Need some help fixing CPC6128 not booting

[ikonsgr]

Finally today the desoldering gun arrived along with the 6845 replacement!
I solder the new 6845 but unfortunately still, no picture at all 
Now, according to the service manual the "no picture at all" symptom can caused by faulty:

Z80
6845
40010
RAM
CRYSTAL

CPU,6845 and 40010 i use are 100% ok (i know for sure because i have a second amstrad cpc where all these work ok) so next,is crystal. I put a new 16Mhz crystal but still same symptom. So the only thing left is ram!
The problem is that none of my ram chips seems to get hot, in fact one row doesn't get hot at all, and the other is heated very little!
Because ram change is a lot of work,and it would be very disappointed to have no success again, i was wondering if there is anything else on amstrad motherboard that could cause the "no picture at all" symptom. For example, could a  faulty 74lsxx chip cause this symptom?
And if i end up on replacing the ram chips, since i have no heat difference among chips meaning there is no way to know which are the faulty ones,i was wondering, if i replace the 1st bank only (i think it's the right row of chips closer to the 40010, as these seem to get hotter on both amstrad cpc's i have) would i get any picture even if there is a faulty ram chip on the second bank of 64k?

[TotO]

I hope that you have put sockets for doing better tests without the need to unsolder each time, else you may damage the main board.

[ikonsgr]

Of course i will use dip sockets! I used a 40pin dip socket for replacing the 6845 too!
Btw, is there a chance that a faulty rom chip could cause the "no picture at all" symptom or not?

[TotO]

As I know, a faulty ROM let the screen grey with a black border.

[ikonsgr]

I desolder the 1st ram bank, and now i'm waiting for the new ram chips... 

Btw, the desoldering iron makes very good job, apart from rather fast and easy desoldering, i managed to remove the old 6845 and all of the 8 ram chips without breaking or cutting any legs,so i will be able to test some of the old ram chips to see if there are any good ones!   

[RobertM]

Finally today the desoldering gun arrived along with the 6845 replacement!
I solder the new 6845 but unfortunately still, no picture at all 
Now, according to the service manual the "no picture at all" symptom can caused by faulty:

Z80
6845
40010
RAM
CRYSTAL

CPU,6845 and 40010 i use are 100% ok (i know for sure because i have a second amstrad cpc where all these work ok) so next,is crystal. I put a new 16Mhz crystal but still same symptom. So the only thing left is ram!
The problem is that none of my ram chips seems to get hot, in fact one row doesn't get hot at all, and the other is heated very little!
Because ram change is a lot of work,and it would be very disappointed to have no success again, i was wondering if there is anything else on amstrad motherboard that could cause the "no picture at all" symptom. For example, could a  faulty 74lsxx chip cause this symptom?
And if i end up on replacing the ram chips, since i have no heat difference among chips meaning there is no way to know which are the faulty ones,i was wondering, if i replace the 1st bank only (i think it's the right row of chips closer to the 40010, as these seem to get hotter on both amstrad cpc's i have) would i get any picture even if there is a faulty ram chip on the second bank of 64k?

Hi, The diagnostic path (chip replacement) in the manual is selected to cover only the most probable faults. You say you have "no picture at all" ... if you have a box or window where the picture normally goes then the problem is most likely very different to problems when you have "no raster". Most of the chips "have to work" for there to be any sign of life so it is possible that a smaller chip is at fault (but this is not most likely).

If you have a stable box or windows on the screen then the crystal and oscillator chip are fine.

A faulty ROM can cause various faults depending on what is wrong with it. If you have a stuck address or data bit then you will get nothing! If it is just corrupted later on in the address space you will get the blank screen with the box as that is one of the first things the CPU sets up.

The RAM BANKS shares all signals except the data bus and /CAS so a stuck address bit on bank 2 can cause a fault. The CPC doesn't need bank 2 to run. On this thread someone gave instructions on how to remove the PAL(HAL) and use jumpers to disable bank 2 but remember a stuck address bit in bank two will still cause a fault.

