4
\$\begingroup\$

I’m working on a new PCB design using an Mstar LCD controller (MST9812Q1). I can’t link or post a datasheet as everything has “confidential, for internal use only” plastered all over it.

The controller, among other things, basically takes a HDMI signal and converts it to LVDS to drive an LCD panel.

Mstar reviewed my schematic and told me that I need 0 ohm resistors on the HDMI bus between the connector and the LCD controller to “damp” the signal. Mstar was not very clear as to why they are needed. I’m guessing this is for tuning the impedance should there be a problem or for EMI attenuation?

I’d rather not do this as I want to keep the trace impedance as consistent as possible. I already have a common mode filter with ESD protection (EMI8142MUTAG) on the HDMI bus which adds 6 ohms of channel resistance. http://www.onsemi.com/pub_link/Collateral/EMI8141-D.PDF

Of all the application notes I’ve read on HDMI, none have said to add these 0 R resistors to the HDMI bus, and other HDMI product’s I’ve inspected have not included them either.

I have talked to my PCB manufacturer and gotten recommended trace widths and spacing for a 100 ohm differential impedance.

My questions are

  1. What function are these 0R resistors meant to serve?
  2. Will the EMI8142MUTAG provide enough damping?
  3. If I am careful about my trace impedance, how likely am I to need to add “damping” to the HDMI signal?

Below is the HDMI portion of my schematic with 0 ohm resistors added. 0 ohm resistors on HDMI bus EDIT: I’m not using bus elements for power and ground nodes. They are ‘medium’ sized wires. It is a drawing style I was taught at a previous company and have kept because I feel it makes the power rails clearer to identify in the drawing.

EDIT2: I've just had a response from Mstar (well, not Mstar directly as they are forcing us to go through a Taiwan distributor). "There are for EMI purposes. If you do not care about passing EMI then you can take them off." So, considering I already have common mode filters on the port which also have 6 ohms of channel resistance, I guess I can safely not fit the 0R resistors?

EDIT3: I've just had another response from Mstar. They are pretty insistent about adding the 0R resistors. "Please still need you to put back the 0R. It is not just for EMI and it also for impedance test so we can fine tune it when signal is not so good. This recommend from MStar."

\$\endgroup\$
8
  • 2
    \$\begingroup\$ Extract the answer from Mstar because they said it they should justify it. Anything else is a guess. Be brave. \$\endgroup\$
    – Andy aka
    Commented Aug 25, 2016 at 13:05
  • 2
    \$\begingroup\$ I suspect that their 'review' consisted mostly of comparing your circuit to their reference design & then pointing out where they differ - not much else. \$\endgroup\$
    – brhans
    Commented Aug 25, 2016 at 13:06
  • 3
    \$\begingroup\$ "cargo culting" \$\endgroup\$
    – Daniel
    Commented Aug 25, 2016 at 13:46
  • \$\begingroup\$ I am concerned that you're using bus elements as power and ground nodes in your design. My gut says that shouldn't work in Altium. \$\endgroup\$
    – Daniel
    Commented Aug 25, 2016 at 13:48
  • \$\begingroup\$ @Daniel Just a guess, but judging by the connection dots I think those are just 'medium' sized wires (instead of the default 'small'), not busses. \$\endgroup\$
    – brhans
    Commented Aug 25, 2016 at 14:04

2 Answers 2

5
\$\begingroup\$

In a perfect world, 0 Ohm resistors wouldn't do anything.

In reality, they only have parasitic effects — namely, primarily a parasitic impedance (says Vishay).

Now, there's two different possible reasons the people at MStar recommend using 0R in your signal lines:

  1. The person you've talked to isn't that involved in the design of PCBs or the IC and was under the misconception that where, for example, their demo board had 0R resistors for flexibility, you should have such, too, or
  2. You actually need these parasitic effects.

Option 2. is actually the more interesting one. It clearly indicates that MStar can't or didn't guarantee a good impedance matching for their differential endpoints – otherwise, without any doubt, the "plain" differential transmission line with the characteristic impedance their datasheet claims would be the optimal connection.

So, please do two things: Ask back; be polite. Engineers and support people are human, too! Also, report back. Having a non-perfectly matched real impedance isn't terrible, but it of course calls for knowledge of that fact if you want to build a reliable device.

