Question about SATA_refcLk

Shouming Li18

I use LVDS oscillator(not clock buffer) to generate clock for sata_refclk. Schemetic :

But i found sata does not work. And the reason is that the lvds clock signal quality is not good(measure at CPU pin). See waveform :

Could you please support me to solve this problem. Thank you very much.

over 10 years ago

0 Titusrathinaraj Stalin over 10 years ago

TI__Guru** 116100 points

Hi Shouming,

Please refer to the below links and cross check that you have violated any electrical spec.

http://processors.wiki.ti.com/index.php/OMAP-L138/C6748/AM1808_Hardware_Debugging_Guide#Electrical_Issues

http://processors.wiki.ti.com/index.php/OMAP-L138_Hardware_Design_Guide#Designing_the_Clocking_Subsystem

What is the SATA frequency that you are operating ?

Where did you measure the SATA clock exactly ?

http://e2e.ti.com/support/dsp/omap_applications_processors/f/42/t/88797.aspx

0 Senthilkumar Srinivasan over 10 years ago

TI__Mastermind 43660 points

Hello Shouming,

The problem seems to be on the PCB layout routing. Please check the layout for proper routing of SATA_REFCLKP/N.

Is the LVDS oscillator provided with proper filters for Vdd ?

Regards,

Senthil

0 Senthilkumar Srinivasan over 10 years ago

TI__Mastermind 43660 points

Hello Shouming,

Have you followed the SATA Routing specifications and AC coupling requirements as mentioned in the device datasheet ?

Regards,
Senthil

0 Shouming Li18 over 10 years ago in reply to Senthilkumar Srinivasan

Prodigy 80 points

Dear all:

Here is my design.Please check thank you very much.

7455.problem_en.docx

Next document is writed in Chinese. This is a very strange phenomnon.

7002.picture.docx

0 Senthilkumar Srinivasan over 10 years ago in reply to Shouming Li18

TI__Mastermind 43660 points

Hello Shouming,

Could you please answer the below questions.

Why 5V supply is connected to the SATA connector pin 7 ? It should be connected to ground per standard SATA connector pinout.

Why C110 decap is connected in common to 3.3V and 2.5V rails ?

Try remove the AC coupling caps and check the clock output at oscillator end using differential probe ?

Regards,

Senthil

0 Shouming Li18 over 10 years ago in reply to Senthilkumar Srinivasan

Prodigy 80 points

Question about SATA pin7: This is recommended by SATA DOM company(Innodisk). Here pin7 is the power pin.

Question about C110: Here is the compatible circuit for 3.3V or 2.5V power supply Oscillator. Now Bead(FB100) is used for 2.5V.And it is not soldered. I only use 3.3V.

While removing AC caps. the waveform as following picture(Measured by differential probe):

0 Senthilkumar Srinivasan over 10 years ago in reply to Shouming Li18

TI__Mastermind 43660 points

Hello Shouming,

I guess the the problem is related to the ground issue. Especially the SATA controller doesn't get proper ground reference.

Please check for proper ground connection, assembly issues and dry soldering of the device.

Regards,

Senthil

0 Shouming Li18 over 10 years ago in reply to Senthilkumar Srinivasan

Prodigy 80 points

I soldered 5 boards. And they face the same question. So the dry soldering may not be the issue.

And after checking the PCB file, i found that the gound pins are well connected.

And when i connected the negetive signal output from lvds oscillator to 1.2V(output from LDO), then clock is ok and SATA DOM works well.

But i can find the reason.

Hehe!

0 Shouming Li18 over 10 years ago in reply to Shouming Li18

Prodigy 80 points

From my side, does this problem associated with the common voltage of my OSC?

0 Senthilkumar Srinivasan over 10 years ago in reply to Shouming Li18

TI__Mastermind 43660 points

Hello Shouming,

Could you please share the clock waveform when you connect 1.2V on CLK_N pin.

Regards,
Senthil

0 Senthilkumar Srinivasan over 10 years ago in reply to Shouming Li18

TI__Mastermind 43660 points

Hello Shouming,

I just had a look into the oscillator (NBXSBA010LNHTAG) datasheet. As you mentioned in the above post, this is not a LVDS oscillator. This is LVPECL output oscillator and it requires termination at the receiver end for proper operation.

Please refer Figure 4 in oscillator datasheet for termination recommendation.

http://www.onsemi.com/pub_link/Collateral/NBXSBA010.PDF

This is the reason behind when you connect 1.2V (Termination Voltage, Vtt = ~Vdd - 2.0V) on CLK_N and the interface was working well.

When you use any device, please go through its data manual and strictly follow their recommendations without fail.

Regards,

Senthil

0 Shouming Li18 over 10 years ago in reply to Senthilkumar Srinivasan

Prodigy 80 points

the waveform is measured at the pin of oscillator(with active differential prob)

0 Senthilkumar Srinivasan over 10 years ago in reply to Shouming Li18

TI__Mastermind 43660 points

Hello Shouming,

As mentioned above, try adding the termination resistors on clock lines and check the SATA functionality.

Regards,

Senthil

0 Shouming Li18 over 10 years ago in reply to Senthilkumar Srinivasan

Prodigy 80 points

Soory, the osc in my schematic that i provided to you, is not the real chip i soldered on my board.

I am sure that the osc soldered on my board is a lvds osc. I have checked with the vendor.

And not only connect the negetive signal to 1.2V, but also to other voltage (lower than 1.2V and high than Gnd), the sata dom can work well too.

0 Senthilkumar Srinivasan over 10 years ago in reply to Shouming Li18

TI__Mastermind 43660 points

Hello Shouming,

The problem persist on the oscillator side only. You are required to feed the clean reference clock to the SATA controller for proper operation.

In our reference boards, the SATA interface is working fine without any issues. We used clock buffer for reference clock.

I noticed the common voltage of your oscillator output is not correct. You have to work with oscillator vendor for any termination recommendations in specific and try to implement the same in your design.

Thanks for your understanding.

Regards,

Senthil

0 Shouming Li18 over 10 years ago in reply to Senthilkumar Srinivasan

Prodigy 80 points

Dear

I have found that the quality of the oscillators that bought from Net are bad.

And the oscillators that made in TaiWan were tested in my board, the result is that sata dom works well.

So the problem has been solved.

Thank you for your support.

0 Titusrathinaraj Stalin over 10 years ago in reply to Shouming Li18

TI__Guru** 116100 points

Hi,

Sounds good.

Thanks for your update.

Processors

Processors forum

Question about SATA_refcLk