Maybe I measured wrong before, everything is possible and probable, but today I did more measurements.
Lifted 5 pin of the MIC,JP8 1-2;4-5 85k;3-5 69k
5 to far end of R20 85k but never stabilizes, oscillates up/down 0,3
5 to nearest end of R21-71k
Lifted pins 3 and 5
5 to far end of R20-89K
Value of the resistors in circuit;
JP8 1-2
R20-85k;R21-69k
JP8 OFF
R20-145k oscillating a bit.
R21-89k sometimes going of scale, tested a lot of times.
JP8 2-3
R20-94k;R21-90k;R69-70k
The far end of R20 is connected to the Socket, do the current goes through R20?.
Pin 4 is also connected to the Socket.
The 3v-5v connects the vout to the Socket or connect the 5v of the AT connector to the Socket.
There's something odd with this board, I'm not very intelligent but, following your schematic those resistor readings are very weird, without jumper on JP8, why R20 is not 160k, no more or less....
Measured resistance between the far ends of the resistors with JP8 off;
R20-R21=154K;R20-R69=198K;R20-R68=315K;R20-R67=191K;R21-R69=190K (but after a minute it still increases slowly).
R21-R68=203K,R21-R67=175K
R69-R68=212;R69-R67=185K
R68-R67=195K
So the nearest end of,R21,R69,R68 and R67 is connected, but also the far end of then and R20 is connected somehow.
Always taking the reference of the Mic as nearest or farthest.

Hoping wrote on 2021-11-30, 15:44:

Maybe I measured wrong before, everything is possible and probable, but today I did more measurements.

Are these with the 5V-3V jumpers in place? Probably best to leave those off for the moment. As you've found, they connect the output to the socket, and at the moment that just makes things more complicated. We want to get your 3.3V on pin 4 of the regulator first.

Lifted 5 pin of the MIC,JP8 1-2;4-5 85k;3-5 69k 5 to far end of R20 85k but never stabilizes, oscillates up/down 0,3 5 to neares […]
Show full quote

Lifted 5 pin of the MIC,JP8 1-2;4-5 85k;3-5 69k
5 to far end of R20 85k but never stabilizes, oscillates up/down 0,3
5 to nearest end of R21-71k
Lifted pins 3 and 5
5 to far end of R20-89K

Is that measuring from the lifted pin 5, or from the PCB hole for pin 5?

The far end of R20 is connected to the Socket, do the current goes through R20?.
Pin 4 is also connected to the Socket.

That sounds right, Pin 4 is the output, which goes to one side of the 5V-3V selector block, with the middle of that block going to the CPU. It also goes through R20 to Pin 5 (Adjust). A very small amount of current (<1mA) goes through R20, then through R21 and finally to Ground. The Adjust pin is connected between R20 and R21 and can sense the voltage at that point, which is set by the ratio of R20 and R21. The regulator should then adjust the output voltage so that the voltage at the adjust point (between R20 and R21) is at 1.25V above Ground.

There's something odd with this board, I'm not very intelligent but, following your schematic those resistor readings are very weird, without jumper on JP8, why R20 is not 160k, no more or less....

I think it's partly because everything is in circuit. So the test current from the meter can go through the component you're measuring, but can also go other routes, which means it measures a lower resistance than just the resistor you're testing. In some cases it'll be higher. For example, measuring between the far ends of R21 and R69, current can go through R21 to Ground, then from Ground back up through R69, so they will be connected. If JP8 is removed then it should measure as the sum of the two resistors.

I'll have a look through what you've measured and see if I can understand it.

Is that measuring from the lifted pin 5, or from the PCB hole for pin 5

I've had the 5v-3v jumper in 3v position, and I used the pin1 hole of the LT instead of the pin 5 hole of the MIC to prevent touching the legs. My hands aren't very steady and I have to take the measurements more than once to be sure.

