MAME cabinet, video one more time
Returning to video issues for the final time, hopefully.. This post has been under the works for quite a while because instead of simply installing the VGA converter (and being happy with the cabinet), I decided to perform more tests and see if I could possibly tweak the quality a bit further.. In the light of these tests, it seems that I prematurely passed judgement on the quality of the GeForce 6600 tv-out when actually the motherboard I have is more the culprit! Anyway, let’s first start with the a bit of video of the Geforce 6600 s-video tv-out..:
Here, the card is running 720*576 50Hz (PAL), the signal timing values have been modified by hand to match the PAL spec. Prior to this, the output was set to 800*600 60Hz (VESA), and this made the colorful ghost pattern visible especially around window edges above more messy (sort of), contributing to a blurrier overall image. Put short, the s-video output quality you see in the video clip above is about as good as I can make it by editing signal timing values. Composite output doesn’t have the colorful distortion but then again, it looks about the same as when you add some flicker (blur) filter over the s-video output.
After hitting a dead end with the 6600, I ordered the VGA converter. As I much hoped for, feeding the VGA output of 6600 through the converter actually improved picture quality. The outputted picture is slightly rotated to left, but the converter has only basic size/position adjustments. Weirdly enough, with this configuration the s-video quality is only marginally better (similar ghosting, but less of it), whereas composite looks way better. This got me thinking what else might be there with the s-video that’s causing this? Sure, the composite output was now at a (somewhat) satisfactory level so I should ditch s-video over composite and move on.. But still, I just couldn’t leave it to rest :)
I then decided to see whether I could get some additional gfx cards for testing. The ones I was able to borrow were a Asus Geforce 7500 LE and a MSI Radeon HD2400. Much to my surprise, the s-video output on both was even messier than on the 6600. Now wtf?! If gfx cards and even a external video converter exhibit similar distortions on their s-video output, what else is there to blame than the motherboard? Power supply can be ruled out of the equation, as I did test powering up the converter (requires +5VDC) from both the PC PSU and a separate AC/DC transformer. No luck there and the power supply feeds also look pretty ripple-free. So what more to check for?
As a added bonus, I was unable to get the tv-out of the 7500 working using the latest Forceware drivers (270.something). A bit of downgrading tests, and version 169.something proved a success. Anyway, with the fault starting to point towards the mobo, it was (finally?) time bid “bye bye” (or “fuck off”?) to s-video.
Whilst comparing the picture quality with some test bitmaps, I also noticed a completely new and very faint repeating noise pattern visible on some colors (most apparent on large solid color surfaces). This pattern seems to be related to SATA bus transfers; when data is copied from one drive to another, the appearance frequency of the noise increases. I didn’t spot this at first, but just happened to bump into it as I had a rather lengthy copy running on the system. Something to do with the transfer bursts, but I’m definitely not going to start analysing that.. So yeah, mobo looking even more “classy”. Thanks a fucking bunch, Foxconn.
I did quickly check the composite output of the 6600 with a oscilloscope, but being no expert with tv signalling I was unable to determine anything conclusive. The color burst on the back porch doesn’t look as strong (300mVpp) as in the signal reference picture above, but that’s about it. How the hell I would go about boosting this burst portion anyway?!
Amidst scoping the signals, my thoughts wandered back to the Geforce settings / resolution tests. This time around though, it occurred to me that (when testing tweaks between 800*600 & 720*576) the appearance of the distortion had slightly changed with the resolution. So what if I just simply drop the resolution somewhere even lower?
Well, 320*240 didn’t really work detail-wise ;).. But go up to 640*480 and the system still remains usable plus there’s minimum ghosting on-screen. Large solid color surfaces are still a bit of a problem, but I guess the SATA flickering can be minimized with a careful choice of frontend theme (read: make it black). Using this resolution and composite output there’s hardly any difference between the converter and the TV-out, so I figured I’ll just disconnect the former from the system altogether to get rid of the slight picture rotation. At this point things started to look like as if I can finally move on..
.. and shortly after, the 6600 crapped out \:D/. The symptoms it showed looked a lot like BGA solder joint problems, with repeating pixel junk patterns getting worse on-screen as the card warmed up. Instead of binning the card straight up, I figured I might just as well ask the SMD guys at work to run the card through the vapour-phase oven. Not that I had much to lose anyway. And besides, the resoldering process has worked on similar occasions before.
No idea what brain fart occurred, but this time around (and for some reason unknown) I didn’t think of removing the on-board capacitors for the resoldering process although I did do so with the connectors. So into the oven the card goes and shortly after, BANG BOOM! One capacitor blew up forcefully enough to displace the adjacent VRAM chip from its solder pads. Oh well, that was fun :D
Come to think of it, these damaged caps look maybe even slightly classier than the one on the header of this blog. So maybe this could be a good time for a header re-design. Both are “my DIY” anyway ;)
But ok, minus getting a new graphics card, it’s now _finally_ on to frontend setup and whatnot. Big thanks to Antti and Hanski for helping with the gfx cards!