ZXpand Maths prob and a display problem

[prolog]

Hello wonderful people
Hope you are all well. Managed to have some quality time with my zx81s over the holidays and while at play noticed some interesting issues - some of which jigged a memory of an old problem with some zx81s back in the 80s but I have truly forgotten what the issue was. I have run through the forums as one issue is ZXPand related but I couldn't find a thread (might be a ROM/Basic issue though...)

So I have 4 zx81s and 2 ZXpands work normally with all of them except on one, NedZed, with either ZXpand attached I get 7**3 = 480.8751253 instead of 343 and SQR 49 comes out as 7.14356... Without the ZXpand Basic gives the right results also correct with ram packs etc. I do recall an early floating point issue with a batch of zx81s back in the 80s, but I can't recall what it was exactly. I haven't opened up NedZed to see what issue he is as I don't want to break the warranty etc...

And there's also FredZed who works fine with the ZXpand and for whom the Basic maths is correct every time. However, FredZed has some intermittent drop out on the CRT display and also a very slightly wobbly picture as if needing more tuning - with the ZXpand attached the wobble is very slightly worse. It's an issue 3, I am thinking ULA ....??

grateful for all/any suggestions!

cheers
Jeff

[RWAP]

That is interesting about NedZed - I wonder if that has one of the small boards on it to correct the square root error, but when the ZXpand pages in its own ROM, the small board affects the square root returned by the ZXPand!

[Moggy]

Although referred to as the SQRT bug it was thought to be attributed to a rounding error in the DIV routine but as the most common example of ROM testing was to check the SQRT of .25 it acquired ( erroneously in my opinion) the SQRT bug moniker.

The modification Rich refers to (pictured) nullified the effects of three ROM bytes by causing a DAA instruction to be executed thus correcting some aspects of the bug. These three bytes are absent in the improved ROM but the DIVision routine remained unaltered and the zeddy still holds some computations incorrectly.
Dr Frank O'Hara wrote an interesting article in ZX Computing on this very theme.

As a curious aside a game written in basic that I used to play used a SQRT computation to open doors on various levels and when used with a modded Zeddy the doors would only open half way, remove the mod and replace the ROM and the doors opened fully leading me to believe that the mod never fully cured the problem and at best was a temporary fix.


In the case of the ZXpand these modified Zeddies tend not to play ball, I agree with Rich that a combination of Charlie's overlay ROM based on the newer improved one and a SQRT modded zeddy using the old ROM is a bad mixture in as much as the mod on top of the CPU is now trying to nullify three bytes of the improved overlay ROM hence the mathematical corruption.

I had a couple of these modded Zeddies and by removing the mod and using with the ZXpand had no further trouble, though to use on its own requires a replacement ROM in the zeddy, in my case the "Shoulders Of Giants" ROM burnt on to an EPROM.

[prolog]

NedZed.jpg

(1.96 MiB) Downloaded 337 times

Thanks Rich and Moggy! (hope you are well both incidentally) - yes, well I've cast warranties to the wind and have now opened NedZed and do find a most peculiar arrangement inside (that I've not seen before but have distantly heard of) and so here it is - You both have the explanation so thanks for that! Also the ULA is welded to the board so I can't just unplug and put in FredZed...shucks!

Thanks again

btw the other problem does sound like ULA no?
Jeff

[Moggy]

Yes this is a modded Zeddy.

The ULA pictured there I can say with 100% certainty is the oldest of the known three ULAs used,
on the underside will be printed 2C158E. the 8119 on top is week 19 year 1981.

The differences between these ULAs ,apart from missing back porch signal, is as far as I'm aware (but with no proof whatsoever) nothing more sinister than power consumption and possibly timing differences , though everyone one of these early ULAs I have come across over the last thirty odd years had a tendency to put a slight weave or breathing effect on the display when used with B/W CRT TVs and when they were about to expire would cause the top 2 lines of the display to bend to the right then implode some short time later.

I would add these are my experiences only and no doubt a guru will be along shortly to point out my errors.

[Moggy]

For your display problems the item listed here may be the answer though reading the manual it seems to imply it has only been tried with the the 2 later ULAs.

http://www.sellmyretro.com/offer/detail ... -ZX81-3712

I have one on a ZX80 and the picture is superb plus it's a cheaper more modern and I would say a more reliable bet than a replacement ULA.

[prolog]

Hi Moggy hmmm interesting! SO, a Victorian ULA in NedZed as well!... though NedZed's display is ok - FredZed gives me a slight warping at bottom left of crt screen and and occasional black snow drift horizontally across the screen (none of my other zeddies do this) - I think FredZed has a Ferranti branded ULA, so a later one, I'll check..

Yes I saw the little board - you are using it on a flatscreen presumably Moggy?

Jeff

[Moggy]

Hi Moggy hmmm interesting! SO, a Victorian ULA in NedZed as well!... though NedZed's display is ok - FredZed gives me a slight warping at bottom left of crt screen and and occasional black snow drift horizontally across the screen (none of my other zeddies do this) - I think FredZed has a Ferranti branded ULA, so a later one, I'll check..

Yes I saw the little board - you are using it on a flatscreen presumably Moggy?

Jeff

Hi Jeff.

Rgardless of ULA fitted the PAL signal the zeddy gives out is somewhat out of whack as they say and whilst nearly all B/W and some colour CRT sets are fairly forgiving , modern LCDs and the like no so much.

Your issue 3 most likely has the most up to date ULA hence any picture at all on a flat screen but your flat screen is having trouble latching on to a dodgy PAL signal hence the corruption.

The little board as you describe it works a treat on both types of display and is a good investment as time marches on and CRTs bite the dust.

The ULAs were I believe solely Ferranti manufactured unless any one knows otherwise.