Temporal Drift: This Real-Time Clock is on FIRE!

[@matthewday7565]

There should normally be capacitors of about 33pF from each side of the crystal to ground which help to keep it from jumping to a harmonic - can usually get away without them, but a shift in the AC parameters of the 4069 allowed it to mode jump. the characteristics of CMOS biased into linear mode are definitely not tested by your programmer

[@Stoney3K]

That could also point to a failed capacitor as the root cause right?

[@bitsundbolts]

Just because someone asked, I did replace the one capacitor below the crystal. That didn't make any difference. I did not test without the capacitor though. Thanks for your explanation!

[@TheMovieCreator]

A broken signal path, bad solder join, sketchy bond-wire, etc.. may also act like a series capacitor, creating a highpass filter. In such a case the fundamental frequency of the crystal may be so supressed that an overtone will oscillate instead. Using filters to select for overtones is a well known technique in radio-technology, for instance to create frequency-multipliers.

[@T3hBeowulf]

That was fascinating, thank you! When you first measured ~80KHz from the region of that crystal, I was pretty sure that was the problem. It's not often that either the crystal oscillators or inverters go bad. Great fix! Regarding your 486 donor board with a "Low Insertion Force" socket, I have one of those too and I honestly never considered that I could just replace the socket. (The To-Do List grows.)

[@bitsundbolts]

Hehe, yes, you just need to make sure there's enough space for the new socket. It might not be possible to add the lever mechanism. I was a bit surprised to see 80KHz on the RTC module because I tested the inverter chip that's connected to the RTC chip. I'll retest that chip one more time before I take it back to the scrapyard. But it was definitely the issue. I can't explain why things seem to have gotten slightly better after replacing the crystal though.

[@mamuttme]

​@@bitsundbolts Crystals require an oscilating circuit, which contains capacitances of a few up to 50-ish picofarads. The exact value depends on the actual crystal and the construction of the oscilator (even how the chip used in the circuit is designed). Since that particular oscilator is build on linearised inverters, it was also kind of depending on the internal capacity of the 4069 itself. A different chip had a different capacity which made the oscilator run fine. The first attempt with changing the crystal also changed the capacity and the requirements to start the oscilations. Also a note on testing logic chips in a programmer: programmers mainly test the function in a static (or low frequency) test, no stressing the chip, no working at the edge of the electrical parameters etc. Which needs to be considered when getting a passed test result. The oscilator with linearized inverters works a bit on the edge of the acceptable parameters of the chip and is also putting the chip outside of the regular TTL voltage levels, so one chip might behave differently than the other. Thats why its not a well beloved circuit among designers.

[@bitsundbolts]

Thanks for all this great information! So, the difference in CheckIt (after replacing the crystal) could be due to different parameters of the crystal itself - not the frequency, but other values. In any case, I am testing the board a few days later, and the time is spot on.

[@geremychubbuck3730]

Great video as usual. Keep up the great work 👍👍👍

[@bitsundbolts]

Thank you!

[@OtherWorldExplorers]

I think your computer still hungover from Y2K. 😂

[@bitsundbolts]

Haha, definitely! Y2K bug

[@David_Ladd]

Nice catch :) Normally in my past repairs I have found the problem to be the crystal. Great job :)

[@therealjammit]

The "input/output" pins are basically an amplifier. Even though the crystal is really good at resonating at one frequency they need some amplification, so the "input" is the signal out of the crystal and the "output" is the amplified crystal signal being sent into the crystal to continue to make it resonate. It's similar to feed back with a microphone and speaker. The microphone is input and the speakers are the output. The screech coming from the speakers is the resonance of the room. Edit: the 32,767 Hz crystal is very common in many electronic devices.

[@bitsundbolts]

Thanks for this info. There's always something new to learn! Also, OSC2 had a much nicer/perfect square wave. I should make a short showing of the signals arriving at the RTC chip now.

[@SimonSideburns]

Very informative and interesting to find it wasn't just the crystal but also the inverter. Now it begs the question with regard to any differences in the benchmarks performed in the previous repair video. Hopefully there will be a follow-up at some point.

