Page 1 of 1

Re: New Spectrum 48k

Posted: Tue Jun 30, 2020 3:54 pm
by peloritano
payty wrote: Tue Nov 12, 2019 6:10 pm I have a few boards that are not working properly but the PCB is not damaged and can be fixed.

So I am trying to build this one with less recovered components.
I am very curious about the results. I know a guy that built one of these but has some noise on the video signal.
Have you completed the job? I also had problems with video signal disturbances ...

Re: New Spectrum 48k

Posted: Tue Jun 30, 2020 8:00 pm
by payty
I've got most of the parts in place but haven't powered it up yet.

today I have received the nebULA module that goes in place of the original ULA.

Re: New Spectrum 48k

Posted: Mon Jul 13, 2020 11:04 am
by payty
For the lower ram I have ITT 4116-4N chips which are 250ns. The Sinclair repair manual states that the lower ram should be 150ns.

So far I have used the 250ns chips to repair some of the computers from my collection which had dead lower ram chips. And they appear to be working (tested them 30-60 min with the Retroleum test ROM). At least 5 computers have been repaired like this.

Should I expect to have any problems because of the different specs?
Are there any 'real world' tests that I should perform? I am thinking about certain games that could not work because of these, or other situations.

Re: New Spectrum 48k

Posted: Mon Jul 13, 2020 12:02 pm
by 1024MAK
About the speed rating / access time of chips: typically a manufacturer will aim for a design speed when manufacturing the chips. However in production, some will work okay at the design speed, some will be slower and some may be faster.

Hence before putting the chips into their encapsulation, they are graded. The faster chips are then sold at a higher price and the slower chips are sold at a lower price (often there being far more slower chips than faster chips).

As time goes on, improvements in the production process normally result in more higher speed chips compared to lower speed chips but in order to continue selling the higher speed chips at a premium, some are not always marked up and sold at their actual real speed rating.

You should also understand that the testing done is fairly vigorous, but the resulting grading is very conservative.

Hence a lot of 1980s and 1990s chips can be ‘overclocked’ far more than people expected.

So if they work, then that’s fine. But do keep in mind that you are not the first person to have tried using slower DRAM chips. It does not always work for some makes / batches of 4116 (or equivalent) DRAM chips. The same applies to fitting slower DRAM chips in the ‘upper’ (extension) RAM positions.

In terms of ‘real world’ tests, in a ZX Spectrum (just like most of the microprocessor systems of this era), the Z80 and the memory always runs at the normal system speed. There are differences in timing between the Z80 fetching an instruction and reading / writing data, or reading from or writing to an I/O device. But it’s doing all of that anyway as a result of the ROM code.

The device in a ZX Spectrum that really sets the required speed of the ‘lower’ RAM is the ULA. As it needs the RAM to be fast enough to grab the screen data on time, every time.

So in practice if it runs a diagnostic / memory test okay, then try it with any machine code games you have. Preferably games that either use the floating bus or which have fancy multicoloured effects or draw in the border area.

Mark

Re: New Spectrum 48k

Posted: Mon Jul 13, 2020 5:21 pm
by payty
Thanks a lot for detailed explanation!

Fortunately, the Retroleum rom has a floating bus test, but I will test the computer with some games as well.