HDMI Buffer

[jbaum81]

Hey guys,

I have a computer project that utilizes an Artix 7 based FPGA Dev board (Mercury II) running a VDP directly driving the TMDS lines for the HDMI Signal. It works great, however I did not account for the voltage that is getting back fed into my board due to the pullups on the HDMI line. I figured a buffer of sorts should solve the issue so I set out and found this https://www.mouser.com/ProductDetail/NXP-Semiconductors/PTN3363BSMP?qs=kzneCeLdMY5ikkzAy7x%252BKQ%3D%3D

I've attached my circuit for reference. On the FPGA side I see normal signals, on the other side of the decoupling caps I see the analog signal, but the chip outputs nothing. If anyone has any advice I'd really appreciate it, thanks!

[niconiconi]

I don't have any experience with HDMI but let's start with common sense.

Do you see even any sign of life from this chip at all? In particular, its the datasheet, "Section 7.7.2 Read operation" suggests that you can send I2C commands to the chip to read its ROM. Does this chip respond to your command? If you haven't tried yet, better to give it a try and see whether you get a response. If you get one, the chip is at least up and running. If there's no response, you have an electrical issue in its external supporting circuit or a soldering issue.

Other times, defective solder joints can also cause mysterious problems. How is the board assembled? Hand soldering gives inconsistent result, reflow soldering with stenciled solder paste works better but occasional defects are not uncommon, finally some PCB prototype services now offer in-house SMT assembly. QFN/BGA is especially difficult to troubleshoot as the joints cannot be inspected or probed. A microscope helps a bit, but it cannot distinguish working-but-bad-looking and non-working joints. One trick to check a suspected I/O pin fault is isolating the chip from all external circuitry, and perform a diode measurement to VCC with a multimeter. If the I/O pin's solder joint is not defective, you should see the chip's ESD protection diode drop.

Another possibility is that the external circuitry contains a mistake. Compare your schematic with the reference design and check each pin one by one. Since I have no experience I have no comment (I took a quick look at your current-setting reference resistor and all the output enable inputs, but didn't find any obvious problem).

Bringing up ASICs is often a frustrating experience, it's not working because the external circuitry is likely incorrect, but there's no feedback and all you have is a dead chip. But sooner or later it'll be alive after you have exhaustively checked everything.

P.S: Clock and reset/power sequencing are often the cause of problems, but in this case, they cannot be. You better looking at elsewhere. As the datasheet says, this chip "does not retime any data. It contains no state machines. No inputs or outputs of the device are latched or clocked. Because the PTN3363 acts as a transparent level shifter, no reset is required".

[wasedadoc]

1.  From the first para of the datasheet and with my emphasis:

"PTN3363 is a low power, high-speed level shifter device which converts four lanes of
low-swing AC-coupled differential input signals to DVI v1.0 and HDMI v1.4b compliant
open-drain current-steering differential output ..."

2.  By connecting the HIZ_EN pin to zero volts you are enabling the internal 50 ohm terms and thereby double terminating the input signals.

3.  Granted calling them so will not cause the circuit to not function but C1 to C8 are not "decoupling capacitors".

[niconiconi]

1.  From the first para of the datasheet and with my emphasis:

"PTN3363 is a low power, high-speed level shifter device which converts four lanes of
low-swing AC-coupled differential input signals to DVI v1.0 and HDMI v1.4b compliant
open-drain current-steering differential output ..."

Aha, it's probably the problem.

2.  Granted calling them so will not cause the circuit to not function but C1 to C8 are not "decoupling capacitors".

+1. The OP clearly meant to say "AC coupling capacitors".

[wasedadoc]

Also see the "double term" problem added in my edit above.

[niconiconi]

1.  From the first para of the datasheet and with my emphasis:

"PTN3363 is a low power, high-speed level shifter device which converts four lanes of
low-swing AC-coupled differential input signals to DVI v1.0 and HDMI v1.4b compliant
open-drain current-steering differential output ..."

Aha, it's probably the problem.

