HyperRAM RWDS Refresh Collision Timing

Tip / Sign in to post questions, reply, level up, and achieve exciting badges. Know more

cross mob
GeFu_3059156
Level 1
Level 1
First like given 5 sign-ins Welcome!

I have a question about the RWDS timing during the CA interval where simulation disagrees with the datasheet.

The simulation shows a HyperRAM write cycle where a refresh collision has occurred. RWDS changes in the middle of the CA interval signaling 2x latency. The datasheet only shows RWDS valid on CS# low.

If it is important to know when RWDS should be sampled to guarantee proper latency.

Infineon S27KS0642 1.8V 64Mb 200MHz
S27KS0642GABHI020

Clock is 125MHz. Initial latency set to 5. Variable latency enabled

 

RefreshCollision.PNG

Infineon memory model with the correct `define selected

Model.PNG

0 Likes
1 Solution
lock attach
Attachments are accessible only for community members.
Shubham_D
Moderator
Moderator
Moderator
10 likes received First question asked 50 solutions authored

Hi @GeFu_3059156 ,

 

I apologize for the delay in response.
Please find the updated model with logic bug fixed. Please review and let me know if it solves the issue. 

 

Thanks,
Shubham

View solution in original post

0 Likes
6 Replies
Shubham_D
Moderator
Moderator
Moderator
10 likes received First question asked 50 solutions authored

Hi @GeFu_3059156,

 

Could you please clarify significance of the data packets present at the starting on Data line?

Could you please confirm RWDS in Test bench is not high at any instant?

Could you please provide the write timing parameters?

 

Thanks,

Shubham

 

0 Likes
GeFu_3059156
Level 1
Level 1
First like given 5 sign-ins Welcome!

Thank you.

Line 1 is just a system clock for reference. 125MHz

1. The data at the leading edge of the diagram is finishing up a read cycle. You can see the final CK and CS# going high.

2. I can confirm that the memory and not the FPGA is driving RWDS in the timing diagram with the exception of the trailing edge of the diagram where the FPGA drives RWDS low (unmasked) as part of the write cycle.

3. Timing parameters. Please let me know what you are looking for.

CK# is 125MHz. Initial latency set to 5 in the control register. Variable latency enable in the control register. Simulation has verified reading and write to memory and control registers correctly with specified latencies.

In my experiments, I have added extra clock cycles to CS# high,width, and low timing to see if it affects when RWDS changes state.

These refresh collisions happen very infrequently. I find them by watching refresh signals internal to the HyperRAM model. My goal is to verify that the FPGA handles variable latency correctly.

Based on my experiments, RWDS low Z follows CS# falling but RWDS variable latency indication follows CK.

I can move the FPGA's RWDS sample point but the datasheet does not dimension this position .

 

 

 

 

 

0 Likes
Shubham_D
Moderator
Moderator
Moderator
10 likes received First question asked 50 solutions authored

Hi @GeFu_3059156 ,

 

I apologize for the delay in response. 

we will fix the model and released the updated version by next Friday.

 

Thanks,
Shubham

0 Likes
lock attach
Attachments are accessible only for community members.
Shubham_D
Moderator
Moderator
Moderator
10 likes received First question asked 50 solutions authored

Hi @GeFu_30591563059156,

 

Thanks for pointing out the issue. We reviewed the model and was able to replicate the issue. Please find the updated model with logic bug fixed. Please review and let me know if it solves the issue which you have seen in the earlier version of model.

 

Please find the attached model with this response.

Thanks,
Shubham 

0 Likes
lock attach
Attachments are accessible only for community members.
GeFu_3059156
Level 1
Level 1
First like given 5 sign-ins Welcome!

 

I have incorporated the new simulation model into my testbench and see the same RWDS anomaly where simulation differs from the datasheet.

The datasheets shows RWDS status being valid on the falling edge of CS. The simulation (read cycle shown) shows RWDS during refresh collision changing state on the rising edge of the second clock pulse (or thereabouts).

It is important to my implementation that RWDS is sampled at the correct time to handle the extended refresh.

So when is RWDS on refresh collision supposed to be valid and when should the memory controller sample RWDS?

 

Capture.PNG

Capture2.png

GeFu_3059156_1-1675710886330.png

 

0 Likes
lock attach
Attachments are accessible only for community members.
Shubham_D
Moderator
Moderator
Moderator
10 likes received First question asked 50 solutions authored

Hi @GeFu_3059156 ,

 

I apologize for the delay in response.
Please find the updated model with logic bug fixed. Please review and let me know if it solves the issue. 

 

Thanks,
Shubham

0 Likes