Black lines between colours (demo effect)

[Stefan jL]

My ST has the habit of most of the time show black lines between colours in special demo effects (don't know the technical term for it )
It does not always happen, what could it be? RAM problem? From what i know it has always behaved like this... se pics for examples:

[simbo]

i think this is ram related ive seen it before in a machine last year had faulty ram in the video area however
or one or more chips pins are maybe oxidised open it up and press all the chips in tight to there sockets if this isnt the case
and youve checked the psu out then its morelikely the ram has a flaw

[Nyh]

My ST has the habit of most of the time show black lines between colours in special demo effects (don't know the technical term for it )
It does not always happen, what could it be? RAM problem? From what i know it has always behaved like this...

It is a wake-up mode of the shifter. The same problem creates black pixels in spectrum pictures.

Hans Wessels

[Stefan jL]

So it's a bad shifter then? And yes Spectrum 512 pictures can look messy sometimes.

[Nyh]

So it's a bad shifter then? And yes Spectrum 512 pictures can look messy sometimes.

No, nothing is wrong with your computer. AFAIK it just happen with most ST's now and then. Switching the computer on and of will most of the time boot the computer in the other wake-up mode.

Hans Wessels

[MiggyMog]

could it be a bit of both?

it looks like an extreme example of the wake up state issue which causes speckles in Spectrum 512 pictures.

Even with my speckle affected STEs I don't get that problem.

If it's broken memory then it seems strange that the program runs at all. Even if it was localised to where the screen memory you would expect it to crash when moving screen blocks around. The fact that the program likely only changes one colour register for the plasma effect & seems to succesfully do it for most of the VBL points to a timing problem rather than an actual storage problem?

Could it be caused by PSU or problems with cap's and crystal timings simbo?It seems the memory can read/write?

It definately looks like the colour switch is taking longer than normal?

From what I read somewhere about the shifter it stacks up the colours & shifts them in sequence so that it is possible fo r one colour say, colour 1 to show what is held next in line (colour 16?) which is why the wake up state causes speckles.

What machine are you using? if it is an STE you could swap the SIMMS.

Intresting one.

[Nyh]

If it's broken memory then it seems strange that the program runs at all. Even if it was localised to where the screen memory you would expect it to crash when moving screen blocks around. The fact that the program likely only changes one colour register for the plasma effect & seems to succesfully do it for most of the VBL points to a timing problem rather than an actual storage problem?

It is not a memory problem. The problem happens when a colour register is altered at the same time as the shifter is displaying screen data and has to use that colour register.

It definately looks like the colour switch is taking longer than normal?

My guess is the color change is off by one clock cycle.

Hans Wessels

[Greenious]

Yes, that particular effect is most likely caused by wakeupmodes.

This thread in the 68000 coding subforum, explains it pretty good.

It is a very interesting "flaw", if my Atari had it, I dunno if I really would like to "fix" it.

[insanity]

My ST has the habit of most of the time show black lines between colours in special demo effects (don't know the technical term for it )
It does not always happen, what could it be? RAM problem? From what i know it has always behaved like this...

I used to have this happen, but only on my 1.02 TOS ST... The 1.0 TOS ST was fine... I seem to remember it
being a corrupt intro for an Automation CD by Monster Inc. Jupiter Masterdrive on it?!

Interesting though, I thought it was just my ST playing up!

[PaulB]

Wasn't there a Sync program that came with Spectrum 512? It was supposed to be auto booted and would stop these lines from showing up when displaying Spectrum pics. Maybe I'm thinking of something else!

[unseenmenace]

Wasn't there a Sync program that came with Spectrum 512? It was supposed to be auto booted and would stop these lines from showing up when displaying Spectrum pics. Maybe I'm thinking of something else!

I think that program just let you know whether your ST was correctly in sync or not. As far as I'm aware the only cure is to power off and then on and test it again to see if its in sync or not.

[Stefan jL]

Another "problem" with my ST is the fullscreen menus of "European demos" and Nostalgic-o- it's just shows graphic garbage.

Earlier versions of Nostalgic-o- did not even start the fullscreen menu and i got a message that i was using "PaCifiST" and i should try it a real ST
Maybe this is a sync problem then?

[Nyh]

Another "problem" with my ST is the fullscreen menus of "European demos" and Nostalgic-o- it's just shows graphic garbage.

Same 'problem'. If it gets in the right wakeup mode all will be fine.

Hans Wessels

[MiggyMog]

Get it in the wiki guys, you might get hit by a bus tomorrow & us stupid people will have no-one to ask

[stimpy]

Do you have to power off/on the ST to fix this, or does a reset work?

Reason I ask this is because not all parts in the ST are reset when the reset button is hit.

