Let's repair a mystical issue with an old 386 mainboard

hello and welcome i think it's time to make  another repair video and if you don't mind   today i'd like to present one really interesting  issue i never had something like this before so i   am very excited as well maybe you remember this  3d6sx mainboard which i used for overclocking   experiments unfortunately this main board was  not very cooperative as i tried to clock the cpu   above 25 megahertz but still i really like this  mainboard first of all because i repaired and   modified it already many times and meanwhile it  feels very familiar but also because i use it as a   slow test platform for a various hardware and  furthermore it is kind of special because it   accepts simultaneous usage of sim and deep  memory which allows to use for example   5 megabytes of ram it was very unusual back then  and if this board would be able to handle higher   frequencies it would be really awesome anyway let  me show you what's wrong with it some time ago   i was working on one of my projects  and i used this mainboard for testing   i didn't do anything special actually just some  signal measurements on an iso card suddenly the   system became very slow it worked actually kind of  reliably at least i didn't notice any functional   issues but the system became super unresponsive  let me start phil's benchmark pack you already   can see how slowly the display was redrawn  let's take a look at the landmark speed test and look at that cpu clock was recognized as  1.387 megahertz despite that this machine is a 16   megahertz 386sx also down here look how weird this  scale jumps up and down between something around   12 megahertz and some values below scale and up  there the speed behaves also super strange i will   not even try to show you any game since it takes  forever to start one like 15 minutes to get for   example commander keen running believe me it is  not because the tour is off or some trivial thing   there is something wrong with this  board so let's see what it could be   if you follow my channel you know that for my  repairs i try to use simple equipment as much   as possible but this time i think i will need  an oscilloscope to find out what's going on   first i would like to check out the  crystal oscillators and see if they   all generate proper frequencies i can see  here are two and another one is up there   first of all if you use an oscilloscope for such  measurements don't forget to switch it to the 10   times mode i have a switch on the probe for that  and on the oscilloscope i have a switch as well   both must be set to 10x the reason is that the  crystal oscillators generate a very weak signal   which will be pulled down by the probe if used  in a normal mode in that case neither will you   be able to measure anything nor will your system  continue to work properly since the oscillator   would be disturbed too much by 10x an additional  resistor in the probe will be activated and so   the probe pulls down the signal 10 times weaker  as in normal mode anyway first of all let's see   what we get on the main oscillator which supplies  the frequency for the cpu as it all the 386 sx on   this main board is a 16 megahertz version which  must be supplied with the doubled frequency   or 32 megahertz and as you see it seems to work  properly let's see if it arrives at the cpu itself   the datasheet tells that the clock input is  on the pin 15. let's see what we get there   and there are 32 megahertz as well so the  cpu should at least run at the right speed   okay this crystal oscillator up here has  an inscription 8000 on it i guess that   means 8 megahertz i'm currently not sure what  this crystal is responsible for exactly but   looks like it delivers 8 megahertz  indeed so i assume it is all right   the next one should deliver 14.3 megahertz  and is probably the general purpose clock   which goes among others to the iso bus and  also here everything seems to be in limits   during my overclocking experiments i realized  that there was a crack in the old cabs at this   oscillator i replaced them back then and i thought  that the problem could be here around but it looks   good so far the next crystal oscillator is here  near the keyboard controller such a small cylinder   is usually a 32 kilohertz oscillator which  is used for the real-time clock generator   on the old main boards it is usually placed  somewhere near the battery since the oscillator   has to run from the battery to maintain the  time and calendar when the pc is turned off   hmm i can't measure anything  there are some sporadic impulses   but that is probably due to my shaky hand but i  don't see any frequency generated by this crystal nope i can't measure any clock  neither on one leg nor on the other let's see where the connections are going to  following the traces the crystal oscillator is   connected to this toshiba dc 4069 ic you see  the crystal itself is not a clock generator   its output is a sinus wave which has to be  converted into a square wave which can be   used then as a clock afterwards so this ic is  an inverter and is basically responsible for   conversion of the sinus wave from the crystal  into a square wave furthermore it acts as a   buffer which enforces the signal and makes it  more stable well it's hard to tell now what is   going on my assumption for now is that the crystal  oscillator is broken let's visit the bios and see   if the clock is running at all if this crystal  is broken the clock should not count the seconds   what does this have to do with the strange  behavior of the system you might ask   well real time clock generates interrupts and a  lot of things in an 80 system depends on it if   the clock doesn't work the interrupts will not  come and the operation system could fall into   a lot of time outs and other error states  these can end up in various strange effects   the clock seems to work but the seconds  are counted very unequally this is odd   oh look at that the second stopped  at 29 seconds after the boot   so my assumption about the real-time clock  is maybe right i'm not sure if it is the   crystal what