Page 1 of 2 12 LastLast
Results 1 to 10 of 13

Thread: Help with signal integrity

  1. #1
    Join Date
    Dec 2010
    Location
    SF Bay Area
    Posts
    788
    Post Thanks / Like

    Default Help with signal integrity

    I have received my controller boards back from the fab and populated them for testing. The board seems to work well. However, pixels it is (attempting to) drive are not. I think I've isolated the problem to signal noise between the board and the pixels.

    I am driving 32 channels of WS2811 pixels at the "fast" 800kbps rate. With the bit timing for the self-clocked protocol that translates to a signal frequency of about 1.25MHz. I am using 8 RJ45 jacks for 4 channels each, with the 4 twisted pairs each containing signal and ground - 8 jacks total.

    I'm using Cat5 cable to transmit the signal from the RJ45 jack to a distribution board, where the four channels are separated out from the Cat5 to individual signal wires to each channel.

    The result was somewhat chaotic. Pixels lite, but not in the color that I instructed. it's an entirely new system design, and after ruling out one thing after another, I put a logic analyzer on the signal line to the pixels and saw that there was not a clean signal. The high signal line state seems solid, but when in the low signal state, there is a lot of noise/bounce/crosstalk -- enough that it causes the logic analyzer (sampling at 200MHz) to see transitions to high when it should be low. (The ground lines of the twisted pairs are tied to the pixel power supply).

    I disconnected the pixels and looked at the unloaded output from my board, and it is spot-on. So my controller/logic/firmware seems good.

    I then hacked a short cable of untwisted wire from the RJ45 jack to the pixels and that made things much better. So it seems that my interconnect is the problem**.

    Many others are driving WS2811 pixels for long distances. How should I be doing this? I am using TI 74ACT574 octal FFs as line drivers, rated at 24mA output (max sustainable rated at 50mA). I have 1uF and 0.1uF decoupling capacitors for each. I also realized that I made a mistake and used the minimum trace size on my PCB from the drivers to the RJ45 jacks - 6 mils. I wonder if that may be part of the problem?

    Any help/suggestions on how to make this a reliable design would be greatly appreciated.

    ** I understand how cat5 cable signaling works. It requires a balanced signal which I don't have. I *hoped* that by pairing the signal line with ground in the twisted pair I'd at least not have different signals in one twisted pair. But despite the twist ratio, it looks like I have signal cross-coupling in the cable.
    Last edited by ags0000; 03-31-2018 at 01:43 PM. Reason: fixed incorrect frequency

  2. #2
    Join Date
    Dec 2011
    Location
    UK S80 postcode
    Posts
    1,439
    Post Thanks / Like

    Default Re: Help with signal integrity

    Hi, some thoughts.

    My first, according to the data sheet the 74ACT574 has tristate outputs which I've had problems with in the past. Whilst tristate outputs are fine when enabled, if disabled the pin floats and picks up noise which I found to be enough to imply a logic level. Given that the resting state for WS281x IC’s is data line low I cured this with a 10K pulldown resistor at the controller end.

    Are you using an output resistor in series with the data line? You can still get ringing on twisted pair cables. 50 ohms seems to generally considered optimum but given that the 74ACT574 can only supply 24mA you might want to consider something like 220 ohms. This would prevent more than 24mA being drawn if the output ever became shorted to ground.

    Note: If you fit both pulldown and output resistors the pulldown goes first and the output goes second.

    Let us know how you get on.
    Last edited by Barnabybear; 03-26-2018 at 05:56 AM.

  3. Thanks ags0000 thanked for this post
  4. #3
    Join Date
    Dec 2012
    Location
    Hudson MA
    Posts
    3,913
    Post Thanks / Like

    Default Re: Help with signal integrity

    Most really long distance drivers use a "differential" signal between the transmitter and the receiver. In your design you are still subject to approximately a 20' limit. With differential drivers you can go hundreds of feet. RS485 uses a differential tx/rx pair to get data across large distances.


    2019 - Just moved into a new home (yet another change of plans). Will be dim but not dark. Too much to do at the new place to leave time for a show.
    https://www.youtube.com/channel/UCyX...ttrsZNARkUce0Q

  5. Thanks ags0000 thanked for this post
  6. #4
    Join Date
    Jan 2016
    Location
    Lizella, GA
    Posts
    1,484
    Post Thanks / Like

    Default Re: Help with signal integrity

    Also, While other are able to go 50'-80' using Cat5, they are only sending one output's data over the Cat5 not 4 outputs
    Matt

  7. Thanks ags0000 thanked for this post
  8. #5
    Join Date
    Dec 2012
    Location
    Hudson MA
    Posts
    3,913
    Post Thanks / Like

    Default Re: Help with signal integrity

    Quote Originally Posted by Ruppro View Post
    Also, While other are able to go 50'-80' using Cat5, they are only sending one output's data over the Cat5 not 4 outputs
    AKA Crosstalk


    2019 - Just moved into a new home (yet another change of plans). Will be dim but not dark. Too much to do at the new place to leave time for a show.
    https://www.youtube.com/channel/UCyX...ttrsZNARkUce0Q

