Author Topic: Flickering!!!!! Help....  (Read 2923 times)

Offline gadgetsmith

  • Hero Member
  • *****
  • Join Date: Apr 2014
  • Location: Charlton, NY
  • Posts: 989
  • Kudos: 14
Re: Flickering!!!!! Help....
« Reply #30 on: December 05, 2016, 11:44:13 AM »
Have you tried using only one power supply? If it's a 350w supply, one should be plenty for 800 pixels. My thinking is their might be a ground noise issue.

Sometimes only using injection of the + has solved some issues like this for me, but in most cases i do have + and - at every injection point.

Offline FIRECOP252

  • Full Member
  • ***
  • Join Date: May 2014
  • Location: SIKESTON
  • Posts: 200
  • Kudos: 1
Re: Flickering!!!!! Help....
« Reply #31 on: December 05, 2016, 06:54:52 PM »
Have you tried using only one power supply? If it's a 350w supply, one should be plenty for 800 pixels. My thinking is their might be a ground noise issue.

Sometimes only using injection of the + has solved some issues like this for me, but in most cases i do have + and - at every injection point.

Yes I did try using 1 power supply and got the same results but switched back because with my additions I was worried to overwork just the 1 power supply.  Thanks for the info will check into the ground noise issue also.
LANCE

Offline FIRECOP252

  • Full Member
  • ***
  • Join Date: May 2014
  • Location: SIKESTON
  • Posts: 200
  • Kudos: 1
Re: Flickering!!!!! Help....
« Reply #32 on: December 05, 2016, 07:12:18 PM »
Some random thoughts FWIW...

As this is a megatree, 'data distance' should not be a huge issue provided the nodes are laid out smartly and the driving equipment is in or near the base of the tree.

Ideally, you would want plenty of grounds / power injection on decent gauge wire with relatively small runs.  The seemingly best set-up would be a control / power box at the base of the tree with equal wires running out to all points.  The key is that you want balance between the strings and controller.  You don't want any string or power injection point to be an extremely long run, or even notably longer than the rest.  You also don't want one power supply running ahead (or behind) the other voltage wise.

Have you tried a test sequence?  Either make one, or put the Pi in test mode (if applicable - honestly, I don't know much about that specific driver) The best sequence to me would be fade in / fade out each separate color and then fade in / fade out white.  Some sub-observations of that:

(further note I will use 'tend to' and 'seems to' a lot, as there can be an exception for almost every case)

Issues with red pixels seem to be associated with over voltage.  Red is generally the lowest voltage LED, so possibly more sensitive to input voltage 'too high' compared to green and blue LEDs.

White lock-up tends to be a ground issue.  Once all pixels switch on white, the power line droops to the lowest voltage level and ground line is loaded to the highest voltage level, while the high impedance data line remains relatively stable.  The ground line floating up can make the node unable to interpret data which means it stays locked on white.

Rainbows and flickering tend to be corrupt data, though it is unclear why some sequences work for you and some don't. I tried to eyeball the tree and estimate power consumption (patterns like 'snowflake' are generally pretty low power as most of the tree is black with only a few intermittent nodes lit, though other patterns which turn most nodes on bright and white or nearly so, can draw the most power)  Again, this power draw can expose imbalances in power distribution, high resistance joints, etc which can make the nodes behave unexpectedly.

Lastly, I would take a close look at your 'butterfly' effect if that is giving the trouble.  Pausing your video around the 19-20 second mark, I suspect this is the butterfly.  There is a very distinct section of the tree which is out, but at the bottom of that section, you seem to have 2-3 nodes on each strand which are working correctly.  (possibly 2-3 at the top too?)  I'm taking those nodes to be a couple at the beginning of one 100 node section and a couple at the end of the same section.  (and if the top 2-3 are OK, that would be the middle 4 -6 nodes)  Bad data would typically take out the whole string, not exactly the 'middle 96' nodes of multiple strings.  So it is almost as if that is what the tree is being told to do...possibly some set-up option incorrect, or some obscure little thing causing exactly that space to blank out.  It's also curious that the entire tree seems to flash red / green a few times.  Again, corrupt data tends to be really random, it would be very unexpected for 800 nodes, 2400 channels to ALL corrupt to a single color.

Anyway, the RGBW test pattern should narrow things down ...power, ground or data issue based on what you see.  Or if the test mode works fine, then it almost has to be a configuration issue.

