P2020 Processor GPCM with DLL Bypass Read/Write Timing

cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

P2020 Processor GPCM with DLL Bypass Read/Write Timing

Jump to solution
10,212 Views
fmlgb
Contributor II

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.

Labels (1)
1 Solution
9,968 Views
ufedor
NXP Employee
NXP Employee

Your understanding is generally correct.

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

View solution in original post

12 Replies
9,969 Views
ufedor
NXP Employee
NXP Employee

Your understanding is generally correct.

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

9,964 Views
fmlgb
Contributor II

Thank you very much.

0 Kudos
Reply
9,964 Views
ufedor
NXP Employee
NXP Employee

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 

0 Kudos
Reply
9,961 Views
fmlgb
Contributor II

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 signalGPCM.png.
Can you give me a further explanation on the attached figure?


Thank you.

0 Kudos
Reply
9,959 Views
ufedor
NXP Employee
NXP Employee

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:

RT.jpg

9,962 Views
fmlgb
Contributor II

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?

0 Kudos
Reply
9,962 Views
ufedor
NXP Employee
NXP Employee

> 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.

9,962 Views
fmlgb
Contributor II

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

tAWCS

tCSWP

tAWE

tWEN

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

tAWCS

tCSWP

tAWE

tWEN

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:

pastedImage_2.png

Is my description and timing diagram right?

I will be looking forward to your reply.

0 Kudos
Reply
9,963 Views
ufedor
NXP Employee
NXP Employee

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

0 Kudos
Reply
9,964 Views
fmlgb
Contributor II

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 tCSRP

PLL bypass write timing:

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

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?

pastedImage_1.png

                                                                                Figure 1

pastedImage_2.png

                                                                                  Figure 2

0 Kudos
Reply
9,964 Views
fmlgb
Contributor II

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

tAWCS

tCSWP

tAWE

tWEN

tWC

1

0

11

1

1.5

1.25+2* SCY

2

0

2.75++2* SCY

Write timing diagram as following:

pastedImage_1.png

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

0 Kudos
Reply
9,967 Views
fmlgb
Contributor II

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

0 Kudos
Reply