[@bitsundbolts]

I believe that the benchmark values were unaffected by this problem. Unless a benchmark uses the time from the RTC (which doesn't make much sense since I believe the DOS time can be accessed much easier), the results should be valid. But clearly, there is a possibility - as seen in CheckIt.

[@MisterMsk]

Thumb nail was fire!

[@tobias_off]

Great job again. I like the way you troubleshoot the things and showing it to others. If there is no more video in 2024, dann wünsche ich einen guten Rutsch ins neue Jahr. 😉🍀

[@bitsundbolts]

Viele Dank! Ebenfalls ein gesundes neues Jahr!

[@Ale.K7]

What an interesting issue! Great repair!

[@bitsundbolts]

Thank you!

[@protox07]

Have a happy holidays Bits und Bolts

[@bitsundbolts]

Thank you! Same to you!

[@r4z4m4t4z]

did you test to see if doom ran three times faster before repairing? anyways, well done figuring it out and showing your path.

[@bitsundbolts]

I think the test results are valid as it seems that DOS has its own mechanism to track time. That is why I believe DOOM will show the correct frame rate at the end of the benchmark.

[@Clancydaenlightened]

The real time clock has nothing to do with the cpu clock It's basically just for time of day and calendar functions You could simply remove the whole rtc chip but the date and time would reset to bios default each time you power off or unplug It would allow you to bypass hardcoded timebombs in software like trial periods and other things Nowadays with nist and internet time, having rtc is really only needed for initial setup, then the computer will just grab rtc from the network, and rely on rtc only if no network is available

[@PROSTO4Tabal]

I'm dissapponited, I thought RTC exploded lol glad you fixed the board

[@bitsundbolts]

Haha, I had enough explosions this year! Let's see what happens next year 😅

[@envoycdx]

Interesting, nicely repaired though :)

[@germanskena6537]

@BitundBolts. Thank you for the video, i think in replace the chip in the 1st shot, but we are not the same and we all have opinion of how to do the things done. You always finish the job with excellent results. I was thinking, you can make a video of the scrapyard and wich kind of computers you find there? could be nice a video of HUNTING there. Well I also wants to add. Happy New Year. I wish you the best for the 2025. You are doing an amazing job, continue doing in that way.

[@bitsundbolts]

Thank you so much! I wish you also a happy New Year 2025! Regarding replacing the chip vs the crystal. I fully understand. I made the decision to replace the crystal first because I only had to desolder two pins instead of the IC. Anyway, it was the IC that was faulty.

[@commodore71]

Cool, and thank you for the film tip, Interstellar next😀

[@paulbrooks4395]

There's probably more to it than was explained, but I thought it was pretty cool when I first learned in a video that all quartz crystals oscillate at 32,768 Hz.

[@bitsundbolts]

Oh, not all crystals resonate at that frequency. Just this 32.768 barrel shaped one is most likely of that type. There is another one (kind of flat) on the board with 14.318 MHz.

[@ruben_balea]

The datasheet for the W85C178 is available and it has 128 bytes of RAM, I suspect the W85C168 was the same but with only 64 bytes of RAM like the MC146818A on the original PC You can use a RTC with 128 bytes of RAM to replace one with only 64 bytes, the extra bytes will remain unused and the computer won't care.

[@bitsundbolts]

Thanks for the info. I have a question regarding that memory. Is that the one where the BIOS is storing the data for the settings? I find it weird to have this integrated in the RTC chip, but it would explain so much. I once replaced some Dallas RTC Models (the ones with integrated batteries). The boos settings changed whenever I changed the Dallas module. It made testing quite simple. Instead of updating BIOS settings, I just had to swap the RTC module. Well, that was just a very specific usecase for me at that time.

[@ruben_balea]

@@bitsundbolts Yes, computers used the extra memory on the RTC chip to store the BIOS settings because it was already there. The MC146818 had that extra RAM because it was designed to be used with CPUs and MCUs that could use some external RAM but without the complications of using a conventional RAM chip with all the extra circuitry associated with it.

