Flickering w/ LCD running off 12v rail - help!

[Linus]

In an attempt to lower power consumption, I connected my 17" widescreen LCD to the (unused) 12V motherboard connection. It works, but not without flaws - there is a minor but very noticeable flicker in horizontal lines across the screen. After a little experimentation, I've discovered the following:

Affects the flickering:
Color on screen - solid black and white don't show the flicker but everything else does
Brightness setting - intensity of flickering varies from unnoticeable at 100% to very noticeable at 0%

Doesn't affect the flickering:
Voltage of 12v rail - as reported in 8rdavcore, voltage varies from 11.977 to 12.768v, depending on how much I overclock
Several layers of aluminum foil around connections inside computer

I'd rather not run it at 100% brightness all the time because it's too bright at night and it takes away any power consumption gains I might have made. Anyone know what causes this flickering or other things I could try to fix it?

See this thread for background on my system: Small & Flexible: Back-of-LCD variable power consumption

More details on what I did: I measured the power produced by the LCD's stock brick at ~12.07v. Snipped the cable between the brick and the LCD connector (there were cylindrical shielding things on both ends of this cable, could this be related to the problem?) and put a female molex on it, being sure to observe proper polarity. Replaced the motherboard connector on the PSU (my motherboard doesn't need one) with a female molex. Every connection was soldered and covered with electrical tape or heat-shrink tubing.

[Devonavar]

Chances are you're seeing ripple on your +12V line (high frequency variance that doesn't show up unless you have special tools). 8rdavcore is useless for measuring the +12V rail; it measures the voltage after it's passed through the motherboard. It won't tell you anything reliable about the PSU itself. To get proper voltage measurements you need a multimeter, and to measure ripple you need a scope.

I can't think of too much in the way of solutions. The proper electronic procedure would be to insert an appropriately sized capacitor into the circuit, but that's going a bit beyond most people's DIY skills.

[Linus]

I wondered if a capacitor might do it. I assume the desired capacitance would depend on the frequency of the ripple, though, and I don't have access to a fancy multimeter or oscilloscope.

I had enough DIY skills to solder on molex connectors, so I assume I could handle capacitors. Could I just try different capacitances and see if anything helps? Would anything endanger my LCD? If this might work, how would the capacitor need to be wired and what values might I start with?

Sorry for all the questions, but I'd really like to get this licked.

[Elixer]

The mostly likely reason this is happening is due to voltage ripple in your power supply. ATX specs up to 120mV of ripple on the 12V line from your power supply. Your LCD probably needs to recieve power with much less ripple. Yes a capacitor may help but my guess is that you'll have to settle with using your included LCD power supply. The power supply already has a capacitor on the 12V line and with all the noise from the computer load it's going to look messy pretty much no matter what you do.

If you still want to insall a capacitor it should be safe, and what you're looking for is a capacitor with as large of a capacitance (in uF) as possible. Make sure this capacitor is rated for more than 12V and be absolutely sure you get the polarity correct when you install it.

[GOATMAN!!!]

I'd get a voltage regulator too.
Here's a little diagram I whipped up for you. It isn't too complex.


The voltage regulator will have a code something like 7812 or LM340T12. Make sure you get one that will take enough current for your monitor, allthough it wont hopefully have to do too much work. If you are worried, you can buy clip on heatsinks for them.

Electrolytic capacitors are dirt cheap, and the voltage reg would set me back a couple of dollars AUD at the most, so it wont be too expensive. To make it easier to layout, get some veroboard too.

I just realised, on that diagram, I didn't mark it, but the right hand pin of the voltage reg will be the output pin, and the pin that goes to the earth is marked as common or ground. The left hand pin will be the input pin. Put it this way, as long as it is a 7812/LM340T12series, the heatsinky bit will be at the top of the diagram.

Hope that helps. After that, you shouldn't get the slightest ripple.

[Elixer]

I'd get a voltage regulator too.
Here's a little diagram I whipped up for you. It isn't too complex.


The voltage regulator will have a code something like 7812 or LM340T12. Make sure you get one that will take enough current for your monitor, allthough it wont hopefully have to do too much work. If you are worried, you can buy clip on heatsinks for them.

Electrolytic capacitors are dirt cheap, and the voltage reg would set me back a couple of dollars AUD at the most, so it wont be too expensive. To make it easier to layout, get some veroboard too.

I just realised, on that diagram, I didn't mark it, but the right hand pin of the voltage reg will be the output pin, and the pin that goes to the earth is marked as common or ground. The left hand pin will be the input pin. Put it this way, as long as it is a 7812/LM340T12series, the heatsinky bit will be at the top of the diagram.

Hope that helps. After that, you shouldn't get the slightest ripple.

I don't mean to burst your bubble, but this won't work. The 7812 needs at least 14V input to reliably supply 12V. You could put in a 7810 and get a almost noise free 10V line coming out, but that wouldn't do you much good. There is always a voltage drop across the regulator. I know the 78xx series has a minimum drop voltage of 2v to function properly and the LM340 needs ~2.5v. With your current design you'll get less than 12V out and the only noise reduction will be from the capacitors. This design would work if you had a higher input voltage or if you needed a lower output voltage but is ineffective in its current state.