It may be helpful for you to offer a clear and concise description of your fault. The very helpful people here cannot give advice that is any better in quality and accuracy than the quality and accuracy of the fault description. It may also be helpful to know what test equipment you have access to.

If you position the main board in front of you as it would be when using the CPC then bank 1 is on the right and bank 2 is on the left.

If you go as far as buying a programmer for the ROM then choose one that can also set the RAM if you want to test them.

Also, to others, can someone please confirm the BANK positions. I would hate to be the cause of someone pulling the wrong bank out.

[ikonsgr]

Thanks for your answer RobertM.
Regarding ram banks, according to service manual, ic's 127-134 are the 1st bank, and ic's 119-126 is the 2nd one (on both cpc's i have 1st bank is the one closer to the 40010, the right one as you see the mobo from the front).
But if ,as you say, any of the 16 ram chips is faulty i get no picture, changing only the 1st bank will not help much if i have faulty chips on both banks...

Now about my symptom, when i say "no picture at all" i mean LITERALLY NOTHING! 
There is no "window", no colour change on the screen, it's like having video cable disconnected!
As i have a second amstrad cpc ,which functions ok,i can compare the "thermal" behavior of various chips.
The only difference i found out is, that on good amstrad, ram chips are getting hotter compared to the ones on the faulty cpc which barely get warm (especially the 1st bank). What i also note is a difference in the buzzer noise when you turn up volume at max.Good cpc, gives a steady "buzz" while faulty one, has also a repeated "ticking" sound. And this noise remains exactly the same  when i removed the 1st ram bank.
I also make some tests removing 40010 or cpu and what i found out is that noise pattern is changing depending on what chip was removed!
It might be like a "sound diagnostic" where you can findout the faulty ic by hearing the noise pattern! Something like the error beep codes on a pc! 


P.s. Since video ram is placed at address #C000 and it's 16kb long, it means thar all video ram is between byte 49152-65535. That means that all video ram is at the 1st bank.So, if i change all chips of 1st bank, then shouldn't i get something on screen (other than nothing at all....) , even if a ram chip of the 2nd bank is faulty?

[RobertM]

OK, The fail point is very early or immediately at power on. While RAM could be the problem I would forget about it for now.

My first suspect would be that there is a short or low resistance somewhere and the 'tick' is the result of the power supply in the monitor shutting down and restarting (what monitor are you using). My second suspect would be a resistance in the power supply line some where.

The first point I would test in this situation is the on board 5 Volt rail. If you have a multi meter then check that the 5 Volt rail is actually 5 Volts. First disconnect the 12 Volt line and Floppy Drive. The 12 Volt line is only used for the floppy. The specification is 5 Volts plus or minus 10 percent (4.5 - 5.5 Volts) however in practice it should be between 4.8 and 5.2 Volts.

Try to test the 5 Volts after the switch. There have been reported problems with the switch after aging. I take the switch connector off the main board and use a jumper block in the socket instead. The ground is easy to find on the board as it is the larger thicker areas that are most common. Most of the smaller chips have 5 Volts on one pin at one end of the chip and ground on one pin at the other end and other side. The circuit diagram will help and look up the pinout of any of the smaller chips. Google the number with 'IC' after the number and select images.

https://www.google.com.au/search?q=7400+ic&source=lnms&tbm=isch&sa=X&ei=TtRbU6C6OM2kkQWX6IDYCg&sqi=2&ved=0CAYQ_AUoAQ&biw=1024&bih=616

All the smaller IC's starting with 74xx have Vcc and ground like this. Vcc is 5 Volts.

If you don't have a multi meter then plug in the little power light making sure none of the circuit on the back of the little board it is on touches anything conductive or the main board. While listening to the 'tick' sound see if the LED light dims briefly or flashes briefly when you hear the 'tick'. If it dims when you hear the 'tick' then disconnect the main board from the monitor as it may case damage to the PSU power supply in the monitor.

How often (fast) is the tick sound and is it consistent or random?