I think it's partly because everything is in circuit. So the test current from the meter can go through the component you're measuring, but can also go other routes, which means it measures a lower resistance than just the resistor you're testing. In some cases it'll be higher. For example, measuring between the far ends of R21 and R69, current can go through R21 to Ground, then from Ground back up through R69, so they will be connected. If JP8 is removed then it should measure as the sum of the two resistors.

That means that the 3.3v circuit is not isolated from the rest of the board right? So that makes everything more difficult to understand I think.
I think I damaged the board somehow because There's no voltage on the nearest end of JP8, I have the pins 3 and 5 of the MIC lifted.

I go on a bit long here, but eventually realise that some of the measurements are caused by a capacitor charging up. I'll leave what I originally wrote as it at least shows what I was thinking. The short version is that I still can't see anything wrong, and the regulator should be putting out around 3.4V ( 1.25*(1+91/160) ).

Try putting JP8 on 1-2, remove the 5V-3V jumpers. Then fit the spare regulator (maybe there is a problem with it) and check the connections (so pin 1&2 to +5, 3 to Ground, 4 to right-hand column of 5V-3V jumper block, 5 to bottom row of JP8. If those are all ok, turn on and see what the voltages on each pin are.

I've added a load resistor/capacitor to my earlier diagram, which hopefully makes it a bit clearer about some of the problems of measuring components in circuit.

Hoping wrote on 2021-11-30, 15:44:

Lifted 5 pin of the MIC,JP8 1-2;4-5 85k;3-5 69k
5 to far end of R20 85k but never stabilizes, oscillates up/down 0,3
5 to nearest end of R21-71k

So 3-5 should be the same as 5 to nearest R21 (pin 3 is Gnd, and the nearest ends of R21,R69,R68,R67 are all Gnd). I'd consider 69k and 71k to be close enough to look reasonable. Equally, 4-5 is the same as 5 to far end of R20 (4 is Vout, which goes to the far end of R20). You've got both of those as 85k. The oscillation you see might either be slightly noisy probe contacts.

I'm going to guess If we assume the existence of a ~100k resistor between Vout (pin 4) and Ground, then that more-or-less matches with your readings. Measuring from 3(ground)-5(adjust) is measuring across R21 (91k), but if current can flow from Vout to Ground then it's also measuring from Adjust to Vout (through R20, 160k) + Vout to Ground (through my guessed at resistor). So effectively it's measuring 91k in parallel with 160k+100k, which is always going to be less than 91k. If you were measuring with the 5V-3V jumpers in place then maybe that's that cause of the numbers not matching the resistor values.

On the spare regulator, is there a resistance between pins 3&4 (Ground-Vout)?

Lifted pins 3 and 5
5 to far end of R20-89K

So that's equivalent to measuring 4-5. Measured resistance has gone up a bit, but I think it still suggests there's a resistance between Vout and Ground.

Value of the resistors in circuit;
JP8 1-2
R20-85k;R21-69k

So they're as before.

JP8 OFF

This one's interesting because it breaks the circuit with that external resistance I think is confusing the results

R20-145k oscillating a bit.
R21-89k sometimes going of scale, tested a lot of times.

So these are not much closer to the actual resistor values. Going off scale could be interesting. It could either be a dirty probe contact, or maybe a dry solder joint. It might be worthwhile just reflowing the solder there.

JP8 2-3
R20-94k;R21-90k;R69-70k

I think that makes sense. R21 is now disconnected so reads it actual value. R20 is now in parallel with R69+my guessed Vout-Ground resistance. R69 is bigger than R21, so R20 appears to go up.

There's something odd with this board, I'm not very intelligent but, following your schematic those resistor readings are very w […]
Show full quote

There's something odd with this board, I'm not very intelligent but, following your schematic those resistor readings are very weird, without jumper on JP8, why R20 is not 160k, no more or less....
Measured resistance between the far ends of the resistors with JP8 off;
R20-R21=154K;R20-R69=198K;R20-R68=315K;R20-R67=191K;R21-R69=190K (but after a minute it still increases slowly).
R21-R68=203K,R21-R67=175K
R69-R68=212;R69-R67=185K
R68-R67=195K
So the nearest end of,R21,R69,R68 and R67 is connected, but also the far end of then and R20 is connected somehow.
Always taking the reference of the Mic as nearest or farthest.

