LVDS EMC emitted noise

Hi,

We’re developing a HMI using the colibri imx8x with Iris carrier. using the X7 lvds connector to communicate with a 3rd party screen. we’re currently running Torizon OS with a docker container displaying a test image.

We had our first radiated emissions test yesterday and saw a large amount of noise at harmonics of our pixel clock/the LVDS clockrate (52MHz).

I expect the issue lies more with the screen and cabling but could you confirm the following;

  1. that the LVDS bridge, pcb traces, and X7 connectors conform to the 100Ohm differential impedance requirement for lvds signals
  2. that there isn’t an adjustable drive strength possible with the Iris lvds bridge (THC63LVD827)

Are there any further suggestions that you can make from the toradex side to improve our performance?
I understand that it’s difficult to make any direct suggestions as the radiated emissions are so dependent on integration.

Many thanks,
Ben

Hi,
Just bumping this, is it possible to get hardware support on this issue?
Many thanks,
Ben

Hello @benp,

It is not unusual to see EMC issues in new designs. So I would like to get more information on that.
The best would be if we set up a tech call.
You can schedule a call here.

Best Regards,

Matthias

1 Like

Hi @matthias.tx ,

Thanks for your message, I’ve booked a call in with you.
Please let me know if there are any further details that you require.

Best regards,
Ben

Hi @matthias.tx

I’ve been probing the board with both a scope & using a proximity noise probe. I noticed that the emitted 52MHz (and harmonics) signal from the pixel clock was strongest over the “LCD_PCLK_WR” net (clock in for the RGB to LVDS converter, IC9 on the iris schematic).
I have attached a screen shot of this noise (R&S fpl1003, HZ-15); the actual signal captured via the scope (200MHz bandwidth, spring ground connection) looks okay (no distortion, no ringing image also below)

I’ve seen on other posts that users have been recommended to alter the drive strength ( AN5078.fm (nxp.com) , DSE, SRE)
are there any device tree or pin settings that you could recommenced to try changing to reduce this pixel clock noise?

Hello @benp,

Do you have an OSC plot showing the allowed max peak so we can see how much we are over the limit?

Best Regards,

Matthias Gohlke

Hi @matthias.tx ,
I will message you with some sample plots and more details.
The above is just a near field probe measurement so is just indicative of the noise source rather than a calibrated, reference measurement.

Hi @matthias.tx ,

Just checking that you’ve received the plots I sent over via direct mail?
It would be great to get some momentum behind this as our path forward isn’t clear at this point.

Pulling a question out into this thread;

Hi,

A related point I just wanted to check; is spread-spectrum available on the 18-bit rgb bus?

Hello @benp,

I checked with the software team and there is no spread spectrum setting for the RGB bus on the imx8x.
Also, the drive strength setting in the DTS file is already on the low setting.
We will escalate this to R&D to check what we can do.

Bes Regards,

Matthias

1 Like