16K/48K Lower RAM (4116) replacement.
You may want to read this thread for background information on the quest to reach this point.
Following quite a lot more measuring, reading of datasheets and experimenting, I have found an SRAM based replacement for the lower RAM that works in all 16K/48K Spectrums (I've tested it in an awful lot) is possible.
The circuit is basically the same as the one discussed in the thread linked above, but there are a couple of important requirements to note:
1. As previously described, whatever mechanism is used to invert /RAS to drive the latch clock input must be reasonably fast (at least for a high to low input transition) as in most Spectrums falling /RAS also triggers the multiplexors to present the column address to the RAM and so there isn't much time to latch the row address. Extreme measures aren't needed here, however. A 74HC/74HCT inverter is good enough.
2. If the circuit is to work reliably in an issue 2 Spectrum with a 5C ULA, consideration must be given to the voltage levels that appear on the lower RAM address lines. In particular, in some cases a logic high may be represented by approx. 2.3V. This happens when the voltage is pulled down by a logic low at one of the multiplexor outputs. This isn't so much of a problem with later Spectrums, I assume due to 6C ULAs being able to source more current and some models also have higher value resistors in series with the address lines. All this means that a 74HC series latch will not work; a 74HCT one will (it uses the same logic levels as 74LS). The SRAM must also recognise 2.3V as a logic high on its address inputs. I haven't been able to find one that uses original 74 logic levels, but there are many with V(IH) specified as 2.2V and in extensive tests these have always worked.
I noted in the previous thread that the specified access time of the SRAM didn't seem to make much difference. Further testing hasn't changed this. I've tried SRAM ICs with access times from 10ns to 70ns without any problems. So it seems there isn't much risk of problems due to the SRAM latching an address or data too soon, before the bus is stable or de-asserting its outputs too quickly or causing interference due to rapid switching times, etc.
The end result of all this is the ZX Spectrum Lower RAM Replacement Module, pictured below:
It can simply be plugged in in place of the 4116s.
I should have some available to buy very soon. In the meantime, I've got just a few I'm willing to send out to addresses in the UK for free. Please send me a private message if you would like one. The only condition is that you try it in as many Spectrums as you can and report back on any problems you encounter. I will give priority to those who made useful or interesting contributions to the previous thread. Some of the modules I'll send out have been used in component testing/experimentation and so the soldering may be a bit messy etc, but they all work in the issue 2 Spectrum I've been using for primary testing.
Following quite a lot more measuring, reading of datasheets and experimenting, I have found an SRAM based replacement for the lower RAM that works in all 16K/48K Spectrums (I've tested it in an awful lot) is possible.
The circuit is basically the same as the one discussed in the thread linked above, but there are a couple of important requirements to note:
1. As previously described, whatever mechanism is used to invert /RAS to drive the latch clock input must be reasonably fast (at least for a high to low input transition) as in most Spectrums falling /RAS also triggers the multiplexors to present the column address to the RAM and so there isn't much time to latch the row address. Extreme measures aren't needed here, however. A 74HC/74HCT inverter is good enough.
2. If the circuit is to work reliably in an issue 2 Spectrum with a 5C ULA, consideration must be given to the voltage levels that appear on the lower RAM address lines. In particular, in some cases a logic high may be represented by approx. 2.3V. This happens when the voltage is pulled down by a logic low at one of the multiplexor outputs. This isn't so much of a problem with later Spectrums, I assume due to 6C ULAs being able to source more current and some models also have higher value resistors in series with the address lines. All this means that a 74HC series latch will not work; a 74HCT one will (it uses the same logic levels as 74LS). The SRAM must also recognise 2.3V as a logic high on its address inputs. I haven't been able to find one that uses original 74 logic levels, but there are many with V(IH) specified as 2.2V and in extensive tests these have always worked.
I noted in the previous thread that the specified access time of the SRAM didn't seem to make much difference. Further testing hasn't changed this. I've tried SRAM ICs with access times from 10ns to 70ns without any problems. So it seems there isn't much risk of problems due to the SRAM latching an address or data too soon, before the bus is stable or de-asserting its outputs too quickly or causing interference due to rapid switching times, etc.
The end result of all this is the ZX Spectrum Lower RAM Replacement Module, pictured below:
It can simply be plugged in in place of the 4116s.
I should have some available to buy very soon. In the meantime, I've got just a few I'm willing to send out to addresses in the UK for free. Please send me a private message if you would like one. The only condition is that you try it in as many Spectrums as you can and report back on any problems you encounter. I will give priority to those who made useful or interesting contributions to the previous thread. Some of the modules I'll send out have been used in component testing/experimentation and so the soldering may be a bit messy etc, but they all work in the issue 2 Spectrum I've been using for primary testing.
Post edited by Zorn on
Comments
I know I've got at least one machine with a lower memory fault, so this replacement module could work out for me... what sort of cost did you have in mind?
I've had a few responses to my offer to send out some free ones. If you're one of those people, check for messages from me. I might be waiting for you to tell me your address.
Most of my 16K and 48K Spectrum's have operational 4116 and work fine :smile:, but I do have two with faults :sad: (but I have not confirmed the cause of the faults yet). So may be in need of some "replacement 16K RAM" ;)
Mark
Repair Guides. Spanish Hardware site.