I checked the HDMI spec (again, no HDMI experience), and it's clear that in HDMI's signaling scheme, the pull-up is provided by the far side. So no pull-up resistors are required at the transmitter. But sure, if you are probing it with no load, you won't see anything. The datasheet shows the use of 50-ohm pull-up at the far-end of the coax cable for jitter measurement.

https://engineering.purdue.edu/ece477/Archive/2012/Spring/S12-Grp10/Datasheets/CEC_HDMI_Specification.pdf

[niconiconi]

Also see the "double term" problem added in my edit above.

  This must be the problem.

If HIZ_EN is LOW, all input pull-up resistors should be removed.

The datasheet says,

50-ohm termination to VRX(bias) if HIZ_EN = LOW.
>100 k-ohm termination to VRX(bias) if HIZ_EN = HIGH.

And the block diagram clearly shows the internal pull-up termination.

Hopefully the OP didn't damage the chip's internal bias circuitry during this bus conflict...Damage should be impossible, as the pull-up are installed at the left side of the AC coupling capacitor.

[wraper]

I checked the HDMI spec (again, no HDMI experience), and it's clear that in HDMI's signaling scheme, the pull-up is provided by the far side. So no pull-up resistors are required at the transmitter. But sure, if you are probing it with no load, you won't see anything. The datasheet shows the use of 50-ohm pull-up at the far-end of the coax cable for jitter measurement.

https://engineering.purdue.edu/ece477/Archive/2012/Spring/S12-Grp10/Datasheets/CEC_HDMI_Specification.pdf

It depends on which HDMI spec. Since version 2.0 it requires termination resistors on the source side too.

[wasedadoc]

Hopefully the OP didn't damage the chip's internal bias circuitry during this bus conflict...

The external input terminating resistors are on the other side of the coupling caps from the internal ones.  The external ones have no impact on bias and cannot cause any damage.  However the double term will reduce the input signal swing.

[niconiconi]

Also see the "double term" problem added in my edit above.

  This must be the problem.

If HIZ_EN is LOW, all input pull-up resistors should be removed.

Wait a minute... Not that simple.

The chip can be used in two ways.

First, if the transmitter has a push-pull output, the circuit is wired according to datasheet Diagram 1. "Typical HDMI/DVI level shifter application system diagram". You only use AC coupling capacitors between the transmitter and the buffer with no pull-up resistors. In this case, HIZ_EN should be connected to LOW to enable on-die termination, allowing the chip rebiases itself using the internal 50-ohm pull-up/termination (see datasheet 7.1.4 TMDS high input impedance termination).

Alternatively, if the transmitter has an open-drain output, the circuit is wired according to datasheet Diagram 2. "Typical HDMI redriver application system diagram" At the left side of the AC coupling capacitor, 50-ohm pull-up / termination to 3.3 V is used to allow the HDMI transmitter to function as its designers intended. In this case, HIZ_EN should be connected to HIGH to disable on-die termination, as the signal has already been 50-ohm terminated externally.

It's up to the OP to decide which case it is. But since an FPGA is used, my guess is case 1. So, set HIZ_EN to low, and desolder all pull-up resistors. Alternatively, set HIZ_EN to high, and keep all pull-up resistors soldered. If you're free to choose open-drain/push-pull on the FPGA, case 1 is better as it uses fewer parts with on-die termination, and has better signal integrity because there's no impedance discontinuities from the soldering pads (delete them from PCB layout).

[wasedadoc]

"biasing", "rebiasing" etc is not the correct terminology.  The coupling capacitors are present in both configurations.  The internal or external resistors are only for terminating the high frequency signals.  Nothing to do with "biasing" the receivers.

[niconiconi]

The internal or external resistors are only for terminating the high frequency signals.  Nothing to do with "biasing" the receivers. Nothing to do with "biasing" the receivers.

You're right. AC coupling is always used and the receiver always provides its own bias voltage in both cases. if on-die termination is used, the 50-ohm resistors serve double duties, as both a connection to the bias voltage, and as an RF termination. If terminated externally, the receiver is still connected to the bias voltage via >100kΩ resistance, only without RF termination. I stand corrected.