issue 6a faulty

Comments

• gorski

  March 2009 edited March 2009

  I make simple program.
  
  10 for i = 32768 to 65535
  20 poke i,170
  30 if peek i <> 170 print i
  40 print ".";
  50 next i
  
  program run and exit without errors ?
  
  Greets,
  Gorski
• Phu

  March 2009 edited March 2009

  Change.
  The.
  Resistors.
  
  I know nobody believes me, but if you change the memory on the spectrum and put 4164s or 4464s in you MUST increase the resistance of resistors R1 to R8 (the data bus separators). 330 ohm is NOT enough. It needs to be AT LEAST 820 ohms.
  
  The reason is the data is "leaking" from the screen to the upper memory. As the Spectrum enters the memory test cycle it writes certain values to the upper memory. Because the resistors are too low, these values get contaminated by screen data coming from the lower RAM to the ULA. When the spectrum reads the values back they are wrong, so it fails the memory test.
  
  This is why RAMPTOP is so low and it starts up like a 16K machine. You'll find resetting it makes it initialise more 48K-like, because the values in the screen are now set to be more friendly towards reading the expected values back out of the upper memory. But you'll find any 48K games won't load or run properly.
  
  -- Richard
• gorski

  March 2009 edited March 2009

  Thanks Phu :)
  I will try tomorrow and report result.
  Greets,
  Gorski
• MrCheese

  March 2009 edited March 2009

  Phu wrote: »

  I know nobody believes me

  
  I believe you.
• gorski

  March 2009 edited March 2009

  status report: (not lucky)
  
  1. I get 8 x 830ohms = http://forest.no-ip.org/6a/IMG_7541.JPG
  2. when i put on pcb and check with multicet i get 790ohms = http://forest.no-ip.org/6a/IMG_7543.JPG
  
  no changes at all :(
  i will try with 1kohms but not have now
  we talk for r1 to r8 ? = http://forest.no-ip.org/6a/IMG_7537.JPG
  
  Greets,
  Gorski
• Phu

  March 2009 edited March 2009

  Yes, those are the right resistors. Its worth upping to 1K - I had to do this on a board to get it working - I found 820 ohm wasn't quite enough.
  
  Its normal for resistors to read slightly different in-circuit. The multimeter (which I notice you're running in diode test mode rather than ohmeter) will be measuring the resistance of all circuit paths between the two points, which can increase or decrease the total resistance. You'll probably find its an unstable reading too.
  
  I notice there's a number of repairs done to the board. Have you checked all of those wires are correctly connected ?
  
  -- Richard
• gorski

  March 2009 edited March 2009

  Hi Phu,
  
  [cut]
  The multimeter (which I notice you're running in diode test mode rather than ohmeter)
  [cut]
  
  oops .. :)
  
  [cut]
  I notice there's a number of repairs done to the board. Have you checked all of those wires are correctly connected ?
  [cut]
  
  about repairs:
  i drop 5 memory track until i desoldering ram 3 in low memory and 2 in hi memory
  i change today wires because i found very thiny current wires but not have changes, low memory is fine even wires and all is good. in 16kb sinclair is rock stable, loading games and etc.
  hi memory i drop one of linear track, i put 2 wires, i measure from ic15 to ic22 and i get very fine 0 in multimeter for this track.
  
  i put socket for cpu i hope all is fine with this (why when i put nec cpu d780c (brand new) sinclair not start?) i have old one z80a cpu and new one i'm with new one now.
  
  put socket for 74ls04n ( i put socket for all stufs who i have new)
  i drop one track and here .. i think repaired
  
  about rom i received with mod for 27128 all be right with this but i make mod from http://www.speccy.org/trastero/cosas/JL/ROM-EPROM/ROM-EPROM.html
  this working always fine for me. this is my fav mod for eprom and put my 27256
  
  that's all about repairs not small number but i hope i'm lucky and all is fine with this ..
  oops i forgot about tr4 ztx650 nothing interesting i simple changed this one with another ztx650
  
  i hope that though my bad english you understand about 70% of what i want to say :)
  I will continue after weekend. I want all sinclair pcb in the world to work fine :) i have bad experience with issue 6a for second time.
  