[simbo]

after fully reading about this issue
i see lots a fairy storys about wakeup states {smoked italian sausages}

i recon its gota be the startup voltages of a cold tired out psu current resoviours {capacitors etc}
same old story

and moreover the amount of current avalible {cold current}

so i dug out my stf and did some tests and sure enough mine had no issues and was an older machine

now ive reworked the psu on this machine replaced all the capacitors AND the optical isolator

on another stf i hadnt done i was able to again service the psu and replace the caps and the isolator
and low and behold the problems were gone capacitors dry out with time and heat/use
till there is no liquid electrolyte in them and there values will reduce hence they dont hold as much current resoviour

in working around the board i also added more capacitors to the +5v rail at the mmu +5v and the shifter 100uf@16v
and replaced C179 on the main board next to the psu socket 4700uf
it seems rarely gets changed by even professional repairers
but it will dry out the same as the others only faster becouse its bigger
so replace it
a full service like this needs done every 10 years maximum !!!!
this seems to stabilise the video mode wakeup rubbish although i think its just the 32 mhz clock slightly off frequency
as this machine also had crystal drift so i replaced the crystal Y2 32mhz {32.084988MHZ}
also calibrating the video clock y3 4.433618MHZ using C98 {small green or red trimmer} pal machine ntsc has different values
{i use a shaved down plastic knitting needle or you can use a brass trim tool NOT a screwdriver}

if the +12v rail or the +5volt rail is a low voltage lower than 5.1 or 12v
or has any noise on it above 50mv ripple / noise this will effector both clocks greatly as one syncs with the other
the 32mhz being controlled by varicap diodes

i bought ten of these crystals from knight crystals lab in england for £15 + post
but you can get the new unused original atari part from BEST electronics for $10

it makes sense the current creaps up with heat as most people report its heat related

and most atari capacitors inorganic low tempreture types exhibit less capacitance till they warm thru
while i was there i also checked what voltages the psu produces
and the +5v rail was 4.73 although the +12v rail was 12.1 i can live with 12.1
but 4.73 this is not good enough you should have at least 5v i set it always for +5.2v {disk loading} when it can be adjusted or look for the reason it isnt if it cant {usualy the 1uf caps or smaller non polerised types

so i changed the rectifier also and this seems to have fixed the low voltage {see diagram below part DB1 replace it}

it appears the older rectifier {which still worked fine} had a 20 or so MEG R leak across one of its diodes
this will cause a poor quality 350v DC chopper supply
and also polute the output rails with paracitics and transiants {that gets dumped to earth mostly}

so i cleaned all the sockets for the psu cable with a burnish brush and replaced the mains cable
and re checked all
and it appears to have fixed this issue

hint.....perhaps its just a psu thing you have

if you look in the simbo faq there are circuits for the stfm and there is a guide to service it
i used high tempreture 105deg caps for all and the new bridge rectifier was a 2amp type and not the older round one or squareish package type
that was in it
its interesting that this psu wasnt the same as my older stf first discussed
so it maybe that the later psu design was not as well made as ive found before
.....
also check the actual on off switch sometimes it can provide lesss than 240v zzzzzzz to the rectifier and this damages it
i always use another switch {the wall} to turn on/off the atari

the total cost to do this was about £15 or so i buy in bulk so it was more like about £6

[Nyh]

Another "problem" with my ST is the fullscreen menus of "European demos" and Nostalgic-o- it's just shows graphic garbage.

Same 'problem'. If it gets in the right wakeup mode all will be fine.

I was wrong on this point!

Found my Mega ST in a wake-up mode that doesn't display the main menu of the Nostalgic-o-Demo correct, but... the colors were displayed correctly. So the must be another wakeup mode. I never payed much attention to this problem but now I will watch it more closely.

Hans Wessels

[simbo]

listen crew

This must be on reflection and seeing the interest
some sort of carefully thought out practicle joke

man its not 1/4 /2008 yet
if hardware has tempreture related problems
it is faulty
period ....!

these stf and mega machines are well past there even 7 years in use stage {10 - 30,000 hrs uptime on a good mains supply }
where the capacitors are now DRY and devoid of elecrolyte fluid
if you havent serviced the psu and main board for the 7 -10 years {or less} then DO IT

im really shocked someone writes a piece of software to measure a state of a bus colussion
when obviously there is a problem with the psu not providing clean correct voltages {the blood of the machine}
with enough red and white cells {the current and special timing switchmode supplies provide...
frequencies called harmonics that digital machines need to function clean}

this type of problems cermounts to a insulin dependance senario in the body

so basicaly fix the dam psu and forget this wakeupstate rubbish !