  9. #6
    Join Date
    Dec 2010
    Location
    SF Bay Area
    Posts
    788
    Post Thanks / Like

    Default Re: Help with signal integrity

    Quote Originally Posted by MartinMueller2003 View Post
    Most really long distance drivers use a "differential" signal between the transmitter and the receiver. In your design you are still subject to approximately a 20' limit. With differential drivers you can go hundreds of feet. RS485 uses a differential tx/rx pair to get data across large distances.
    Understood - that would be the bullet-proof design. With 32 channels, using RS485 would add substantial cost to the system (if needed, then I'll just have to accept the cost). I am only running about 10-12 feet, and believe others run this length without RS485 or problems - but I'm not sure if anyone is using Cat5 for multiple signals in a single cable.

  10. #7
    Join Date
    Dec 2010
    Location
    SF Bay Area
    Posts
    788
    Post Thanks / Like

    Default Re: Help with signal integrity

    Quote Originally Posted by Barnabybear View Post
    Hi, some thoughts.

    My first, according to the data sheet the 74ACT574 has tristate outputs which I've had problems with in the past. Whilst tristate outputs are fine when enabled, if disabled the pin floats and picks up noise which I found to be enough to imply a logic level. Given that the resting state for WS281x IC’s is data line low I cured this with a 10K pulldown resistor at the controller end.

    Are you using an output resistor in series with the data line? You can still get ringing on twisted pair cables. 50 ohms seems to generally considered optimum but given that the 74ACT574 can only supply 24mA you might want to consider something like 220 ohms. This would prevent more than 24mA being drawn if the output ever became shorted to ground.

    Note: If you fit both pulldown and output resistors the pulldown goes first and the output goes second.

    Let us know how you get on.
    Right - there are two (possible) issues: ringing (on a single line) and crosstalk between lines (or just plain noise). The serial resistor may help with ringing. The example application circuit provided by the manufacturer for the WS2811 calls for a 33Ω resistor, so maybe that would be best for me to try...

    I like the tri-state so I can isolate the drivers from the lines to protect them. Though I've never had a problem damaging the drivers in a previous design, I thought this would be a good idea, and would allow me to work on the pixel lines without having to power-down the controller entirely. I use a power-up sequence that turns the pixel supply on, clears the stored values, then enables the output. The issues I am seeing are with the outputs enabled. Nonetheless, the idea of pulldowns on the outputs is not a bad idea. I'd like to get the noise/ringing issue resolved first before considering that change.

  11. #8
    Join Date
    Dec 2010
    Location
    SF Bay Area
    Posts
    788
    Post Thanks / Like

    Default Re: Help with signal integrity

    Quote Originally Posted by Ruppro View Post
    Also, While other are able to go 50'-80' using Cat5, they are only sending one output's data over the Cat5 not 4 outputs
    So here's the dilemma: I have 8 "banks" of 4 signals, each bank going to a different distribution point, about 10' from the controller. RS485 seems overkill (and expensive) for such a short run. Bundling multiple signals (4) in a single cable (Cat5) may be causing the problem with crosstalk.

    I thought of using IDC w/ ribbon cables, but if this is the problem with the Cat5 approach, I'm not sure if ribbon will be better - even if I separate each signal from the other with a ground wire. I've also never heard of using ribbon cable outside of an enclosure (not that that means it can't be done) and in this application it will be exposed to weather.

    Another alternative may be to run separate cables for each of the signals, but that will mean a much bigger board with 32 individual connectors - and the individual cables will still be running in parallel (though I could try to separate them somewhat).

    Pulling the distribution points into the controller box would make for a big box, but worse will require much more copper (and more voltage drop - using 5v pixels) getting power and signal out to every pixel strand. I suppose I could leave the power distribution in the remote/satellite boxes and have the signal lines all come to the controller... Seems much less elegant that being able to connect both signal and power for each pixel strand with one connection in a single place.

    Isn't anyone running multiple signals between enclosures (exposed to weather) w/out RS485?

  12. #9
    Join Date
    Dec 2012
    Location
    Hudson MA
    Posts
    3,913
    Post Thanks / Like

    Default Re: Help with signal integrity

    rs485 is not as sensitive to cross-talk. FYI the F48 is using differential drivers. You could look at those to see what needs to be done to make it a compact system. You also don't need to use RS485. There are other differential drivers available on the market. Some with pretty good chip density.


    2019 - Just moved into a new home (yet another change of plans). Will be dim but not dark. Too much to do at the new place to leave time for a show.
    https://www.youtube.com/channel/UCyX...ttrsZNARkUce0Q

  13. #10
    Join Date
    Jun 2015
    Posts
    75
    Post Thanks / Like

    Default Re: Help with signal integrity

    Quote Originally Posted by Ruppro View Post
    Also, While other are able to go 50'-80' using Cat5, they are only sending one output's data over the Cat5 not 4 outputs
    I used three pairs on a single CAT5 ~ 50' without any issue on F16v3, controller only provide data, not power.

Page 1 of 2 12 LastLast

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •