Computer ZX-777
ZXNet echo conference «hardware.zx»
Hello Mick
Mic> In my understanding, this is so. Processor memory access cycle 3 or 4
Mic> beat, naturally without WAIT. Phase H1 =0 (~1.75 MHz) contains 2
Mic> processor clock cycle. Naturally, he manages to either read out or
Mic> write. The remaining 1 or 2 measures appear on the video, but this is no longer
Mic> scary. So, if you increase only the processor frequency, then in phase
Mic> H1 =0 all 3 or 4 bars will fit, but on the video you need to set WAIT.
The principle is something like this. The Z80 processor usually processes commands in 4 clock cycles
CLK, or a multiple of four. Of course, there are commands executed in 7, 15 clock cycles,
but rounded it is a multiple of 4. In Spectrum everything is implemented in such a way that
4 clock cycles of the CLK frequency, we could access the memory 2 times with the video controller (for
byte of pixels and byte of attributes) and 2 times by process. Since in 4 clock cycles, usually
only one clock cycle is occupied by the processor and the possibility of turbocharging has appeared -
the second clock cycle at 3.5 MHz with a probability of ~50-80% turns out to be idle for memory.
Thus, when increasing the processor frequency to 7 MHz, we get
performance is at 150-180%, compared to the usual 3.5 MHz. Most
good (greater performance gain) the Pentagon is turbocharged, due tofeatures of its circuitry (buffering of video controller data).
Hello, CHRV
Hurray, I saw the turbo in action on the z84006, although not for long. The processor seems to be broken.
At first it works as it should for a minute, then it works for about five minutes in basic 48, and then
is completely cut off (it also behaves in normal mode). In short -
The speed of the keyboard is impressive, but I'll remove the snow on the screen and need a disk drive
try. But this is after I replace the processor. Our 1858VM1 in quantity 3
pieces did not start in the turbo at all. Percent z80A - I tried to show something there, but
lost his temper.
Hello Mick
Mic> Our 1858VM1, 3 of them, did not start in the turbo at all. Percent
Mic> z80A - tried to show something there, but failed.
I use in Turbo2+ z0840008
Hello, CHRV
CHR> I use z0840008 in Turbo2+
Well, about this matter - in the last letter on the soap, I already agreed with you.
:)
Another question - there is snow on the screen, maybe I missed something? Maybe it depends
series of logic chips (155, 555, 1533...)?
Hello Mick
Mic> Well, about this matter - in the last letter on the soap I’m already with you
Mic> agreed. :)
Mic>
Mic> Another question - there is snow on the screen, maybe I missed something? Maybe this
Mic> depends on the series of chip logic (155, 555, 1533 ...)?
Apparently wait in the turbo is not working correctly. Or it doesn't work at all.
Hello, CHRV
CHR> Apparently wait in the turbo is not working correctly. Or it doesn't work at all.
That's right, it doesn't exist at all. I use a full turbo - all frequencies work with
memory (H0, H1, RAS) are increased 2 times, fortunately statics allow such speed
work. Moreover, it works normally and does not go astray.
It seems I found the reason for the snow, it is not recorded correctly in the turbo
video information into the IR16 registers, such as a loop of one or two pixels, is obtained.
I'll think further.
Hello Mick
Mic> So, if you are interested, please contact me - I will send it.
Or maybe it’s worth putting it on the “List of Domestic...”?
Hello Mick
Hurray, thanks to Roman Chunin I now have 8 MHz processors - now work on
turbocharging will resume. While I was waiting for them, I finished restoring the circuits from
this computer. There were two of them. Conventionally, I called them option 1993 (scheme
which I had) and the 1994 version (sent by Conan). In short, I painted them in
pdf version. So, if anyone is interested, please contact me and I will send it.
Hello spensor
spe> Or maybe it’s worth putting it on the “List of Domestic...”?
I can go there too, I’m just “throwing in the bait” - in the sense that if there are no objections -
I'll throw it in.
Hello Mick
While I was fiddling with the turbo, by the way, the frequency of the processor and memory was doubled
Impressive - the screen is cleared instantly. But there are still some shortcomings. A
I just can’t run it on the imported version of the Z80 (Z084008). Removing the turbo - there
didn't work either. But by returning the old processor clock generator, namely
the RAS/ signal and CLKZ were equal - miraculously it worked. But in this situation, static
does not work stably. Now the driver is such that the frequency CLKZ is ahead of RAS/
half a beat. Moreover, Z0840008 refuses to work, Z840006, which
I thought it was broken, it was unstable. However, our domestic 1858VM1 works
It’s stable with both the old driver and the new one. So the question is - who is the thread
will tell you information about our 1858VM1, preferably with diagrams to compare how they are
are different. Are the bourgeois people really so capricious - I’m already starting with ours
good :) :)
Hello Mick
I continue to scream into the void :mad: . Where are you Z80 experts who are in the forum
great multitude. Or is this just empty talk.
I give food for thought - after doing some research I found
1858VM1 - analogue of the regular Z80 without letters;
T34VM1 - analogue of similar Z80A
German U880 - something similar in architecture to Z8400B.
I drew conclusions from the following attachment.
All processors worked according to the old scheme.
Only 1858VM1 and Z80 worked with the new one.
Z80A, Z0840008 - black screen.
Z8400B and U880 - worked, but with constant amnesia and were very eager to
BASIC 48.
In total, it was studied
1858VM1 - 3 pcs.
