RiscPC Modifications

by James R Miller


These notes document some modifications you may find worthwhile to apply to your Acorn RiscPC. They are: #include <std.disclaimer>
You modify/repair your computer at your own risk. I cannot be held responsible if you die of electrocution because (say) I forget to tell you to turn off the mains electricity supply first. Likewise, "use a fine tipped temperature-controlled soldering iron" does not mean a fireside poker will do instead. Nor that you might even void your machine's warranty, nor that the EC (or CE) thought-police might arrest you for ignoring their precious edicts. Oh, and I don't warn you about how static electricity which you can't feel or see, can easily destroy semiconductor devices, or more insidiously, damage them so that they unexpectedly fail later. Earthing straps are not just a trendy fad. OK? Trembling? Let's go.

Restoring Full Video Bandwidth

Why?

If you have a RiscPC with the second generation motherboard, the maximum video bandwidth displayed is about one half of what it should be. The affected board is part number 1208,000
and *Status SoundSystem returns 16bit.

What this means is that screen modes like 1024x768 suffer, and 1280x1024 or finer have only half the indicated horizontal resolution. The money you spent on that gorgeous 17" monitor was, in effect, substantially wasted. You deserve better. A lot better.

Surely you're joking, mister?

Sorry, I'm not. I bought my RiscPC 600 + AKF85 monitor in October '94. From first switch-on it took my breath away, and in its present StrongARM etc incarnation it still does every time I use it, which is roughly 24 hours/day.

In January '97 I acquired a second RiscPC (on behalf of a friend), switched it on for testing and couldn't believe what I saw. Had I really bought the same object that I'd been using so happily these last few years?

The immediate difference was screen quality. Everything slightly blurred; the textfile icon (FFF) had been changed; each line of the black pseudo-writing was just a smudged horizontal line. The desktop fonts were shrouded in faint dark mist. BASIC listings, white writing on blue paper in my set-up, had colour aberrations where white met blue. Hard to read.

The machine went straight back to the dealer, and was replaced without question, though with the comment that "they couldn't see anything wrong" which is pretty much an industry standard response nowadays to nebulous issues like quality.

The replacement machine was just as bad. Obviously a generic problem. So I started an investigation. Why was the video quality so lousy? I checked some other newish RiscPCs. Same problem. Time for some science.

The Facts

It was obviously a video bandwidth problem. I took my Oct '94 RiscPC and a mid '96 RiscPC to an electronics lab, and with the help of a test program (1K), looked at the signals coming out of the video connector using a Hewlett Packard spectrum analyser. Oh dear! The cause of our poor video quality was immediately apparent.

The system was set up to generate the most demanding video signal from the older machine. That is 1600x1200 at 55 Hz frame rate. The pixel rate for this mode is 139 MHz, used to generate the widest possible spectrum for the tests. I could have run 1280x1024 at 75 Hz as well.

Video Spectrum Plot
Spectrum of RiscPC video, early motherboards and later motherboards.
Horizontal 0 - 200 MHz, 20 MHz/div, vertical 10db/div.

MK I = early motherboard type 197,000; MK II = later motherboard type 1208,000

The MK I trace has a classic Sin(x)/x profile, where x = pi*f/140, f in MHz, and indicates that the video signal emanating from the video system has rectangular pixel shape (in the time domain), and is not further processed in any way. In other words, it's pure video.

The MK II trace shows that the video signal of current RiscPCs has been heavily filtered to remove the higher frequencies, from about 80 MHz upwards. Subtract trace II from trace I and you have the filter's amplitude response; already 6db down at 45 MHz, and all but destroyed at 80 MHz and upwards.

These are the very spatial frequencies that give the sharpness and clarity to the image. In consequence, screen modes that demand a pixel rate of about 80 MHz or more are degraded, and even those in the region 60-80 MHz are slightly smudged. That is, adjacent pixels get "smeared" into each other. Transient colour aberrations are due to different group delay characteristics between the three filters.

Investigation of the old and new motherboards quickly located additional filtering components close to the VIDC20 chip and some more by the video connector. This was later confirmed with reference to a circuit diagram (schematic) kindly provided by Acorn.

