I recommend a POST code diagnostic card. They aren't expensive, and you won't cry if blows up. It will also display an LED for the 5V rail.

You'll still need a video card, so many buy a cheap one. The Chinese Rage XL cards on eBay all require 5V and won't boot without it.

Lots of cheap Rage XL PCI graphics cards for Socket 7 on eBay and Aliexpress

NachtRave wrote on 2023-02-19, 12:37:

LOL. That was just a random piece of solder wire that had fallen there. I just didn't catch it in time for the photo, haha. :3 […]
Show full quote

Sphere478 wrote on 2023-02-18, 22:09:

What is with the wire laying on the mobo where you tap for the psu?

🤣. That was just a random piece of solder wire that had fallen there. I just didn't catch it in time for the photo, haha. :3

Well, I worked up the courage to test it... And, I got a post beep! I immediately turned it back off, cause, me'sa scared. :3

I guess that means that I just now need to hook up a monitor to it and test but hearing that post beep and not seeing any magic black smoke pop off as soon as it was turned on is an amazing sign.

I mean, without the mods, it wouldn't even post beep. It would do a vid-card-not-detected beep code, so I am definitely thinking this may actually work. I just need to hook it all back up now and make it nice. I also had an AWE64 Gold audio card that I wanted to put alongside this orpheus.

Awesome news! Fingers crossed! ^ Post card is a good idea 😀

Oh, it posted just fine, I just didn't have a monitor hooked up. I've been knee deep in coding for other things so haven't had as much time to devote to this just yet.

Next step is getting my scope on it and checking out that waveform under load. Now that it's posting, it's just a matter of seeing how the ripple pattern looks when gaming. I am super hopeful that the Nvidia 5500 card will win out though, so far so good!

Great news! :p

So it looks like it definitely boots up and everything, but it also looked like it may have some drop out after running for a short while. Here’s where the current waveform stands on the scope. Might experiment with adding a few extra caps on top of the 0603s I have.

By drop-out I mean screen turns off and you get the beep-code for missing video card upon reset. First time I ever saw a computer do that but I guess that's how it works when your video card says nopity nope. I take that it isn't filtered enough? Hmm...

Any tips/help with what settings I should turn on with the scope? Tried to the picture when it had finished doing an averaging window run (so you see some valid avg/min/max readings)

Another option is to start work on another proto-board for a pentium style cap-field for filtering. Basically just a bunch more of the small ones. A field! A field of caps to count! Uh ha ha ha! =p

Update: Okay, so definitely think I have narrowed it down to simply just heat - those LDOs produce a lot of it! It gets pretty spicey hot, so I suspect that it may be dropping out due to such. May need to design a little fan mount for it. :3

That's very interesting. It looks like it has power but there is a compliance issue with the signaling. I wonder how it even decides to reject that.

You could build the pico psu adapter and just run a wire from it. Or use a male sata pigtail. Or use calamitylime’s power supply: (see pcb projects in sig)

Again, I’ll remind everyone, the capacitors probably aren’t going to be a one size fits all, and most people probably are fine to just not use them.

Each card will have its own noise that different caps may be ideal for.

What can we do to measure for and select the appropriate capacitor value?

Kahenraz wrote on 2023-02-26, 04:49:

What can we do to measure for and select the appropriate capacitor value?

Install on desired motherboard, install desired card. Hook up to an oscilloscope and see what frequencies you are seeing your noise at.

Long story short, different capacitors will respond to different frequencies differently.

It is best to get your reads from as close to the item you are trying to stabilize as possible. Perhaps from the card even.

Here is a quote from a convo I was having with a friend of mine who understands this a little better.

There is also the try it and see approach. If you buy a capacitor assortment you could play with the values till you had the cleanest signal.

Is it not ideal to have a similar design to the ISA Voltage Blaster?

megatog615 wrote on 2023-07-17, 10:12:

Is it not ideal to have a similar design to the ISA Voltage Blaster?

Many mobos don’t route the signals

Sphere asked me to post this here:

I made a mini PCI Voltage Blaster for +3.3V. This is useful for those +5V keyed cards that need +3.3V to work. Like the "new" Chinese made ATI Rage XL PCI cards all over eBay.

As noted above, many +5V only boards do not route the +3.3V pins. So you'll need to bodge a few of the +3.3V pins to each other on the back of the mobo.

Also, this mini version is only good for about 500mA max. Won't work for PCI GPUs that need a lot of power.

