Scored a damaged socket 7 on ebay

[wierd_w]

I got the needful to actually power the board on earlier today.

 

Dropped the TNT2 in the PCI slot, dropped in the RAM and CPU; Gussied up the power supply with the ATX->AT converter cable, and wired up a mechanical lever switch to drive it with.

 

System powers on, but display is illegible from rainbow snow.

 

So, I rule out the video being bad, by dropping it in the P4 machine I have.  Works just fine.

 

I suspect that the blown inductor is involved in keeping the power rails clean on the motherboard. There was probably a dead short in one of the slots at some point, if I were to hazard a guess.  The snowy video happens regardless of which slot, so I am pretty sure it is power rail related, and that it is actually off the PCI bus, not the ISA bus.

 

After the 26th i have lots of time off. I will manually circuit probe the PCI slot lines against the solder pad of the blown inductor then.

[Keatah]

..Or maybe its directly, serially, involved in supplying the power. Just a thought there.

 

[wierd_w]

Possible.  I will see where it traces to.  I will include the pins on the P4 and P5 connectors in the probing.

[wierd_w]

OK, I decided to probe today.

 

The upstream end of that inductor is directly connected to the +5v rails (both 8bit and 16bit) on the ISA slot, and to all of the +5v rails on the P4/P5 connectors.

 

Clearly this inductor is involved in the power regulation going through the motherboard, and is vitally important to the health and wellbeing of any device plugged into it. Given there is basically 0ohm resistance on the P4/P5 +v5 rail connectors, it is probably safe to say it is likewise implicated on the PCI bus's +5v rail as well. (and that this is likely a bypass style noise filter)

 

Since the downstream end is against common ground and a capacitor, I once again suspect this is power noise filtration of some kind.

 

If you are there Keatah, and have a suggestion on what to put in there, I would be happy to order an appropriate discrete inductor. The verbiage I am reading suggests that an inductor in the range of 1uH to 2uH is commonly used. (this would be in line with my guess, based on the size of the original blown package)

 

If you want, I can TRY to put the cap against the LCR meter I just got in the mail...

(Leaving that tiny thing in circuit, I put the LCR meter on the 2uF setting, and got .120 as the reading on the cap against ground.)

 

Investigating to see if this was a decoupling filter (and not a bypass filter), I tested the upstream side of the inductor against ground, but the capacitance reading maxes out the meter on all settings. (it only goes to 2kuf)

Edited July 24, 2020 by wierd_w

[Keatah]

5 hours ago, wierd_w said:

OK, I decided to probe today.

 

The upstream end of that inductor is directly connected to the +5v rails (both 8bit and 16bit) on the ISA slot, and to all of the +5v rails on the P4/P5 connectors.

 

Clearly this inductor is involved in the power regulation going through the motherboard, and is vitally important to the health and wellbeing of any device plugged into it. Given there is basically 0ohm resistance on the P4/P5 +v5 rail connectors, it is probably safe to say it is likewise implicated on the PCI bus's +5v rail as well. (and that this is likely a bypass style noise filter)

In relation to what?

0 Ohms from the 5v rails to?

 

Are the rails themselves going directly to the Molex power connector's 5v pin?

 

[wierd_w]

It would seem so. There is essentially 0 resistance from the +5v pins of the P4/5 connectors either to each other, or to/from the +5v rails of the card slots. (Or from these rails to/from the upstream end of this blown inductor.)

 

Full elaboration:

 

AT Power Pin #12 to Pin #11: 0ohm

AT Power Pin #12 to Pin #10: 0ohm

AT Power Pin #12 to Pin #2: 0ohm

ISA 8bit slot pin location B29 to AT Power pin #12: 0ohm

ISA 16bit slot pin location D16 to AT power pin #12: 0ohm

ISA 8bit slot location B29 to ISA 16bit slot location D16: 0ohm

...

 

Etc.

 

The inductor in question appears to be hung on the end of this seemingly parallel arrangement, with a (inaccurately measured because still in circuit) .120uF cap tacked at the end, between it and the common ground.

 

I have both the board and my meter handy, if you want further probing done.

Edited July 24, 2020 by wierd_w

[Keatah]

So it goes like this.

 

5v ISA to inductor to capacitor to ground. Right? The capacitor is in series with the inductor?

[wierd_w]

Correct.

 

Edited July 24, 2020 by wierd_w

[Keatah]

So that means this inductor circuit's purpose is to short out a certain frequency (or band surrounding a certain frequency) and bring it to ground. All the while leaving the DC portion alone. Agree?

 

The nearby OSC pin is a 14MHz signal. But that is for some older EGA boards' dot clock. ISA cards have a BCLOCK of anywhere between 4.77MHz to 20MHz, with 8.3MHZ being typical.

 

I wonder what frequency your board resonates at? A 'scope could tell us. Then we'd calculate the value of the capacitor and inductor to come up with filter that'd short the noise to ground. And use LSpice to verify it. In practice, the circuit covers a range of frequencies with the target (most effective) being in the center. A peak.

