The pitfalls of using cheap chinese power supplies

whats better 1000uf or 4700uf, @50v ??
as i noticed some flickering in my incan's, they run off a seprate (non ray) 37v power supply bank
i will turn off the 12v an the 24v transformers i got an see if that solves the fault my end.
 
TimW said:
Might be overkill.... ? :-\ (1000uf I would understand)
I'm going with the radiated emi from proximity to the supply theory .
Click to expand...
If you are wanting to use the electro to reduce the switching noise, choose a Cap with as low ESR as possible, Jaycar Comic stuff won't hack it. You are better off with a Pi network.
I tried the AM radio trick. Its screaming for 30cm from mine! Holy Farady cage batman!
Click to expand...
At 30cm, did you turn the radio around to find the directionality of the emmited noise?

A couple of turns on a ferilte ring is another tool used for making the stuff quite.

BTW in a past life I designed Flybacks and Push Pull PS. PS noise was always one of the more significant time eater in Radiated and Conducted noise testing.
 
kool-lites said:
TimW said:
Might be overkill.... ? :-\ (1000uf I would understand)
I'm going with the radiated emi from proximity to the supply theory .
Click to expand...
If you are wanting to use the electro to reduce the switching noise, choose a Cap with as low ESR as possible, Jaycar Comic stuff won't hack it. You are better off with a Pi network.

Hi Matt - As fate would have it I needed to replace some caps in a monitor power supply the other day. Turns out Jaycar do have a specific line of low ESR hi temp electros. Who would have thought it? I like the ferrite core as a practical solution as well. Lets see how it goes.
Click to expand...
 
There is no easy answer to your question without knowing the freq. of the EMI being caused, you need to test if you have an 'In noise' or an 'Out Noise'. The 240v 50hz wave can be made noisy by the power supply without good filtering on the input. Or if the noise is on the 24v coming out, it is generally at the switching rate of the transistors/fets on the output upwards of 24Khz, which can also be the hardest to remove.

Filters are generally PI filters (with some intricut designs) using both capacitors and inductors to remove noise, capacitors on their own do not clean the signal enough.

digikey and mouser have an extensive range of inline filters available for mains and outputs to clean them up.

Realistically, it pays to spend the extra $20-30 at the time of purchase of the supply.
 
regencyxmas said:
There is no easy answer to your question without knowing the freq. of the EMI being caused, you need to test if you have an 'In noise' or an 'Out Noise'. The 240v 50hz wave can be made noisy by the power supply without good filtering on the input. Or if the noise is on the 24v coming out, it is generally at the switching rate of the transistors/fets on the output upwards of 24Khz, which can also be the hardest to remove.

Filters are generally PI filters (with some intricut designs) using both capacitors and inductors to remove noise, capacitors on their own do not clean the signal enough.

digikey and mouser have an extensive range of inline filters available for mains and outputs to clean them up.

Realistically, it pays to spend the extra $20-30 at the time of purchase of the supply.
Click to expand...

I think the issue is on the output side as i have used a high quality input filter on the supply but this made no difference. The issue looks more to be with the noise caused by the FETs and the location of the power supplies, but ive decided that the best cure is to use good quality power supplies as the cheaper Chinese power supplies let off a lot of EMI around the power supply itself. So Ive gone with the meanwells which have got a great reputation and was the difference with Ryans noise issue last year.

So I'm in agreeance here that it pays to invest in the backbone of your display and purchase decent power supplies
 
One other thing I found and spoke briefly about with a few on chat was the extension leads running from the boards to the lights.

I had a major flickering issue with a Star with 40 x 2801 pixels on it. Had 4 core alarm cable on it for Data and Clock and figure 8 18AWG cable for the power. I tried to remove the flicker by every means possible, checking each pixel, injection power and replacing what was thought to be bad pixels. None of it worked

What did work was removing the extension and going for 6 core alarm cable, 2 wires for pos, 2 for neg and data + clock. Plugged it in and straight away flicker was gone. There was a lot of interference in the other configuration. I noticed this when I moved the cable a few times and things got slightly better. I don't know why this happened but it did.

I am now getting a longer run of cable without loss of signal and less null pixels.

I do however agree that these power supplies are making a fair bit of noise around them.
 
Mike said:
There was a lot of interference in the other configuration. I noticed this when I moved the cable a few times and things got slightly better. I don't know why this happened but it did.
Click to expand...


All power cabling leaks electrical field and this will change in intensity proportional to current draw and cable type. Given the fig8 cable is untwisted the direction of this EM field will be consistent and can push the voltage of your data cable around. The 6 core cable will have an overall twist in it, over distance this should mix up the field direction and reduce the influence it has on data lines.


Also, if you had 4 core cable previously and were only using 2 conductors for data the other 2 conductors could have been floating antennas picking up any stray field and coupling it into your data lines by proximity. You may have had some improvement if you had grounded the spare conductors.


The 6 core cable has power in closer proximity to the data so at first you'd think it should be worse, but maybe you're getting some benefit from distributed capacitance between data and power/ground conductors.. I'm not entirely sure why it worked in your situation but clearly the effect was significant.