is broken but let's replace  it and see if we will get any further and here it is it has no markings on it which  would tell us what kind of crystal this is   however as i already mentioned for the real-time  clock usually a 32 kilohertz crystal is used and   this can be found on almost all main boards i will  salvage a replacement crystal from another dead   main board in my spare parts box this main board  seems to have what i need by the way time goes on   and the old hardware gets more and more expensive  i don't even talk about the retro hardware itself   but about the replacement parts this crystal is  definitely not a good example but many other parts   don't get produced anymore and they have to be  salvaged from scrap however meanwhile even this   crap became incredibly expensive this is a hobby  channel which should carry the expenses on its own   unfortunately it seems not to work anymore and  i feel uncomfortable having to mention it but   if you would like to support my channel i have  a patreon and a paypal account which you can   use to help me to keep this channel running so  meanwhile the replacement crystal is in place   let's power up the system and see  if the clock is ticking properly now no that doesn't look good now it  already stopped after 8 seconds   well i didn't check if the replacement crystal  is working but i think the probability that both   crystals are broken in the same way is minimal so  there must be something else what is wrong with it   off off-camera i made some measurements and  the crystal oscillator didn't generate any   signal just as before however during my  investigation i stumbled upon something   interesting look what happens when i turn on the  power right after the power up our crystal seems   to generate proper frequency of 32 kilohertz but  the amplitude decreases after a while it starts   at about five volts peak to peak and falls down to  1.7 volts where the oscillation stops eventually   such a behavior looks like a discharging of  a capacitor for me let me explain what i mean   this is a regular sinus wave how it should  look like periodically and with always the   same amplitude the vertical axis represents  the voltage over the time on horizontal axis   the job of a crystal oscillator is to generate  such a sinus wave our wave's amplitude on the   other hand seems to decrease over the time  and looks more like this now if i draw a line   over the local maximums of our wave i'll get  something like this and this blue line looks   very much like a capacitor discharge i have  to keep this in mind for the further analysis   i think the crystal is all right but there is  some other oddness around it first of all there   is this capacitor nearby which is connected to  the crystal so maybe it is bad and damps the   oscillation after a while let's replace it by  another one and see if it makes any difference   i think i'll also remove the inverter ac and  try another one in case it is not a capacitor the capacitor has an inscription 22 k i'm  actually not sure what capacitance it needs to be   k sounds for me like kilo 22 000 of what i guess  22 000 picofarad or 22 nano farads in other words   usually the capacitors near such a crystal should  be more like 22 picofarad as far as i know but   that k irritates me well i guess i'll try to use  22 nanofarads first and for the ic i'll install   a socket to be able to swap the ic later if i  need to okay everything is in place i'll use   the same ic for now as before to check out if  the capacitor changes anything on the result i'm measuring now directly on the inverter  i see that is why there is a square wave and   not a sinus as it comes from the crystal yeah  and unfortunately nothing has changed the clock   signals flattens and disappears after a while just  as before so the capacitor didn't change anything   on that result okay let's change the ic by another  one which by the way i again got from scrap yeah and nope the same behavior so neither the  capacitor nor the ic seem to be the culprit again off camera i made some further  investigation and i think i'm a step further   first of all i would like to come back to the  capacitor i googled a bit and found that the k   in 22k doesn't stand for kilo but for 10 percent  precision so the capacitor is a 22 picofarad as i   expected it to be from the beginning since another  capacitor and the crystal didn't change anything   i put back the original parts just to exclude  the possible errors by using wrong values that's   why you see the blue capacitor again back in  place the problem is obviously somewhere else   i'd like to confess an absolute beginner's mistake  which i previously made i didn't check the voltage   on the suspicious tc4069 ic so a reminder number  one if you have a suspicious part always check the   voltage first and this is what i found so far when  i turn on the power the voltage on the named ic   rises to 5 volts and then drops down to  under 2 volts resulting in terrible ripple   let's step back and see how it works as i said  the frequency for the real-time clock is generated   by the crystal oscillator which swings with 32  kilohertz its sinus wave is then sent through an   inverter or buffer which generates the square wave  which is then sent into the real-time clock i see   on many main boards like the one in this video  the real-time clock functionality is integrated   directly into the chipset and in case of dedicated  real-time clock i see the inverter is often   integrated into the ic but this all doesn't play  a role now important is that to maintain the time   and calendar the crystal the inverter and the  real-time clock i see have to continue to work   even if the pc is turned off therefore there is  a battery on the main board which maintains the   power for the slower power components if needed  even over the years but when the pc is turned on   there is no need to drain the battery so there  is some smart switching circuitry which detaches   the battery and pulls the required power from  the power supply since the inverter gets its   power through the magical switching circuitry  probably the culprit