[GOATMAN!!!]

Lol, I didn't really think about that, but it doesn't need quite 14v. I've had decent results @ around 2A with a heatsinked 7812 coming off a dirty sub 13v transformer . But yeah, you would need more. Hmm, perhaps you would get enough off going between the +3.3 and -12v. However, I don;t have much experience with PSU's and am not sure if this is possible.

BTW, you aint bursting my own bubble. It aint my lcd thats not going to work...

[Elixer]

Lol, I didn't really think about that, but it doesn't need quite 14v. I've had decent results @ around 2A with a heatsinked 7812 coming off a dirty sub 13v transformer . But yeah, you would need more. Hmm, perhaps you would get enough off going between the +3.3 and -12v. However, I don;t have much experience with PSU's and am not sure if this is possible.

BTW, you aint bursting my own bubble. It aint my lcd thats not going to work...

hahaha, yeah they'll work a little lower, but the manufactorer specs them 2V higher, that's what I was basing my comment of off. I too have ran them a little lower with some results, a 7806 with 7 volts, but the output voltage dropped as more current was pulled, had some interesting effects on the audio amp pulling from the supply when the volume was really turned up. The +3.3 volt and -12 idea would work if enough current was available on the -12 volt line but I'm thinking a 17" LCD is going to pull 3A+ and I believe the -12 volt line is only rated for 1A in the ATX power supply specification. There might be enough current if the power supply had a -5 volt line (though few new power supplies have them). You could go from +12V to -5V. I don't know if the -5 volt line has enough current though. You'd probably want at least 4A available. I say just use to stock adapter and just think about how much power is saved over using a CRT monitor.

[Linus]

Thanks for the ideas. I'm definitely not looking to get too complicated here, especially by building in a regulator that would probably lower the efficiency.

Is the stock brick a regulated power supply?

If so, another thing I may try is buying a good 12V regulated power supply and switching to a DC-DC board inside the PC.

[Linus]

I'm going to go ahead and try the simplest & cheapest (capacitor) solution. I assume the positive end of the capacitor wants to go to the 12v off the PSU; does the negative end want to go to the LCD's 12v input or to the ground?

[Elixer]

I'm going to go ahead and try the simplest & cheapest (capacitor) solution. I assume the positive end of the capacitor wants to go to the 12v off the PSU; does the negative end want to go to the LCD's 12v input or to the ground?

The capacitor should be hooked in parallel with the power supply. So the plus side of the capacitor should go to +12 and the minus to the ground of the power supply. Then the +12 to the positive input on the lcd and ground to the minus input of the lcd as you had done before.

[Mr Evil]

Even simpler than a capacitor would be to wrap the +12V wire around a ferrite torroid. This creates an inductor, which impedes AC.

[Linus]

What's a ferrite torroid? Are those the shielding "bulges" on either end of some cables?

[Mr Evil]

They are pretty much the same thing, yes. The bulges you see on some cables just clip over the cable, whereas a toroid (please excuse my spelling errors - only one r in toroid) is a doughnut shape that you need to wrap the wire around. Toroids will generally be more effective than the clip-on ones.

You should be able to buy one from an electronics shop, or pull one out of some old equipment.

EDIT: Here's a picture of the sort of thing:

Motherboards often have several which could work, near the CPU.

[Linus]

Okay, this is what I bought:



First question - which wire is positive and which is negative? Do I remember right that the longer one is positive?

Second question - just to make absolutely sure... I need to hook up the positive wire on the capacitor to both the +12V from the PSU and the +12V input on the LCD, and the negative wire goes to both the PSU ground and the LCD ground? And as long as I observe proper polarity, there's no danger of mishap?

[Morse]

Hi Linus

Yes, traditionally the longer leg on a cap is positive. Also, the jacket will have a "-" sign that is in an arrow pointing to the negative lead; it's clearly visible in your illustration.

Sprague makes good caps, so that's not a problem, though 680uF might be a bit on the smallish side without modelling it. Make sure that if it's in a high temp area of your case that the cap's temp rating is not exceeded (usually 85C and 105C are available on electrolytics).

The "+" leg goes to the positive line (measure with a DMM and refer to ground if you are in any doubt!) and the "-" leg goes to the ground. As far as "mishap", any cap can fail under the right circumstances, but Sprague is a good brand and you're allowing some overhead for excessive voltage on the line with your choice of a 16V part.

FWIW, you can combine caps and inductors into a "passive power supply filter" (commonly called a "PI" filter due to it's resemblance to the Greek letter). You would have a cap going to ground then an inductor in series with the B+ (positive voltage leg) and follow it up with another cap going to ground. Traditionally you make the first cap a smaller value than the second. You could also put a very low value, low ESR cap (say a .01uF mylar or polypropylene cap) in parallel with each of the other caps (a "snubber" cap like this will provide a path to ground for high freq ripple that a higher capacitance high ESR cap may have trouble passing).

