max 1.6 [get_ports {ftdi_data_*}]
# min delay is -4.8 (from FTDI docs) + 0.2 data wires + 0.2 clock wire = -4.4
set_output_delay -clock [get_clocks clk_ftdi] -min -4.4 [get_ports {ftdi_data_*}]
--- End code ---

The FT601->FPGA figures (set_input_delay) are easy to meet either way.

The FPGA->FT601 figures (set_output_deay) are practically impossible to meet because of the very long hold requirement. Of course, if you use "hold" instead of "minus hold" for set_output_delay -min (as in the example application) the timing analysis will succeed every time.

StanleyChien:
Hi, NorthGuy:

Thanks for your reply!   :)

I have the same opinion with you on the min value of input and output delay.

I think your SDC is correct unless we misunderstand the FT601's datasheet.

Navigation

[0] Message Index