[@synestetic1687]

Good video!!! Just your words confuse oscillator with resonator. Oscillator is a feedback amplifier that is wired to constantly change its state. Inverters are useful for that. Then there is a resonator such as a crystal in the tank that stabilizes the oscillator to a certain frequency. So the metal barrel thing is a resonator, specifically a crystal. And the electric circuit with the inverters and whatnot, they are the oscillator. The oscillator oscillates at an unreliable frequency, but when its feedback is connected to a crystal, you get a stable frequency.

[@synestetic1687]

Also, crystals are finicky. They need to be loaded with just enough capacitance in the circuit so that the crystal operates at a good point. Something in the circuit has aged and changed the operating point. Not sure what it is, but certainly it seems that the new crystal with the new inverter oscillator is a good match.

[@bitsundbolts]

Thanks for clearing that up. Yes, it was not clear to me until your explanation. Thanks 👍 And yes, it seems like the new pair works well together. So far, the time has been kept perfectly.

[@tobias_off]

That's the comment I would posted also if you had done it. However, the inverter is used to build the oscillator circuit, it also converts the sine wave crystal output to a square wave the digital logic wants to have. I think it could be the Input capacitance of the inverter gate what drifted over the years and pulled the crystal off of it's nominal frequency. Perhaps a closer look on the on the capacitor near the crystal could not hurt. If it is only for decoupling the chip everything is fine, but if the designer involved it in the oscillator circuit I would check its healthiness.

[@MidnightVisions]

I agree with the other poster saying there should be 2 capacitors in series with the crystal to buffer and control the crystals speed, about 22pf to 30pf. All crystal oscillators Beed them. If they are missing, or not installed, this time circuit will drift again.

[@MidnightVisions]

Need, not Beed (autocorrect)

[@bitsundbolts]

There is one 30pf capacitor right next to the crystal. It is there and it is working. The clock on the board works perfect even days later.

[@skonkfactory]

That crystal isn't an oscillator, it's just a crystal. The inverter chip and the resistors are used to make it oscillate.

[@bitsundbolts]

Yes, I learned that from all the comments regarding this. Thanks

[@abunk8691]

Never seen an RTC go that fast due to a bad part though so it was interesting to see fixed. Bad RTC time drift bugs me as I had a Gigabyte B350 motherboard (yes something that recent) that drifted so bad to the point that it was 5mins behind within a few weeks. The regular sync to internet time was not frequent enough to help correct the time and I was late on quite a few online classes and deadlines due to that false time fooling me. Good thing that motherboard is long dead and has been replaced so I'm not dealing with that mess anymore.

[@abunk8691]

To make fun of that Gigabyte board I mentioned. I have an 8 USD digital clock I got around the same time and its maybe 10 seconds behind in a year. I've only seen wall clocks on dying AA batteries be behind in time as badly as that Gigabyte board I had. (It wasn't even the CMOS battery since I've tried replacing that and it didn't fix it)

[@bitsundbolts]

Keeping the correct time without syncing seems to be a real issue. I am surprised by the behavior of that gigabyte board. Maybe something on that board was faulty. A capacitor or a chip that is out of spec. Very weird and shows what this can cause: missing appointments.

[@abunk8691]

@@bitsundbolts That was only one of the things faulty with that board haha. There were other stuff that was wrong with it like randomly USB disconnects, not always reading all the RAM, sometimes just powering on to a black screen with no POST beep, the VRMs cooking themselves with a Ryzen 5 in good airflow, etc. It was a weird one as I thought it was only driver issues initially until I used Linux to test and had the same issues. Thankfully all my issues were resolved with a motherboard swap.

[@PileOfEmptyTapes]

@@abunk8691 "the VRMs cooking themselves with a Ryzen 5 in good airflow" - now THAT is actually not entirely uncommon with boards of this generation. I think what happened was that they used pad-printed thermal paste of very lousy quality that went rock-hard within short order, entirely defeating the purpose of VRM heatsinks.

[@omfgbunder2008]

Your clock still looks like it's drifting, the time between seconds isn't very monotonic.

[@Megatog615]

this is common on amibios. it often doesn't display the seconds perfectly

[@Def-cd6bm]

Funny, I've just finished replacement of a dead 48MHz smd crystal on a modern AM4 mobo. 32,768KHz are still in use today as well.

