iraito wrote on 2024-06-22, 00:11:

Let me add one common issues that plagues some of my PCs, it's those horizontal speckles, i think it's electrical noise bleeding from other cables or devices connected nearby but if anybody knows this specific issue i would appreciate it.

What interface, monitor and system (specs) are you using for this?

Since the problem is inside the chip, there's nothing you can do really. It's faulty and needs to be replaced.

Ozzuneoj wrote on 2024-06-22, 00:46:

iraito wrote on 2024-06-22, 00:11:

Let me add one common issues that plagues some of my PCs, it's those horizontal speckles, i think it's electrical noise bleeding from other cables or devices connected nearby but if anybody knows this specific issue i would appreciate it.

What interface, monitor and system (specs) are you using for this?

VGA, LCD (but that's from a capture card) the specs: CPU: Intel® Pentium® 4 CPU 2.65GHz - L2:512K - Northwood
Motherboard: American Megatrends Inc. - M935D Rev 2.0
RAM: 512 DDR-SDRAM - PC-2100U single channel 133Mhz
GPU: GeForce 4 TI 4200
Soundcard: Sound Blaster Audigy Platinum EX

It's a noise present even on some consoles, hence the electrical issue.

Let me add that this issue is present only in the capture but not on the LCD, I tried connecting everything directly bypassing any switch or splitter and the noise is still there, I'm also using shielded cables and really thick ones.

Can't hurt to wash the card.
Normal boring soapy water with a tooth brush and a long sun dry.

You'd be surprised. I've fixed completely dead modern GPUs with a simple wash.
My theory is that conductive dust caused issues.
I don't know for sure, but I know the results have NEVER made it worse.

Yeah I've actually had a problem before with some fine organic dust, seemed like superfine wood or maybe paper/cardboard dust, like what maybe happens in an office. Anyway, that seemed to be hygroscopic, drawing moisture from the atmosphere and it was making a slightly conductive bridge between pins on an i/o chip causing malfunctions. I guess it packed in there dry, randomly from airflow eddies, but when it got the slightest amount of moisture in it, it jammed in real hard, and while loose damp dust wouldn't have a very low resistance, the combo of the pressure from it swelling and the gaps between pins being very small brought it down around a kiloohm or so, which was disruptive.

myne wrote on 2024-06-22, 12:29:

Can't hurt to wash the card. Normal boring soapy water with a tooth brush and a long sun dry. […]
Show full quote

Can't hurt to wash the card.
Normal boring soapy water with a tooth brush and a long sun dry.

You'd be surprised. I've fixed completely dead modern GPUs with a simple wash.
My theory is that conductive dust caused issues.
I don't know for sure, but I know the results have NEVER made it worse.

I tried with alcohol, no dice. But at least it was squeaky clean 😁

I find water, the universal solvent, is better.

Alcohol for the rinse before drying.

There was a bargain price PCI TNT2 on eBay showing the same symptoms as OP.

I was close to buying that TNT2 after researching these issues, and the TNT2 was ultimately bought by @giantenemycat.

My opinion is that the root cause is more likely in the analog stage Barium Titanate capacitors than in any RAMDAC.

Principally, a RAMDAC generates each pixel sequentially left-to-right. That means full-line darkening (pixels to the left) would require a RAMDAC to do time travel and I'm not buying that - so it's not the RAMDAC.

What I'm seeing in all the photos is whole lines darkening (or brightening) to match an average. Looking really closely at the artefact, I think each line is showing the average brightness of the preceding line. To me that looks like some kind of moving average bias in the analogue stage after the RAMDAC.

I have done some reading and I think the fault can be better described as "baseline shift/droop in the analogue stage of resistance and capacitance".

Healthy PCBs stay within 0.0v for deep black to 0.7v for bright white, but afflicted cards are drifting. Sometimes the voltage drifts too low (e.g. -0.3v to +0.4v - darkened) or drifts too high (e.g. +0.3v to 1.0v - brightened).

According to one manufacturer of ceramic capacitors, Barium Titanate capacitors are known to show DC bias and ageing characteristics - and that is a very elegant explanation for everything we know about this issue. The challenge now is in identifying analog stage Barium Titanate capacitors, testing, and replacing them.

My hypothesis suggests all affected cards should produce a near flawless output of 3D rotating camera scenes where the brightness of the preceding line is continually changing thus pulling the voltage up and down and blending together.

That sounds like a very plausible explanation.
Well done.

If there are capacitors that cause this (stuff I saw in the first post with darkened sections) it will be caused by something around the RAMDAC VREF/IREF bits if it was in the video card itself. The capacitors on RGB lines are single digit pF value class 1 ceramics that lack DC offset related effects (they are not using the mentioned material) and don't even have enough capacitance to show effects that span more than a pixel or two (depending on resolution). But bad references for the DAC can cause what is coming to be shown to get modulated according to what was shown, as it will pull the reference higher or lower when it cannot hold its value for whatever reason.

---

When I see this darkened bits problem it is usually caused by the display and its input clamping stage. There signal is usually AC coupled and right after sync pulse the incomign signal is sampled to take a snapshot of what is supposed to be black, and this voltage is used to push things toward what rest of the circuitry uses as black, and this diffece of voltage is stored in that AC coupling capacitor. If that capacitor is under or oversized you get this type of darker and brighter bands... When the value is too low, the voltage across the capacitor changes too much throughout the line and black level shifts, and especially on big contrast changes. This is not seen on the photos, and such a problem gets "fixed" every beginning of a line. When the capacitor is oversized, the sampling window can be too small to change the capacitance enough in one go and the effect is more gradual and can persists several lines, including the left side of image.

This goes into details of the problem and some of the solutions : https://www.analog.com/en/resources/technical … eo-signals.html

It can be that the mechanism mentioned by MattRocks is in the monitor itself, usually these AC coupling capacitors are class 2 types rather than class 1 for cost reasons, and they show lonlinear effects aswell as aging related capacitance loss and increased sensitivity to DC bias. There is a fix, which is to heat up the capacitor to soldering temperatures which resets the crystalline structure for lack of better description. Some information is found here : https://www.hitano.com.tw/technical/aging-of-class-ii-mlcc/

Surely an aged capacitor is an undersized capacitor?

Those who report the issue have multiple VGA cards and multiple displays - all report that it's the VGA card that is misbehaving.

That is generally the case yeah. Quick solders can fix the issue, purely because of the heat involved which has the ability to recondition a typical ceramic close to its original performance level. Outright replacements are not always going to be necessary. If I find a card that exhibits this type of a problem, I'll be sure to experiment and share the findings ~