TMS570LC4357: EMIF Wait is ignored at 70C temp.

Hi Koichiro-san,

I never heard this problem before. 

At EMIF reading failed, did you read back the value of CE2CFG register and AWCCR register to make sure that the EW bit is set and WP0 bit is cleared?

Hi QJ,

Customer did not take register dump after the failure.
But it is not likely register values are corrupted because software only configured them once after power up.
And the issue still observed after power cycle (i.e. EMIF registers are initialized again).

Thanks and regards,
Koichiro Tashiro

The timing parameters of nWAIT signal (setup time, period, and delay) in your excel file look fine. 

Are other parts on the board (memory, crystal, power supply, etc) industrial grade (up to 85C) or automotive grade (up to 125C)? The upper temperature limit for commercial parts is 70C only.

Any improvement if they increase the EMIF Async setup/strobe? 

HI QJ,

QJ Wang said:

Are other parts on the board (memory, crystal, power supply, etc) industrial grade (up to 85C) or automotive grade (up to 125C)?

Customer uses Industrial grade.

QJ Wang said:

Any improvement if they increase the EMIF Async setup/strobe? 

Customer checked it briefly. Adding +1 strobe solves the issue.

Thanks and regards,
Koichiro Tashiro

HI Koichiro-san,

Good, thanks Koichiro.

It is understandable that the memory device needs larger setup time and strobe time at higher temperature (memory becomes slower). If using the same setup/strobe, does shorten nWAIT period help?

Hi QJ,

I am not sure I understood your point correctly.
The access is read, so Hercules side needs to extend STROBE period based on nWAIT input, right?
Why slower operation at memory side results in such behavior on Hercules side?
It seems the memory asserts nWAIT signal at the same timing(36.25ns from nCS2) at 70C and RT.

Thanks and regards,
Koichiro Tashiro

Hi Koichiro-san,

I'd like to know that if a smaller nWAIT pulse width works at 70C

Hi QJ,

QJ Wang said:

I'd like to know that if a smaller nWAIT pulse width works at 70C

Customer wants to know why this information is needed. It takes time for them to change nWAIT pulse period.
They suspect the root cause of the issue resides in Hercules side not the memory side.
If datasheet timing parameters needs to modified for high temperature, they need updated values.

Thanks and regards,
Koichiro Tashiro

The strobe width is programmable from 0 to 0x3F. The strobe width of a read operation is mainly determined by the data output delay of the Async memory. 

What EMIF timing parameters in TMS570 datasheet need to be modified?

Hi QJ,

I understood that longer strobe period is needed if the memory gets slower. And the strobe width is programable.
But the issue is that EMIF does not extend strobe even nWAIT is active at 70C.
Because nWAIT signal assertion timing seems the same both 70C and RT, it implies EMIF cannot detect nWAIT properly.

So my guess is EMIF parameter #14, "tsu(EMOEL-EMWAIT) : Setup Time, EMIFnWAIT asserted before end of Strobe Phase" 
minimum needs to be longer at 70C. 


Customer observed adding +1 cycle in EMIF strobe solved the issue.
This agrees to above assumption.

Thanks and regards,
Koichiro Tashiro