16K RAM Pack Schematic / Chip Layout

[TMAOne]

Awesome diagram PokeMon.

I don't have any 74HCT02's in my junk collection, so I ordered one. With only 6 connections "external" to the chip socket, I think I'll just wire it up and let the chip float free rather than try to breadboard it. Unfortunately I didn't keep any bad 4116's to do a test with. But since PokeMon has made it so easy for me, I definitely want this tool in my arsenal. I might have to microwave a 4116 and then play Russian Roulette with it to test the tool!

I have:

• - a couple of working packs, one still out of the case with 3 socketed RAM chips to function as a chip tester
  - one pack with good voltages, 8 sockets, and 8 tested good RAM chips, but the pack still doesn't work
  - two packs with bad voltages in spite of replacing everything in sight (electrolytics, the transistor, zeners, other diodes). One of those heats up the 4116's. (Not sure if it's all 8, but it's more than one.)

I will play with the $8 logic probe when I get it, and try to compare the open working pack with the one that seems to (by elimination) have a flaw in the logic chips. Maybe that way I can narrow it down to 1 chip without really getting in too deep. I have a scrap computer board with dozens of 74 series chips. I have no idea what it did in its original life, but it sure is one big collection of gates to act as a donor board now!

The two packs with no -5V, and low +12, I don't know. Maybe bad 4116's are dragging the voltages as RetroTechie suggested near the beginning of this thread. What I really need to do is separate the voltages from the chips and then diagnose, as he also suggested. Unfortunately I struggle with clinical depression and my concentration is poor, so I tend to go after this idle hobby obsession in fits and spurts, and put it back on the shelf often. But I keep taking it down off the shelf too! Stubborn that way. As I've said before I don't need these packs to work, I just want the sense of accomplishment from fixing them that I don't get anymore from my job, which was mainframe software developer.

Meanwhile my working ZX81's and Timex use the 32K mod, or my ZXpand. Okay, I'm rambling again--stop.

[sirmorris]

I'd be happy to make you up a board - Making circuit boards is my new favourite hobby It could sit as an extender between the zeddy and the rampack in question.

I imagine this kind of tool would be useful to a few people. I wonder if it would be worthwhile getting some PCBs manufactured? Probably home production on demand would suffice.

C

[siggi]

Several years ago Kai Fischer used a similar adapter to debug faulty ram packs (see old issues of ZX-Team Magazin). AFAIK he had also a data bus driver on his adapter to isolate the ram packs databus, because the faulty ram pack could always drive the data bus (even when not selected!), preventing boot of the Zeddy ...

Siggi

Edit: here is the link to the ZX-Team-Magazin 2/2002, where Kai describes his test adapter at page 16 (in German language!)
http://forum.tlienhard.com/magazin/2002/magazin_202.pdf

[PokeMon]

I suppose one could also mod the address decode logic in a RAM pack, such that the 16K is mapped to a region normally not used (and internal 1K active as usual). For example 8000-BFFFh and/or C000-FFFFh areas.

Then you could run plain BASIC proggies to determine which RAM chip(s) in the pack are faulty.

C000-FFFF is not good as you have the internal 1k mirrored to C000-C3FF.
If you disable internal RAM in this region you maybe do not see the test results on screen.

[PokeMon]

AFAIK he had also a data bus driver on his adapter to isolate the ram packs databus, because the faulty ram pack could always drive the data bus (even when not selected!), preventing boot of the Zeddy ...

Siggi

Could be, but I think this has not much probability. I think the internal output driver of the 4116 is operated with the (always working) 5V supply only and think this should work even when internal structures are damaged due to wrong/broken/faulty +12V / -5V supplys. Even if the Zeddy doesn't start at all with added RAM pack than you know there is a big trouble inside.

[gozzo]

TMAOne - if you don't have any known duff 4116's, try killing one by powering it up with all power lines except the -5v, this supposedly causes quick destruction !(or just reverse the polarity of all ! ) I may have some in a while as I have few duff rampacks and speccy boards to sort out as and when I get time !! But yes, some sort of rampack tester would be a HUGE help - I also had the idea of an adaptor to 'move' its address range to 32768-49151, with buffering/current limiting so the zx81 could run normally but access the suspect pack to test it... also had an idea of a stand-alone rampack tester using a simple computer circuit based on a z80, with LED indicators displaying the result as OK, or whatever bit/chip is faulty !

[TMAOne]

Quite the lively thread; I love it.

I'd be happy to make you up a board

That's a very gracious offer sirmorris. In the short term I still think I'd like to wire it up experimentally to prove the concept. If you can afford to wait until my chip arrives and I have time to play, anything I discover could be rolled into your new product.

if you don't have any known duff 4116's, try killing one by powering it up with all power lines except the -5v

That's a good idea, but I'm not so sure the legend is true. One board I have has one socketed chip, I measure +0.6 on the -5V line, and the one socketed chip has survived several power-ups. These little guys seem more resilient than rumored. In fact at the beginning of this adventure I had a pack that had -5 voltage problems I didn't know about, and the whole thing worked anyway except for 1 bad bit out of 16K bytes! Heroic little turds, some of them,...

Realistically, I'm not even sure I need a "known bad" chip to test with. As long as the setup works with a good pack, and I can run the basic program POKEing and PEEKing, I think it would safe to assume that a bad chip would throw off one bit of either 0 or 255. I think I'll try that first.

Waiting for the postman again,...

[gozzo]

If you wish you can replace the whole lot of 4116's with a Hitachi HM4816 (I have been told they were used in the BBC computer..???) which is pin compatible EXCEPT, the existing +5v and -5v lines are disconnected(not required), the +12v should be disconnected and connected to +5v instead (no more rampack power oscillator required - wahey!), it should then work as normal..(hopefully- never tried it myself, just read about it somewhere!) could also use 4164's as well in the same way, just leave the existing +5v as it is OR take it to ground instead (still need to remove the -5v(possibly?) and change the +12v to +5v(a must!) NOTE: I say that you need to remove the original +5v and -5v to use the HM4816 - this may not really be necessary as those pins are listed as 'no-connection' on the data sheet I have, but if it was me, I'd rather not chance it !

[1024MAK]

On ZX Spectrum's (which use these 4116 as well), typical failures are excessive current on the +12V line, which causes problems.
Each faulty chip causes one data bit to produce an incorrect result across most, if not all of the 16K address range.

Mark

[TMAOne]

My 74HCT02 chip arrived today, so I wired the circuit up, with a known good ZX81 and RAM pack.

It won't boot for me.

Admittedly, the way I've implemented it is far from elegant, but I've checked the connections and they seem to be correct. The RAMCS and A14 connections have been "broken" by using cellophane tape over the edge connector. Many of the connections at the chip are to each other, and I'd swear PokeMon even made it easy by laying out which gate to use where so that the socket pins are clustered to facilitate simply soldering them together.

Power-up only produces a white screen.

Any thoughts on how to debug?