ALL the machines {hardware} your using is unstable if its psu hasnt recently enough been serviced
and is devoid of i would say about 20-40 cc of electrolytic fluid that has evaporated from the resoviour
your towns are starving and the hose pipe ban was in place long ago in terms of the cold current you have avalible

servicing a machine is a MUST

these type of subtle timing problems are a sad reflection on whats to come for you all in hardware time left
if you dont change the caps and other components that are exposed to stress
like the bridge rectifier and optical isolation
a component that comunicates with an led
and therefore wears out much like laser leds etc etc led signs and lights 50,000 hrs and thats it gubbed
as for the caps about 30,000 hrs on a good day

so the caps wear out and like anything with voltage rails full of crap will blead from its viens

i think the point is
if youve reached this point
then you need a wakeupstate yourself and so does your psu ...

what total rubbish this topic is and not anything i wouldnt expect from tired psu's

for mega machines its more likely the psu is all thats to blame as there is only the main shifter clock and its an oscillator block
the 16mhz clock is provided by the shifter and further divisions made
BUT they also suffer from instability on switch on a known problem with this type of clock {falcons suffer from it also }
low output from its buffer and also the crystal drifts by a few khz

even if this is the case just a change of 32khz at 32mhz is {32/8=4 :. total drift @ 4mhz = +- 32khz min X8 at 4mhz or 128khz
becouse of the way the chips divide the frequency using exclusive or gates so its 256khz total range for just 32khz errror or 3.2% of 32mhz
over ten years these type of oscillators will drift by as much as +-10% !!!!!!

that is enought to confuse a bus and cause colussions for sure !!! every bus right the way down the chain
and together with a bad psu start up will be slow and the machine jerky
till its warm
and this lenght of time increases the longer you fail to address the cause

remember yer old grandad used to put the tv in the cupboard for a few days and try it again {magic pixes used to come and breath life into the caps etc..}

it was simply in need of a service and the parts that wear out changed !!!!



the final solution

to fix it once and for all once you fix the psu in mega machines {this doesnt apply to stf or ste etc just mega and falcon}
change the oscillator block for a high current sink buffer type
then isolate its supply pin +5v except the decoupling cap...
make a small regulator stage using a 78L05 500ma 5v regulator
and dirrive the +5v from the 12volts rail
adding some capacitors like 470 uf in input and 220 uf on its output
this will ensure the clocks will always be clean and run from +5v no matter what
as old ttl clock are very sensitive to lower than 5v supplies
forinstance reducing the voltage to 4.5 you will loose half its output
if infact if it even keep's oscillating

at peek current use
the +5v rail will dip to as little as 4.3v or less and if you have hard disk drives even lower
when there is bad capcitors {current storage resoviour that buffers this voltage swing}
i use atx psus to avoid problems with atari here ....but even these can be low rail voltage again with age

for stf machines again after a full psu and board caps replace and service

simply the 12v rail provides power to the 4mhz interlace with the 32mhz main clock
so.... turn the 32mhz crystal y2
the other way round {take it out and turn it round} solder it back
then youll see
now replace it as it will just drift back faster as crytals do with time....

[MiggyMog]

Ok we were talking a while back about diagnostic tools. I think NYH has discovered a method of diagnosing the symtoms with software, so this is not so much useless as it could reveal the hardware fault, seems we we could use a new benchmark type utility which checks timing between chips etc.

I think that some STs were shipped with crappy components though, my STE had a bug from day one with SPU/SPC and the one I got since has the same fault.

Simbo, maybe you should sell a PSU maintenance kit with parts,instructions/video

Or an adaptor for an ATX PSU, or that one Techi Alison was pointing out before..

I am armed only with a multimeter, a soldering Iron (likely of the wrong rating) and a small brain

[simbo]

hum

i do agree its a good tool for diagnosing hardware issues
but so is a hand full of caps once every 5 years
and it should take anymore than running a game
to see the issue that isnt normal to the platform its designed in
to realise there is a hardware issue

software fully depends on chipset timing as the software contains its own timings setup
on hardware thats aok


so its 50 50 and yes its usefull i suppose



a kit to fix psu's etc...

do you think people would want a kit of bits to do this ???

if so then possibly if i got some help i would ...if a few vendors were around the place and a standard agreed on by machine and psu
there are so many versions i think 15 or more accross the platforms

i think the first thing i would do
is make a board to replace the oscillator crystals with a system programmable phase locked loop

used in 286 and above machine's and macs since 1993
but this was only hardware adjustable later machines p3 and above and powermacs
were able to trim there clocks from software subsystem directives

i think youll find the bug you have with ste's is to do with software and not hardware
and the way the blitter get used or not... and im sure this exists within the falcon / TT also

the ste machines use a slightly different clock arrangement from the st series and this maybe isnt cattered for in software

can you send me some full details of this issue
and ill give it a go on my ste machines {thathave been heavily upgraded }
and see exactly this issue

im keen to help resolve these niggle for sure !!!!

