Tech Blog – Removing AGA Display Trashes
by Grzegorz Kraszewski, 16 May 2016
Many AGA Amiga users have experienced irritating flashing display "trashes" appearing in higher resolution AGA display modes like DblPAL or Multiscan. In general opinion these artifacts are caused by AGA chips overheating, being loaded with larger amount of video data to process.
This opinion is so common, that even Jens Schönfeld in user manual for his Indivision AGA scandoubler, recommends attaching a radiator to the Alice chip if trashes appear on the screen. Unfortunately it does not work, as I've learned myself.
In some Amigas the problem does not appear immediately after turning on. Trashes are visible after a quarter or so. This indeed suggests some relation with temperature. Then after roughly 45 minutes trashes disappear, while it is obvious that AGA chips are not becoming colder. Putting a radiator on the Alice chip just delays things a bit, trashes appear later and disappear later.
Attaching a radiator to a chip in PLCC84 package is not easy. One can try thermal conductive glue. I have used thermal conductive adhesive tape, but it turned out to be impermanent solution. As there is little space between Alice and the keyboard, cooling the chip is even more difficult.
Browsing the Internet for alternative solutions to the problem, I've encountered an old archive on the Aminet. It has been uploaded by Denis Sotchenko in 1997. The document states that trashes appear on the display only if RGB components of any displayed colour have at least 23 bits set in binary form. Particularly white colour has all the 24 bits set, and almost any Workbench palette contains the white colour. A slight change from (255, 255, 255) to (254, 254, 254) is almost unnoticeable, but surprisingly fixes the problem.
I have been rather skeptical to this theory, but well, trying it is easy. I've adjusted the Workbench palette and... trashes vanished and never came back, even after 16 hours of keeping Amiga turned on. Very good, but not fully satisfying. First, not every software allows for adjusting its screen palette. Secondly, the fact that my Amiga cannot display any colour without trashes is a bit uncomfortable.
Fortunately the archive describes another solution to the problem – a hardware modification. The author suggests that shorting E127R resistor on the mainboard with a piece of wire fixes the problem as well. Resistance of the element is 27 ohm. The resistor togehter with E127C capacitor constitute a lowpass filter for CCK clock signal. This signal is sent from Alice to Lisa and is used for clocking pixels in video output. The filter dampens oscillations on edges of the signal. As a sideeffect it introduces a delay. It seems that the delay is too long for display modes using 35 ns pixel time (like DblPAL, PAL SuperHighRes or Multiscan) and the chipset runs a bit unstable. How it is related to number of "ones" in binary color representation? Seriously, I have no idea, the document does not explain it either.
While I'm not a fan of hacking around Amiga mainboard, I've decided to try it. The software trick worked, so hardware one could be valid too. Unfortunately E127R is placed on the bottom side of the mainboard, so it has to be pulled out of the case and shield (including unscrewing all the ports on the rear). One can find this resistior under the Alice chip, as shown on the first photo.
I've simply placed some soldering flux on edges of the resistor, added some solder, then soldered a wire across the element, as shown on the second photo.
Then I've reassembled the Amiga, turned it on (ufff, it still works...). I've set white colour in the Workbench palette back to full white. Yeah, it works. Have tested my Amiga for a few hours and got no trashes on the display at all. Tests included special high resoultion display modes useful with Indivision AGA: HighGFX (1024×768) and Xtreme (1280×1024).