Noise in the DVI Output of ADV7511W

Hi GuenterL,

We have few more details to share regarding the observation.

The complete system block diagram is attached. It is summarized below:

1. There are 2 backplanes in a rack.

2. Backplane 1 has Power circuits and Backplane 2 has 14 Cards.

3. VGA IN and DVI IN is given on front panel of card 14.

4. The Card 14 has FPGA as controller (ALTERA ARRIA V).

5. The DVI OUT from FPGA are connected to Backplane.

6. DVI OUT signals are connected to the DVI monitor through Cables which runs through couple of Circular connector.

7. The schematics of VGA IN/DVI IN and DVI OUT circuits in Card 14 is also attached.

We observe the static sparkling white dots only on certain portion of the monitor.They do not change their positions after restart or reprogram of the system.

The visibility of dots varies with intensity of Primary colors (Red, Green and Blue). With the dark intensity, we observe very less dots. If we reduce the intensity of the color (for example shaded RED color, Shaded green color), more sparkling white dots are observed. Can you suggest us how can we eliminate 100% of dots in all color combinations?

There was a change in the observation when we experimented with the CLOCK signal between FPGA and ADV7511WBSWZ (that is signal FPGA_DVI_OUT1_CLK in schematics).

The number and intensity of dots were reduced when CLOCK INVERSION is performed for clock FPGA_DVI_OUT1_CLK.

We took out the Card 14 from the system and connected to a plane Test JIG that has only DVI OUT connector, Backplane connector and POWER circuits. We gave VGA IN to Front panel of Card 14 and observed DVI OUT from DVI connector of Test Jig. We do not see any dot or noise in this case.

Let us know how this issue can be addressed and improved.

If you suspect on Noise issue at the analog side, then please suggest how it can be removed/reduced using any standard filters

If you suspect on digital side, then what are the possible corrective methods?

Thanks & Regards,

Nanjunda M

I'm glad this is a simple design

1) Lets look at the FPGA to ADV7511 connection first.

     a) Try, if possible, different FPGA drive strengths.  Our ADV7842 has 3 different levels which we sometimes have to tweak depending on the board. This allows the drivers to more closely match the board impedance.

     b) Try, if possible, different clock to data phase relationships.  The ADV7842 has 16 different phase clock to data setting.  Depending on the board we've had to tweak the phase delay to match properly.  See https://ez.analog.com/message/89211#comment-89211

2)  I assume the ADV7441 and ADV7511 power decoupling is clean and done similar to the reference designs.  Noise can be injected due to weak/poor decoupling.

3) Since the noise appears to be static, if the noise was on the input to analog section you should be able to scope it out and see it on the analog input signals.

4) Is there anyway you can go directly from card 14 to a monitor and bypass the cables.  I'd like to totally eliminate the 2.5 meter run for testing.

5) Is there away to ground out the analog inputs on the cable and just have the FPGA generate a solid color?  I want to see if the analog input is coupling noise into the TMDS signals or vice versa.

6) What format are you running at.  I looked at the cable data sheet and it does not specify the max DVI format it is spec'd for.  inter-pair skew looks kind of high but depends on the format timing.

Hi,

Thank You for the reply.

With respect to the points 1.a and 1.b, we are checking with our software team, will get back to you soon.

2) Decoupling is clean and designed as per reference design.

3) We will connect the DVI Input directly to the Monitors and observe the DVI OUTPUT once. Later we will update the observations.

4) We will check for connection of DVI OUT of Card 14 directly to DVI Monitor. Later we will update the observations.

5) We had done this before. We disconnected the DVI/VGA Input source from the card 14. later generated solid color patterns from FPGA (3 basic colors : Red, Green and Blue). No noise was observed on the DVI display.

6) We are running SXGA(1280x1024) @60Hz. The 1.5 meter cables used in between the circular connectors are individual single ended cables and are not twisted. They are also not matched in length properly. Because of this reason can we expect any such noise patterns on DVI display ?

There are some length difference of around 10 to 50 mm within the differential pairs in the cable set of 1.5 meter between 2 circular connector.

We also observed the cable lengths between the differential pairs. Whether this type of length mismatches cause any similar noise effects ?

Regards,

Nanjunda M

SXGA60 has a clock rate of 108MHz.  If the intra-pair skews are 50mm different, this could cause a problem.  From the math from here, HDMI Layout Guideline , I calculate intra-pair skew should be less then 20mm.  This can cause what appears to be noise depending on the data channel to clock channel length ratios.

Hi,

Thanks for your feedback.

We took out the card from the system and made a test jig as shown in the attached figure.

We had probed and captured the POWER, INPUT DATA LINES, INPUT CLOCK and OUTPUT lines of ADV7511 in Card 14. Everything looks very clean and fine.

Later when we Touch the DVI OUT Data line Tx1+ and Tx1- only on corresponding  pins of connector P2 in card 14 through finger , we observed that all the noise on DVI display were cleared. We observe not a single dot on the display when we touch the Tx1+ and Tx1- pins of DVI OUT interface on P2 connector, if we remove the finger the dots re appear on the screen. Same observation was found if we touch the wires of Tx1+ and Tx1- in the DVI Cable set connected to Rare side of P2 connector in the backplane.

We assumed that the Tx1+ and 1- is picking up the noise at the P2 connector side of card 14, we tried connecting a common mode choke (90 OHM) in series with the Tx1+ and Tx1- (by cutting the traces on backplane card). But no improvement was found.

In card 14, there is an RS422 differential pair routed in between the pins for Tx1+ and Tx1- of P2 . Assuming that these lines causing noise over the DVI signals, we disabled the RS422 IC. We observed improvement in the noise level (the number of dots were reduced).

Finally, we covered the DVI Cables (that is connected to Rare side of P2 as shown in figure) with Copper tapes. All the noisy dots disappeared from the DVI display. This is how the issue is currently solved. We are not sure whether we will be able to reduce the noise in the system by the same procedure (by shielding the cables properly with copper tapes). We will get back to you very soon with the result after checking in the final system.

Regards,

Vijetha

Finger touching a trace adds a small amount of capacitance to the line.  This capacitance maybe just enough to filter off any noise that is getting injected into the line.  Since the copper tape works this implies the noise source is electrostatic in nature, not magnet coupling.  I'd look for high dv/dt signals running close to the cable. 

Hi,

We integrated the card in the system. The block diagram explains the situations after integrating the card 14 in the system.

I have added block diagram of two situation.

• Discontinuity at one location on external cable: Noise reduces completely
• Discontinuity at two location on external cable: Noise still remains.

Please can you look at the block diagram and comment on observation on noise.

As per your below reply, touching pins add small capacitance.. can adding high frequency capacitor, btn pins will be helpful?

Any suggestion on attending this type of noise?

Regards,

Vijetha

Twisted pairs like this will introduce impedance mis-match and allow a point for noise to couple in.  I once had a system where a bad ballast in the overhead fluorescent light was injecting noise into the system.  (i.e. The system would only work in the dark)

The twisted pair probably does not maintain the 100 Ohm differential spec'd for HDMI. 

Another possible noise source is a ground loop.  If the cable has a shield, try disconnecting the shield at one end or the other.  This assumes there is still a ground wire in the cable.

Also is the P2/J2 and RTM/RTM connector rated for 1GHZ 100 Ohm differential?

Hi,

In the above system, for situation 2,Please let me know whether placing a repeater ( for ex: MAX3815A) near DVI connector in RTM card ( RTM J2 in above fig) solve the problem.

Regards

Madhu