If my guessed at resistor/load is real then I think these look reasonable, they're mostly adding the R21 to R67 resistors. So R21 to R69 is 91k+100k, or R68 to R69 is 110k+82k (ish). The interesting ones are the R20 ones. They have to go through my guessed resistor, then back up R21/69/68/67.

Which looking at those finally makes me realise that this guessed at resistor is actually probably a capacitor gradually charging up as you take your measurements.

Hoping wrote on 2021-11-30, 17:22:

I've had the 5v-3v jumper in 3v position, and I used the pin1 hole of the LT instead of the pin 5 hole of the MIC to prevent touching the legs. My hands aren't very steady and I have to take the measurements more than once to be sure.

Always a good idea to find the easiest place to take a measurement from.

That means that the 3.3v circuit is not isolated from the rest of the board right? So that makes everything more difficult to understand I think.

Yup. In circuit measurements can be a pain. Being able to disconnect bits of the circuit (like with that jumper block) can be very useful.

I think I damaged the board somehow because There's no voltage on the nearest end of JP8, I have the pins 3 and 5 of the MIC lifted.

Ah, if 3&5 are lifted then it's not going to work anyway. Probably want to have all the pins connected what applying power.

On the spare regulator, is there a resistance between pins 3&4 (Ground-Vout)?

Infinite resistance between pins 3-4, tested on the soldered one and on the others I have (I still have the first ones)
The capacitor is C32, it's positive lead is connected to the far end of R20 and of course the negative lead is connected to ground, and the nearest end of the other resistors is connected to ground.
Just like in your schematic.
Maybe this capacitor is failing and causing all the troubles?

Hoping wrote on 2021-11-30, 18:35:

Infinite resistance between pins 3-4, tested on the soldered one and on the others I have (I still have the first ones) The capa […]
Show full quote

On the spare regulator, is there a resistance between pins 3&4 (Ground-Vout)?

Infinite resistance between pins 3-4, tested on the soldered one and on the others I have (I still have the first ones)
The capacitor is C32, it's positive lead is connected to the far end of R20 and of course the negative lead is connected to ground, and the nearest end of the other resistors is connected to ground.
Just like in your schematic.
Maybe this capacitor is failing and causing all the troubles?

Ok. Don't think the capacitor is causing the problem, mostly because it's not causing a short between Vout and Ground. From your measurements it's apparent resistance varies (probably depending on how much charge is stored), but it's generally about 50k. So it's not causing too much current to be drawn through the regulator.

I'd reconnect the regulator pins you lifted (or swap it for a spare), remove the 5V-3V jumpers, fit JP8 1-2, then turn it on. Measure voltages from Ground to all 5 pins, ideally using the actual pins. If the power supply has molex connectors then you can stick the ground probe in the ground on the connector, then you only need to worry about getting the red probe on each pin of the regulator. Pins 1&2 should be +5V, Pin 3 should be 0V, Pin 4 should be 3.4V and Pin 5 should be 1.25V.