This is open source: https://github.com/chadr/PCI-Voltage-Blaster

I made 10 of these up when I was testing out stuff. If anyone wants one lmk. Just cover postage and I'll send one or two out.

Very nice. :p good job! 👍

I soldered two wires to my 8400gs rev2 to the back of its pcb where the pins of two electrolytic caps connect to the 3.3V supply line of the pci bus, and connected them with a wago clamp to a 3.3V wire from one of the SATA connectors of the power supply. Worked instantly and seems stable, but I haven’t stress tested it so far. Also the wires might break off one day and damage something else. But it’s nice to see that it works for now on my asus P2B. I would suggest to make a small pcb for a pci to sata adapter. In the long term new power supplies make more sense anyways and they should all have SATA as well.

Long time no talk. I recently came back to this issue because I was determined to get the Nvidia 5500FX card I have working, and I am happy to report that I finally have achieved stable 3v3 with an AT power supply using a custom 3v3 board based on a LT1584-3.3. I used two 220uF bulk poly caps, a 10uF tantalum, 22uF tantalum, a 0.1uF ceramic, and two ferrite beads for filtering, and was able to get down to 90~130mV of ripple on the scope. Couldn't get it to work on 12v rail, but was able to get it to work on 5v rail without issue.

The Nvidia 5500FX I am happy to announce likes this setup much better, but the heatsink I am using (a 50mm lengthed one) gets really hot. Definitely need to properly cool these linear voltage regulators if you go this pathway. I definitely recommend instead soldering the wire off the ATX line rather than going to the trouble that I have with an AT power supply, alas. It's done.

Thanks for all the help, I've done a lot of studying and learning electronics to get to this point and feel like I've come full circle now that I have a working 3v3 powered PCI port.

NachtRave wrote on Yesterday, 06:26:

Long time no talk. I recently came back to this issue because I was determined to get the Nvidia 5500FX card I have working, and I am happy to report that I finally have achieved stable 3v3 with an AT power supply using a custom 3v3 board based on a LT1584-3.3. I used two 220uF bulk poly caps, a 10uF tantalum, 22uF tantalum, a 0.1uF ceramic, and two ferrite beads for filtering, and was able to get down to 90~130mV of ripple on the scope. Couldn't get it to work on 12v rail, but was able to get it to work on 5v rail without issue.

The Nvidia 5500FX I am happy to announce likes this setup much better, but the heatsink I am using (a 50mm lengthed one) gets really hot. Definitely need to properly cool these linear voltage regulators if you go this pathway. I definitely recommend instead soldering the wire off the ATX line rather than going to the trouble that I have with an AT power supply, alas. It's done.

Thanks for all the help, I've done a lot of studying and learning electronics to get to this point and feel like I've come full circle now that I have a working 3v3 powered PCI port.

I wonder if 3.3V 7A is overkill for just single PCI slot?

roytam1 wrote on Yesterday, 13:11:

I wonder if 3.3V 7A is overkill for just single PCI slot?

Not really. It's actually the same kind of amperage the entire 3v3 in a PCI bus is spec'ed out as. The one I tried before was only a 4A 3v3 regulator (or something to that effect) and it wasn't enough. Would overheat and kill itself trying to power that Nvidia 5500FX card to the point it was entirely unusable - the 7A version feels a lot more manageable. I let the issue sit for the last two years as I did more other electronics work and put it on the back-burner until I was more comfortable in my electronics knowledge to try and tackle it again.

The bigger issue now is just the heat output. I have two additional aluminum block pieces coming in the mail to extend the piece I have with what I would describe as "fins," but that piece right now gets really freakin' hot as is. Hot enough to almost burn you if you touch it, and I'm hoping I don't need to 3D print some sort of bracket to install a blower fan on it to keep it in check. Definitely want as large a thermal mass as possible when using these kinds of linear voltage regulators.

Either way, the tantalums and the ferrite beads mixed with a large bulk cap and several smaller nF ceramics is enough to get the ripple down to the 100mV range, which seems like good enough for the Nvidia 5500FX to initialize into high graphics mode with 3D acceleration. I don't want to test it for longer sessions until I have more work done around cooling the darn thing.

Also, don't use the 12v rail on an AT system to power this. Won't work. Even if the voltage regulator is supposed to be able to take 12v (and does so off a separate desk power supply), the 12v off the AT supply is too much for it. Had to use the 5v rail to get it to even go. The 12v rail only let the 3v3 regulator get up to 2.9v under load, and 1.2v under no load (1.9v with output pin lifted). Eventually gave up with the 12v rail and went to the 5v rail without any issue so far.