CalamityLime wrote on 2022-06-05, 19:28:

3.3 from the pico at could be handy. You'd want a nice stable connector […]
Show full quote

3.3 from the pico at could be handy. You'd want a nice stable connector

---

Some simple feedback:

Pads:
Make the pads for the ATX connector larger (and the AT connector if there's space). The pads you have are on the smaller side, I think you'll find that solder will ball up on the conductor itself without a good contact to the pad. I've had issues with tiny pads doing this on even leaded solder.

Traces
I do see that you have nice thick traces for the -12 and -5 volt rails but to make room for the pads here are some numbers that might help:
You have the thick traces at around 50mil (1.25mm) which is good for around 1.4amps but most negative rails are about half an amp to an amp.
You're not going to see a whole amp on either the -12 or -5 but to accommodate an amp you want around 32mil which is about 0.8mm.

For PWR_OK
It's a signal line so you don't need much. To accommodate say half an amp (can't find a proper spec but it feels right) you want a trace about 0.35mm

I'm giving trace thickness measurements based on 1oz copper, if you're using 2oz copper than you can half the thickness.

Hope that helps a bit.

I had to find some of information on trace thickness for the USB-2-232-KBD to deal with RGB keyboards. Ultimately it came down to fitting the thickest trace I could fit.

Might play around with the pads and traces. Thx for input.

Here is a pic of a soldered unit (prev version) latest versions were (re)done by me, but original was from wiretap.

If you wanna play with it, the thread is posted above, you could post your version. It’s open source.

I’ll probably put one of these together myself soon. And report results about the pads. 😀 ( I already ordered)

Sphere478 wrote on 2022-06-04, 10:15:

I think those three voltage jumper positions have one side tied together yes?
If so, next assumption is this,
The other side all goes to the same place but not before passing through a resistor that is different for each location?

Indeed, that's exactly the typical setup for 486 and Pentium voltage regulators.

Are the higher voltage settings higher or lower resistance? If higher resistance gives lower voltage, then do this, Grab a hdd l […]
Show full quote

Are the higher voltage settings higher or lower resistance?
If higher resistance gives lower voltage, then do this,
Grab a hdd led wire and solder a pot to it, plug it into it in to one of the jumper spaces and test the voltage at different settings.
You now have a way to set lower voltages. If My theory is correct. And no need to modify the board.

Whether higher resistance means higher or lower voltage depends on the mainboard design. The output voltage is set by an voltage devider. On some boards, the high-side resister is jumper-switchable, on other boards, the low-side resistor is jumper-switchable. You describe a working solution for boards with require higher resistances for lower voltages. The typical approach for boards that use lower resistances for lower voltages is to set multiple voltage jumpers at the same time, so the lower effective resistance of both resistors in parallel applies. There is at least one wide-spread socket 7 board that gets close enough to 2.2V when you jumper both 2.7V and 2.8V at the same time to run newer K6 processors in-spec.

Hello.

About the idea of pumping the 3.3v from the ATX psu to the CPU IO and then to the step down converter for CPU vcore.
As I mentioned in the thread about the ATX pico, I was thinking that the 3.3v line should be controlled by a mosfet. The logic being that the 5v power good line of the PSU goes high once the power rails are stable. It would be nice to have the cpu get power once the psu rails are stable and not before.

So I was looking into the topic and spec sheets and I came up with parts and a potential circuit.

2 p-channels in parallel to help cut the resistance and spread any heat HSU60P02.
Resistors to pull them high thus keeping the mosfet off, protection diode for any reverse current.
Then an 2n7000 n-channel mosfet that is switched on by the 5v power good signal that should pull the gates of the p-channels to ground thus sending power to the CPU.

What do you think?

I never made a circuit with a mosfet before so I could very easily be missing something important. It's nice to actually feel like I'm starting to understand mosfets, so even if the circuit is a waste, may have learned something from it at least.

CalamityLime wrote on 2022-06-07, 00:38:

The logic being that the 5v power good line of the PSU goes high once the power rails are stable. It would be nice to have the cpu get power once the psu rails are stable and not before.