Good Luck

Corey,

I kind of did the idea of a control box with everything in one big box including both PS's the Pi and the distro boards at the center pole of a 180 tree.  I used 18ga power/ground wire and each run is 8' from the PS.  I am using 14ga wire from the PS to the distro boards and the 18 ga from there out to the pixels. 

I have put the Pi in test mode and single colors such as R/G/B no issues start doing mix colors like purple or white or other mixes and the issue arises.  As far as white lock up I am not sure but if thats the case I made need a ground for each of the 50 ct strings. 

As for the 19-20 second mark that is actually a marquee with a strobe effect in the middle which is why the outer pixels are working.  If you look back a bit further I have 2 other videos that I posted and one of them has the butterfly effect in it and it shows how bad it is flickering. the strobe is the reason for the almost all red or green when I have the tree strobe inside the marquee.

I also got to talking to a friend of mine who has a very similiar set up in a matrix and is not having any issues but he told me I might try turning down the brightness of my pixels to cut power consumption so I went in and cut it down to 40% and this took almost all of my flicker out.  I could even run the butterfly effect with no flicker.  With this its telling me I must need more grounds than i thought was good.  Would you guys agree with this?  that since I was able to turn the power down it took most out that it might be the ground issue?  Also I think my pixels have 20ga wire instead of 18ga  this might be a consideration also. 

Thanks for the info!

Offline corey872

  • Supporting Member
  • ******
  • Join Date: Oct 2013
  • Location:
  • Posts: 490
  • Kudos: 16
    • Link to the sale...
Re: Flickering!!!!! Help....
« Reply #33 on: December 05, 2016, 09:02:57 PM »
If you suspect the power side, the big keys are proper voltage, solid ground and having everything equal / balanced (especially with two separate power supplies.

For proper voltage, some have reported ~11.5-11.7V seems to be good and a solid 12.0V causes issues.  I've not noticed that with my lights, but every year, every node type, every manufacturer can sometimes be different.  If your nodes are sensitive to being right at 12V, being in a megatree may actually hurt there as you only have very short power runs to the strings and likely minimal voltage drop.  If they were on an eve or similar, you might have 20, 30 or more feet of lead wire which would provide some natural voltage drop.

For solid grounds, the node can only interpret data with respect to what it sees as 'local ground'.  There is no real 'power' on the data line, so that signal is usually a solid 0-5V square pulse.  But bright colors, white, etc dump a lot of power on the ground line and it may float up to 1V or more at the node.  (Just the same way power may be 12V at the supply and droop to 11V at the nodes, the ground line might be 1V at the nodes and drop to 0V at the supply)  If the ground at the node is floating at 1V, then the node interprets the data signal as a -1 to 4V square pulse.  At a 1.5V floating ground, the node would see a -1.5 to 3.5V pulse and that is where issues would start as the nodes are generally spec'd to need a 3.5V data signal minimum.

For balance - a bit hard to say, there.  I've never heard great things about paralleling power supplies...some claim to get away with it, others claim the world would end, your teeth rot out and you go blind...and that is before you even plug them in.  The easiest test I can think of would be to hook a volt meter to a solid ground on your Pi (where the data is generated and referenced), then start probing other grounds at the nodes, other power supply, etc. (this would be with the tree lit using a 'problem' sequence)  If you see any more than ~1 volt across the grounds, that is likely too much.

Lastly, just some back of envelope scratching, 800 nodes x 60mA per node (all full on white) = 48 amps, 48A x 12V = 576 watts.  2x 350W supplies 'should' cover that - assuming these are straight 350W / 12V supplies.  If they are 350W 'computer' / ATX supplies, they might only have 10-20A available on the 12V rail and that might not be enough to cover the very bright / full on sequences.

Apologize in advance if this has been detailed in prior posts - I scanned through them fairly quickly.

Corey
 2018 uSC, Afterburner, uAmp co-op - pending May-June 2018.  Remaining boards are now FOR SALE

Offline taybrynn

  • Sr. Member
  • ****
  • Join Date: Sep 2013
  • Location: Castle Rock, CO
  • Posts: 364
  • Kudos: 8
    • RockinChristmas Facebook Page
Re: Flickering!!!!! Help....
« Reply #34 on: December 12, 2016, 01:26:59 PM »
Wow, I just tried lowering my tm1804 port to 40% on a longer 30' run that flickered on white and the flicker disappeared on 40% brightness !  The rest are at 75% with no flicker.

 

Back to top