T34VM1 - 1 piece.
Z80 - 1 piece.
Z80A - 1 pc.
Z8400B - 1 pc.
U880 - 2 pcs.
Z0840008 - 5 pcs.
From all this I conclude that some durations are longer on the 1858VM1
than the Z80A. So, experts, tell me which ones? I hope for an answer.
File: clkz.rar http://zx.pk.ru/attachment.php?attachmentid=2805
Hello Mick
Mikhail, I would look at the listed problems from a different angle. The point is
that the likely causes of failures are differences regarding buffering (or its
absence) for Z80 from various manufacturers. I came across a review article (no longer
I’ll remember where, but maybe others can tell you) for different Z80s. In particular
internal buffering of CLK can lead to delays relative to it all
processor output signals. And buffering the address and data buses gives
specific delay of only these signals. In addition, technology (not to be confused with
topology) Z80 could be different, which also led to variations in characteristics
(for example, due to different levels <0> and <1>). In a critical time period
mode, all this could come out (and did come out before) in the form of various failures. And
about increased (compared to passport) frequencies is a separate conversation.
Hello Conan
Con> Mikhail, I would look at the listed problems from the other side.
Con> The fact is that the probable causes of failures are differences concerning
Con> buffering (or lack thereof) in Z80 from various manufacturers.
Con> I came across a review article (I don’t remember where, but maybe others
Con> will prompt) on various Z80. In particular, internal buffering of CLK
Con> may cause delays in all output signals relative to it
Con> processor. And buffering the address and data buses gives a specific
Con> delay only these signals. In addition, technology (not to be confused with
Con> topology) Z80 could differ, which also led to scatter
Con> characteristics (for example, due to different levels <0> and <1>). In critical
Con> from the point of view of temporary mode, all this could get out (and did get out
Con> previously) in the form of various failures. And about the increased ones (compared to
Con> passport) frequencies have a separate conversation.
Thank you, you just responded, my brain is already racing. But I'll explain all this
tested at normal frequency. At elevated temperatures they can only workZ0840008(black screen) and Z8400B(intermittent amnesia). It only occurs to me
one assumption is that the durations of RD/ and MREQ/ in 1858VM1 are longer
longer than the Z80A. And perhaps this is why the data flies past the processor. But
this requires confirmation.
Hello Mick
Today I started shortening INT - damn it, I’ve already plowed half the circuit with modifications.
By the way, as I understand from the Z80 documentation, the duration of INT should not be shorter than itself
the longest Z80 command in terms of clock cycles, namely 23 clock cycles. Well for a simpler one
processing probably did 32 clock cycles (3.5 MHz). In short, we shorten them in the turbo
doubled.
But before that I decided to play a trick - I ran the test from Caro version 1.1 - the result
measurements on all pages of 139776 parrots: v2_finge; .; Judge for yourself
turbo capabilities. After I shorten the INT, I will also make a measurement.
Hello, The Exploited
The> well, everything is quite obvious - 224*312*2=139776
The> on different clones in different ways. from 130 to 170% generally.
I wish I could measure them in cycles with the same test for comparison.
The> some work fine, some falter. the question is not understood.
The> in general, in turbo it’s nice to watch 3D demos with correct synchronization
The> (refresh, dogma, Voodoo...)
Well, for example, 7Reality works correctly (synchronized), but Illusion,
shock - it wobbles (the effects are eaten).
So I asked, is this how it should be or is it just me.
Hello Mick
that's how it should be
Hello Mick
Today I decided to try different processors in turbo mode. I installed 1858VM1 and T34VM1
- damn they bastards, they earned it, but the bourgeois Z80, Z80A and U880 didn’t even
turned on.
Here is the paradox: either ours are so clumsy, or theirs are too capricious.
But it’s still nice that the processors operate at a higher frequency.
Hello Mick
Well, everything seems to have made all the changes to the diagram. Therefore, I am posting my findings.
Maybe someone will use them. So I present to the public the updated
computer "ZX-777 128Kb Real Turbo".
You can kick your feet and say that this cannot be - but it really is
really works ;) .
File: ZX-777 2006.part1.rar http://zx.pk.ru/attachment.php?attachmentid=3368
File: ZX-777 2006.part2.rar http://zx.pk.ru/attachment.php?attachmentid=3369
Hello Mick
Oh yeah, I forgot, in a fit of joy, to inform you that this computer can function
only with memory, access time no more than 20ns. It may work at 35ns -
I didn't have any of these in stock. Memory 55ns and more will definitely work
there won't be.
Now the idea is to use PAL AT16V8B in the memory manager in order to still expand
range of chips used and expand memory up to 512Kb.
I've cooked up the firmware, now all I have to do is solder it all together. To access memory
Above 128Kb I plan to use bits 6 and 7 of the 7FFDh port (someone else’s standard).
Hello Mick
Damn, PAL turned out to be interesting things. All multiplers (KP13 and KP11), which
work with memory stuffed into PAL. I'll drive for a few more days - I'll post mine
developments. Only for now it’s working again with fast memory - there’s still some left
push the logic of working with memory - maybe then times will even out. But in general
interesting.
Hello Mick
Mic> - it also has one PAL.
But is this EPM PAL?
I would like this firmware myself - my green board is lifeless
Hello Mick
I was looking at the Scorpiona diagram - it also has one PAL. By the way, not at
Who doesn't have PAL firmware lying around? It's interesting to see how people used to decide
circuit design problems.