That sounds like a good idea (preventing the CPU to get overvolted by cutting off the supply until it is stable) at first glance, but depending on the mainboard, it's not. When the power good line goes high, the mainboard might expect that all components already have stabilized power and may immediately start operation of the system, for example by deactivating the RESET signal. If the CPU just starts getting powered at that point, it might not be able to perform a clean power-on reset and behave strange. It highly depends on the board, how long after power good going active it takes until the board expects everything to be in perfect working condition, so your logic might work on some boards, but it might fail on other boards.

CalamityLime wrote on 2022-06-07, 00:38:

2 p-channels in parallel to help cut the resistance and spread any heat HSU60P02. Resistors to pull them high thus keeping the m […]
Show full quote

2 p-channels in parallel to help cut the resistance and spread any heat HSU60P02.
Resistors to pull them high thus keeping the mosfet off, protection diode for any reverse current.
Then an 2n7000 n-channel mosfet that is switched on by the 5v power good signal that should pull the gates of the p-channels to ground thus sending power to the CPU.

What do you think?

I never made a circuit with a mosfet before so I could very easily be missing something important. It's nice to actually feel like I'm starting to understand mosfets, so even if the circuit is a waste, may have learned something from it at least.

Your circuit looks a lot like a bipolar transistor circuit modified for MOSFET operation. You can simplify it a lot. But before just simplifying: Drop the 10K resistor between the 2N7000 and ground. MOSFETs are voltage controlled, so you want to have "source" connected to a stable potential like ground to get a clear definition about what a certain gate voltage in respect to that potential does. In your circuit, nearly two thirds of the 5V supply (i.e. 3.2V or the like) drop at the source resistor of that MOSFET, so only 1.8V of the 5V power good signal remains between gate and source. That is likely not enough turn the MOSFET on completely.

Now to the simplifications: the gates of both power MOSFETs is in parallel anyway, so you just need one pull-up resistor. 10K looks like a good choice, though. You can omit the left diode, because all standard MOSFETs have a diode like that integrated by design (the "body diode"). It's even part of the MOSFET symbol to remind you. Relying on the body diode to carry high current over long time is bad style, but it is perfectly fine to use it for discharging the 3.3VP line in the case of powering down the computer.

The right diode seems pointless, too. The inductance of your coil should be low enough that the energy stored in that coil can safely be dumped into the output capacitor in case the processor decides to not need any current at the moment (e.g. HLT instruction), so you don't need any freewheeling diode. For discharge at power down, the low DC resistance of the coil is good enough to provide a current path, no need for that diode.

To protect the 2N7000 (those bastards are very ESD sensitive), you might want to add a small cap (like 100pF) between gate and ground, especially if the gate resistor you use is above a like 33Ohms. That gate resistor and the cap I propose form a low-pass (i.e. delay the turn-on, which might be problematic, see my other post), so don't go too high with the gate resistor if you want quick reaction to power good. And finally, as MOSFETs are voltage controlled, you want something to pull down the gate of the 2N7000 when there is no 5V supply. Relying on the low impedance of the 5V line will indeed work if the circuit is permanently attached to the mainboards 5V line, so no immediate action required - but generally a pulldown resistor (like the 10K pull-ups at the power FETs) is a good idea.

Hello!

Sorry for being quiet, suddenly had the house to myself so I just enjoyed living as a hermit in the quiet for a bit.

Quick update on the tag ram:

I soldered little wires to the chipset. Not the nicest of soldering but should be fine, I got 2 identical tag ram chips from electomine on ebay and JLCPCB finally shipped the tag ram boards I made.

I also socketed the keyboard controller IC and I removed the din port because it was looking a little destroyed.

About the circuit,
I'll have to look into the point about the board expecting the system to have power already when power good line goes high.
And thanks for the input on the circuit 👍🏻

About the RTC
I ordered a ds 1685-5 from Farnell, might be fun to experiment with that, especially if the ds 12885's I ordered are random fakes.

I also ordered 2 D45VH10G pnp fets to replace the d45h2a's of the power regulation on the motherboard, if they are need replacing that is. I think they are drop in replacements.
I still have yet to order a cpu.

