Getting more out of your graphics card.


Status
Not open for further replies.

Firefox

New Member
Feb 15, 2004
716
0
0
Bedok
This is a simplified guide to modifying a graphics card for better image quality.

This guide assumes that the user following it is capable of using a soldering iron/ gun and is fairly proficient with it. The user should be comfortable with working with SMT/ SMD components.

Note that any modification will void the warranty (if any) and I'm not responsible for any damages/ injuries/ loss of life incurred.

CRT users are much more likely to note a difference than LCD users. This is only applicable for analog outputs. ie. VGA or DVI-I connections
It'll not improve DVI-D (digital) connections.
It'll also be much less effective on cards using high quality components on the analog output.
Examples are Matrox cards which already utilise high-quality components and more harm than good would be done if this mod is done on them.

First of all, I'll explain why this modification can improve image quality.
Almost, if not all, video cards have a low-pass filter on the VGA output. This filter serves to remove RF interference.
A low-pass filter's job is to remove all frequencies above the designated cut-off point.
Any interference that is of a higher frequency than this cut-off point will then be eliminated before reaching the display. This is in the ideal world.

However, the components used to create this filter usually have varying values from listed. Usually +/- 20% or +/- 10%.
If you set the cut-off point to exactly the bandwidth of the card, you'll find that in some instances, part of the video information is cut-off as well. This will reduce image quality.

Cheaper cards use cheaper components which are less stable in value and so they may deviate too much and cause the filter's cut-off point to fall below the video bandwidth.
Better cards like Matrox cards have better components that are tighter in tolerance and are much less likely to do so.
Other factors include heat & vibrations but these are not faults of the manufacturers.

So what you can do is to either rebuild the filter with high-quality components (which is unlikely as it takes too much effort and money) or to remove it and hence remove the possibility of it reducing video bandwidth.
This guide teaches you how to identify the components to remove.

This is a close up of the low-pass filter (formed of inductors and capacitors). It is usually formed out of 9 capacitors and 6 inductors.
Some cards may leave out a set of 3 capacitors and or inductors, so take note.
This filter is usually located on top/ below the card but always very close to the vga output connection.

originalcloseup4kh.jpg



The inductors are boxed in red and the capacitors are boxed in green. You'll note that inductors are usually labelled with a 'L' prefix and capacitors are labelled with a 'C' prefix. It's kind of an inductry standard. In this instance, my card was labelled with 'SL" & 'SC' instead.

originallabelled1sg.jpg
 

Excellent work Firefox! A big thanks! :thumbsup:
 

The idea is to remove the inductors and capacitors. Following which, you'd jumper across the pads (first pad of 1st inductor to last pad of last inductor) using wires.
Do not jumper across the capacitors. That can and will cause irrepairable damage to the card.

If you're uncertain as to which is the first pad or last pad of the inductors, simply jumper across individual inductor pads. You'll end up with 6 wires instead of 3 as in the picture.

I used 24awg, teflon insulated, 99.99% pure, solid core silver wire. You need not do so and are unlikely to find it sold locally (not to mention it'd cost you at least S$5/ ft at audio shops if you can find them).
I'd recommend using just insulated copper solid core wire. This is sold as magnet wire.

moddedcloseup8co.jpg



And the clearer picture of where the low-pass filter was located. Note the VGA connector directly to the right of the filter.

modded9hd.jpg




I'm not formally trained in EE but I just have a knack for it. Those who're trained in EE may be able to improve on this. I don't have the proper equipment (LCR meter) to check the inductance and so I couldn't rebuild the filter. Those who do, may wish to do so instead.


If you feel that the card is too expensive to risk such a modification. Or you're uncertain of your ability to perform the modification.
Do not proceed to modify the card!
 

Anyway, I'm neither from nor currently in Temasek Poly. I just got the post-it notes from a relative lecturing (i believe in some Business course) there.