Test equipment that would help from cheapest upwards -

Digital MultiMeter (ebay $5)

Logic Tester (ebay $11)

If you have the money then -
DSO (ebay $66)

I am NOT recommending the sellers above. The links are just for example pictures.
If you buy a DSO then you don't need the Multimeter of Logic Tester. The DSO I linked to is the bottom of the range and it's usefulness is very limited as a DSO when working on computer board clocked at 4 Mhz. You ideally would want a DSO of 10 Mhz or 20 Ms/s or better. Let me know if you want a DSO and I will look for something suitable. I don't have one unfortunately.

If you get a logic tester then this is the diagnostic path after verifying that you have a stable 5 volt supply.

Test that /RESET on the CPU pin 26 is close to 5 Volts (high and not pulsing). If not replace R101, R111, R145, IC117, IC110 in that order testing as you go so that you don't replace more than needed. or ask for mor detailed info.

Test that you have a constant pulse signal on the CLOCK pin 6 of the CPU. If not test IC117 pin 4. If not pulsing then change IC 117. IC 117 is 74HCU04 and needs to be that so don't use the more common 74HC04.
By 'constant pulse signal' I mean that the 'PULSE' light stays on.

Test that you have a constant pulse signal on the /INT pin of the CPU. If so then CPU and gate array and all associated addressing (glue logic IC's) should be good and most probably the 6845 is ok as well.

Use the working board to compare voltages on CPU /RFESH /M1 /INT /MREQ /NMI /IORQ /BUSRQ, /RESET /READ /WRITE

With a Logic tester they should be -
/RFSH (28) constant pulse.
/M1 (27) constant pulse.
/INT (16) constant pulse.
/MREQ (19) constant pulse.
/NMI (17) permanently high.
/IORQ (20) constant pulse. I should verify this.
/BUSRQ (25) permanently high.
/RESET (26) permanently high.
CLOCK (6) constant pulse. Shown as the slashed 0.
/RD (21) constant pulse.
/WR (22) constant pulse.
/WAIT (24) permanently high.
/BUSRQ (25) permanently high.
/BUSAK (23) permanently high.
/HALT (18) permanently high.

Other places -
Any (only test one per bank) /RAS (pin 4)BANK 1 and BANK 2 constant pulse.
Any (only test one) /CAS (pin 15) BANK 1 constant pulse (CAS0).
Any (only test one) /CAS (pin 15) BANK 2 permanently high (CAS1). I should verify this. Not sure if the refresh uses CAS or RAS.
All DO pins (one per chip pin 2) on BANK 1 constant pulse.
All DI pins (one per chip pin 14) on BANK 1 constant pulse.
All Address pins in BANK 1 (pins 5, 6, 7, 9, 10, 11, 12, 13) constant pulse.
If you have a DI data pin that is not pulsing then that RAM chip or the CPU is faulty.
If you have a DO that is not pulsing then that RAM chip or IC115 or the gate array are faulty.
If you have an address pin that is not pulsing then test two pins on the CPU if it is A0 then test A0 and A8 if it is A6 then test A6 and A14 etc on the CPU. If pulsing on the CPU then several chips could cause this.

[RobertM]

If the clicking is getting slower when you remove the larger gate array or CPU then that greatly increases the chance that it is a power problem. It could even be the monitor at fault especially if the two CPC are slightly different models.

The location and BANK of the video RAM is determined by the CPU programming the smaller logic array IC118.

With the RAM BANK 1 removed, I suspect,  you should still get the box on the screen unless there is a stuck data or address line in BANK 2. With all RAM (both banks) removed then I expect you would get the box on the screen.

[ikonsgr]

First, thanks for you answers and concern mate! 
Second,i'm not using an amstrad monitor i use a 5volt power supply and a scart cable to a tv.
Third, i have many multimeters and i've already checked supply pins on many ic's  on the board (z80,6845,40010,ram chips, and some 74xx) and all seem to have the right voltage.
Fourth, the ticking sound is continuously  with a period of , i say, about 4-5ticks/second. When the 40010 is removed no ticking sound is heared only a low noise "whistle".

A digital oscilloscope would be a marvelous tool indeed, but at the moment i can't afford for something like that. The logic tester on the other side, seems more interesting, and if i could make all those checks you mention, then i might get one!
Now, since i already removed the 1st bank of ram chips and ordered new ones,i will wait until i get them, and if i still have problem then i might get a logic tester to check all the signals you mention.