Super glue the wire down to keep that from ripping up a trace

Sphere478 wrote on 2022-06-13, 02:50:

Super glue the wire down to keep that from ripping up a trace

The photo was taken immediately after I got the wires in place. I was honestly stunned I got them on.

I used some tape to give it some temporary support.
I think I'll add a bit of epoxy or even solder mask stuff.

I finally ordered some cpu's

I ordered 2 different ones. A basic pentium mmx of some speed and an AMD k6-2 500 if I'm feeling brave.

grab a k6-2+ 570 and mod it.

I have a question,
Would it be so bad if I bought a simple ready made step down buck converter off of amazon for vcore voltage?
Thanks to how the motherboard is layed out I don't even need to mod the board for that, just pump 2.2v into a header pin.

Something like this say
https://www.amazon.co.uk/Step-Down-Converter- … s%2C64&sr=8-117

Just modded for fixed resistors.

I checked several of those ready-made adjustable VRMs when I worked on my Gigabyte GA-586HX

Gigabyte GA-586HX (Rev. 1.53)

Many modules were supposed to do 5,6 or 8A. When I checked the datasheets of the used regulators it turned out that in the voltage-range that´s relevant for us when there´s 5V input and for example 2V output, the maximum current wasn´t 8A as claimed but just 2A.

You should check ther data carefully before ordering the modules.

CalamityLime wrote on 2022-06-16, 16:07:

I have a question, Would it be so bad if I bought a simple ready made step down buck converter off of amazon for vcore voltage? […]
Show full quote

I have a question,
Would it be so bad if I bought a simple ready made step down buck converter off of amazon for vcore voltage?
Thanks to how the motherboard is layed out I don't even need to mod the board for that, just pump 2.2v into a header pin.

Something like this say
https://www.amazon.co.uk/Step-Down-Converter- … s%2C64&sr=8-117

Just modded for fixed resistors.

You could go all out and get a lab power supply. Super adjustable, super stable.

Btw:

Re: vrm module project thread roundup, share ideas, make new designs

I was looking into other solutions,
There is the murata units and i was looking at them wondering if i could make a mounting pcb that fits where the old PNP transistor and it's heatsink sat.

I done some initial investing and i think 4 layer pcb using the front and back to handle the bulk of the current with the middle layers having any signal lines for resistors and ground, should let me make a PCB with 1oz copper that may take around 10 amps okay.

I'll have to look into some more measurements but eh, initial idea. Prefer to make sure the board works first but it was fun to look into today.

CalamityLime wrote on 2022-06-17, 03:05:

I done some initial investing and i think 4 layer pcb using the front and back to handle the bulk of the current with the middle layers having any signal lines for resistors and ground, should let me make a PCB with 1oz copper that may take around 10 amps okay.

There´s already a solution for that:
Gigabyte GA-586HX (Rev. 1.53)

majestyk wrote on 2022-06-17, 05:12:

There´s already a solution for that:
Gigabyte GA-586HX (Rev. 1.53)

Yes, i had spotted that, thats where i saw the murata modules to begin with.
But i think i want something different for the shape and how it mounts than what was in that thread.

I was planning to explore that idea a bit today but something came up. All i got done was the footprint for the 12 amp module.

I do have a question though, what do I do with the sense lines? Just connect them to vout and gnd? Seems too simple for the pins to be exposed like that.

Looking at the gerber files in the other thread, I think I do nothing with the sense lines unless I'm doing something smart.

I started putting together the PCB I had in mind, hopefully in it's current state the idea comes across.
The plan is to replace the PNP transistor and mount the replacement using the holes for the transistors heat sink. I'd have to run a wire to ground but other then that it's fairly drop in.

Love it!

Alrighty!

The back of the board has the resistors and the resistors have values. The table to the side is upside down since I think that's how it'll be looked at most once it's installed.
The resistor network is a simple thing. Default the trim to gnd resistance is set for 1.8v, so hopefully if the switch breaks it won't fry anything. Flicking on the switchs just adds another resistor in parallel to decrease the resistance and up the vcore voltage.

The module is off to the side to make routing a thick trace out vout possible.

I'm wondering what to do with the blank space to the right, a power good LED? A canvas for graffiti?