[@johannesbakkedal1801]

You could also start by measuring the supply voltage to the inverter IC. If it is low, that could also explain the wrong frequency

[@bitsundbolts]

I didn't check that. I'm not sure what voltage is expected, but I have a feeling that it is probably 5v. It would be good to know anyway - just in case there is a future repair like this.

[@GeoNeilUK]

I think alot of reasons why we see a lot of videos like this on channels like yours (and why a lot of original hardware is appreciating in value) is more and more old ahrdware isn't just aging out of functionality but more and more remaining hardware was deemed BER (Beyond Economic Repair) and just sat buried for years only to be repaired decades later when prices and rarity now make those repairs economical again!

[@bitsundbolts]

Yes, prices have increased, but I don't see the economic value in the repair - at least for me. I would have to live in Europe or the USA.

[@EpsilonsReviews]

“I hope it doesn’t invalidate my previous benchmarks!” - Retro computing in a single sentence…

[@bitsundbolts]

Yes, it is always an issue with old and flaky hardware.

[@VladoT]

I am more inclined to believe that it was a bad connection on a chip's pad, pin or via and not the chip itself.

[@bitsundbolts]

I removed the chip when I freed the board from corrosion. There was fresh solder on the pins and vias. I'm not sure if it could have been that. I've heard those inverter chips can get bad.

[@VladoT]

@@bitsundbolts Yes it could. Mark the chip as marginal and try it in some other working board in the future. It would be interesting failure mode if it's the case.

[@metroid031993]

What can happen with these chips is that the corrosion can actually wick up the legs and actually get into the internals of the package. It can destroy bond wires, etc. Perhaps what we're seeing is some capacitive coupling at low frequencies or something, thats allowing tests to pass, while internally the connection is damaged

[@bitsundbolts]

Interesting thought and absolutely possible that corrosion had something to do with that!

[@guldkrans]

Strange inverting it two times. Is the hex inverter needed to enhance the signal since the crystal is relative far away from the RTC chip?

[@bitsundbolts]

I would like to know this too. I followed one pin of the crystal which seems to go through the hex inverter three times before the signal arrives at the RTC chip. I wouldn't be surprised if the other pin of the crystal goes through the other three gates on the inverter chip.

[@WooShell]

Would be interesting to see how far off the RTC is if you let it sit around until next day.. 2 seconds drift within that short time between boot and checkit run is already quite a lot, I wouldn't be surprised if it was several minutes off overnight.

[@bitsundbolts]

This clock was running so fast! I left the board overnight and it was three to four days later as per the date.

[@WooShell]

@@bitsundbolts I meant after the repair, since checkit still reported several seconds difference between RTC and DOS time (a lot less than before, but still more than I'm used to seeing on those old boards).

[@sebastian19745]

@@WooShell Well, for me, at minute 20 the clock does not seem to pass 1s/s but maybe there is something with the video compressing. However, you are right, 2 seconds drift from boot to the test (that might be few minutes time, max 5) seems too much to me.

[@bitsundbolts]

I need to check this difference on other boards, but those one or two seconds between the DOS clock and RTC clock seem to be fine. I have tested the board now several days in a row and the time is kept perfectly - there is no drifting anymore.

[@aarong9378]

You showed that it went through two inverter gates. I wonder if both were double-triggering?

[@bitsundbolts]

It even went through a third one if I measured correctly. So, three gates. Hm, maybe you're right and those gates triggered simultaneously - which could explain the frequency!

[@tetiwhitton]

How did you attach the heatsink to the CPU? I'm considering doing the same, but unsure on how to do it. Thanks!

[@bitsundbolts]

There is a sticky thermal pad/tape on that heatsink. They usually are quite good and remain on the CPU. I may have to change it though - this CPU is still too hot!

[@UTubeRangerBob]

U1 is probably an SA or SARC 6042A, which might be compatible with the Intel 8042 keyboard controller.

[@bitsundbolts]

I tried different keyboard controllers salvaged from other boards. They all seem to be working fine. I need to get some spares though.

[@the_kombinator]

I noticed that a low battery will lose seconds depending on how low it is. Theoretically, could you not put in enough resistance between the battery (use a pot?) to slow it down enough? That might be a fun experiment.

