SN75DP159: Failing EMC testing

HI Gregory,

What speed data are you running during the test?    These spikes are higher than we typically see for HDMI 2.0 retimers.  As a test, can you try dropping the resolution (4K2K30 Hz or lower) and putting the DP159 into redriver mode, that would help us determine what is sourcing the spikes.  We do recommend 10 pF and 4.7 pF caps on the voltage rails to improve EMC, but that alone is likely not enough to resolve this.

Regards,

JMMN

We were only running 1920x1080/60Hz and 1280x1024/60Hz for our EMC testing. The plot included here is for the 1280x1024 resolution running only. The 1920x1080 plot looks similar, just different fundamental frequency. We did try redriver mode across full range (using these resolutions), and we saw no effect.

Gregory,

That is very strange, I would expect to see noticeable impact from dropping to redriver mode? What does the baseline for this project look like?

I also would not expect to see the large spikes at 600 MHz/ 450 MHz, can you provide your I2C settings from the test?

Regards,
JMMN

JMMN,

Sorry for the delayed response - I thought this thread went dark.  The baseline, with HDMI disabled, shows no violators.  See below.

Here is a snippet of the U-boot I2C commands we used to try to reduce the EMI.

//U-Boot commands to try to reduce HDMI EMI
//

//top HDMI port driven by HDMI1 I2C4 addr (7-bit): 0x5E
//bottom HDMI port driven by HDMI2 I2C4 addr (7-bit): 0x5D

//set active bus to I2C4
i2c dev 3

//Address 0x0c
//adjust VSWING_DATA (Data output swing control)
//[7:5] 000 - Vsadj set
//[7:5] 100 - decrease by 30%
//[7:5] 101 - decrease by 21%
//[7:5] 110 - decrease by 14%
//[7:5] 111 - decrease by 7%
//adjust VSWING_CLK (Clock output Swing Control)
//[4:2] 000 - Vsadj set
//[4:2] 100 - decrease by 30%
//[4:2] 101 - decrease by 21%
//[4:2] 110 - decrease by 14%
//[4:2] 111 - decrease by 7%
//[0:1] 00 - No de-emphasis
//[1:0] 01 - 2-db de-emphasis

//try -30%, -30%, 2-db (0x91)
i2c mw 5e c 91 1
i2c mw 5d c 91 1

//Address 0x0A
//set to redriver mode from default redriver/retimer
i2c mw 5e a 38 1
i2c mw 5d a 38 1

Hi Gregory,

I just wanted to confirm that after you changed the swing setting and/or the redriver / retimer settings, you set the apply_rxtx_changes bit in Register 0Ah. The changes don't get picked up until that bit is set or HPD toggles.

Is the DP159 sitting close to the source? Do you see the EMI drop in half if only one DP159 is running?

Rgds,
JMMN

I saw the datasheet indication of the need to set the apply_rxtx_changes bit, but we noticed that the HDMI display changed even without applying this, so I stopped. We'll need to investigate this further. But NO, there was not appreciable reduction in EMI when only one DP159 was running.

The apply_rxtx_changes bit doesn't need to be set on every change, but going from retimer -> redriver mode and back it does.

When one of the DP159s was not running, was it in powerdown mode or just idle?

Regards,
JMMN

When one of the DP159s was not running, the HDMI cable was unplugged.

Hi Gregory,

Your results don't match with any of the EMI testing we have done on this part. I'm getting feedback from additional engineers now.

Rgds,
JMMN

Hi Gregory,

One of my team members noticed that the baseline for your testing is very high.  This is a baseline we took during EMI testing and I checked on several 3rd test reports and they are showing similar baseline results.

Thanks for this key observation. Will investigate this further.