At last... it's working, reflowed everything, desoldered C32 and tested with the ESR meter and it seemed ok so I resoldered it, by chance I've found a defective jumper cap, it is missing an small piece of metal on one side, maybe it did a false contact so I replaced it.
Also when I was resoldering the MIC regulator the copper of pin 3 on the board flew complete and I've put a wire between the nearest end of the resistors(GND) and pin 3.
Now I have four voltages available at pin 4.
JP8
1-2=3.34V
3-4=3.20V
5-6=3.01V
7-8=3,61V
Some voltages aren't the ones expected but they are perfect for my aim, and on the positive lead of C32 the voltage is 0,02v lower, and on the socket it's 0,03v lower, so 3.01v on pin 4 is 2.98v on the socket.
I'm very very happy with these voltages because my aim was to undervolt and it worked better than expected, undervolt plus overclock, I can't ask for more.
Now it's working at 120 MHz (3*40) although it says 100 MHz at bios screen, at 3.01 Volts. Tested with Quake 1.08 timedemo1. 8,4fps at 100 mhz and 10,7 at 120 mhz, with sound enabled. I can't believe that the CPU works so well even after being overvolted to 5V before.
And you were right about the heatsink because it reaches 39,6 degrees Celsius after twenty minutes of running Quake wile the CPU only reaches 16,9 degrees. and the ambient temperature is very low at this moment.
So the specs of this computer are:
Intel DX4 100Mhz at 120Mhz.
Ram 8mb 70 NS, this needs an upgrade to 16mb at least.
Quadtel S3 805 2mb VLB
Unknown model VLB IDE I/O controller, I've never researched the model because it has all the jumpers documented on the board.
Creative CT2940
And an unbranded CDROM, at least it doesn't have anything printed, only something written in Japanese I think, maybe Mitsumi.

I don't have enough words to thank you snufkin.
Thanks a lot.
And I think that we documented the 3 volts circuit of this board fairly well. 😉

Congratulations, glad it paid off. Nice that you've been able to undervolt and overclock at the same time. And I think you've ended up pretty close to the expected voltages; based on your measurements of the resistors then the measured voltages are all <0.7% from the calculated value. Even going by the colour bands then the furthest out is 160:91 which ends up 2.4% out and that's because you managed to find a 91k resistor that measured 92.7k (technically out of spec for a 1% tolerance resistor). I think the jumper was probably making a contact as we could see a difference in the measurements if it was fitted or not. My guess was the Pin 3 Ground connection was the problem and the jumper wire has fixed it.

I'm assuming that CPU@16.9C is a bit lower than it was before, that sounds like it's at room temperature? Of course, now that you know how the regulator works you can always change one of the resistor to drop the voltage even lower (e.g. keeping the 158.6k R20 and putting in a 120k would take the output down to ~2.88V. I've no idea what the undervolting capability of the DX4 is. 40C for the regulator is fine, lots of headroom for warmer weather. It would have probably worked bolted to the motherboard, but it would have worried me.

Never tried playing Quake on a 486, 11fps sounds almost playable if there aren't too many grenades around... Think it was the first game I played after getting a P166.

Now it is running at default,100*33 3.3v. I'm not planing on runnig Quake on this computer, only for testing purposes, and the heatsink of the mic after an hour was really hot, so after so much work I'm not going to risk damaging anything. The Cpu has a big heatsink with thermal paste and a fan, overkill for an 486. And I noticed that the mic gets hotter as de voltage decreases.

Hoping wrote on 2021-12-01, 13:27:

Now it is running at default,100*33 3.3v. I'm not planing on runnig Quake on this computer, only for testing purposes, and the heatsink of the mic after an hour was really hot, so after so much work I'm not going to risk damaging anything. The Cpu has a big heatsink with thermal paste and a fan, overkill for an 486. And I noticed that the mic gets hotter as de voltage decreases.

That's normal since the mic has more power to dissipate as heat, since there is more voltage drop over the mic
for example :
input voltage 5V, output voltage 3.3V, voltage drop 1.7V, current drawn (for example 3A) means power dissipated by mic is 1.7V * 3A = 5.1W
now lower the voltage 0.3V. output voltage 3V, voltage drop 2V, same current drawn, 2V * 3A = 6W

btw, this is only true for linear regulators.

weedeewee wrote on 2021-12-01, 13:55:

That's normal since the mic has more power to dissipate as heat, since there is more voltage drop over the mic for example : inp […]
Show full quote

Hoping wrote on 2021-12-01, 13:27:

Now it is running at default,100*33 3.3v. I'm not planing on runnig Quake on this computer, only for testing purposes, and the heatsink of the mic after an hour was really hot, so after so much work I'm not going to risk damaging anything. The Cpu has a big heatsink with thermal paste and a fan, overkill for an 486. And I noticed that the mic gets hotter as de voltage decreases.

