Checks the sanity of the SDRAM module on MiST and MiSTer systems
Add the JTFRAME git submodule following the standard procedure. Read the README file in it to learn about how to compile JTFRAME projects in general.
In summary, you will need to be doing this from a linux workstation and from the project root folder type:
. setprj.sh
jtcore sdramThat is all
The FPGA will constantly fill the first 8MB of each memory bank with pseudorandom data at a fast pace, but without bank interleaving. Then it will read it back at full speed; that means 96MB/s at 48MHz operation. It will stop during vertical blanking to issue autorefresh commands. Each memory filling is checked four times. Then a new round begins.
If a problem occurs, the screen will turn red, the LED will blink and a high pitch tone will replace the normal one. If only part of the screen turns red, it means that the problem only occured in one bank.
Use the macro ONEBANK to simulate only with a single bank. This speeds up simulation.
Phase inversion of the SDRAM clock is done in MiSTer by using the altddio_out primitive. This method does not seem to be so different from using phase shifting at the PLL.
SDRAM clock path delay examples:
| Clock | DDIO/PLL | Min | Max | Delta |
|---|---|---|---|---|
| 48 | DDIO | 6.2 | 11.6 | 5.4 |
| 48 | PLL | 6.8 | 12.9 | 6.1 |
| 96 | DDIO | 4.2 | 8.5 | 4.3 |
Clock in MHz, time values in ns.
I don't think instantiating a DDIO cell changes the clock path delay. Quartus seems to promote the PLL output to a global clock net anyway, so there is no reason why the delay should be different.
On top of the delay, the PLL will add a given precise phase shift, and the DDIO will add a fixed 180º shift. In practice, it is not possible to control the phase of the SDRAM clock using either method. Feeding back the delay to the PLL is the only way of doing it and that seems to require an extra pin of the FPGA. The pin must be left unconnected to the PCB as well so it doesn't get loaded.
At the end, it is the synthesis tool that checks that SDRAM I/O constraints are met. Sometimes the extra 180º provided by DDIO may provide better STA results, on other occasions, it will a given phase shift produced at the PLL.
The jotego nickname had already been used by other people so on some networks you will find me as topapate.
Contact via:
You can show your appreciation through