The pictures are in B&W because I don't own a digital camera (used to but not anymore) and I happened to have spare exposures on my roll of T-MAX 400 (hence the grain) loaded in my camera at the time.
 

is the stage you are modifying the output stage?

if so, then the monitor will be receiving high frequency output from the graphic card. it's equivalent to removing the anti-imaging filter. this may cause some high frequency imaging on the display. this can be regarded as distortions.

good filters are hard to design. ill-designed or cheap filters usually have poorer performance. the reason why some data are cut-off is because of the relatively gentle roll-off of the transition band of the filter.
 

Even though I am an electrical engineer I have never worked on VGA cards before, so what I have to say here will be partly experience and partly educated guessing.

The filters are installed to reduce electro-magnetic interference (EMI).

Since the bandwidth of video signals are rather high (going from D.C. to >100 MHz, depending on resolution/refresh rate setting and the image being displayed) a very good quality VGA cable with ample shielding is required in order to meet FCC limits on radiated emissions.

To qualify VGA cards with lower grade VGA cables card manufacturers have to add filters to modify the output signal's bandwidth. When it comes to choosing between lowering performance or adding cost to meet agency requirements, I think the choice is clear for low cost manufacturers. Lower quality filter design/components will have wider tolerances in their characteristics, as well as more non-linear distortion (dielectric losses).

The filters also help to prevent ringing or reflections of the signal if the imedance does not match well with the monitor input port.

If you remove the filters, you will recover the portion of the signal that was originally reduced by the filters, which will most likely show up as better colours, contrast and sharpness. However, you also stand to gain possible interference with other equipment in the system or in the room because of higher levels of emmisssion. Also, if there is poor impedance matching, there might be other distortions introduced to the image (ghosting?).

While I would not recommend removing the filter altogether, if you decide to do so, see if you can get a higher quality VGA cable.

If possible, find out the value and type of capacitors used in the filter circuit. Try switching to smaller values as well as changing them to higher grade capacitors, such as NPO (2%) type if they are Z5V or X5R type, instead of removing them altogether. For the ferrite bead, maybe replace them with metal oxide resistors in the <100 ohms range.

I stress again, I am just guessing base on my own experience. :p
 

Just discussed with my colleage and we both agree that to send the worst signal through the VGA cable, one can generate a full-screen image of alternating black and white pixels.

With such an image the R, G and B signals will have to alternate between highest and lowest output intensities going from 1 pixel to the next, and this can really stress the VGA card. :devil:

If the screen resolution is 1280X1024 with a refresh rate of 80 Hz, the signal bandwidth starts from the base frequency of 50 MHz (1280 * 1024 * 40) and the harmonics go much higher into hundreds of MHz.

Anyone wants to generate such a signal for screen testing? :devil:
 

the test sounds quite crazy... :sweat:
 

mervlam said:
is the stage you are modifying the output stage?

if so, then the monitor will be receiving high frequency output from the graphic card. it's equivalent to removing the anti-imaging filter. this may cause some high frequency imaging on the display. this can be regarded as distortions.

good filters are hard to design. ill-designed or cheap filters usually have poorer performance. the reason why some data are cut-off is because of the relatively gentle roll-off of the transition band of the filter.


Like I said, one can always re-build the filter with better parts and circuit design.. This is a cheap card and I'm not too bothered with trying to source the parts out for a better filter.
If I do, it'd be with 1% Polypropylene or Polystyrol or Silvered-mica caps..
My experience so far has been good..
This mod is for the adventurous.. If there's no issue with EMI/ RFI, one can get Matrox level of image quality with a moderately cheap card (<$100).
 

Firefox said:
Like I said, one can always re-build the filter with better parts and circuit design.. This is a cheap card and I'm not too bothered with trying to source the parts out for a better filter.
If I do, it'd be with 1% Polypropylene or Polystyrol or Silvered-mica caps..
My experience so far has been good..
This mod is for the adventurous.. If there's no issue with EMI/ RFI, one can get Matrox level of image quality with a moderately cheap card (<$100).

nice experimentation, btw :thumbsup:
 

roygoh said:
If possible, find out the value and type of capacitors used in the filter circuit. Try switching to smaller values as well as changing them to higher grade capacitors, such as NPO (2%) type if they are Z5V or X5R type, instead of removing them altogether. For the ferrite bead, maybe replace them with metal oxide resistors in the <100 ohms range.