That's normal since the mic has more power to dissipate as heat, since there is more voltage drop over the mic
for example :
input voltage 5V, output voltage 3.3V, voltage drop 1.7V, current drawn (for example 3A) means power dissipated by mic is 1.7V * 3A = 5.1W
now lower the voltage 0.3V. output voltage 3V, voltage drop 2V, same current drawn, 2V * 3A = 6W

btw, this is only true for linear regulators.

It seems to me a very inefficient energy conversion since it wastes almost 50% in heat.
If I find myself in such a case again, I will try to use more modern technology, it may even be easier.

Hoping wrote on 2021-12-02, 17:48:

It seems to me a very inefficient energy conversion since it wastes almost 50% in heat.
If I find myself in such a case again, I will try to use more modern technology, it may even be easier.

Oh, they're horribly inefficient. But they're also cheap, small, quiet (electrically I mean, no switching noise), reliable (if they don't overheat) and easy to include in a design. If you don't need it to be adjustable then you really just need the regulator and a capacitor. Switch mode stuff needs power transistor, gate drive, controller, big inductor and capacitors. If you know you're only going to draw an amp or two, and the voltage drop needed isn't too big, then regulators still have a use. Going to a lot of effort to save a couple of watts when the whole system (including monitor and like) is burning over a hundred doesn't usually make sense. I think they disappeared for CPU power supplies when the currents got too big to be able to dump the heat (P2 onward maybe?).

Although I'm a little out of touch, so maybe there are now self contained switch mode regulators.

I was thinking, if it ended not working, on use a DC-DC converter based on an XL4005 converter, they are cheap and fairly small and a lot more efficient, and maybe also more stable in the output voltage. If I've connected this to ground and to pin 4 taking out the resistors and the MIC, I guess it would work.

Another day and another problem with this board, or, again I'm doing something wrong. It doesn't charge the battery, there is 0.0V on the battery.
When they gave me this computer, the only bad thing I saw was the battery that was already leaking a bit, so I took it out and came up with the wonderful idea of soldering some cables to the solder points of the battery and using a CR2032 battery without any modification But now I know this was wrong, just like putting a DX4 on 5 volts. What is done is done and now that I want to put a new rechargeable battery in it, I find that there is no voltage at the soldering points of the battery.
According to how little I manage to understand and from what I have read in the forum it is possible that one of the diodes D3 and D6 fails, or that both fail, I think that D3 is the problem since on one side it has voltage and on the other nothing, and the side with nothing is connected to the battery, but it also surprises me that D6 has the same voltage on both sides.
Already tried bridging the pins 2&3 of the external battery connector but nothing.
And another confusing schematic, I'm really bad at this.

Hoping wrote on 2021-12-03, 17:13:

I think that D3 is the problem since on one side it has voltage and on the other nothing, and the side with nothing is connected to the battery, but it also surprises me that D6 has the same voltage on both sides.

It might actually be the other way around. The diodes stop the flow of current, so D3 might be stopping current from wherever it's connected from flowing to the battery, so you read a voltage one side but not the other. Either that or it's failed open. D6 being the same both sides sounds a bit odd as there would normally be a voltage drop across the diode when it's conducting, so maybe it's failed short. Won't really know with tracing what goes where. Don't think the 2032 can have damaged the board, I think the problem is more the board trying to charge the non-rechargeable battery.

Can you post photos of the front and back of board, showing a little more, maybe from BIOS chip to the power connector? I can follow it mostly from Deksor's photos on ultimateretro ( https://www.ultimateretro.net/en/motherboards/7821 ) but a brighter higher res version of the area might help.

Already tried bridging the pins 2&3 of the external battery connector but nothing.

Ah, I was hoping that was it...

And another confusing schematic, I'm really bad at this.