I tackled them about this heavy-handed video change, and in their explanation they admitted that the need to embrace European Directive EN55022-B EMC requirements had driven them somewhat unwillingly to this solution. I was not able to discuss this engineer to engineer, but I did wonder whether the real need had been simply to suppress the residual pixel clock noise apparent at 139 MHz; that could be reduced by improving the associated VIDC20 Phase Locked Loop. What I do know is that in devising such a crude solution, the baby was thrown right out with the bathwater.

Thus it is that one of the chief glories of the early RiscPCs, outstanding video, is no more and has not been since the motherboard design was changed.

Even more astonishing is that no-one appears to have noticed!

Restoring the Video Bandwidth

The modification applies only to second generation motherboards, part number 1208,000. Simple to state, it's less easy to implement.
  1. Remove three identical chip capacitors adjacent to VIDC20 pins 37-41
  2. Remove three filters adjacent to the 15 way D-type video connector
  3. Replace three filters with wire bridges.
Components to remove

The location of these components is marked on the diagram. Drawfile 5K.

Caution. If you do not know how to re-work surface mount devices on printed circuit boards, do not bother to read beyond the end of this sentence; take the machine to a dealer or person who has the right tools and experience.

Repeat. This job is difficult. If you are not very skilled with soldering irons and fine tools, take the computer to someone who is.

I suggest the following tools:

Strip the machine down to the motherboard so that you can easily work round about the VIDC20 and video connector area.

(1) Using the big tip in the iron, heat both ends of a chip capacitor simultaneously; it'll hopefully stick to the bit and is thus removed. Otherwise the toothpick will be needed to dislodge it from the board. Check that a solder bridge is not left where the capacitor(s) were, or anywhere else. Clear up any debris immediately.

(2) Removing the filters is difficult because, unlike the chip capacitors, these devices have three soldered terminals. There are the "live" tabs, one at each end, but the third "ground" tab is in fact a strip running the full width underneath the device, and pokes out at each side, i.e. the device appears to have 4 tabs and is really well stuck down. If you do not have the proper SMD desoldering tool, then you have to remove the filters either by rocking them gently from side to side using the pliers until they work free, or scrunching them to bits and clearing up the mess.

Do not pull upwards or you will simply rip tracks off the board leaving you a very difficult repair job. Do not force anything. If the filter shows no sign of coming off the board, and if it doesn't happen at once, and it probably won't, gently crunch it up. Vacuum up the debris every few seconds. Using the fine tip bit, desolder the residual debris and use the solder-sucker. Vacuum clean. Check for solder shorts. This process is nerve wracking; I still find it draining even after having modified quite a number of machines. Take a ten-minute break between each filter smashing.

(3) Now bridge the live pads (where the filters were) with a piece of fine wire. Use insulated wire to prevent it shorting to the transverse ground track. Since these wire bridges are barely 3mm long, good soldering skills are needed to affix them. [Note: you only bridge where the filters were; do not bridge the ex-capacitors].

Check and re-check; then reassemble the machine and switch on. Hopefully you will still have a picture, hopefully still in normal colours.

The Bottom Line

If you and your computer survive the operation, you will be rewarded with the crystal clear display that you should have had all along:

 Big FFF icon


Repairing Broken Front Flap Springs

The RiscPC's elegant lines are achieved with a roll-front door cover that's held open or closed by a spring.

The spring is secured to the case by a peg that has a propensity to snap off. Then the door stays permanently flopped open

To repair this you have to provide a new anchorage at exactly the same point as the old peg offered.

How to re-fix spring

You will need:

  • M3x6 screw and nut
  • M3 solder tag
  • 3.3 mm (or 1/8") drill bit and hand drill.

    Remove the slice of case that has a broken pin.
    Drill a 3.3 mm hole as indicated in the diagram. Drawfile 5K.
    Secure the solder tag under the shelf, and affix the spring.

    That's all. Should take 5 minutes.


    There have been  millions of visits to this page.
     Contact

    Created: 1997 Dec 20 -- Updated: 2005 Oct 20