Page 1 of 1

Replacing a 4116 VRAM with a more reliable 4164 - Courtesy o

Posted: 13 Jun 2014 14:31
by Dave
Hi Folks,

as we know, the most common video problem with the MTX is failure of 1 or more of the ITT 4116 VRAMs.

These, and other 4116, RAMs are notoriously unreliable and users of other machines have found various ways of replacing them.

Courtesy of Martin, this page describes one way of doing it on the MTX :- ... mavram.htm


Re: Replacing a 4116 VRAM with a more reliable 4164 - Courte

Posted: 11 Jul 2014 21:45
by Crazyboss
Great work Martin and Dave.

Could it even be possible to remove all the 4116 and replace them by a single chip?

Could replacement mean quicker access to the VRAM ? or is it the X-Tal to decide the speed?

Re: Replacing a 4116 VRAM with a more reliable 4164 - Courte

Posted: 11 Jul 2014 22:25
by Martin A
The VDP is designed to accept dynamic ram, fitting static ram is possible but messy, and needs several extra chips. to handle the multiplexed address bus See ... 8+VRAM.pdf

Two 64k by 4 bit wide dynamic ram might work, but the refresh could be an issue, 4116 and 4164 used 128 refresh cycles in 2ms, the 61464 datasheet I found says that one uses 256 in 4ms, so may not work. The combined D and Q signals are another potential issue with the 4 bit wide DRAM

And that wont affect the vram access speed at all.

It might reduce the current consumption and the heat in the case though.

Replacing 4116s with modified 4164s reduces the power consumed by the VRAM by about half.

Re: Replacing a 4116 VRAM with a more reliable 4164 - Courte

Posted: 12 Jul 2014 12:14
by 1024MAK
Ahh, refresh. In hindsight do you think Zilog spent some time kicking themselves over the decision to only have a 7 bit refresh counter?

The big difference between 128 cycle / 2ms refresh and 256 cycle / 4ms refresh is that the first is 7 bit (uses A0 to A6 of the address bus) and the later is 8 bit (uses A0 to A7 of the address bus).

As the VDP in the Memotech is designed to work with a 16k DRAM bank, this is not a problem as long as the correct address lines are used on the larger DRAM chips. As it does not matter if the area that the VDP cannot access is not refreshed ;)
Note that the correct address lines may not be what you think they are, as it depends on the internal layout of the chip.


Re: Replacing a 4116 VRAM with a more reliable 4164 - Courte

Posted: 05 Aug 2015 07:51
by protek
Nice work guys! I've done the same hack on one of the Spectrums that needed the lower RAM replaced. I had an easy access to 4164 chips, so I went and modded those.

Re: Replacing a 4116 VRAM with a more reliable 4164 - Courte

Posted: 06 Aug 2015 14:36
by protek
The use of 4164s is a cheap and relatively easy way to replace the faulty 4116s but the Spectrum hobbyists have a more elegant solution too:

Of course this can't be directly plugged into an MTX due to different arrangement of chips but I bet the same schematic could be used to make a PCB that would sit on top of 2x4 chip arrangement.

Re: Replacing a 4116 VRAM with a more reliable 4164 - Courte

Posted: 06 Aug 2015 14:50
by Dave
If there is any demand (i.e., for more than 1), I would do this


p.s., I had thought about doing this previously, the thing that put me off was the fact that the user would have to remove the existing VRAMs. That is maybe not so much of a problem if the machine has a VRAM fault and is broken anyway, but I'm not sure how many people would want to do it as a pre-emptive measure.

This has also just reminded me of something that I read on this page a while ago . . . . .

At the bottom of the page is the following . . .


Memotech MTX512

Richard Gellman ("splodge" at the site "") writes: "Thought you might like to know, its been used successfully on a Memotech MTX 512 as a replacement for the main RAM. The only caveat is that the wires on LK6 need to be set horizontal and EAST. By default they are set WEST, which prevents the mod from working. LK6 is reminiscent of the RAS or CAS delay selector on the later ZX Spectrums (used to delay either RAS or CAS to suit DRAM timings). The parts used were a 74HC574, 74F04 and AS7C512-15 64kx8 SRAM. All surface mount parts."

There's probably an easier way to use SRAM for MTX512 main memory, but apparently the MTX512 schematic is Missing In Action ??

TMS9918 video RAM

4116 DRAMs were often used as video RAM connected to a TMS9918 (or related type) videochip. This thread on the forum made it clear that you can't use this circuit for TMS9918 video RAM. The TMS9918 sends address and VRAM write data over the same lines, reads VRAM data over another set of lines (using the 4116 separate 'Q' outputs), and uses some 'non-obvious' signal timing to make this work. An SRAM as video RAM is surely possible, but would likely require an address- and data latch. And it's unclear at this point how to arrange / control these latches. To be continued..?

So, more research required I think

Re: Replacing a 4116 VRAM with a more reliable 4164 - Courte

Posted: 07 Aug 2015 05:51
by 1024MAK
A suitable SRAM can be interfaced to the main board without too much hassle. No need for a multiplexer chip or latches/ register. If a daughter board was designed, it would plug into all the DRAM sockets AND the existing address multiplexer socket.

Then all required data and address lines would be available.

Alternatively wires (maybe a IDC DIL socket and ribbon cable could be used if these can be bought at a sensible price), could be used so a smaller (cheaper) PCB could be used.

The DRAM used for the video RAM can be replaced with a SRAM, but a number of 74 series logic chips are needed to do the signal conversion. So while 4116 chips are still available, it's not really worth it.