  Greets,
  Gorski
• roko

  March 2009 edited March 2009

  Phu wrote: »

  Its worth upping to 1K - I had to do this on a board to get it working - I found 820 ohm wasn't quite enough.

  I believe 100% that you had this experience, and now are helping others out.
  My Spectrum48K (version 3b) runs since 1984 with 4164 and without changing the R's. That is also a 100% true experience.
  Appearently you had an impedance problem on the data bus, and it lookes like Gorski's problem might be similar.
  But I think that changing from 330 to 820 ohm should at least give some sign that a solution is coming. When nothing changes
  this might be the wrong road. I am curious to see the Gorsky-results.
  
  @Gorski: your simple test program will not find defect memory bits that are permanently set, when they equal a bit that is set
  in binary 170. In your actual situation it would be more interesting to know if reading 10000 times from 32679 would give 255
  each time, after that number is poked there of course.
  In your situation I would temporarely solder a few decoupling capacitors (10 - 100uF) on several positions on the board, between
  +5V track and Ground. This way you could catch (with some luck) possible dips (glitches) in 5V due to weak contacts or bad tracks.
  
  Just theoretically: I remember the 6a has a soldered jumper for RAS/CAS timings, and think this is done to accomodate different
  brands of memory chips. What would happen when 'fast' and 'slow' chips are mixed?
• gorski

  March 2009 edited March 2009

  Thanks for advice's Roko i will try all. report after Monday.
  
  I have issue 3b with 3 types 4164 at same time
  =
  http://forest.no-ip.org/3b/IM000880.JPG
  I never have ANY problems with this sinclair and not changed resistors.
  
  jumper = http://forest.no-ip.org/6a/jumper.jpg
  i not know what must be other position to change RAS/CAS stuffs ..
  
  Greets,
  Gorski
• roko

  March 2009 edited March 2009

  gorski wrote: »

  jumper = http://forest.no-ip.org/6a/jumper.jpg
  i do not know what must be other position to change RAS/CAS stuffs ..

  The jumpers seem to affect the video RAM only, and are therefore not relevant for your problem.
  BTW: C stands for CAS, R is RAS. On your board C-in is directly connected with C-out and R-in goes to R-out via 12 nsec delay by two LS04 gates.
• gorski

  March 2009 edited March 2009

  Hi All,
  
  fast report
  
  no changes with 1k resistors = http://forest.no-ip.org/6a/IMG_1228.JPG
  on this picture i not like how c59 and c60 looks, also r13 and r15 have marks
  may be not bad idea to change c55 till c62?
  and change and all blue condensators = http://forest.no-ip.org/IMG_1227.JPG
  
  after this i will give up and ready to sent if anyone have to take a look to this bad pcb ..:)
  
  Greets,
  Gorski
  
  
  p.s. i put one 10mF around to r90 but not have result, i not have knowledge to do this fine, i not know what is good place mean to put this 10mF and what after this .. :)
  i think to go back to 470ohms R1-R8
• roko

  March 2009 edited March 2009

  gorski wrote: »

  after this i will give up

  Thank you very much for testing the resistor theory.
• gorski

  March 2009 edited March 2009

  Hi All,
  I want to report for successfully repaired this issue 6a pcb.
  I make very stupid error, lost connection between leg 15 ic15 and ic16 and all is from this ... :)
  loading games and etc. all is right and I'm Happy :)
  Thanks Phu and Roko for tips.
  
  http://forest.no-ip.org/issue6a/IMG_1259.JPG
  http://forest.no-ip.org/issue6a/IMG_1262.JPG
  pictures of happy pcb :)
  
  Greets,
  Gorski
• MrCheese

  March 2009 edited March 2009

  Always good to see a spectrum that is back from the dead. Excellent.