is somewhere there the   circuitry which switches between the battery  and the psu looks always similar to this one   it has two bipolar transistors usually 3904 and  3906 plus some resistors diodes and capacitors   such circuitry is usually located right under  the battery and by the way suffers first from   a battery leakage anyway to check if  there is a problem in the circuitry   i will disconnect the main board from  the power supply and solder a wire from   the battery directly to the tc4069 inverter ic  this way we will bridge the switching circuitry   and hopefully get a steady 3 volt voltage on the  ic let's see if we will get our clock stable then   and here we go we have around 3  volts on the vcc of the inverter ic and the crystal delivers steady 32 kilohertz sinus  wave now that's great and is a strong sign that   the culprit is somewhere in this circuitry for  switching between the battery and the power supply   i removed that barge wire  again and a long time ago   i modified this main board to be able to  use a cr2032 battery just to be sure i   removed previously installed diode so the  board can be now only powered by the pcu   let's measure the voltage again on the inverter  and see if the voltage drop is still the same yeah as you see it starts at 5 volts and then  drops slowly down to 1.7 volts which is probably   too low to keep the clock up and running what  do we have in this area i already checked all   these simple parts and all the resistors  and diodes seem to be all right there are   those two transistors which can be bad but since  the voltage drops slowly i still have a feeling   that it could be some capacitor do you remember  this image every time you see a voltage slope like   this one somewhere keep an eye on the capacitors  and these two guys here are coincidentally sitting   on the trace which goes directly to the vcc pin of  the tc4069 ic where i measured the voltage slope   so i think i will start with them and if they are  okay i will continue with the both transistors   i remove the first capacitor it is  only a bypass capacitor and it is   responsible for the voltage stabilization it  is better to have this capacitor in place but   the main board should actually work without it as  well at least for the test run it should be okay   and would you look at that i guess we found  the culprit the voltage on the inverter   is now absolutely stable at slightly under 5  volts and what does the crystal oscillator say   yep also here we are getting the expected  32 kilohertz perfect and it was a capacitor   indeed it is a 4.7 microfarad one and it is  suitable for up to 16 volts let's test it at least the tester detects it as a capacitor  you can see it on the c in the end of the line yeah it takes quite a long time but  eventually the tester says that it is faulty unfortunately neither do i have a new one nor  did i find a 4.7 microfarad capacitor in my scrap   however i found here a 6.8 microfarad for up  to 25 volts i call it good enough in this case   the tesla confirms that it is okay tantalum  capacitors like this one are polarized and   inserting them wrong way around will end up most  certainly in an explosion so pay attention to   the polarity this tantalum capacitor has a plus  on one leg so the other one is obviously ground   on the board there is this white line on one  of the contacts which usually is a ground   but if you are not sure switch the multimeter  into continuity mode and check which one of   the contacts is connected to the ground in this  case the one with the white line is ground indeed   and the other contact is the vcc trace which  goes to the inverter ic as i mentioned before   and here it is the new green capacitor the  protection diode to the battery is also get back   in place the inverter ic dc 4069 is now socketed  and the original blue capacitor and the crystal   are both back where they were i guess i can now  put back the battery as well oh and i will insert   four megabytes of ram which i removed before  i started with the repair as i said this board   can handle deep and seam memory simultaneously  and i use it with four times one megabyte sim   and eight times 128 kilobytes deep modules  to get to five megabytes of memory in total so the graphics card and an ide controller are now  back in place and the system is ready for testing   let's check the voltage on the inverter  once again yeah 4.8 volts slightly low but   good enough i guess now the  frequency on the crystal   32 kilohertz just perfect in bios the  clock is ticking we are approaching   third minute already that's fine but the seconds  are ticking slightly weird every third second   seem to come faster than the other two i will  have to observe it if this is an issue somehow back in dos let's see what landmark speed test  has to say now this looks very promising cpu   clock is at 15.9 megahertz and the  benchmark is stable at 21.2 points   and look at the clock up here the seconds seem  to tick equally i guess this was just a refresh   issue in bios and the clock is probably working  fine but i will keep an eye on it in the future well the main board seems to work normally now it  is at least as fast as a 16 megahertz 386sx can be   but as i said i use this main board when i need  a slower system anyway so it is just as fast as   i need it to be honestly i never had a case like  this one before and i found it really interesting   who would know that the system can become slow  because of a faulty capacitor on the power   supply for the real-time clock generator i didn't  expect that tantalum capacitor is usually just   explode leaving you with a shorter hardware but  this was very unusual and exciting case for me   i hope you feel the same and i'd be very  glad to have you on my channel again   please don't forget to like dislike write your  comments and share my channel if you like to   i would very much appreciate your help feedback  and support so far thank you and goodbye

Channel: Necroware

Views: 31,700

Keywords: retro, hardware, soldering, repair, review, nerd

Id: koL3T5OAlIk

Length: 24min 1sec (1441 seconds)

Published: Fri Oct 08 2021