[@bitsundbolts]

Oh, you mean to reduce the voltage applied to the circuit. Interesting. No idea if that would work, but I think I don't want to desolder that inverter chip again.

[@djdoo]

Does the first crystal work? Maybe it was the chip from the beginning

[@bitsundbolts]

I'm sure the crystal was fine. Only the chip was at fault. I tried the crystal first because I only had to desolder two pins. That wasn't the issue - so, I had to get my desoldering station out and remove the IC anyways.

[@AndyFletcherX31]

Those unbuffered 4069 are very fragile, they only need a sniff of static to kill them :)

[@bitsundbolts]

Who knows what happened to this chip. I also don't really remember where I got this board from - I think it was from the scrapyard, but it might also have been a donation from a viewer.

[@snap_oversteer]

Very interesting failure, I don't think I've ever seen failed timing/CMOS circuit before.

[@bitsundbolts]

I was also surprised, but I have seen others fixing the RTC circuit before. Sometimes it's the crystal, sometimes it's the RTC chip - well, and in this case, it was the 4069 inverter chip.

[@Pulverrostmannen]

I was funny enough just struggling with the same crystal circuit myself in a 286 system, it also ran at wrong speed but it turned out the circuit require above 4volts to run properly and I had replaced the barrel battery with a CR2032 after repairing extreme motherboard damage from the leak it had caused. since everything was corroded around the board the clock circuit was included in the damage and anything could been wrong. But I replaced everything without result except I have ordered new inverters and they seemed better than the original one. but they also drifted in speed. the solution was easy in the end and it was to use a 4,2V lithium battery instead. the new hex inverter is good to about 3,6volt so it gonna be close to a discharged cell. but it had me for a loop before i figured it out

[@bitsundbolts]

Good to know and thanks for sharing. I usually use an external battery pack (3x1.5v). That gives 4.5v at the battery connector. Then come a few diodes - I didn't check the voltage that finally arrives at the RTC chip, but I think it is around 4v.

[@Pulverrostmannen]

@@bitsundbolts ​​⁠ no problem 😊 I think this specific circuit is a bit different since the hex inverter is part of the oscillator and does not because of this only rely on the crystal and a counter. The HEF4069UBP is a Schmitt trigger and requires a certain voltage to switch and the oscillation stops below a certain voltage. And if it is the wrong operation voltage the frequency drift massively too. I had to use my scope and bench supply to research what was going on with different voltages. Since my computer is a “laptop” It does not have connector for external battery so I had to make my own and I connected a 4,2V pouch cell to it. It’s a miracle this computer even lives again from all that damage. It is the worst I ever repaired with over 20 broken traces and components everywhere. Even the sockets and connectors were gone. Stupid Vartas right?

[@bitsundbolts]

Yeah, those rechargable batteries caused so much damage to old hardware. Great that you could fix that laptop and figure out what was wrong with the RTC circuit. It must have been such a relief once you figured out that there was an issue due to know voltage. I don't know if I would have figured that out.

[@AK-vx4dy]

18:59 LIF sockets are fine, yo😢u just need proper tool, it looks like comb angled 90

[@bitsundbolts]

Yes, I have seen them, but never had one of those. I might have to make one myself once. I still find those sockets a pain to work with.

[@AK-vx4dy]

@bitsundbolts zif were complicated and pricey. I don't know how those were done but have some kind of "click" for CPU not to fall from vertically mounted motherboard. For me PLCC are worse, DIP also are scary in some way ;) with LIF i have some fear when CPU were thin ceramic.

[@Gersberms]

I had no idea DOS kept its own time!

[@bitsundbolts]

This was new to me as well not too long ago! But it looks like DOS counts properly and is immune to the RTC bug. However, the problem arises when the RTC tells DOS the time during the boot process. DOS counts perfectly, but with the wrong time. Anyway, that issue is now fixed! A couple of days have passed and the time is still spot on!

[@user-qf6yt3id3w]

A timer in the PIT chip generates a clock signal at 18.2 Hz. This causes interrupts and DOS uses those to keep track of time. I know this because if you reprogrammed the PIT chip to generate a different, usually faster, clock you had to make sure you called the original DOS handler or DOS's internal time value got messed up.