But i was wondering, since i don't have a power supply problem and all the basic ic's are 100% ok  (z80, 6845,40010 are all tested an working on my other cpc) is there any chance to have problems to any of the signals you mention?

p.s. i just ordered this logic tester for only 6euros! So now i will be ready to make a full "check up" of my cpc!

[Bryce]

Amstrad monitors use a Sanyo L78MG to regulate the 5V supply. This chip has internal thermal shutdown and over-current protection. If the current is too high or it gets too hot it shuts down completely. They are very robust and survive even a complete short circuit. If there was a short or very low resistance on the CPCs 5V rail the regulator would shut down almost immediately and won't go back on as long as the short remains. So there wouldn't be any ticking and the LED would never light.

Bryce.

[RobertM]

Hi,
    That's the same Logic Tester I bought for this. It's cheap but not great. I also just last night ordered the desolder tool that you bought.

With my last post I was focussed on what happens before you get a box on the screen.

The first thing that needs to happen is that the Z80 has to read and run code from the ROM. For this it needs a clean data and address bus and the right control signal inputs so there is the place to start testing. It will run code from the ROM even with bad RAM and get to the point of showing a box on the screen if enough of the other parts are working.

If you have the correct control signals to the CPU and the buses are OK and you can see activity on all the address and data pins then then you can move on to other testing. If however one CPU pin is not as expected then which pin that may be is a big clue as to where the problem is. If it IS running code from the ROM and the other major chips are working then the best bet is to change the ROM but first check the control signals to the ROM. IF the ROM control is wrong then it's the GLUE logic (smaller chips and small HAL) at fault. If the ROM control signals are right then it's most probably the ROM.

The fault you have is more difficult than the one I have. I haven't had time for it lately but my case is very likely to be RAM or ROM. In my case I am getting a box on the screen but the /INT signal at the CPU is totally wrong.
I have replaced CPU and the larger gate array. Getting the box on the screen means that the CPU is reading and running code from the ROM. The fact that the box is blank and not a garbled random mess of pixels indicates the CPU is also writing zero's to the Video RAM. The box also means that the larger gate array is working and that the Z80 is also able to set the registers in the 6845 so the 6845 is probably OK to.

After the CPU gets the blank box on the screen it then has to set up jump tables in RAM. Before it does this it is NOT dependant on RAM providing the RAM is not interfering with the buses to the CPU. Once the tables are set up it needs the RAM and address glue logic to work (some smaller chips and the HAL). This is why I don't think the fault in your case is RAM unless it is interfering with the buses to the CPU.

I will be removing all the RAM on my CPC. I expect I will still get a blank box on the screen with no RAM installed. I will test that and let you know. After that I will install BANK 1 only and see if it then works. If it doesn't then my next step is the ROM. I am also replacing the 6845 and CPU even though I don't think that is the problem. I have only changed the larger gate array with a new part. I used an old Z80 CPU. I have a new 6845 and Z80 on order. The Z80 I ordered is one of the modern 10Mhz versions. The old Z80 in there seems to be working so I wanted to verify that the newer Z80's will work in this board.

[RobertM]

Hi Bryce,
            Thanks once again for your help. The ticking sound has got me confused lol. I thought it may be transitions on the power rail getting through the audio amp. It may however be the CPU initializing the Audio section in which case the CPU would be looping at around 4 Hz. 4 Hz is an unusual frequency for a CPU clocked at 4 MHz and having a 16 bit address space but still possible.

[ikonsgr]

I just finished testing  signals with a voltmeter of ram chips of bank 2 on faulty amstrad cpc.
The ras/cas signals of ram chips seem to be at the same level as with the working cpc (most probable this means that 6845+40010 are ok which i already know anyway  )
All Dout are constantly at ~5volts, and Din are at 2-2.5 volts except one that gives 0.6volt and another that gives 3.7volts!
On the good amstrad cpc, all Din (of all 16 ram chips) are giving ~1.7volt and Dout ~0.2-0.3volt
I also check all the cpu signals and seem to be ok.
Of course i have removed the ram chips of bank 1,so i don't know how this can affect the signals (although Din and Dout are separated).

