PowerQUICC ProcessorsのトピックRe: P2020 Processor GPCM with DLL Bypass Read/Write Timing

P2020 Processor GPCM with DLL Bypass Read/Write Timing

fmlgb — Fri, 17 Apr 2020 06:49:13 GMT

When looking at the P2020 QorIQ Integrated Processor Reference Manual (Rev. 2, 12/2012), I notice that when the local bus is in bypass mode, eLBC drives new address, data, and control signals effectively on falling edges of LCLK, but continues to sample synchronous read data on rising edges of LCLK to maximize the set-up margin for reads. However, the precise timing relationship between TA signal and read-write signal is not clearly stated.
I need to do an accurate timing analysis, but I don't know the specific timing relationship between TA signal and read-write signal, and which rising edge of LCLK is the sampling point for reading data when TA signal is asserted.
The register settings in my design are as follows:
BR：0xd0001001

OR：0xffff0ef5

Any help would be much appreciated.
Thanks.

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

ufedor — Fri, 17 Apr 2020 11:41:55 GMT

Please refer to the P2020 QorIQ Integrated Processor Hardware Specifications, 2.12.2.3 Local Bus AC Timing Specifications for PLL Bypass Mode:

https://www.nxp.com/webapp/Download?colCode=P2020EC&location=null

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

fmlgb — Mon, 20 Apr 2020 01:28:02 GMT

I have read the p2020ec, but I still haven't figured out the exact timing relationship between the sampling point of reading data and the reading signal. At the same time, the time when writing data becomes invalid is related to the accurate timing of writing signal.
Can you give me a further explanation on the attached figure?

Thank you.

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

ufedor — Mon, 20 Apr 2020 04:31:46 GMT

P2020 QorIQ Integrated Processor Reference Manual, 12.4.1.7 PLL Bypass mode:

"In PLL bypass mode, eLBC drives new address, data, and control signals effectively on falling edges of LCLK, but continues to sample synchronous read data on rising edges of LCLK to maximize the set-up margin for reads."

In the P2020 QorIQ Integrated Processor Hardware Specifications, 2.12.2.3 Local Bus AC Timing Specifications for PLL Bypass Mode:

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

fmlgb — Mon, 20 Apr 2020 05:41:30 GMT

Thank you for your reply.

According to your description, the sampling point for reading data should be the rising edge of lclk before the LOE is invalid, and at the next falling edge of lclk, LOE becomes invalid. Is my description right?

In addition, how about write timing?

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

ufedor — Mon, 20 Apr 2020 06:49:46 GMT

> Is my description right?

Yes.

> In addition, how about write timing?

Please refer to the Figure 40 in the P2020 QorIQ Integrated Processor Hardware Specifications, 2.12.2.3 Local Bus AC Timing Specifications for PLL Bypass Mode.

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

fmlgb — Mon, 20 Apr 2020 08:30:48 GMT

If Twen = 0, is writing data and writing signal invalid at the same time?

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

fmlgb — Thu, 23 Apr 2020 06:12:56 GMT

In my design, OR register is set as follows:

AM(0~16)	17~18)	BCTLD(19)	CSNT(20)	ACS(21~22)	XACS(23)
0x1fff8	00	0	1	11	0
SCY(24~27)	SETA(28)	TRLX(29)	EHTR(30)	EAD(31)
1111	0	1	1	0

According to P2020RM （Rev. 2, 12/2012） Table 12-191. GPCM write control signal timing parameters as follows：

TRLX	XACS	ACS	CSNT	t_AWCS	t_CSWP	t_AWE	t_WEN	tWC
1	0	11	1	1.5	1.25+2* SCY	2	0	2.75+2* SCY

But because TRLX = 1, so LCSn and LWE signals are negated one cycle earlier during Writes. The parameters in the above table will change to the following：

TRLX	XACS	ACS	CSNT	t_AWCS	t_CSWP	t_AWE	t_WEN	tWC
1	0	11	1	1.5	0.25+2* SCY	2	1	2.75+2* SCY

According to the above description, I have drawn the following timing diagram：

Is my description and timing diagram right？

I will be looking forward to your reply.

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

ufedor — Thu, 23 Apr 2020 14:40:29 GMT

You need to use timing data from the corresponding row of the "Table 12-191. GPCM write control signal timing" without any modification.

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

fmlgb — Fri, 24 Apr 2020 02:42:24 GMT

If I use timing data from the corresponding row of the "Table 12-191. GPCM write control signal timing" without any modification.

TRLX	XACS	ACS	CSNT	t_AWCS	t_CSWP	t_AWE	t_WEN	t_WC
1	0	11	1	1.5	1.25+2* SCY	2	0	2.75++2* SCY

Write timing diagram as following:

If it is the above timing, write data and write signal are invalid at the same time. so the hold time of write data will be insufficient.

And according to your comment, I am confused about the following description in P2020RM:
TRLX is 1:
Relaxed timing on the following parameters:
• Adds an additional cycle between the address and control signals (only if ACS is not equal to
00).
• Doubles the number of wait states specified by SCY, providing up to 30 wait states.
• Works in conjunction with EHTR to extend hold time on read accesses.
• LCSn (only if ACS is not equal to 00) and LWE signals are negated one cycle earlier during
Writes

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

fmlgb — Mon, 27 Apr 2020 06:37:09 GMT

According to your help and my understanding of the P2020RM, I think the reading and writing sequence is as follows:

PLL bypass read timing：

The sampling point for reading data should be the rising edge of lclk before the LOE/LCS is invalid. This time is determined by the parameter t_CSRP.

PLL bypass write timing：

Write data is invalid at the end of the t_WC. This time is determined by the parameter t_WC.

Is my understanding right?

I have another question：

As described in P2020RM， LCSn (only if ACS is not equal to 00) and LWE signals are negated one cycle earlier during Writes when TRLX is 1.

The write sequence will change from Figure 1 to figure 2？

Figure 1

Figure 2

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

ufedor — Mon, 27 Apr 2020 10:57:20 GMT

Your understanding is generally correct.

Please consider that it is convenient to create new Community question/thread for the new question.

Re: P2020 Processor GPCM with DLL Bypass Read/Write Timing

fmlgb — Mon, 27 Apr 2020 11:18:17 GMT

Thank you very much.