 

You could guess the inductor might be around .8uH through 9uH. Various calculations say so. Most of these parts are at a 30% tolerance rate anyways. So your suggestion of 1-2uH is within range.

 

https://www.translatorscafe.com/unit-converter/en-US/calculator/series-rlc-impedance/

 

 

[wierd_w]

Sadly i dont have a scope.  If i did, i would remove the cpu and ram, along with the video card, power it on and do a capture off that part of the board to see what the voltage noise and current ripple look like.

 

I am not going to invest in one just to probe a 40$ board.

 

I will be clever and order female dupont connectors along with some inductors in that range, then walk to the middle by trying them on the dupont connectors i will put down, until i am happy with board stability.

[Keatah]

Sounds like a plan. If you have the ability to change bus speeds or duty cycles in BIOS, change them and test again.

[wierd_w]

Changing bus speed changes the characteristics of the snow somewhat. Still all garbled up.

 

I am gonna throw in the towel.  I soldered in the female pin headers (and tested continuity over my solder job, so I know it's good), and have tried everything in my assortment from 1.1uH to 150uH, with some improvement (if you can call it that) in the ~40-50uH range.  I would clearly need a scope to properly see what is going on here.  It is likely at this point that either the associated cap, or some other cap in this circuit is damaged, but not obviously so.

 

It's not really worth digging this far down the rabbit hole.  I will just get a different socket 7 from fleabay that has been tested as working.

[Keatah]

Going down these kinds of rabbit holes just makes for frustration and disappointment. And we don't need that.

 

There could even be issues in the chipset. Or anyplace else.

[wierd_w]

Well--

 

Since I bought and received a replacement board for this project, and had the same exact randomly icky colored snow, I decided to order a new (period Appropriate!) PCI video card, just to rule out the TNT2 having compatibility issues with older PCI bus versions.

 

Lo and behold, this presumed dead board (and the new replacement) both work JUUUUUUUSST FINE.  It was that TNT2 board not wanting to work in old vintage PCI slots!  Fancy that!

[Keatah]

On 7/31/2020 at 8:23 PM, Keatah said:

There could even be issues in the chipset. Or anyplace else.

 

I was right!!!! See!!!!

[wierd_w]

That's the thing though; the TNT2 works peachy in the pentium 4 board's PCI slot. I suspect it needs newer revisions of the PCI specification.

 

The Diamond Stealth 2000 2MB PCI  (S3 Virge DX) I popped in there is period appropriate, and works just peachy.  Doom runs quite nicely.

 

 

I am waiting on some serial headers and a serial mouse (along with a 6 wire PS/2 header and port connector, that I can move the IDC pins around in the connector to match what the manual says the pinout of the motherboard connector is.

 

I intend to 3D print an appropriate IO shield that exposes the AT keyboard hole, a serial and parallel cutout, and the PS/2 mouse hole, then attach all the header connectors to their appropriate spots. 

[DragonGrafx-16]

5v vs 3.3v PCI?

[wierd_w]

Kinda sorta.

 

(from wikipedia)

The first version of PCI found in retail desktop computers was a 32-bit bus using a 33 MHz bus clock and 5 V signalling, although the PCI 1.0 standard provided for a 64-bit variant as well. These have one locating notch in the card. Version 2.0 of the PCI standard introduced 3.3 V slots, physically distinguished by a flipped physical connector to prevent accidental insertion of 5 V cards. Universal cards, which can operate on either voltage, have two notches. Version 2.1 of the PCI standard introduced optional 66 MHz operation.

 

 

I suspect the TNT2 card is a PCI 2.1 card, using 3.3v, and driving at 66mhz.   The RAM in the card is not getting a proper reference signal from the bus, and experiencing DRAM refresh corruption. (theory.)

 

Since the card has two notches, it is a universal card, so it must have voltage regulation circuitry. I suspect that the issue is it really does not want to operate at 33mhz speeds. 

 

 

 

 

[wierd_w]

I have the socket 7 fully up and playing games now.

 

Has an SD->IDE drive hosting the main system, running win98se (which is configured to boot to DOS first. The board has USB on it, and I have a 6-in-1 card reader in the 3.5" bay in the front. Win98 gives access to that, which allows me to easily bring files over without digging inside the case.), with 128mb of RAM, a now fully functional PS/2 mouse port (had to order a connector bracket from Amazon, then fiddle with the pins inside the dupont connector so they were in the right order, but works fine now!), serial ports, and an LPT port, along with a soundblaster 16 PnP.

 

 

So.

 

Now, I have an unneeded socket 7 board.  I don't have any more RAM or another CPU to throw in, but the board does work, and has an HX chipset. (I am rocking the VX chipset that has the dropped in inductor, as it is working just fine. The older chipset boards are slower, but they have much less crap in the adapter rom region, which means better hardware UMB availability with UMBPCI.)

 

I know for a fact that the "brand new" EDO RAM you can get for KORG synthesizers works just fine in it.

 

Anyone need or want it?

Edited September 2, 2020 by wierd_w