If you're looking for detailed pointers on values and freeware modelling programs for passive power supplies, you could go over to the DIYAudio forum and ask around; the valve ("vacuum tube") guys there are highly skilled and practiced at the design of passive filters like that. I'd help out more myself, but my workshop has been disassembled for the time being, and with it my computer modelling programs and reference archives. Without my modelling programs, I'd hesitate to guess about the frequency of ripple that is resonant on any particular set of values or the voltage drop to be expected.

Good luck and all the best,
Morse

[Mr Evil]

Sounds like you know what you're doing.

In addition to the length of the leads indicating polarity, also note the arrows down one side with '-' symbols in them, which point out the negative lead.

EDIT: Too slow. Be careful if you do try a pi filter though, since inductors + capacitors together can resonate, which can increase noise if it's at the wrong frequency and not well damped enough.

[Morse]

Good point Mr Evil*;

Oscillation in an RFI pumped pi filter could be a significant issue, which is why I would definitely model it before trying. Most filters I used to design were for audio freq circuits fed from AC mains, so aside from an X2 rated safety cap and a toroidal RF choke on the way into the trafo, there wasn't that much thought to oscillation req'd in the passive PS filter til I got to the amp stage. Whole new ballgame there, though.

Definitely model before building and test for oscillation with a scope before trusting that pi filter idea!

OTOH, a simple large value cap going to ground (preferably with a .01uF mylar in parallel with the big cap) shouldn't present any such issues (and simpler always is better, so if it's good enough....).

All the best,
Morse
*PS OT, I envy you your location in the UK - medical bills are eating me alive in the USA!

[Linus]

I tried the 680uF capacitor, and could discern no change in the screen's flickering.

So the question becomes what else I can try. It was noted that 680uF was a little on the low side as far as capacitance goes; is something bigger at all likely to help? The electronics parts store is a long ways away and I could have sworn that 680uF was the biggest they had in the lower voltages.

What information would I need to model a proper circuit to remove this noise? Assuming my PSU is complying with the ATX spec, ripple is <120mV, but I have no idea of the frequency. I'm not planning on buying an oscilloscope anytime soon.

The stock brick provides 5A, and my PSU supplies only 0.3A on the -12V rail, so it looks like the voltage regulator idea is a no-go.

Any chance that I could find a brick that could run the whole system (PC & LCD) with low enough ripple? I could switch the computer PSU to a DC-DC one and use the Delta Electronics 220W brick. I've never recorded AC draws as high as 220W, so it shouldn't overload the brick.

Surely moving to a battery-based system would remove the ripple - I would lose some efficiency in charging the battery, but would gain mobility (my PC is on a cart that I could roll around the house). Although I know Bluefront had troubles getting his DC-DC PSU to work stably in his low-power system...

Any other thoughts?

[Solid Snake]

I know this thread is a few years old, but I saw so many crazy suggestions that I had to respond. So for reference's sake, here it goes:

1. There is nowhere near enough current on the negative lines to support a monitor.

2. There's a 90%+ chance that power ground and signal ground are tied together, so what happens when you connect your signal cable to your computer when your monitor is grounded at -5v? That -5v hits ground as your signal cable sparks off your video card and your PSU shuts down.

3. Because the monitor pulls so much current, it's subject to a ground loop. This is an undesirable effect that manifests itself in all kind of audio/video equipment when more than one connection to earth ground exists. In audio equipment, it announces itself as a 50/60hz hum. In video, it will appear as a crawling ripple on a video screen. For example, I had to pull out the ground prong of my workstation's power cord because that connection to ground was clashing with the ground on my TV tuner when connected to CTV. In your case, pulling the prong will not help because your ground loop exists between your monitor and your computer.

The power brick that came with your monitor electrically isolated it from earth ground via a transformer. This allowed your monitor to have a floating ground when unplugged. When plugged into your computer, it assumed its ground. Because you are powering your computer from your PSU, your monitor now has two paths to ground, one through power, one through signal. Although they are nominally 0v, you will measure an AC and DC voltage between them. This is most likely where your flicker is coming from.

To solve this problem you must do one of two things:

1. Lift one ground. You can't do this because your video card is not going to like a 5A draw through the signal lines!

2. Resolve the grounds, and minimze the loop. This is really your only choice. You have to make sure the AC and DC voltages between your monitor's signal ground and power ground are as close as possible. You must ground your monitor and video card at the same point. I would suggest the screw that holds your video card in. Take a VERY heavy cable (think car audio), solder a ring connector onto it and fasten it to the case through your video card screw. The 12v cable to your monitor doesn't have to be nearly as heavy.

[Trekmeister]

... ground loop ...

Yep ground loop would be my guess too. Really annoying stuff.

Keep in mind that the ripple from a decent PSU is most likely a fraction of than from the included el-cheap'o wall wart.

Anyway, old post... Hope it worked out for the OP in the end!