It's perfectly possible that MStar has experience we don't have – for example, it might be that HDMI sources typically don't adhere to spec themselves, or having a bit of an inductance to compensate for capacity effects in HDMI cabling etc.

EDIT: you received a reply from MStar's distributore:

Please still need you to put back the 0R. It is not just for EMI and it also for impedance test so we can fine tune it when signal is not so good. This recommend from MStar.

In other words, they tell you through the flower "please don't trust your design skills and/or the specs of our components overly much; it's usual to exchange the 0R for something that compensates for factual mismatch later on".

Which frankly, isn't that bad of advice (considering 0R isn't really expensive, and placing a couple of SMD resistors in an automated process isn't either), but also means that

  1. specs are to be handled with a grain of distrust and
  2. you can't use resistor networks (like the ones I infer from your schematic), but should use individual resistors so that you can correct the individual differential pairs if necessary.
\$\endgroup\$
8
  • \$\begingroup\$ I've just had a response from Mstar (well, not Mstar directly as they are forcing us to go through a Taiwan distributor). "There are for EMI purposes. If you do not care about passing EMI then you can take them off." So, considering I already have common mode filters on the port which also have 6 ohms of channel resistance, I guess I can safely not fit the 0R resistors? \$\endgroup\$
    – philby
    Commented Aug 28, 2016 at 22:39
  • 1
    \$\begingroup\$ "EMI Purposes" is about as specific as "They are there for, you know, reasons!"... \$\endgroup\$ Commented Aug 29, 2016 at 7:04
  • 1
    \$\begingroup\$ I just had another reply from Mstar. They are pretty insistent about adding the 0R resistors. "Please still need you to put back the 0R. It is not just for EMI and it also for impedance test so we can fine tune it when signal is not so good. This recommend from MStar." \$\endgroup\$
    – philby
    Commented Aug 30, 2016 at 2:17
  • 3
    \$\begingroup\$ AHA! So they are intended to be stand-ins for non-zero resistors if the design impedances turn out to be inadequate. \$\endgroup\$
    – Daniel
    Commented Aug 30, 2016 at 3:22
  • 1
    \$\begingroup\$ Look how much engineering and customer service time they just put in because od crappy documentation! \$\endgroup\$ Commented Aug 30, 2016 at 11:12
0
\$\begingroup\$

If I am careful about my trace impedance, how likely am I to need to add “damping” to the HDMI signal?

The trace impedance won't do much if the receiver chip is mismatched due to whatever reason. Such resistors are almost always added for tweaking purposes, MStar's advice is solid in that they are not joking about it, but that doesn't mean that there's some magic to these resistors. They are purely CYA (cover your ass) and strongly suggests that their typical customer support process outcome doesn't identify root issues, only workarounds. This may well be because their average customer can't effectively participate in root issue identification, but it may well say something about the usefulness of the support process.

When the chips are in a particularly hot and price competitive market like HDMI LCD panel receivers, it's not unusual that they'll be rushed and everyone will fine-tune things only to pass EMC and/or HDMI compliance and call it a day - no single person may have any clue about why things got "good enough". If some test lab happens to have a good engineer, they may be the only ones with a full picture in their mind of exactly what went on. It depends on the particular lab, some lab "chains" are not the best at retaining talent, and the market seems to place no demand for them to have the talent. So, in a short time, engineering turns into magical thinking, people use hacks that may be completely inapplicable anymore, etc. It can be a real mess. The customers who implement the chips in their designs are often a big source of trouble as well.

When the product is <$100 delivered to US online stores, nobody will be putting in the time to really figure it out unless they got a fire burning under their seat. They'd go out of business in a split second otherwise.

That's fairly par for the course when it comes to such markets. It's a rat race. Silicon gets released as long as anyone willing says "good enough", and often e.g. HDMI compliance pre-tests are taken to mean that the chip is "officially" OK whereas all it may mean that this particular chip in that particular test fixture with that particular cable happens to work. People often take some "official" result, like one furnished by a test lab, to mean much more than it really means, and try to use a test lab success to shed responsibility (our chip is compliant! our chip is compliant! nananana, we won't listen, our chip is compliant!). That's how that goes all too often in real life.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.