It's an ok start, seems to be matching what I've got so far.

Can you post photos of the front and back of board, showing a little more, maybe from BIOS chip to the power connector? I can follow it mostly from Deksor's photos on ultimateretro ( https://www.ultimateretro.net/en/motherboards/7821 ) but a brighter higher res version of the area might help.

I couldn't get the right angle but I hope they are good enough, as a note if it may be relevant, in circuit, D3 and D6 give the same reading, .632.
Edit: forgot the photos, I can believe it.....

Well, I can't figure out that charging circuit. As best I can tell, charge leaves the battery on the dark blue line. The only place that seems to go is pin 2 of J3 (it also goes to the pad for D1, but that's not fitted). To run off of the battery then Pin 2 and Pin 3 should be shorted with a jumper. Following the cyan line, Pin 3 of J3 looks like it goes to pin 1 of D3 and pin 2 of D2. D2 is actually a 0 ohm link, which then connects to Pin 1 of J3. So Pin 1 and Pin 3 should measure as shorted together. That means the only place for current to leave is through D3. D3 then goes to pin 3 of the clear CMOS jumper JP13. So that's the battery supply. Putting a jumper on 2-3 of JP13 (now on the reddish purple line) you can see the middle pin then connects to R26 (I think a 10ohm resistor) from where it runs (light blue line) to the various things that the battery needs to power (like the inverter gates to make the real time clock crystal oscillate).

I guess Q1 and Q2 are to do with switching in the main board supply when the computer is turned on, and isolating the battery backed bits when the computer is off.

So that all seems pretty self contained. D6 seems to be something to do with the board level power good signal.

The problem I have is that if current from the battery can only leave via D3, then there's no way for current to enter the battery, it'd be stopped by the diode. So how does the battery get charged?

The only way I can see at the moment for it to work is if D3 is a zener diode. That way, if the board supply voltage was high enough above the battery voltage then the zener would start to conduct and allow the battery to charge. But if that's the case, we don't know what the breakdown voltage is.

Maybe try putting a 1k resistor across where the battery would go, turn on and see if there's a voltage across it. I'm wondering if you'd only see a voltage if the (possible) zener is conducting.

When you used the CR2032 were the CMOS settings saved and the date and time correct?

I've soldered terminals to the battery soldering points and attached a rechargeable 3.6v nicd battery. When powered OFF, the battery's voltage is on both sides of D3, and it decreases slowly, the battery is old and doesn't hold a charge for long time but it still charges when connected to a 3v power supply, and there's also battery's voltage on one side of C40 and C39; and when powered ON there's 5 volts on one side of D3 and battery's voltage on the other, so since D1 is not present, there are 11,48v on one on of the soldering points of D1, and the battery is on the odder side, and D3 seems to be the only way to battery's positive I think that D3 is defective, but following your explanation, what kind of diode is D3?.... When powered on there are 4.90v on one side of D3 and I guess the battery needs 3.6v for charging, so the diode needs to drop around 1.3v right?
Also tried with a 1K 1% resistor and 0.0 volts on the battery soldering pads.
Early I forgot to mention that D6 goes to the middle pin of JP13 but not a direct short, the multimeter reads .698 with the negative prove on the middle pin.
The RTC and CMOS work at 3.6v or 4.9v?, or there's also something not right there, maybe D6 doesn't work properly and doesn't drop the voltage enough, and so, D3 also don't work properly.
I don't remember if when I had the CR2032 fitted it saved the CMOS settings because that was around eight years ago,time passes by so fast, I think that the RTC was slowing down but not sure.
Edit: forgot to mention that the board was running with the JP13 jumper on the 1&2 position, I didn't it notice until now, maybe that did some damage?

Updated the schematic with the connections that I could follow of the Q1 and Q2 transistor.
Q1:2N3904; E=0V,B=0,7V,C=4,22
Q2:2N3906;E=5.09,B=4,22V,C=4,87V.
Edit: corrected one mistake in the photo.