[@MonochromeWench]

That is how dos could keep time on the original IBM PC and XT that did not have a RTC and why DOS asks you to enter the date and time when booting if you don't have an autoexec.bat. If you have an autoexec.bat Dos assumes that it sets the correct date and time and doesn't ask the user. On AT class machines with an RTC, newer dos versions read the RTC but still asks to enter date and time cause that code was notremoved to maintain compatibility with systems with no RTC.

[@user-qf6yt3id3w]

@@MonochromeWench Exactly.

[@felixokeefe]

Testing logic chips using a programmer will only tell you if the thing works electrically. It tests at a very low frequency. You would need to build a circuit on a breadboard. Feed a known signal in from a signal generator and measure it's output with the oscilloscope.

[@watchmakerful]

Even more, the programmer tests these chips by feeding purely digital signals. A quartzed generator is an analog circuit and requires correct functioning of the chip in the whole range of input signals, not only with two digital levels.

[@bitsundbolts]

That's what I've been told before as well. I understand that it's probably too much effort to implement such tests for each chip on the programmer. Good to know that it's not guaranteed that a chip is ok just because the programmer test says so.

[@watchmakerful]

@@bitsundbolts I had a similar issue with an old Soviet clock. The clock ran fine on grid power (10.5..11 V), but intermittently stopped on the backup battery (9 V). The problem was in one of its frequency dividers which became stuck at 9 V.

[@bitsundbolts]

Wow! Technology from Soviet time...

[@leonardoliveira]

In that case of use you are dealing with analog domain reality which mean tolerances are tight.

[@gaborszucs8935]

Was the beginning a Scrap Computing reference or its just me..... :P

[@bitsundbolts]

Haha, right - I didn't think of that, but you are absolutely right - similar (but only for this video).

[@Lazydiv]

Pause the video at the boot screen intro to do something. Forget it for a few minutes. Get back to the computer and have a mini heartattack over wtf happened to the computer while i was gone... Maybe my teacher was right about my head only being usefull to smash rocks with.

[@bitsundbolts]

😅

[@monad_tcp]

I've seen clock skew even higher once, minutes would fly like seconds.

[@bitsundbolts]

Wow. That frequency must have been extremely high!

[@sparky4insano]

repair maor!

[@bitsundbolts]

Yes, I still have a non-working 286 OCTEK FOX II.

[@pvc988]

I wonder why did they go with their own oscillator when RTC chip has one already builtin. And RTC chip probably does it using less power than the inverter circuit.

[@bitsundbolts]

Maybe those all-in-one modules weren't around at that time. I think those Dallas modules became popular with Pentium boards.

[@SidneyCritic]

I have had a stuck crystal on a 4th Gen that wouldn't boot. I accidently did a diode test across it, ie, it applied a slight volage, and now the thing is fine - lol -.

[@bitsundbolts]

Haha, wake-up call!

[@watchmakerful]

Where is that 2 MOhm resistor? This value does not make any sense! If it is in parallel to the oscillator, it must be 10..33 MOhm. If in series, typically 100 kOhm to 1 MOhm.

[@bitsundbolts]

Those two resistors in the thumbnail. One is 2 MOhm the other one is 1 MOhm. Based on the color bands, this is correct.

[@MlokKarel]

Crystal oscillators are very much heat sensitive. Desoldering one that is soldered to a PCB is next to certain to boil it to death 😢

[@bitsundbolts]

It looks like the one I transferred is working. The one I removed from the donor board only had a metal rod to keep it in place - it wasn't soldered. But then I soldered it to the board, but I used low-melt solder for that.

[@awilliams1701]

It wasn't just the chip it was the crystal too.

[@bitsundbolts]

That is definitely a possibility. Unfortunately, I no longer have that crystal.

[@kaunomedis7926]

Sorry, you do not know what are you doing.

[@bitsundbolts]

I never said I know what I am doing. I am trying to understand what could be wrong without an electrical engineering degree. The result is absolutely satisfactory for my skill level since the board work after replacing the faulty chip.