RDK2.8 NSF

Hi,

Can you please look into below link and see if helps?

http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/717/p/214886/780896.aspx#780896

Regards,

Brijesh Jadav

Hi, Brijesh

Accroding to the thread,  I change   nsfProcCfg->staticFrameNoise[i] = 0;  to  nsfProcCfg->staticFrameNoise[i] = VPS_NSF_PROCESSING_CFG_DEFAULT;

It seems there also has no change.

By the way, the dup Link need to be inited?  Before we not add dupLink init, sometimes there was error about dup, sometimes not.

So we add the dupLink init, there was no more error. Is it a bug?

Hi, Badri. Thanks very much for your help.

We just add MULTICH_INIT_STRUCT( ).

Can you help me about the NSF?

Adding MULTICH_INIT_STRUCT( ) is good.It will initialize create params to default values. Probably all the create params were not initialized and thats why you were seeing instability.Adding MULTICH_INIT_STRUCT is the right solution,

Regarding NSF how are you checking NSF is taking effect or not ? Do you have noisy content from camera ?  The fix done by Brijesh resolved the issue in the other post so we are sure that is the fix. Are you seeing some effect if you set nsf params to invalid values as mentioned in the post ?

Hi, Badri.

I use the usecase of capture-display. We see the image on screen and it looks like not take effect.  Before we use EZDK5.05, just the same usecase, the effect of NF is obvious.

Now there has obviously noise on the screen.

I set the nsf params to zero. but there is no change.

You are setting     nsfPrm.bypassNsf                 = TRUE right ? We will try out cap display with nsf enabled and update you on Monday .

Hi, Badri. Thanks very much for your help.

I set nsfPrm.bypassNsf    = TRUE and also set it to FALSE. The result looks like same.

We use RDK2.8.

Have a good weekend.

Hi Shary,

Is your picture really noisy? if yes, then only it would help. Noise filter would the noise from the noisy images. If there is no noise or it is very minimal, you would get the almost same output with or without noise filter enabled?

Regards,

Brijesh Jadav

Hi, Brijesh

Is there any change of your output if you set bypass TRUE or FALSE?

the output image quality  is very bad, there has most noise and it is vague. So it can not be used.

How do you set?

Hi, Badri.

How about your cap display? the nsf is working?

Hi, Brijesh.

I also have another problem, can you help me?

If we use cap-nsf-dei-display, the image in screen is OK. But if we use caputer-nsf-encode, the image in screen shiths left a few pixels. what's the problem?

Sorry I was unable to try out yesteray.Will update you by end of day today.

Can you share image showing the shift when nsf is present. In your capture link do you have enableSDCrop set to TRUE ? Can you try disabling it and checking if there is any difference

Hi, Badri. thanks vey much.

I set  enableSDCrop to FALSE, it's OK now. Thanks.

wait for your update.

Hi, Badri.

I test three days ,  the nsf also seems not work.  do you have any advice?

I tried enablign nsf in multich_vcap_vdis.c usecase and I see it is getting enabled. I dont see any artifacts as you mentioned. Can you confirm you are using PG2.0 816x ? Also can you add a Vps_printf in funtion

ti_tools/hdvpss/packages/ti/psp/vps/hal/src/vpshal_nf.c

VpsHal_nfSetConfig()

Vps_printf("\n Enter NSF HAL Set Config");

You should see the debug print on the console.

If possible pls muigrate to the latest RDK 3.5 release. There are a number of HDVPSS driver fixes since RDK 2.80 so maybe that is why you are seeing the issue.,

Hi, Badri. Thanks very much for your help.

We use 8168, ourself board, I'm not very clear about the PG2.0.

I add the Vps_printf in VpsHal_nfSetConfig( ); my manager will want to use RDK2.8, so ....

whether I set bypass ture or false, the printf is also printed in console.  Is there anything I set wrong?

Hi, Badri.

Do you use RDK2.8 when you test? Before we use EZSDK5.05, the hdvpss is also hdvpss_01_00_01_37. the RDK2.8 is same.

why the result of NSF is not same?

The HDVPSS package is different between EZSDK and RDK even though the name may be same. You can check the PG version when you boot up.The kernel will print the PG revision on the console.

Hi, Badri. Thanks very much.

The PG version is the CPU GP Version? If right, we use 2.0.

And if we need to use RDK2.8, how can I resolve the NSF problem?

Hi,

If nsf is externally or internally bypassed or all config parameters are set to 0, in that case nsf will not do anything. can you share nsf register settings? I will check what could be wrong.

Regards,

Brijesh Jadav

Hi, Brijesh. Thanks very much for your reply. the value of register is as following: 4810c200: 00000801 04380780 00000808 10101010 ......8......... 4810c210: 0000060d 000000ad 00000f0f 00000000 ................ 4810c220: 00001fff 1fff1fff 00000000 00000000 ................ 4810c230: 00000000 00000000 00000000 00000000 ................ 4810c240: 00000801 04380780 00000808 10101010 ......8.........

HI, Brijesh.

I test EZSDK5.05, the register is :

4810c200: 00001805 04200780 00000000 00000000    ...... .........
4810c210: 00000600 000000ad 00000f0f 00000000    ................
4810c220: 00000000 00000000 00000000 00000000    ................
4810c230: 00000000 00000000 00000000 00000000    ................

I found some register are different with our RDK. Is something wrong?

Thanks.

Hi, Badri.

The PG version is the CPU GP Version? If right, we use 2.0.

the register is :

4810c200: 00000801 04380780 00000808 10101010    ......8.........
4810c210: 0000060d 000000ad 00000f0f 00000000    ................
4810c220: 00001fff 1fff1fff 00000000 00000000    ................
4810c230: 00000000 00000000 00000000 00000000    ................

Is there something wrong?

I'm  looking forward your help.

Hi, Badri.

I compared RDK2.8 and RDK3.5 about NSF.

But I found no great substantial difference between them.

There seems no document about the difference.

Can you help me?

The RDK 3.5 release notes has changes from previous release but only new features and API changes are documented. There were a number of bug fixes in HDVPSS driver package. Anyhow based on register dump you shared it looks like all NSF configuration is happening correctly in RDK but you still dont see the effect so I think migrating to RDK 3.5 will not help for this particular issue. It is better to migrate to RDK 3.5 as there are a number of bug fixes from RDK 2.8 which you will face if you continue on the old release

Hi, Badri.

Does the NSF have effect in RDK2.8? How about RDK3.5?

Thanks very much.

Hi, Badri.

I can do nothing about NSF. Can you help me?

The nsf is effect in RDK2.8? How about RDK3.5?

How do you test your cap-display?

I really need your help.

Thanks.

The register settings are correct for NSF so we dont know what the issue is. We have run the HDVPSS NSF test bench and confirmed NSF is taking effect. Can you share either picture of video of input video and the difference you see between EZSDK with NSF enabled and RDK with NSF enabled.