Next, i'm going to solder dip cases and put back the old ram chips to see what measurements i will take.

[RobertM]

All DO's of bank 1 are directly connected to the co-responding DO of bank 2, they are NOT separate between banks but also are NOT directly connected to the CPU data bus. Instead they are buffered by IC114.

The same is NOT true of the DI's. Not only are all the DI's in bank 1 connected to the co-responding DI in bank 2, they ARE also DIRECTLY connected to the Data bus D0 - D7 of the CPU.

So if you have a DI that is stuck as 1(5V) and a DI that is stuck as 0(0V) then the co-responding Dx's of the CPU will also show the same voltage or logic level.

This means that the CPU is NOT able to run code from the ROM. On the CPU the D0-D7 are both output and input.

Test the co-responding Dx's directly on the CPU you should see the same voltages. If you do then remove the socketed chips (including the CPU) one by one testing to see if the problem goes away.

Do the following tests without power. Test the resistance from 5V and ground to some of the 'working' Dx's on the CPU socket as a reference. Then test the suspect Dx's on the CPU socket with all the chips in place. Then repeat removing the socketed chips to see if it goes away.

If it doesn't go away then you are on your way to finding the problem as you have identified the problem and now just need to locate it. That is assuming the resistance of one pin is greatly different to the reference.

If the resistance is very low then you have a short. The short may me in a chip. Unfortunately the data bus goes practically everywhere.

If you have a short it is hard to find as you have to move around the board disconnecting pins as you go. The resistance of the traces on the board is very low so testing from close to the short will give the same result as testing far from the short with a standard multimeter.

If you are looking for a short then the best solution is to use an ESR meter (Equivalent Series Resistance) or low ohms meter. ESR meters are also excellent for testing capacitors in circuit. You will find ESR meters on ebay.

I am probably jumping the gun a bit with the voltage testing as you WILL see different results with the RAM bank removed. However the results you mentioned definitely do NOT look healthy. A logic tester is better for this as it shows transitions. If there are no transitions then you definitely have a fault. The problem is that although the start code of the ROM will still give predictable and similar results to the working CPC for the start period, the start period is likely to be far less than a second after which the results are not predictable for me.

However the resistance tests should still be valid. I am probably writing ahead of where you are as there is a delay between our posts. I am at GMT +10:00.

I hope this helps, and it looks like you are on you way to fixing this definitively now.

I would suggest doing the resistance test before anything else as a low or very low resistance will give you a definitive "go, no go" test.

[RobertM]

I just had another look at the ESR Meters on ebay. Some of them have a resolution of 0.5 OHM.

The one I have used for finding shorts in circuit boards has a resolution of 0.01 OHM or 50 times the resolution.

I don't think a resolution of 0.5 OHM will cut it for you. You will need better than that.

I will wait to hear what resistances you see.

[Bryce]

I just finished testing  signals with a voltmeter of ram chips of bank 2 on faulty amstrad cpc.
The ras/cas signals of ram chips seem to be at the same level as with the working cpc (most probable this means that 6845+40010 are ok which i already know anyway  )

Measuring RAS/ and CAS or any other signals in a CPC with a multimeter is pointless, because they are (or should be) switching on and off quicker than your Multimeter can display.

Bryce.

[Munchausen]

Test equipment that would help from cheapest upwards -

Digital MultiMeter (ebay $5)

Logic Tester (ebay $11)

If you have the money then -
DSO (ebay $66)

A bit more expensive (~£100) but for a scope I got a hantek USB one recently. I've had a really good experience with it, 40MHz dual channel so will do you for most things, there's software for linux available and it comes with a pretty good pair of probes. It isa bit more expensive but worth it if you want to use it for more things and especially if you need something better than a few MHz for a low price (a stand-alone DSO with the same spec would cost at least 3 times the price): Hantek DSO2090 PC Based USB Digital Oscilloscope 100M/s 40MHz Bandwidth 2CH


And you can get a 24MHz logic analyser very cheap too (£10). Uses the seleae logic software: Hobby Components USB 24M 8CH 24MHz Logic Analyser, 1.1.16 | eBay
Test clips (£5): Test Hook Clips Ideal for Logic Analyser 10-way in Yellow | eBay


For a multimeter I have a low end fluke, but I guess a cheap meter is probably good enough for most things (mostly continuity testing you'll need I think).