WoS - can't download? Info here...
former Meulie Spectrum Archive but no longer available :-(
Spectranet: the TNFS directory thread
! Standby alert !
“There are four lights!”
Step up to red alert. Sir, are you absolutely sure? It does mean changing the bulb!
Looking forward to summer in Somerset later in the year :)
Brilliant, just ordered 2....
Frans
2 ordered.
Thank you.
Any comments from people that tried it out?
PS: your module reminds me of something I designed some time ago as a replacement for 4116 too, but with a standard SIMM module :D
http://www.zxprojects.com/index.php/simm-adapter-for-replacing-lower-memory/13-making-it-pretty
Such a thing is on the way. I have already done some tests to prove the concept.
I think I have repaired more Spectrums that have upper RAM faults than lower RAM faults. Perhaps because they contained faults to begin with (those who know what the 4532 is will understand this) and/or because many of their connections are directly exposed on the expansion connector.
Even if it works i may try and change everything anyway just to make it more reliable and maybe even lower in power use.
Yeah, corners were cut, but lets be fair here, machines that died due to cutting corners were very few and far between compared to those that failed due to abuse. The fact a little 48K machine is still working after 30 years is a testament to be honest.
After 30 years the electrolytics in pretty much ANY circuit is going to be suspect, regardless of who built the board in the first place.
You're right to change all the old electrolytics, add new heatsink compound to the regulator/heatsink junction, reseat all the chips etc but let's be fair, there's no need to swop out the RAM, ROM, CPU if it's still working fine after 30 years.
You're much better off adding a heatsink to the ULA chip than changing it for another one for example.
Remember that by messing with a proven reliable for 30 years circuit, all you're going to do is to risk the introduction of faults.
Bottom line - fix it when it's broke, not before.
In other computers there are DRAM chips that fail (various chip types) ULA failures, 74LS series logic chips that fail etc.
Electrolytic capacitors (made from 1970 onwards) should be renewed as they are known to degrade. Renew or replace any other faulty components. But as Death says, it is best to leave the rest alone.
Another point to think about is this: if an electronic component has lasted 30 years and is still okay, it is likely to continue to work okay for many more years.
Most component failures occur in the first days/weeks/years of use (unless there is a design fault or misuse).
Talking of misuse, always be careful when plugging anything in on the expansion edge-connector. If the locating "pin" goes missing and the connector is not 100% aligned correctly, it can kill a Spectrum in seconds.
So always check the the locating "pin" is present.
Mark
Repair Guides. Spanish Hardware site.
WoS - can't download? Info here...
former Meulie Spectrum Archive but no longer available :-(
Spectranet: the TNFS directory thread
! Standby alert !
“There are four lights!”
Step up to red alert. Sir, are you absolutely sure? It does mean changing the bulb!
Looking forward to summer in Somerset later in the year :)
Yes. Please wait for further information...
Oh goodie goodie :)
I would also add the 7805 power regulator as something than I would replace even if it still works. The heat puts a lot of stress to this component over the years and although its a robust one, it is known to fail.
Ah but a brand new one is more likely to fail than one that has proven good for 30 years. The risk is minimal, but as 1024MAK has said, components usually fail within the first few weeks if they're going to. In essence, you are replacing a proven good component with an unknown.
Barring abuse, just making sure that the heatsink compound is in good order is sufficient - and where none exists, use some!
If you really want to reduce the heat generated in the 7805, add two 3A rectifier diodes into the supply line - a lot (too many!) of Sinclair PSU units deliver over 12V ON LOAD (Spectrum only needs 9V), so reducing this a little does no harm and helps the 7805 regulator by having to drop less voltage across it. Just a volt or two can make a huge difference!
:D
I have come across worn off 7805 that cause bad picture a couple of times but never a new one to fail.
RECOMs are indeed a good choise as long as you don't care about keeping the machine original.
Quite true, mainly because we must give up the metal heat-sink. But it's hard to keep the machine original anyway. We already struggle to find things such axial capacitors of the same size values (never mind other aspects) to recap a board, for example, and some just bend some radial caps and go on to their lives. We mod for composite video and throw away the inwards of the modulator. How many of us use the original PSU? The Recom regulator at least it is not a eye sore. Some use cheap dc-dc step down buck converter modules, and those really seem out of place.
DEATH's advice to use two rectifiers to lower the voltage is sensible enough but with it we are not keeping the machine original either. It might be a better choice for aesthetics in the context of a restoration, though.
/Pedro
This is a pro, not a con in my opinion :)
/Pedro
Technicaly speaking it is a pro. Retro preservation speaking its a con. Most of my speccys are preserved as they were with only new caps installed, but I have a couple of them that are "technicly advanced" as in the picture i posted on page 2.
http://www.vintage-computers.it/4116-ram-replacement-module
I don't think anyone needs my approval, it's not an especially unique idea. It isn't a copy of my circuit board layout either, though it appears this one is far more complex than it need be. I shouldn't think that'll affect its operation though.
I wonder whether it has been tested in an issue 2 Spectrum. Careful selection of both ICs is necessary for there not to be a problem. It looks as if the latch is a 74HC series and that won't work reliably in an issue 2.