[Nyh]

after fully reading about this issue
i see lots a fairy storys about wakeup states {smoked italian sausages}

i recon its gota be the startup voltages of a cold tired out psu current resoviours {capacitors etc}
same old story

so i dug out my stf and did some tests and sure enough mine had no issues and was an older machine

Yep, the problem isn't with any ST. Although a good PSU is important I know from experience the wake-up problem isn't just a simple power supply problem. Back in the old good days we had a party at Lord Q's home. It was quite hot and during that party Insh_Allah's ST got problems. Vertical black lines on the screen. Shifter was overheated. As Insh_Allah doen't like his ST to become overheated he attacked the problem in multiple ways. One of them was to turn down the 5V supply voltage to the low side of the TTL level. Somewhere around 4.9 volts if I remember correctly. Experimenting with the power supply voltage didn't cause wakeup problems.

If you theory is correct you should be able to cause wake-up problems by getting your old perfectly working STF, turn down the power supply voltage and by that induce the wakeup problems.

this seems to stabilise the video mode wakeup rubbish although i think its just the 32 mhz clock slightly off frequency
as this machine also had crystal drift so i replaced the crystal Y2 32mhz {32.084988MHZ}

Sorry, I cannot follow this reasoning. How can the master clock being a bit of create the wake-up problems. All relevant timing is derived from the 32MHz master clock. It being a bit of causes all timing to change a bit and should affect wake-up modi.

these stf and mega machines are well past there even 7 years in use stage {10 - 30,000 hrs uptime on a good mains supply }
where the capacitors are now DRY and devoid of elecrolyte fluid
if you havent serviced the psu and main board for the 7 -10 years {or less} then DO IT

Even before the 7 years limit the wakeup problems existed. When the Antic released the Spectrum 512 'true color' drawing program in 1987 the wake-up problems with the colors were already a known issue. This is 2 years after the introduction of the ST.

Your 10 to 30.000 hours uptime is a good guess. I think most ST's even don't come close to the 10000 hours uptime unless they were running 24x7. If a ST isn't running good the first thing to look at the the PSU. But a bad PSU will result in program crashes and (especially) a flickering screen during disk access.

Hans Wessels

[Nyh]

Found my Mega ST in a wake-up mode that doesn't display the main menu of the Nostalgic-o-Demo correct, but... the colors were displayed correctly. So the must be another wakeup mode. I never payed much attention to this problem but now I will watch it more closely.

Well, I got my Mega ST running in the black llines mode and the wrong wake-up mode for fullscreen.

When I first switched on the main menu of the Nostalgic demo wasn't running correctly (the fullscreen intro and the other fullscreens in de demo were OK as were all the fullscreens in the Dark Side of the Spoon by ULM). In the sprite blast demo the color switch errors were visible. I took some pictures.

Resetting the computer doesn't solve the problem. Both fullscreen error and color error continue to exist.

I switched off the computer for a short time (about 5 seconds (note: this is not good for you PSU)). The fullscreen error was still there but the colorswitch error was gone.

I switched off the computer for a bit longer time (about 20 seconds). The fullscreen error was gone but the color switch error was back again.

Conclusion: fullscreen error and colors witch error are completely independent from each other.

Hans Wessels

[Desty]

Wow, the fullscreen error is really bad.

[edit]
Woo, 512 posts!

[MiggyMog]

i think youll find the bug you have with ste's is to do with software and not hardware
and the way the blitter get used or not... and im sure this exists within the falcon / TT also

Yeah the 68000 & wakeup threads identified that certain methods can be used to avoid it occurring on different STs

Dougie Little also mentioned it in the Photochrome help file that he was able to avoid the speckles.

NYH how is the PCS disection coming on for the file formats in the wiki?

[simbo]

hi

now i have a clear idea whats going on i see the issue and how to fix it

the main clock is divided by the shifter to give the 16 mhz this is further divided by the stb to 4 and 8mhz
assuming there is a phase missmatch between the 32mhz clock and the 16 or that the time delay of the 16mhz clock is too allow output mode to be registered
its safe to assume that the shifter is initialised by its reset then the next command loaded to it
then enable clock output

i think the shifter is slow to initalise and initialised to soon before the rest of the chipset
i think supply of the 16mhz clock via a /2 stage using a 74HCT74 flipflop

see attached circuit

this way a dil switch can be used to supply the correct phase of clock and even the other flipflop in the package
used to supply the 4 and 8mhz clocks

however the addition of an added inverter chip would also give the ability to switch the phase of the main ttl 32 mhz
a 74hct 04

its worth some experiment using jt a simple three chip ttl circuit to narrow down the problem with the chips startup

i think corrupted data gets shifted in becouse of a waisted clock cycle
and using a divider then and it with the shifter clock may fix the problem thats loadeding and incorrect control word
to the shifters registers


so to sum up in a mega ste i would definatly sum the 16mhz with the 32 /2
this will give a nice clean 16mhz clock

one flipflop and one and gate