Heh... NPO/ C0G capacitors are not very readily available. I think using non-SMD silvered mica or polystyrol caps (1% or tighter tolerances possible) would be good.. Even 1% polypropylene film & foil (Philips & Wima have some) is possible.

Yeah, replacing the inductors with tight tolerance resistors resistors is a good idea too. But I guess it should be less than 10 ohms each to reduce the voltage drop across them. Probably can get Vishay Dralorics (0.5% & tighter) in that kind of values for cheap.
 

if u can spend $600-$1k on a monitor, why not spend another $200 on a better card rather than risking your whole system? the $ saved is not worth it if u blow your monitor due to high freq
 

quekky said:
if u can spend $600-$1k on a monitor, why not spend another $200 on a better card rather than risking your whole system? the $ saved is not worth it if u blow your monitor due to high freq

I don't think you can blow your monitor.. I'd think that anyone who has spent $1k for a CRT monitor probably has a Matrox or better card.
This is just for home users who have got budget cards and simple CRT monitors but want more out of their graphics adaptors.
Not everyone will buy a Matrox. Not everyone isn't a gamer....
 

Firefox said:
I don't think you can blow your monitor.. I'd think that anyone who has spent $1k for a CRT monitor probably has a Matrox or better card.
This is just for home users who have got budget cards and simple CRT monitors but want more out of their graphics adaptors.
Not everyone will buy a Matrox. Not everyone isn't a gamer....

Hi Firefox...

I like the sharing and the input by so many of you! Please continue to give input as it benefits all. For those who are adverturous please proceed to do the mod. Give it a try and feedback you finding.

Yappy & happy.
 

Anyway, if anyone wants to rebuild a filter, most graphics cards use a RCLCLC filter on the output. You'll see the 3 resistors just beside the filter in the pictures.

To calculate the cut-off frequency, use the formula: Freq. = 1 / (2 * pi * r * c)
Frequency is in Hertz (Hz), c is the capacitance in Farads (F), r is the resistance in ohms and it's interchangable with inductance, Henry (H).

Video cards have a bandwidth equal to half the RAMDAC's. If you have a 350MHz RAMDAC, design a filter for 175MHz or slightly higher. 200MHz for a 400MHz RAMDAC etc..

Note that the resistors/ inductors in each stage will matter for the next stage's filter.
ie. If all the resistors are 10 ohms, the 1st stage's filter will be calculated with r = 10 ohms. The 2nd stage's filter is calculated with r = 20 ohms etc....
 

Firefox said:
I don't think you can blow your monitor.. I'd think that anyone who has spent $1k for a CRT monitor probably has a Matrox or better card.
This is just for home users who have got budget cards and simple CRT monitors but want more out of their graphics adaptors.
Not everyone will buy a Matrox. Not everyone isn't a gamer....
u saying that all gamers will choose matrox? :think:


if someone dun game, i think it's justifyable to pay 5-10% more for a better card (% compared to how much u paid for your whole system). even if (s)he play game, can choose a more branded nvidia card
 

quekky said:
u saying that all gamers will choose matrox? :think:


if someone dun game, i think it's justifyable to pay 5-10% more for a better card (% compared to how much u paid for your whole system). even if (s)he play game, can choose a more branded nvidia card

I'm saying that not everyone is going to be a Matrox user since he/ she may be a gamer.. Heck.. The games industry drives the graphics adaptor industry...

Even if you choose a better branded Nvidia card, you probably won't get Matrox type of quality. You're usually paying the extras for stuff like the HSF, ramsinks and faster spec'ed ram chips. Sad, but true..

And secondly, I've gotten better image quality out of a Geforce DDR as opposed to a Matrox G450 card due to this modification. Which is to say, it's possible to have more for less if you're willing to take a risk.. But then Singaporeans aren't really known to be risk takers to begin with....
 

Status
Not open for further replies.