Good luck with your repairs

[ikonsgr]

@robertm : thanks for your analysis and suggestions. I will make the resistance measurements to all ram chips to see if i can locate the "culprit"! 
@bryce: yes multimeter can only sample about 3-4 values/sec, BUT what you can actually see is the rms (or something like "average") value of the pulsing signal! For instance,i measure pin 6 of cpu to be ~2.5volt and that's because pulses are at 4Mhz beteween 0-5volt and with ~50% duty cycle, which gives an "average" value of ~2.5volt. So, in most cases if i get a value more than 0volts and less than 5volts this means that the particular pin is pulsing and you will get an rms value according to duty cycle of the pulse!  .Ofcourse by using a multimeter you can't distinguish if you have a truly pulsed signal or just a constant voltage, and that's where a logic tester is useful to! 

[Bryce]

That's fine for a clock signal where the duty factor is 50%, but a CAS signal has pulse of about 25ns per clock cycle, that gives a duty factor of about 97%, so assuming the positive level is exactly 5V (which it rarely is) a multimeter would read about 4.88V - 4.9V, which could be a pulse, could be a stable positive level or maybe just the accuracy of the multimeter. So you've no idea what's really happening. Accurate measurement is crucial when trying to repair a computer, otherwise you are just stumbling in the dark and guessing.

Bryce.

[ikonsgr]

I think i made a great discovery! ALL Dout on the good amstrad cpc has only ~25kΩ resistance, while ALL Dout on the faulty one has ~0.4MΩ(resistance i measure between 0v and Dout  on all ram chips i unsoldered is ~2MΩ)!
Mind also that  i removed the 40010 as Dout's has direct connection there, so each Dout is shared between the 2 ram chips in the same row (from bank 1 and bank2),and  the 2ic's (74ls244 and 74ls373) (i check this with the multimeter)
I believe that ram chips are ok (at least none of the 8 chips i unsolder, had a short circuit between Vcc and 0V pins and none was getting hot either) and as i happen to have another 74ls244, i "piggy bag" it above the one on the motherboard, but resistance of all Dout remain the same! So would it be safe to assume that the 74LS373 (the octal latch) is the TRUE culprit? 

[Bryce]

Is there a short or very low resistance between Dout and GND on the RAM ICs? That's where the short would be, not between Vcc and GND.

Bryce.

[ikonsgr]

Is there a short or very low resistance between Dout and GND on the RAM ICs? That's where the short would be, not between Vcc and GND.

Bryce.

On the contrary, there is a rather big resistance (~400KΩ) between Dout and ground, while on the working cpc resistance of Dout-grd is ~25KΩ!
You think that this indicates a problem with all ram chips(as all ram chips have the same measurement), or maybe the problem is with the octal latch?

[Bryce]

No, it doesn't really point directly to anything. Comparing the ICs resistance between a loose part and one in a circuit doesn't really tell you anything unless it really is a pure short circuit. But now that they are out, replace them with known good ones and you've removed another factor from the equation.

Bryce.

[ikonsgr]

The Dout resistance i mention in both cases (fault and working cpc) are with the ram chips ON BOARD not "in the air". The Dout of each pair of ram chips (bank 1 and bank 2) are directly  connected to specific pins of the 74LS244, 74LS373 and also the 40010 (which in both cases was removed from board).
So,when i measure 25KΩ on all Douts of the working cpc and ~400KΩ on the faulty one, doesn't this mean that the problem would be either with ram chips and/or the 2 74LS ic's?
But, since none of the ram chips of the faulty cpc has short circuit (neither got very hot which i think it's a typical symptom of a bad ram chip) will it be more probable the problem lies in one of the 2 74LXXX ic's?