I have worked on an industrial RS485 bus spanning a couple hundred metres that was only communicating to 50% of the modules due to 20 centimetres of Fig8 cable at the start of the run (the tech had used it because it was lying around and made his life easier). The cable was coupling in the electrical field from the 240V supply to the building (there was a box behind me that had "240V 200A" written on it!!) and throwing off the bus voltage to the whole system. As soon as I replaced that sneaky bit of Fig8 with twisted pair, I got all the modules online instantly. That site visit has been a constant reminder of the importance of using the correct type of cable for the job! :)


Geoff
 
Unforeunately I have just discovered the same problem with my Ray 12V 30A power supply.

Setup:
* Power Supply mounted inside old PC case, with 8 way fuse block. One fuse block output to ~1m of cable to...
* PIX8AD
* 1 x 6803 Strip directly connected (no extension) = Perfect
* 1 x 6803 Strip with a ~4m extension (4core alarm cable) = OK at full brightness, but major flicking when dimming, worst at ~10% or 80% brightness points.

Test A - Shorten extension to ~2m = the same.
Test B - No extension = perfect
Test C - Power string (from controller end) from a old PC power supply. = the same
Test D - Power entire setup from PC power supply = Perfect.

Really annoying... I spent so much time making a nice power supply case with volt/amp meter for my Ray PSU :(

Another test I'm yet to try is running the PIX8 on it's own power supply, and use the Ray PSU to power the strips.
 
Update: Putting the Pix8 on it's own seperate 5V PSU solved my issues. (Previously it was powered via 12V from S2, along with the lights.

Now to start praying it doesn't happen once i put long lengths up to the roof!! :-X
 
random said:
Also, if you had 4 core cable previously and were only using 2 conductors for data the other 2 conductors could have been floating antennas picking up any stray field and coupling it into your data lines by proximity. You may have had some improvement if you had grounded the spare conductors.
Click to expand...


Geoff,
I am using 16/4 for my data line, but only using 2 conductors. I think I may have the problem you are describing. A 3rd conductor is transporting the ground from the DMX source, but that gets dropped out at the first stop in the chain. It now sounds like I should use legs 3 and 4 of the 16/4 as ground and couple them together at each drop along the way... I have to re-work my controller harnesses anyways, so maybe that is a good test.


Here is the harness as I have it now:
http://dl.dropbox.com/u/32942193/photo.JPG


The black connectors are power in/out, white/green is DMX in, white/red is DMX out, and the top white is SPI+Power out to pixel.


The dd-100 controllers are rated for 8-24v but seem to work when I feed them 5v power shared off my in/out feed. I am having mixed results, so I am considering adding yet another power connector for dedicated 12v power to the dd-100. I hope I don't have to do that, but I will do what it takes to make it work.


Any suggestions here?


Chris
 
Carrying the ground through is a good idea. I assembled my pixel matrix last night and added a longer cable run, after Eddy's issues I used screened audio cable for the SPI data runs and joined the shield to power ground inside the display. My first pixel in the string shows some corruption which reduces when I ground the other end of the cable (I left it ungrounded at first in case it caused a ground loop issue, however it seems to be better with both ends grounded). Bit of a moot point as the controller will sit on the back of the display anyway, but I thought it would be interesting to try and see what length cable run is possible.


If the controller is rated 8-24v it probably has a linear 5V reg in it, and it won't be working properly if you feed 5V in. It really wants about 2-3 volts above the output voltage to do proper regulation. So while it may "seem to work" being fed 5V, you might still experience issues. Since you're running it out of spec by only giving it 5V, giving it a 12V feed would be the first thing I would try.
 
I have just read thru this post and if I am reading correctly, the problem is not related to the power supplies?
What I couldn't figure, and probably just skipped over it, is the conditions to cause the problem....


We have lots of 2801 strings. Some 12 volts 3, 4 & 8 led modules & some 5 volts Single led modules.
What I can say is, if this is the same symptom we have, is that what appears to fine in a 5 volt system, can be fine or not fine in a 12 volt system, depending on the circuit design of the 12 volt boards.
We see this problem specifically when any two colors for any given pixel are at the same value, usually getting worse at values under 50%. The closer to the power supply, the less of a problem it is. But a 5 volt system with a 15 foot lead cable (power and data) is just fine, but the same thing on a 12 volt system creates color "popcorn" or flashing......


I have not been able to figure out how to design around this problem yet, but I will be studying the problem in January this year. I have tried Shielded, Twisted, ribbon cable etc... nothing seems to really fix the problem, on the products that we have a problem with.


What I find MOST perplexing, is that it is related to TWO or Three of colors being at the same level.


Can I propose a test for you?
Try to recreate the problem with a yellow color that is 100% red, 70% green.... If it is the same problem we see, then it should go away, HOWEVER, if your controller uses proportional dimming, then when that color fades to 0, the two channels will approach the same values and cause the problem as it fades out....


Whyrl
 
You must log in or register to reply here.
Back
Top